PHYSIOLOGICAL	
  AND	
  ECOLOGICAL	
  INVESTIGATIONS	
  OF	
  CLOSTRIDIUM	
  DIFFICILE	
  
	
  

By	
  
	
  
Catherine	
  D.	
  Robinson	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
A	
  DISSERTATION	
  
	
  
Submitted	
  to	
  
Michigan	
  State	
  University	
  
in	
  partial	
  fulfillment	
  of	
  the	
  requirements	
  
for	
  the	
  degree	
  of	
  
	
  
Microbiology	
  and	
  Molecular	
  Genetics-­‐	
  Doctor	
  of	
  Philosophy	
  
	
  
2014	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  
	
  

ABSTRACT	
  
	
  
	
  
PHYSIOLOGICAL	
  AND	
  ECOLOGICAL	
  INVESTIGATIONS	
  OF	
  CLOSTRIDIUM	
  DIFFICILE	
  
	
  
By	
  
	
  
Catherine	
  D.	
  Robinson	
  
	
  
	
  
	
  
Disease	
  caused	
  by	
  Clostridium	
  difficile	
  is	
  currently	
  the	
  most	
  prevalent	
  
nosocomial	
  infection	
  and	
  leading	
  cause	
  of	
  antibiotic-­‐associated	
  diarrhea.	
  	
  It	
  is	
  clear	
  
that	
  the	
  intestinal	
  microbiota	
  plays	
  a	
  role	
  in	
  preventing	
  C.	
  difficile	
  infection	
  in	
  the	
  
absence	
  of	
  antibiotics;	
  however,	
  the	
  mechanisms	
  involved	
  in	
  this	
  protective	
  function	
  
are	
  poorly	
  understood.	
  	
  Since	
  antibiotic	
  administration	
  is	
  an	
  inducing	
  factor	
  of	
  C.	
  
difficile	
  infection,	
  treatment	
  employing	
  antibiotics	
  often	
  results	
  in	
  recurrent	
  disease,	
  
yet	
  it	
  is	
  still	
  the	
  primary	
  line	
  of	
  treatment.	
  	
  Therefore,	
  a	
  central	
  goal	
  of	
  research	
  in	
  this	
  
area	
  is	
  to	
  better	
  define	
  the	
  role	
  of	
  the	
  intestinal	
  microbiota	
  in	
  suppression	
  of	
  disease,	
  
and	
  ultimately	
  develop	
  alternative	
  ways	
  to	
  prevent	
  and	
  treat	
  C.	
  difficile	
  infection.	
  	
  	
  	
  
	
  

In	
  this	
  thesis,	
  I	
  present	
  a	
  novel	
  in	
  vitro	
  model	
  that	
  was	
  developed	
  to	
  study	
  

complex	
  fecal	
  communities.	
  	
  This	
  in	
  vitro	
  model	
  is	
  a	
  continuous-­‐culture	
  system	
  that	
  
utilizes	
  arrays	
  of	
  small-­‐volume	
  reactors;	
  it	
  is	
  unique	
  in	
  its	
  simple	
  set-­‐up	
  and	
  high	
  
replication.	
  	
  We	
  adapted	
  this	
  model	
  to	
  operate	
  as	
  a	
  C.	
  difficile	
  infection	
  model,	
  where	
  
in	
  vivo	
  C.	
  difficile	
  invasion	
  dynamics	
  are	
  replicated	
  in	
  that	
  the	
  fecal	
  communities	
  
established	
  in	
  the	
  reactors	
  are	
  resistant	
  to	
  C.	
  difficile	
  growth	
  unless	
  disrupted	
  by	
  
antibiotic	
  administration.	
  	
  We	
  then	
  go	
  on	
  to	
  use	
  this	
  model	
  to	
  show	
  that	
  newly	
  
emerged,	
  epidemic	
  strains	
  of	
  C.	
  difficile	
  have	
  a	
  competitive	
  fitness	
  advantage	
  when	
  
competed	
  against	
  non-­‐epidemic	
  strains.	
  	
  We	
  also	
  show	
  this	
  competitive	
  advantage	
  in	
  

	
  

vivo,	
  using	
  a	
  mouse	
  infection	
  model.	
  	
  This	
  result	
  is	
  exciting,	
  as	
  it	
  suggests	
  that	
  
physiological	
  attributes	
  of	
  these	
  strains,	
  aside	
  from	
  classical	
  virulence	
  factors,	
  
contribute	
  to	
  their	
  epidemic	
  phenotype.	
  	
  Finally,	
  the	
  metabolic	
  potential	
  of	
  C.	
  difficile	
  
in	
  regards	
  to	
  carbon	
  source	
  utilization	
  is	
  explored,	
  and	
  reveals	
  that	
  epidemic	
  strains	
  
are	
  able	
  to	
  grow	
  more	
  efficiently	
  on	
  trehalose,	
  a	
  disaccharide	
  sugar.	
  	
  Moreover,	
  
preliminary	
  in	
  vivo	
  mouse	
  studies	
  suggest	
  that	
  trehalose	
  utilization	
  plays	
  a	
  role	
  in	
  
colonization.	
  	
  Therefore,	
  the	
  growth	
  advantage	
  conferred	
  by	
  this	
  increased	
  ability	
  to	
  
utilize	
  trehalose	
  may	
  contribute	
  to	
  the	
  ecological	
  fitness	
  of	
  these	
  strains	
  in	
  vivo.	
  	
  	
  	
  
	
  

The	
  in	
  vitro	
  model	
  developed	
  and	
  presented	
  in	
  this	
  thesis	
  could	
  be	
  used	
  to	
  

study	
  many	
  aspects	
  of	
  C.	
  difficile-­‐microbiota	
  interactions	
  and	
  has	
  the	
  potential	
  to	
  
elucidate	
  mechanisms	
  that	
  are	
  important	
  for	
  in	
  vivo	
  resistance	
  to	
  establishment	
  of	
  
disease.	
  	
  In	
  addition,	
  the	
  metabolic	
  investigations	
  described	
  provide	
  insight	
  into	
  
understanding	
  the	
  physiology	
  of	
  not	
  only	
  C.	
  difficile	
  as	
  a	
  whole,	
  but	
  also	
  physiological	
  
attributes	
  unique	
  to	
  epidemic	
  strains.	
  	
  Ultimately,	
  these	
  types	
  of	
  ecological	
  and	
  
physiological	
  investigations	
  will	
  bring	
  us	
  closer	
  to	
  finding	
  better	
  ways	
  to	
  treat	
  and	
  
prevent	
  disease	
  caused	
  by	
  C.	
  difficile.
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

ACKNOWLEDGMENTS	
  
	
  
	
  
	
  
	
  

First	
  of	
  all	
  I	
  would	
  like	
  to	
  thank	
  my	
  advisor	
  Rob	
  Britton	
  for	
  helping	
  to	
  make	
  my	
  

experience	
  as	
  a	
  doctoral	
  student	
  so	
  successful	
  and	
  positive.	
  	
  Having	
  had	
  one	
  child	
  
coming	
  into	
  graduate	
  school	
  and	
  then	
  two	
  babies	
  during	
  my	
  PhD,	
  I	
  had	
  many	
  extra	
  
challenges	
  and	
  responsibilities	
  in	
  addition	
  to	
  those	
  of	
  a	
  graduate	
  student.	
  	
  Dr.	
  Britton	
  
was	
  amazingly	
  supportive	
  of	
  my	
  family	
  obligations	
  and	
  understanding	
  about	
  the	
  
difficulties	
  that	
  having	
  children	
  brings.	
  	
  Above	
  that,	
  his	
  encouragement,	
  support,	
  and	
  
mentorship	
  from	
  an	
  academic	
  standpoint	
  was	
  exceptional.	
  	
  	
  I	
  also	
  want	
  to	
  thank	
  the	
  
other	
  members	
  of	
  my	
  graduate	
  committee,	
  Dr.	
  Gemma	
  Reguera,	
  Dr.	
  Chris	
  Waters,	
  Dr.	
  
Terry	
  Marsh,	
  and	
  Dr.	
  Vince	
  Young	
  for	
  their	
  insight	
  and	
  guidance.	
  	
  
	
  

I	
  would	
  also	
  like	
  to	
  thank	
  all	
  of	
  the	
  Britton	
  lab	
  members,	
  both	
  past	
  and	
  

present.	
  	
  It	
  was	
  a	
  genuine	
  pleasure	
  working	
  with	
  all	
  of	
  them.	
  	
  I	
  couldn’t	
  have	
  asked	
  for	
  
a	
  more	
  collaborative	
  and	
  pleasant	
  lab	
  to	
  be	
  a	
  part	
  of.	
  	
  I	
  want	
  to	
  especially	
  thank	
  Jenny	
  
Auchtung	
  for	
  being	
  such	
  a	
  great	
  colleague	
  and	
  friend.	
  	
  She	
  was	
  like	
  a	
  second	
  mentor	
  
to	
  me	
  and	
  always	
  had	
  helpful	
  advice,	
  drawn	
  from	
  her	
  abundant	
  experiences	
  in	
  
different	
  areas	
  of	
  research	
  as	
  a	
  graduate	
  student	
  and	
  postdoc.	
  	
  She	
  was	
  always	
  willing	
  
to	
  lend	
  an	
  ear	
  for	
  venting	
  about	
  both	
  research	
  and	
  child-­‐rearing	
  frustrations	
  and	
  
offered	
  invaluable	
  support	
  and	
  encouragement.	
  	
  	
  
	
  

I	
  can’t	
  adequately	
  express	
  how	
  much	
  the	
  support	
  of	
  my	
  parents	
  has	
  meant	
  to	
  

me	
  over	
  the	
  years.	
  	
  They	
  deserve	
  all	
  of	
  the	
  credit	
  for	
  making	
  me	
  the	
  successful	
  person	
  
I	
  am	
  today.	
  	
  I	
  am	
  thankful	
  that	
  they	
  instilled	
  in	
  me	
  the	
  work	
  ethic	
  and	
  sense	
  of	
  self-­‐
pride	
  and	
  dedication	
  required	
  to	
  persevere	
  through	
  this	
  process.	
  	
  The	
  encouraging	
  

	
   iv	
  

and	
  reassuring	
  words	
  from	
  them	
  really	
  helped	
  me	
  along	
  the	
  way	
  and	
  they	
  never	
  fail	
  
to	
  express	
  how	
  proud	
  they	
  are	
  of	
  me.	
  	
  	
  
	
  

Finally,	
  I	
  want	
  to	
  thank	
  my	
  family-­‐	
  my	
  husband,	
  Dan,	
  and	
  my	
  three	
  girls,	
  Alex,	
  

Lana,	
  and	
  Veda.	
  	
  They	
  are	
  my	
  real	
  motivation.	
  	
  I	
  truly	
  don’t	
  think	
  a	
  better,	
  more	
  
supportive	
  husband	
  and	
  father	
  exists.	
  	
  He’s	
  my	
  rock	
  and	
  my	
  biggest	
  fan	
  and	
  always	
  
encourages	
  me	
  when	
  I	
  question	
  my	
  abilities.	
  	
  I	
  want	
  to	
  thank	
  him	
  for	
  believing	
  in	
  me	
  
and	
  giving	
  me	
  the	
  confidence	
  to	
  believe	
  in	
  myself.	
  	
  	
  I	
  want	
  my	
  daughter	
  Alex	
  to	
  know	
  
how	
  much	
  having	
  her	
  by	
  my	
  side	
  has	
  meant	
  to	
  me,	
  not	
  only	
  for	
  all	
  of	
  her	
  help	
  with	
  the	
  
little	
  ones,	
  but	
  how	
  she	
  supports	
  and	
  encourages	
  me	
  in	
  her	
  own	
  way.	
  	
  I	
  hope	
  that	
  I	
  am	
  
setting	
  a	
  good	
  example	
  for	
  all	
  of	
  my	
  girls	
  and	
  that	
  they	
  will	
  see	
  that	
  they	
  can	
  
accomplish	
  anything	
  they	
  set	
  their	
  minds	
  to.	
  	
  	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
   v	
  

TABLE	
  OF	
  CONTENTS	
  
	
  

	
  
	
  
LIST	
  OF	
  TABLES	
  ............................................................................................................................................x	
  	
  
	
  
LIST	
  OF	
  FIGURES	
  ........................................................................................................................................xi	
  	
  	
  
	
  	
  
CHAPTER	
  1:	
  Overview	
  of	
  Clostridium	
  difficile	
  Disease	
  and	
  Physiology……….……………..1	
  	
  
	
  
Antibiotic-­‐Associated	
  Diarrhea…………………………………………………………………....1	
  
	
  
C.	
  difficile:	
  the	
  Bacterium	
  and	
  the	
  Disease…………………..……………………………..….2	
  
	
  
	
  
C.	
  difficile	
  infection	
  cycle…………………………………………………………………...2	
  
	
  
	
  
Epidemiology,	
  disease,	
  and	
  virulence	
  factors………………………………...…...6	
  
	
  
	
  
Molecular	
  typing	
  schemes………………………………………………………………...8	
  
	
  
Emergence	
  of	
  Epidemic	
  C.	
  difficile	
  Strains………………………………………………...…..9	
  
	
  
	
  
Current	
  impact	
  on	
  the	
  health	
  care	
  system…………………..…………………....11	
  
	
  
Understanding	
  Disease	
  Development:	
  Colonization	
  Resistance……………………11	
  
	
  
	
  
Bile	
  acids	
  and	
  germination……………………………………………………….…...…12	
  
	
  
	
  
Competitive	
  exclusion………………………………………………...……………….…..13	
  
	
  
	
  
Direct	
  antagonism………………………………………………………………………….	
  14	
  
Understanding	
  C.	
  difficile	
  Physiology:	
  Metabolism……………………………….……...16	
  
Early	
  Characterization	
  of	
  C.	
  difficile	
  physiology………………………….……..18	
  
Studies	
  of	
  C.	
  difficile-­‐microbiota	
  nutrient	
  competition…………….………...19	
  
Hydrolytic	
  enzymes……………………………………………………….………………..22	
  
Proteolysis,	
  amino	
  acids,	
  and	
  stickland	
  fermentation……….……………….23	
  
“Omics”	
  approaches	
  to	
  investigating	
  metabolism….……………………….…27	
  
	
  
	
  
Genomics……………………………………………………………………………..27	
  
	
  
	
  
Transcriptomics	
  and	
  proteomics…………………………………….…….28	
  
	
  
	
  
Metabolomics…………....………………………..……………………….……….30	
  
Autotrophy………………………………………………………………………….….………31	
  
	
  
Summary………………………………………………………………………………….…….……..…..31	
  
	
  
REFERENCES………………………………………………………………………….………...……….34	
  
	
  
CHAPTER	
  2:	
  Development	
  of	
  Single	
  Chamber	
  Human	
  Gut	
  Microbiota	
  	
  	
  Mini-­‐bioreactor	
  
Arrays	
  (MBRA)……………………………………………………………………………….……………………46	
  
	
  
Abstract………………………………………………………………………………….……………...….47	
  
	
  
Introduction…………………………………………………………………………..………………….48	
  
	
  
Materials	
  and	
  Methods……………………………………………………………………………....50	
  
	
  
	
  
Design	
  of	
  mini-­‐bioreactor	
  arrays	
  (MBRA)………………………………………...50	
  
	
  
	
  
Collection	
  of	
  fecal	
  samples……………………………………………………..….…….51	
  
	
  
	
  
Preparation	
  of	
  fecal	
  samples……………………………………………………………51	
  
	
  
	
  
Media……………………………………………………………………………………..………52	
  
	
  
	
  
MBRA	
  operating	
  conditions	
  and	
  sampling………………………..………………53	
  
	
  
	
  
HPLC	
  of	
  SCFA……………………………………………………………….…………………54	
  
	
  
	
  
SCFA	
  Analysis…………………………………………………………………………………54	
  
	
  
	
  
DNA	
  extraction……………………………………………….………………………………55	
  
	
   vi	
  

	
  
	
  

	
  
	
  
	
  

	
  

	
  
	
  

Preparation	
  of	
  16S	
  rRNA	
  amplicons	
  for	
  sequencing…………………….……55	
  
Analysis	
  of	
  amplicon	
  data………………………………………………………….…….56	
  
Comparison	
  of	
  cultures	
  grown	
  in	
  BRMW,	
  BRMW10	
  and	
  BRMG	
  	
  
to	
  the	
  fecal	
  slurry……………………………………………………………..…..57	
  
Analysis	
  of	
  variation	
  in	
  cultures	
  grown	
  in	
  BRMW10	
  and	
  
comparison	
  to	
  the	
  starting	
  fecal	
  inoculum……………………………..59	
  
Results………………………………………………………………………………………………….…..59	
  
	
  
Mini-­‐bioreactor	
  design……………………………………………………………..……..59	
  
	
  
MBRA	
  operation…………………………………………………………….……………….61	
  
Short	
  chain	
  fatty	
  acid	
  profiles	
  of	
  cultures	
  grown	
  in	
  each	
  medium	
  
revealed	
  that	
  BRMW10	
  communities	
  were	
  highly	
  stable	
  and	
  most	
  
similar	
  to	
  the	
  fecal	
  inoculum………………………………………………………..…..62	
  
Comparison	
  of	
  MBRA	
  microbial	
  composition	
  and	
  structure	
  
demonstrates	
  that	
  communities	
  cultured	
  in	
  BRMW10	
  are	
  most	
  similar	
  
to	
  the	
  fecal	
  inoclum.………………………………………………………………...….......67	
  
Comparison	
  of	
  class-­‐level	
  differences	
  among	
  reactor	
  communities	
  
revealed	
  the	
  extent	
  of	
  reorganization	
  of	
  the	
  microbial	
  community	
  
during	
  culture……………………………………………………………….........................70	
  
Examining	
  the	
  microbial	
  community	
  structure	
  in	
  BRMW10	
  communities	
  
at	
  more	
  frequent	
  time	
  intervals	
  revealed	
  how	
  communities	
  diverge	
  
from	
  day-­‐to-­‐day	
  and	
  reactor-­‐to-­‐reactor…………………………………...………71	
  
Discussion……………………………………………………………………………….……………...…75	
  
Comparison	
  of	
  community	
  ecology	
  in	
  BRMW10	
  reactors	
  to	
  previously	
  
published	
  models	
  reveals	
  similar	
  trends	
  among	
  the	
  in	
  vitro	
  models….78	
  
Conclusions…………………………………………………………………………………….80	
  
Ackknowlegements………………………………………………………………………………..….80	
  
APPENDIX…………………..…………………….……………………………………………..………..81	
  
REFERENCES………………………………………………………………………………………...…..87	
  

	
  
	
  
	
  
	
  
CHAPTER	
  3:	
  	
  Epidemic	
  Clostridium	
  difficile	
  strains	
  demonstrate	
  increased	
  competitive	
  
fitness	
  over	
  non-­‐epidemic	
  isolates………………………………………………………….……………..93	
  
	
  
Abstract………………………………………………………………………………………….…………94	
  
	
  
Introduction……………………………………………………………………………….……………..95	
  
	
  
Materials	
  and	
  Methods……………………………………………………………….…….………..97	
  
	
  
	
  
Mini-­‐bioreactor	
  array	
  (MBRA)	
  design	
  and	
  operation…………………..……97	
  
	
  
	
  
Strains,	
  media,	
  and	
  growth	
  conditions……………………………………………..97	
  
Collection	
  and	
  preparation	
  of	
  fecal	
  samples	
  for	
  fecal	
  MBRA	
  	
  
experiments………………………………………………………………..……….….………98	
  
C.	
  difficile	
  invasion	
  and	
  competition	
  growth	
  studies	
  in	
  fecal	
  	
  
community	
  MBRA………………………………………………………..………………….99	
  
Quantitative	
  PCR	
  of	
  tcdA	
  gene	
  to	
  quantify	
  C.	
  difficile	
  invasion…………101	
  
Preparation	
  of	
  16S	
  rRNA	
  amplicon	
  sequencing………………………………102	
  
Processing	
  and	
  analysis	
  of	
  Sequencing	
  Data……………………….…………..103	
  
Quantitative	
  PCR	
  Analysis	
  of	
  Competition	
  Cultures	
  and	
  Calculations	
  	
  
of	
  Competitive	
  Index………………………………………………………..……………104	
  

	
  vii	
  

	
  

	
  
	
  
	
  

C.	
  difficile	
  competition	
  experiments	
  in	
  humanized	
  microbiota	
  mice	
  
(hmmice)	
  ……………………………………………………………………..……….……….106	
  
Results…………………………………………………………………………………..………………..108	
  
Fecal	
  mini-­‐bioreactors	
  (MBRA)	
  provide	
  an	
  in	
  vitro	
  model	
  to	
  study	
  C.	
  
difficile	
  invasion	
  in	
  complex	
  microbial	
  communities……………….………108	
  
MBRAs	
  support	
  complex	
  fecal	
  microbial	
  communities……………………112	
  
Bioreactor	
  community	
  composition	
  changes	
  in	
  response	
  to	
  	
  
Clindamycin	
  treatment…………………………………………..………………..……114	
  
Ribotype	
  027	
  strains	
  exhibit	
  a	
  competitive	
  advantage	
  over	
  non-­‐027	
  
strains	
  in	
  the	
  presence	
  of	
  a	
  complex	
  microbiota…………………….……….116	
  
Ribotype	
  027	
  strains	
  display	
  a	
  competitive	
  advantage	
  in	
  vivo…………121	
  
Discussion……………………………………………………………………..………………...………123	
  
	
  
Fecal	
  MBRA	
  as	
  a	
  model	
  for	
  C.	
  difficile	
  invasion…………………………..……127	
  
	
  
Conclusions…………………………………………………………………………..………129	
  
Ackknowlegements……………………………………………………….……….………….…….130	
  
APPENDIX…………………..……………….………………………………...………………………..131	
  
REFERENCES……………………………………………………………………...…………….....…..142	
  

	
  
	
  
	
  	
  
CHAPTER	
  4:	
  Differential	
  Metabolism	
  of	
  Trehalose	
  by	
  Epidemic	
  Ribotypes	
  of	
  
Clostridum	
  difficile……………………………………………………………………………………..……….149	
  
	
  
Introduction………………………………………………………………………………………..…..149	
  
	
  
Materials	
  and	
  Methods……………………………………………………………………...……..154	
  
	
  
	
  
C.	
  difficile	
  strains	
  used	
  in	
  this	
  study	
  and	
  growth	
  conditions……….……154	
  
Identifying	
  nutritional	
  compounds	
  that	
  increase	
  growth	
  of	
  Clostridium	
  
difficile.	
  ………………………………………………………………………………..…..…..155	
  
Growth	
  experiments	
  further	
  investigating	
  C.	
  difficile	
  utilization	
  of	
  
compounds	
  identified	
  in	
  Phenotype	
  MicroArray	
  plates…………….…….157	
  
Trehalose	
  growth	
  experiments………………………………….…………..………157	
  
RT-­‐qPCR	
  analysis	
  of	
  treA	
  expression………………………….……..……………158	
  
	
  
Cultures…………………………………………………………..…………………158	
  
	
  
RNA	
  extractions…………………………………………….……………………158	
  
	
  
Reverse	
  Transcription……………………………………..………………….159	
  
	
  
Real-­‐Time	
  PCR	
  reactions…………………………………………………….159	
  
Alignments	
  of	
  TreA	
  amino	
  acid	
  sequences	
  from	
  C.	
  difficile	
  clinical	
  
isolates	
  and	
  other	
  Gram-­‐positive	
  bacteria………………………………………162	
  
Construction	
  of	
  a	
  treA	
  knockout	
  mutant	
  in	
  the	
  C.	
  difficile	
  parent	
  strain,	
  
CD630Δerm…………………………………………………………………………………..162	
  
Colonization	
  and	
  competition	
  of	
  CD630Δerm	
  wild-­‐type	
  and	
  treA	
  	
  
mutant	
  strains	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection…….……………163	
  
Quantitative	
  PCR	
  Analysis	
  of	
  Competitions	
  and	
  Calculations	
  of	
  
Competitive	
  Index…………………………………………………………………………164	
  
	
  
Results	
  and	
  Discussion………………………………………………………………..…….……..166	
  
	
  
	
  
Identification	
  of	
  carbon	
  sources	
  that	
  C.	
  difficile	
  is	
  able	
  to	
  use	
  for	
  	
  
growth……………………………………………………………..…….…………………….166	
  
Compounds	
  identified	
  in	
  Phenotype	
  MicroArray	
  plates	
  increase	
  	
  
growth	
  yield	
  of	
  several	
  C.	
  difficile	
  strains………………………..………………168	
  
	
  viii	
  

Characterization	
  of	
  growth	
  phenotypes	
  of	
  C.	
  difficile	
  strains	
  grown	
  	
  
on	
  trehalose.………………………………………………..…….………………………….170	
  
Construction	
  of	
  a	
  treA	
  mutant	
  and	
  analysis	
  of	
  the	
  growth	
  	
  
phenotype…………………………………………………………………….……….....…..172	
  
Alignments	
  of	
  treR,	
  the	
  repressor	
  of	
  trehalase	
  (treA),	
  reveal	
  a	
  	
  
conserved	
  leucine	
  residue	
  that	
  is	
  substituted	
  with	
  isoleucine	
  in	
  	
  
RT	
  027	
  strains…………………………………………………….……….........................174	
  
RT	
  027	
  strains	
  have	
  increased	
  treA	
  expression…………...………………….179	
  
A	
  treA	
  knockout	
  mutant	
  of	
  C.	
  difficile	
  displays	
  a	
  decrease	
  in	
  colonization	
  
levels	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection	
  and	
  a	
  decrease	
  in	
  
competitive	
  fitness	
  compared	
  to	
  the	
  wild	
  type	
  strain……………...………181	
  
Summary	
  and	
  Future	
  Directions.…………………………………………………..…….……187	
  
Acknowlegements…………………………………………………………………………..……….189	
  
	
  
APPENDIX…………………..………………………………………….……………………...………..190	
  
	
  
REFERENCES………………………………………………………………………….……….......…..193	
  
	
  
CHAPTER	
  5:	
  Discussion	
  and	
  Conclusions………………………………………………...………..…198	
  
	
  
Discussion……………………………………………………………………………..…….…….…….198	
  
	
  
Conclusions…………………………………………………………………………..……….………..203	
  
	
  
REFERENCES……………………………………………………………………………....….……….204	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
   ix	
  

LIST	
  OF	
  TABLES	
  
	
  

	
  
	
  
Table	
  1.1.	
  	
  C.	
  difficile	
  metabolic	
  functions	
  related	
  to	
  growth…………………………….……17	
  
Table	
  2.1.	
  SCFA	
  profiles	
  for	
  fecal	
  inoculum	
  and	
  MBRA	
  cultures…………………….……....67	
  
Table	
  3.1.	
  Characterization	
  of	
  Strains	
  Used	
  in	
  this	
  Study…………………………………...….98	
  
Table	
  S3.1.	
  	
  Competitive	
  indices	
  of	
  ribotype	
  027	
  strains	
  at	
  selected	
  time	
  	
  
points	
  after	
  C.	
  difficile	
  inoculation	
  as	
  determined	
  by	
  quantitative	
  PCR……………..……134	
  
	
  
Table	
  S3.2.	
  Primers	
  used	
  for	
  qPCR………………………………………………………………….….135	
  
Table	
  4.1.	
  Characterization	
  of	
  strains	
  used	
  in	
  this	
  study……………………………….……..154	
  
Table	
  4.2.	
  	
  Defined	
  medium	
  ingredients	
  and	
  concentrations…………………………….…156	
  
Table	
  4.3.	
  	
  Primers	
  used	
  in	
  this	
  study…………………………………………………………………161	
  
Table	
  4.4.	
  Compounds	
  that	
  conferred	
  at	
  least	
  a	
  1.5-­‐fold	
  growth	
  yield	
  	
  
advantage	
  in	
  Biolog	
  PM1	
  and	
  PM2	
  plates……………………………………………………….……167	
  
	
  
Table	
  S4.1.	
  Compounds	
  that	
  conferred	
  at	
  least	
  a	
  1.5-­‐fold	
  growth	
  yield	
  	
  
advantage	
  (relative	
  to	
  unsupplemented	
  medium	
  control)	
  for	
  either	
  one	
  
or	
  both	
  strains	
  in	
  the	
  Biolog	
  PM3-­‐8………………………………………………………………….….191	
  
	
  	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
   x	
  

LIST	
  OF	
  FIGURES	
  
	
  
	
  
	
  
Figure	
  1.1.	
  Clostridium	
  difficile	
  infection	
  cycle………………………………………………..……….4	
  
	
  
Figure	
  2.1.	
  Schematic	
  of	
  MBRA	
  design	
  and	
  set-­‐up…………………………………………………61	
  
	
  
Figure	
  2.2.	
  Short	
  Chain	
  Fatty	
  Acid	
  analysis	
  of	
  bioreactor	
  cutures………………………..…64	
  
	
  
Figure	
  2.3.	
  Pearson	
  Product	
  Moment	
  Correlation	
  Coefficient	
  of	
  the	
  SCFA	
  	
  
composition	
  in	
  bioreactor	
  cultures……………………………………………………………………….65	
  
	
  
Figure	
  2.4.	
  	
  Microbial	
  ecology	
  of	
  bioreactors	
  compared	
  to	
  the	
  fecal	
  inoculum………..69	
  
	
  
Figure	
  2.5.	
  Comparison	
  of	
  the	
  class-­‐level	
  distribution	
  of	
  microbes	
  in	
  	
  
bioreactors	
  and	
  the	
  fecal	
  inoculum………………………………………………………………………..71	
  
	
  
Figure	
  2.6.	
  Analysis	
  of	
  changes	
  in	
  community	
  structure	
  across	
  time	
  in	
  	
  
replicate	
  reactors…………………………………………………………………………………………………73	
  
	
  
Figure	
  2.7.	
  	
  Pairwise	
  analyses	
  of	
  changes	
  in	
  comminty	
  structure	
  across	
  time	
  	
  
between	
  reactors…………………………………………………………………………………………..….….74	
  
	
  
Figure	
  S2.1.	
  	
  SCFA	
  concentrations	
  in	
  bioreactor	
  cultures………………………………..…..…82	
  
	
  
Figure	
  S2.2.	
  Pearson	
  Product	
  Moment	
  Correlation	
  Coefficient	
  of	
  the	
  SCFA	
  	
  
profiles…………………………………………………………………………………………………..…………….83	
  
	
  
Figure	
  S2.3.	
  	
  Shared	
  community	
  structure	
  of	
  bioreactor	
  cultures…………………..………84	
  
	
  
Figure	
  S2.4.	
  	
  Class-­‐level	
  community	
  analysis	
  of	
  replicate	
  reactors	
  in	
  different	
  	
  
media.	
  …………………………………………………………………………………………………………...…….85	
  	
  
	
  
Figure	
  3.1.	
  	
  An	
  example	
  of	
  a	
  minibioreactor	
  array	
  (MBRA)	
  used	
  for	
  cultivation	
  	
  
of	
  fecal	
  microbial	
  communities..………………………………………………………………………….109	
  
	
  
Figure	
  3.2.	
  Fecal	
  bioreactor	
  communities	
  prevent	
  invasion	
  by	
  C.	
  difficile	
  	
  
unless	
  disrupted	
  by	
  treatment	
  with	
  clindamycin…………………………………………………110	
  
	
  
Figure	
  3.3.	
  C.	
  difficile	
  proliferation	
  was	
  assayed	
  in	
  fecal	
  bioreactors	
  with	
  	
  
different	
  levels	
  of	
  inoculum	
  and	
  in	
  pure	
  culture	
  under	
  the	
  continuous-­‐culture	
  
conditions	
  used	
  for	
  bioreactors………………………………………………………………..…………111	
  
	
  
Figure	
  3.4.	
  Comparison	
  of	
  the	
  community	
  structure	
  between	
  fecal	
  samples,	
  	
  
mock-­‐treated	
  and	
  clindamycin-­‐treated	
  reactors…………………………………………..………113	
  
	
   xi	
  

	
  
Figure	
  3.5.	
  Community	
  structure	
  changes	
  in	
  response	
  to	
  	
  
clindamycin-­‐treatment………………………………………………………………………………….……116	
  
	
  
Figure	
  3.6.	
  	
  Competitive	
  indices	
  of	
  ribotype	
  027	
  strains	
  relative	
  to	
  non-­‐027	
  	
  
strains	
  in	
  the	
  presence	
  of	
  MBRA	
  fecal	
  communities………………………………………..……119	
  
	
  
Figure	
  3.7.	
  	
  Competitive	
  indices	
  of	
  ribotype	
  027	
  strains	
  relative	
  to	
  non-­‐027	
  	
  
strains	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection………………………………………………...…122	
  
	
  
Figure	
  S3.1.	
  Bacterial	
  abundance	
  does	
  not	
  change	
  significantly	
  in	
  clindamycin-­‐
treated	
  reactors……………………………………………………………………………………………….…136	
  
	
  
Figure	
  S3.2.	
  	
  PCR	
  screen	
  of	
  DNA	
  samples	
  from	
  88	
  strains	
  of	
  C.	
  difficile	
  for	
  	
  
detection	
  of	
  insert	
  containing	
  thyA	
  or	
  the	
  uninterrupted	
  thyX………………………………137	
  
	
  
Figure	
  S3.3.	
  	
  Ratios	
  of	
  ribotype	
  027:non-­‐027	
  C.	
  difficile	
  strains	
  over	
  time	
  	
  
in	
  MBRA	
  competitions………………………………………………………………………………………..138	
  
	
  
Figure	
  S3.4.	
  	
  Comparison	
  of	
  the	
  community	
  structure	
  on	
  day	
  7	
  from	
  	
  
clindamycin-­‐treated	
  reactors	
  used	
  for	
  C.	
  difficile	
  competition	
  experiments	
  	
  
to	
  triplicate	
  mock-­‐treated	
  and	
  clindamycin-­‐treated	
  reactors	
  infected	
  with	
  	
  
CD2015…………………………………………………………………………………………………………..…139	
  
	
  
Figure	
  S3.5.	
  Similar	
  community	
  structure	
  changes	
  were	
  observed	
  in	
  response	
  	
  
to	
  clindamycin-­‐treatment	
  in	
  competition	
  bioreactor	
  communities………………………140	
  
	
  
Figure	
  S3.6.	
  	
  	
  Levels	
  of	
  C.	
  difficile	
  strains	
  across	
  time	
  in	
  mouse	
  model	
  of	
  	
  
infection	
  as	
  determined	
  by	
  plating	
  from	
  fecal	
  pellets……………………………………..……141	
  
	
  
Figure	
  S3.7.	
  	
  Competitive	
  indices	
  (CI)	
  of	
  two	
  competition	
  pairs	
  of	
  ribotype	
  027	
  	
  
and	
  non-­‐027	
  C.	
  difficile	
  strains	
  in	
  the	
  MBRA	
  (circles)	
  and	
  Mouse	
  (triangles)	
  	
  
models………………………………………………………………………………………………………………141	
  
	
  
Figure	
  4.1.	
  Results	
  of	
  growth	
  yield	
  experiments	
  using	
  compounds	
  selected	
  	
  
from	
  Phenotype	
  MicroArray	
  plates………………………………………………………………..……170	
  
	
  
Figure	
  4.2.	
  Maximum	
  growth	
  yield	
  of	
  C.	
  difficile	
  strains	
  grown	
  in	
  the	
  presence	
  	
  
of	
  a	
  range	
  of	
  trehalose	
  concentrations	
  in	
  defined	
  medium…………………………...………171	
  
	
  
Figure	
  	
  4.3.	
  Growth	
  yield	
  of	
  C.	
  difficile	
  strains	
  belonging	
  to	
  several	
  ribotype	
  	
  
Groups…………………………………………………………………………………………………………….…172	
  
	
  
Figure	
  4.4.	
  	
  	
  Growth	
  yields	
  of	
  CD630	
  wild-­‐type	
  (WT)	
  and	
  treA	
  knock-­‐out	
  mutant	
  	
  
in	
  defined	
  medium	
  (DM)	
  supplemented	
  with	
  glucose	
  and	
  glucose	
  disaccharide	
  
sugars	
  (25mM)…………………………………………………………………………………………….…….174	
  
	
  xii	
  

	
  
Figure	
  4.5.	
  	
  Trehalose	
  utilization	
  genes	
  located	
  on	
  the	
  C.	
  difficile	
  chromosome……174	
  
	
  
Figure	
  4.6.	
  Alignments	
  of	
  TreR	
  amino	
  acid	
  sequences	
  for	
  several	
  C.	
  difficile	
  	
  
strains	
  of	
  various	
  ribotypes………………………………………………………………..……………….175	
  
	
  
Figure	
  4.7.	
  Alignments	
  of	
  TreR	
  amino	
  acid	
  sequences	
  from	
  several	
  Gram-­‐	
  
positive	
  bacterial	
  organisms	
  showing	
  conservation	
  of	
  the	
  leucine	
  residue……………176	
  
	
  
Figure	
  4.8.	
  	
  	
  Ribbon	
  diagram	
  of	
  the	
  C-­‐terminal	
  (effector-­‐binding	
  domain)	
  of	
  	
  
TreR	
  from	
  Bacillus	
  subtilis	
  showing	
  the	
  locations	
  of	
  the	
  predicted	
  trehalose-­‐	
  
6-­‐phosephate	
  binding	
  pocket	
  and	
  conserved	
  leucine	
  residue	
  (Leu-­‐169)…………..…..177	
  
	
  
Figure	
  4.9.	
  Expression	
  of	
  treA	
  in	
  several	
  C.	
  difficile	
  strains	
  grown	
  in	
  	
  
DM+25mM	
  trehalose	
  as	
  determined	
  by	
  RT-­‐qPCR…………………………………………….…..181	
  
	
  
Figure	
  4.10.	
  	
  CFU/g	
  feces	
  of	
  wild-­‐type	
  and	
  treA	
  mutant	
  C.	
  difficile	
  infected	
  	
  
Mice……………………………………………………………………………………………………………..……183	
  
	
  
Figure	
  4.11.	
  	
  Competitive	
  indices	
  of	
  wild-­‐type	
  CD630Δerm	
  when	
  competed	
  	
  
against	
  the	
  treA	
  knockout	
  mutant	
  in	
  a	
  conventional	
  mouse	
  model	
  of	
  C.	
  diffiicile	
  
infection………………………………………………………………………………………………….…………186	
  
	
  
Figure	
  4.12.	
  	
  Total	
  CFU/g	
  feces	
  of	
  C.	
  difficile	
  in	
  each	
  group	
  of	
  mice……………..………..186	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  xiii	
  

CHAPTER	
  1	
  
	
  
	
  
	
  

Overview	
  of	
  Clostridium	
  difficile	
  Disease	
  and	
  Physiology	
  
	
  

	
  

	
  
Antibiotic-­‐Associated	
  Diarrhea	
  
	
  
The	
  advent	
  of	
  the	
  antibiotics	
  era	
  in	
  the	
  1940’s	
  is	
  one	
  of	
  the	
  most	
  notable	
  milestones	
  
in	
  medical	
  science	
  history.	
  	
  It	
  enabled	
  the	
  treatment	
  of	
  numerous	
  bacterial	
  infections	
  
that	
  were	
  previously	
  fatal.	
  	
  	
  Unfortunately,	
  the	
  use	
  of	
  antibiotics	
  brought	
  along	
  with	
  it	
  
several	
  unforeseen	
  complications.	
  	
  	
  Perhaps	
  the	
  most	
  significant	
  of	
  these	
  is	
  the	
  
development	
  of	
  antibiotic	
  resistance	
  (1).	
  	
  Not	
  far	
  behind	
  it,	
  however,	
  is	
  the	
  common	
  
side	
  effect	
  of	
  antibiotic	
  administration,	
  antibiotic-­‐associated	
  diarrhea	
  (AAD).	
  	
  Often	
  
the	
  cause	
  of	
  AAD	
  is	
  unknown,	
  but	
  may	
  be	
  due	
  to	
  several	
  different	
  etiologies.	
  	
  	
  The	
  
antibiotic	
  compound	
  itself	
  can	
  induce	
  allergic	
  or	
  toxic	
  interactions	
  with	
  the	
  intestinal	
  
mucosa,	
  or	
  general	
  physiological	
  effects	
  on	
  gut	
  motility	
  (2,	
  3).	
  	
  Even	
  more	
  significant,	
  
however,	
  are	
  the	
  implications	
  of	
  the	
  impacts	
  that	
  antimicrobials	
  have	
  on	
  the	
  
intestinal	
  microbiota.	
  	
  	
  	
  	
  
	
  

The	
  human	
  gastrointestinal	
  tract	
  is	
  home	
  to	
  trillions	
  of	
  bacteria,	
  which	
  can	
  be	
  

classified	
  into	
  hundreds	
  of	
  different	
  species	
  (4).	
  	
  This	
  large	
  consortium	
  of	
  organisms	
  
plays	
  a	
  fundamental	
  role	
  in	
  many	
  different	
  aspects	
  of	
  our	
  health	
  and	
  physiology,	
  more	
  
and	
  more	
  of	
  which	
  are	
  being	
  elucidated	
  all	
  the	
  time	
  (reviewed	
  in	
  (5)).	
  	
  One	
  important	
  
function	
  of	
  the	
  intestinal	
  microbiota	
  is	
  to	
  protect	
  us	
  from	
  intestinal	
  pathogens,	
  which	
  
occurs	
  by	
  many	
  different	
  mechanisms	
  including	
  modulation	
  of	
  the	
  host	
  immune	
  
response,	
  production	
  of	
  inhibitory	
  compounds,	
  and	
  competition	
  for	
  limited	
  nutrients	
  

	
  

1	
  

or	
  binding	
  sites,	
  also	
  known	
  as	
  colonization	
  resistance	
  (6).	
  	
  Antibiotics,	
  particularly	
  
broad-­‐spectrum	
  antibiotics,	
  not	
  only	
  target	
  the	
  desired	
  pathogen	
  but	
  also	
  inflict	
  
collateral	
  damage	
  to	
  the	
  intestinal	
  microbiota,	
  thereby	
  compromising	
  these	
  
protective	
  functions	
  and	
  allowing	
  growth	
  of	
  pathogenic	
  strains	
  (7).	
  	
  	
  	
  Several	
  
pathogens	
  have	
  been	
  implicated	
  in	
  AAD,	
  including	
  Salmonella	
  spp.,	
  Clostridium	
  
perfringens,	
  Staphylococcus	
  aureus,	
  Klebsiella	
  oxytoca,	
  Clostridium	
  difficile,	
  and	
  
Candida	
  albicans	
  (2,	
  8).	
  	
  	
  	
  The	
  most	
  common	
  of	
  these	
  is	
  C.	
  difficile,	
  being	
  the	
  causative	
  
agent	
  in	
  as	
  many	
  as	
  25%	
  of	
  AAD	
  cases	
  (9).	
  	
  Moreover,	
  C.	
  difficile	
  is	
  responsible	
  for	
  
~90%	
  of	
  cases	
  of	
  pseudomembranous	
  colitis,	
  a	
  more	
  severe	
  malady	
  of	
  antibiotic-­‐
associated	
  disease	
  (10).	
  	
  	
  	
  
	
  
C.	
  difficile:	
  the	
  Bacterium	
  and	
  the	
  Disease	
  
C.	
  difficile	
  is	
  a	
  Gram-­‐positive,	
  spore	
  forming,	
  anaerobic,	
  motile	
  rod.	
  	
  	
  Hall	
  and	
  O’Toole	
  
first	
  isolated	
  and	
  described	
  this	
  bacterium	
  in	
  1935	
  while	
  characterizing	
  bacterial	
  
intestinal	
  colonization	
  of	
  healthy	
  new-­‐born	
  infants	
  (11).	
  	
  	
  While	
  there	
  is	
  a	
  high	
  
carriage	
  rate	
  of	
  C.	
  difficile	
  in	
  infants,	
  as	
  the	
  intestinal	
  microbiota	
  matures,	
  C.	
  difficile	
  is	
  
quickly	
  replaced	
  by	
  other	
  strains	
  (12).	
  	
  	
  This	
  process	
  of	
  ecological	
  succession	
  has	
  
been	
  demonstrated	
  in	
  both	
  mice	
  and	
  humans	
  (13-­‐16).	
  	
  Consequently,	
  asymptomatic	
  
carriage	
  of	
  C.	
  difficile	
  in	
  adults	
  is	
  very	
  low	
  (17).	
  	
  	
  	
  
	
  
C.	
  difficile	
  infection	
  cycle.	
  	
  The	
  cycle	
  of	
  infection	
  for	
  C.	
  difficile	
  is	
  unique	
  compared	
  to	
  
other	
  intestinal	
  pathogens	
  due	
  to	
  its	
  obligate	
  anaerobic	
  physiology	
  and	
  complex	
  
interactions	
  with	
  the	
  host	
  microbiota.	
  	
  A	
  simplified	
  overview	
  is	
  presented	
  in	
  Figure	
  

	
  

2	
  

1.1.	
  	
  In	
  general,	
  a	
  perturbation	
  of	
  the	
  intestinal	
  microbiota	
  is	
  required	
  for	
  disease	
  
development,	
  usually	
  by	
  administration	
  of	
  broad-­‐spectrum	
  antibiotics.	
  	
  Therefore,	
  the	
  
intestinal	
  microbiota	
  provides	
  resistance	
  to	
  infection	
  in	
  healthy	
  individuals;	
  functions	
  
of	
  an	
  unperturbed	
  intestinal	
  microbiota	
  that	
  provide	
  resistance	
  to	
  C.	
  difficile	
  infection	
  
will	
  be	
  reviewed	
  below.	
  	
  	
  Since	
  C.	
  difficile	
  is	
  an	
  anaerobic	
  organism,	
  it	
  can	
  only	
  pass	
  
from	
  host	
  to	
  host	
  in	
  the	
  spore	
  form.	
  	
  Therefore,	
  germination	
  and	
  outgrowth	
  are	
  
required	
  for	
  toxin	
  production	
  and	
  establishment	
  of	
  disease.	
  	
  However,	
  it	
  is	
  poorly	
  
understood	
  which	
  of	
  these	
  key	
  steps,	
  germination	
  or	
  outgrowth,	
  are	
  most	
  important	
  
for	
  initiation	
  of	
  C.	
  difficile	
  expansion	
  in	
  the	
  gastrointestinal	
  tract.	
  	
  	
  Once	
  vegetative	
  
growth	
  ensues,	
  exotoxins	
  are	
  produced	
  and	
  cause	
  damage	
  to	
  the	
  intestinal	
  
epithelium,	
  eliciting	
  an	
  inflammatory	
  response	
  and	
  establishment	
  of	
  disease.	
  	
  Spores	
  
are	
  produced	
  during	
  the	
  course	
  of	
  infection,	
  although	
  it	
  is	
  also	
  unknown	
  what	
  the	
  
inducing	
  factors	
  are	
  for	
  sporulation.	
  	
  Nonetheless,	
  formation	
  of	
  spores	
  is	
  essential	
  for	
  
host	
  transmission,	
  since	
  it	
  requires	
  passage	
  through	
  the	
  aerobic	
  environment.	
  	
  Spores	
  
are	
  shed	
  through	
  fecal	
  material,	
  facilitated	
  by	
  the	
  development	
  of	
  diarrhea.	
  	
  They	
  can	
  
survive	
  in	
  the	
  environment	
  for	
  very	
  long	
  time	
  periods	
  due	
  to	
  their	
  inherent	
  resistance	
  
to	
  environmental	
  conditions	
  such	
  as	
  desiccation,	
  UV-­‐light	
  damage,	
  chemicals,	
  and	
  
extreme	
  temperatures.	
  	
  In	
  healthcare	
  settings,	
  environmental	
  contamination	
  of	
  
spores	
  is	
  a	
  major	
  issue	
  related	
  to	
  C.	
  difficile	
  infection	
  control.	
  	
  Spores	
  present	
  on	
  
patient	
  bedding	
  resist	
  removal	
  during	
  laundering,	
  are	
  recalcitrant	
  to	
  alcohol-­‐based	
  
hand	
  sanitizers,	
  and	
  have	
  been	
  shown	
  to	
  spread	
  through	
  airborne	
  dissemination	
  (18-­‐
20).	
  	
  	
  Moreover,	
  spores	
  are	
  often	
  transferred	
  from	
  patient	
  to	
  patient	
  on	
  the	
  hands	
  of	
  

	
  

3	
  

healthcare	
  workers	
  (21).	
  	
  	
  Ingestion	
  of	
  spores	
  by	
  another	
  patient	
  with	
  reduced	
  
intestinal	
  resistance	
  continues	
  the	
  cycle.	
  	
  	
  	
  
	
  

	
  
Figure	
  1.1.	
  	
  Clostridium	
  difficile	
  infection	
  cycle.	
  	
  	
  Changes	
  to	
  the	
  intestinal	
  
environment	
  induced	
  by	
  antibiotic	
  treatment	
  present	
  conditions	
  that	
  allow	
  C.	
  difficile	
  
spores	
  to	
  germinate	
  and	
  outgrow.	
  	
  Production	
  of	
  toxins	
  incites	
  intestinal	
  damage	
  and	
  
development	
  of	
  disease.	
  	
  Undefined	
  factors	
  induce	
  sporulation;	
  spores	
  are	
  shed	
  from	
  
the	
  host,	
  contaminating	
  the	
  environment.	
  	
  Spores	
  are	
  transmitted	
  to	
  new	
  hosts,	
  and	
  
the	
  cycle	
  continues.	
  	
  PMC=	
  pseudomembranous	
  colitis.	
  	
  
	
  
	
  
	
  
While	
  CDI	
  is	
  primarily	
  a	
  nosocomial	
  infection,	
  the	
  recent	
  increases	
  in	
  
community-­‐acquired	
  infections	
  offer	
  the	
  possibility	
  of	
  an	
  alternative	
  reservoir	
  of	
  C.	
  
difficile.	
  	
  Several	
  potential	
  sources	
  of	
  community-­‐acquired	
  C.	
  difficile	
  transmission	
  
have	
  been	
  proposed.	
  	
  Recent	
  work	
  has	
  shown	
  both	
  animals	
  and	
  food	
  as	
  potential	
  
reservoirs	
  in	
  the	
  community	
  (22).	
  	
  Another	
  report	
  suggests	
  that	
  the	
  high	
  rate	
  of	
  

	
  

4	
  

carriage	
  in	
  infants	
  is	
  a	
  contributor	
  to	
  C.	
  difficile	
  community-­‐acquired	
  infections	
  (23).	
  	
  	
  
In	
  addition,	
  a	
  recent	
  study	
  surveyed	
  30	
  houses	
  in	
  an	
  urban	
  city,	
  and	
  showed	
  a	
  high	
  
degree	
  of	
  environmental	
  contamination	
  within	
  the	
  households	
  (24).	
  	
  These	
  are	
  all	
  
valid	
  models	
  for	
  reservoirs	
  of	
  C.	
  difficile	
  outside	
  of	
  health-­‐care	
  environments,	
  and	
  it	
  is	
  
likely	
  that	
  there	
  are	
  multiple	
  contributors	
  to	
  community-­‐acquired	
  spread	
  of	
  C.	
  
difficile.	
  	
  Nonetheless,	
  the	
  basis	
  for	
  the	
  recent	
  increases	
  in	
  community-­‐acquired	
  CDI	
  is	
  
undoubtedly	
  multifactorial,	
  and	
  more	
  work	
  needs	
  to	
  be	
  done	
  in	
  this	
  area	
  to	
  find	
  ways	
  
of	
  preventing	
  these	
  infections.	
  	
  
	
  

A	
  major	
  complication	
  of	
  CDI	
  is	
  the	
  high	
  rate	
  of	
  recurrent	
  disease.	
  	
  As	
  many	
  as	
  

20%	
  of	
  patients	
  will	
  have	
  a	
  recurrent	
  infection,	
  and	
  45%	
  of	
  patients	
  who	
  have	
  had	
  a	
  
second	
  infection	
  will	
  have	
  subsequent	
  infections	
  (25).	
  	
  Some	
  patients	
  have	
  multiple	
  
recurrent	
  infections	
  over	
  several	
  months	
  or	
  even	
  years.	
  	
  	
  The	
  paradigm	
  for	
  recurrent	
  
disease	
  is	
  based	
  on	
  failure	
  of	
  the	
  intestinal	
  microbiota	
  to	
  reestablish	
  appropriately	
  
and	
  restore	
  resistance	
  to	
  C.	
  difficile.	
  	
  Recurrence	
  is	
  the	
  result	
  of	
  either	
  relapse	
  due	
  to	
  
reinfection	
  of	
  the	
  initial	
  infecting	
  strain,	
  or	
  infection	
  by	
  a	
  newly	
  acquired	
  strain	
  (26,	
  
27).	
  	
  Multiple	
  studies	
  have	
  shown	
  that	
  the	
  structure	
  and	
  diversity	
  of	
  the	
  intestinal	
  
microbiota	
  of	
  patients	
  with	
  recurrent	
  disease	
  is	
  significantly	
  different	
  from	
  those	
  who	
  
do	
  not	
  (17,	
  28).	
  	
  Indeed,	
  we	
  know	
  that	
  antibiotic	
  treatment	
  can	
  have	
  both	
  short	
  and	
  
long-­‐term	
  impacts	
  on	
  the	
  intestinal	
  microbiome	
  (29-­‐32).	
  	
  Since	
  antibiotics	
  are	
  
typically	
  the	
  first	
  line	
  of	
  treatment	
  for	
  CDI,	
  this	
  can	
  serve	
  to	
  only	
  exacerbate	
  the	
  
problem	
  by	
  further	
  perturbing	
  intestinal	
  communities,	
  thereby	
  increasing	
  the	
  risk	
  of	
  
recurrent	
  disease.	
  	
  Therefore,	
  much	
  work	
  is	
  being	
  done	
  to	
  try	
  to	
  find	
  new	
  ways	
  of	
  
treating	
  CDI	
  in	
  order	
  to	
  circumvent	
  this	
  problem	
  and	
  reduce	
  recurrence.	
  	
  	
  

	
  

5	
  

	
  
Epidemiology,	
  disease,	
  and	
  virulence	
  factors.	
  	
  The	
  association	
  between	
  C.	
  difficile	
  
and	
  antibiotic-­‐associated	
  disease	
  was	
  not	
  elucidated	
  until	
  1978,	
  when	
  Bartlett	
  et	
  al.	
  
showed	
  that	
  it	
  is	
  the	
  primary	
  causative	
  agent	
  of	
  pseudomembranous	
  colitis	
  (33).	
  	
  	
  
Since	
  then,	
  much	
  work	
  has	
  been	
  done	
  to	
  investigate	
  the	
  epidemiology	
  and	
  
pathogenesis	
  of	
  this	
  organism.	
  	
  The	
  three	
  major	
  risk	
  factors	
  for	
  disease	
  caused	
  by	
  C.	
  
difficile	
  infection	
  (CDI),	
  are	
  antibiotic	
  treatment,	
  advanced	
  age	
  (>65)	
  and	
  exposure	
  to	
  
a	
  hospital	
  setting	
  (34).	
  	
  Any	
  broad-­‐spectrum	
  antibiotic	
  has	
  the	
  potential	
  to	
  initiate	
  
CDI;	
  however,	
  some	
  have	
  higher	
  rates	
  of	
  triggering	
  CDI	
  than	
  others.	
  	
  Clindamycin	
  was	
  
the	
  major	
  contributing	
  antibiotic	
  initially,	
  followed	
  several	
  years	
  later	
  by	
  those	
  in	
  the	
  
cephalosporin	
  group	
  (10).	
  	
  Most	
  recently,	
  fluoroquinolones	
  have	
  been	
  added	
  to	
  this	
  
list,	
  due	
  to	
  the	
  emergence	
  of	
  fluoroquinalone-­‐resistant	
  hypervirulent	
  strains	
  (35).	
  	
  In	
  
addition	
  to	
  the	
  elderly,	
  high	
  risk	
  groups	
  also	
  include	
  immuno-­‐compromised	
  
individuals,	
  patients	
  who	
  have	
  had	
  gastrointestinal	
  surgery,	
  and	
  persons	
  taking	
  
gastrointestinal	
  medications	
  including	
  antiperistaltic	
  drugs	
  and	
  proton	
  pump	
  
inhibitors	
  (36).	
  	
  Although	
  community-­‐acquired	
  cases	
  are	
  rare,	
  the	
  incidence	
  of	
  these	
  
infections	
  have	
  increased	
  in	
  recent	
  years	
  (37).	
  	
  	
  
The	
  symptoms	
  of	
  CDI	
  range	
  from	
  mild,	
  self-­‐limiting	
  diarrhea	
  to	
  severe	
  colonic	
  
inflammation	
  and	
  death.	
  	
  In	
  severe	
  cases,	
  CDI	
  is	
  accompanied	
  by	
  the	
  formation	
  of	
  
pseudomembranes,	
  termed	
  pseudomembranous	
  colitis	
  (PMC).	
  	
  Pseudomembranes	
  
are	
  characterized	
  as	
  regions	
  of	
  the	
  mucosa	
  that	
  become	
  inflamed	
  and	
  covered	
  in	
  
yellowish	
  plaques,	
  composed	
  of	
  cellular	
  debris	
  (dead	
  leukocytes	
  and	
  mucosal	
  cells),	
  
fibrin,	
  and	
  mucin	
  (38).	
  	
  About	
  10%	
  of	
  patients	
  with	
  CDAD	
  will	
  develop	
  PMC	
  (39).	
  	
  

	
  

6	
  

Toxic	
  megacolon,	
  a	
  condition	
  in	
  which	
  complete	
  loss	
  of	
  colonic	
  function	
  can	
  occur,	
  
has	
  associated	
  mortality	
  rates	
  as	
  high	
  as	
  80%	
  (9).	
  	
  
	
  

The	
  primary	
  virulence	
  factors	
  recognized	
  in	
  this	
  organism	
  are	
  toxins,	
  TcdA	
  

and	
  TcdB.	
  	
  Not	
  all	
  C.	
  difficile	
  strains	
  are	
  toxigenic;	
  however,	
  only	
  toxin-­‐producing	
  
strains	
  cause	
  disease.	
  	
  The	
  genes	
  tcdA,	
  tcdB,	
  and	
  associated	
  regulatory	
  genes	
  are	
  
located	
  on	
  a	
  19.6	
  kb	
  pathogenicity	
  locus	
  (PaLoc)	
  on	
  the	
  chromosome	
  of	
  C.	
  difficile	
  
(40).	
  	
  	
  Additional	
  genes	
  include	
  tcdR	
  (a	
  positive	
  regulator;	
  sigma	
  factor),	
  tcdC	
  (a	
  
negative	
  regulator)	
  and	
  tcdE	
  (a	
  holin-­‐like	
  protein).	
  	
  These	
  large	
  toxin	
  proteins	
  (308	
  
kDa	
  and	
  270	
  kDa,	
  respectively)	
  are	
  glucosyltransferases;	
  they	
  inactivate	
  members	
  of	
  
the	
  Ras	
  superfamily	
  of	
  small	
  GTPases.	
  	
  This	
  inactivation	
  has	
  a	
  cytotoxic	
  effect	
  on	
  
target	
  cells	
  by	
  disrupting	
  intracellular	
  signaling	
  pathways	
  involved	
  in	
  interactions	
  of	
  
the	
  actin	
  cytoskeleton	
  and	
  initiation	
  of	
  apoptosis,	
  and	
  induces	
  inflammation	
  (41).	
  	
  	
  
Some	
  strains	
  produce	
  an	
  additional	
  binary	
  toxin,	
  CDT.	
  	
  	
  Although	
  its	
  role	
  and	
  
significance	
  in	
  CDI	
  is	
  not	
  fully	
  understood,	
  there	
  is	
  evidence	
  that	
  it	
  may	
  be	
  associated	
  
with	
  increased	
  mortality	
  (42).	
  	
  
	
  

Several	
  non-­‐toxin	
  virulence	
  factors	
  of	
  C.	
  difficile	
  have	
  been	
  studied.	
  	
  	
  Adhesins	
  

are	
  bacterial	
  cell-­‐surface	
  molecules	
  that	
  facilitate	
  binding	
  to	
  host	
  tissues,	
  which	
  in	
  the	
  
case	
  of	
  C.	
  difficile	
  are	
  primarily	
  mucosal	
  surfaces.	
  	
  These	
  include	
  proteins	
  involved	
  in	
  
intestinal	
  adherence	
  by	
  binding	
  to	
  fibronectin,	
  collagen,	
  fibrinogen,	
  and	
  other	
  
components	
  of	
  mucin	
  or	
  intestinal	
  epithelia.	
  	
  In	
  a	
  recent	
  review	
  by	
  Vendantam	
  et	
  al.,	
  a	
  
literature	
  search	
  resulted	
  in	
  a	
  substantial	
  list	
  of	
  C.	
  difficile	
  proteins	
  that	
  have	
  either	
  
experimentally-­‐demonstrated	
  or	
  putative	
  roles	
  in	
  intestinal	
  adherence	
  (43).	
  	
  Several	
  
of	
  these	
  are	
  cell	
  wall	
  proteins	
  (CWPs)	
  and	
  surface	
  layer	
  proteins	
  (SLPs).	
  	
  	
  

	
  

7	
  

	
  

Flagella	
  also	
  play	
  an	
  important	
  role	
  in	
  bacterial	
  pathogenesis.	
  	
  Not	
  only	
  are	
  

they	
  essential	
  for	
  motility,	
  but	
  can	
  also	
  aid	
  in	
  adherence.	
  	
  Motility	
  has	
  several	
  
important	
  functions	
  for	
  pathogen	
  survival.	
  	
  Being	
  able	
  to	
  navigate	
  toward	
  sources	
  of	
  
nutrients	
  as	
  well	
  as	
  away	
  from	
  detrimental	
  chemicals	
  is	
  advantageous	
  for	
  self-­‐evident	
  
reasons.	
  	
  Moreover,	
  the	
  intestinal	
  epithelium	
  is	
  continually	
  pumping	
  out	
  and	
  turning	
  
over	
  viscous	
  mucin;	
  being	
  able	
  to	
  better	
  traverse	
  this	
  landscape	
  aids	
  in	
  intestinal	
  
colonization.	
  	
  Experimental	
  evidence	
  suggesting	
  the	
  importance	
  of	
  flagella	
  to	
  C.	
  
difficile	
  in	
  both	
  colonization	
  and	
  adherence	
  has	
  been	
  reported	
  (44).	
  	
  	
  	
  
	
  
Molecular	
  typing	
  schemes.	
  	
  In	
  epidemiological	
  studies,	
  molecular	
  typing	
  of	
  
pathogens	
  is	
  used	
  to	
  determine	
  genetic	
  relatedness	
  between	
  clinical	
  isolates.	
  	
  This	
  
information	
  helps	
  determine	
  routes	
  of	
  transmission	
  and	
  sources	
  of	
  outbreaks.	
  	
  
Moreover,	
  it	
  can	
  help	
  to	
  elucidate	
  the	
  spread	
  of	
  clonal	
  lineages	
  among	
  populations	
  at	
  
several	
  levels	
  of	
  resolution;	
  within	
  a	
  small	
  hospital	
  population,	
  regionally,	
  nationally,	
  
or	
  even	
  globally.	
  	
  There	
  are	
  several	
  different	
  typing	
  schemes	
  used	
  for	
  C.	
  difficile,	
  
including	
  Restriction	
  Endonuclease	
  Analysis	
  (REA),	
  North	
  American	
  Pulsed-­‐Field	
  Gel	
  
Electrophoresis	
  (NAP),	
  PCR	
  Ribotyping	
  (RT),	
  Multilocus	
  Sequence	
  Typing	
  (MLST),	
  
among	
  others	
  (45).	
  	
  	
  Each	
  method	
  utilizes	
  the	
  genetic	
  differences	
  between	
  strains	
  to	
  
classify	
  them	
  into	
  different	
  groups,	
  or	
  types.	
  	
  There	
  are	
  regional	
  preferences	
  in	
  terms	
  
of	
  which	
  typing	
  techniques	
  are	
  used;	
  for	
  example,	
  in	
  North	
  America	
  C.	
  diffcile	
  is	
  
commonly	
  typed	
  by	
  NAP	
  group.	
  	
  Likewise,	
  REA	
  and	
  RT	
  strain	
  designations	
  are	
  
common	
  in	
  Europe.	
  	
  These	
  typing	
  methods,	
  particularly	
  ribotyping,	
  are	
  becoming	
  
more	
  common	
  in	
  North	
  America	
  as	
  well.	
  	
  Although	
  a	
  recent	
  publication	
  showed	
  a	
  

	
  

8	
  

strong	
  correlation	
  among	
  typing	
  methods	
  used	
  to	
  classify	
  a	
  set	
  of	
  100	
  international	
  
isolates,	
  a	
  common	
  protocol	
  for	
  typing	
  isolates	
  would	
  be	
  advantageous	
  for	
  tracking	
  C.	
  
difficile,	
  and	
  facilitate	
  epidemiological	
  studies	
  (46).	
  	
  	
  
	
  
Emergence	
  of	
  Epidemic	
  C.	
  difficile	
  Strains	
  
An	
  increase	
  in	
  cases	
  of	
  CDI,	
  marked	
  by	
  significantly	
  higher	
  morbidity	
  and	
  mortality	
  
rates,	
  has	
  been	
  largely	
  attributed	
  to	
  the	
  recent	
  emergence	
  of	
  epidemic	
  strains.	
  	
  
Several	
  outbreaks	
  were	
  first	
  reported	
  in	
  Quebec,	
  Canada	
  in	
  2002,	
  and	
  several	
  others	
  
in	
  the	
  United	
  States	
  between	
  2000	
  and	
  2003	
  (35,	
  47).	
  	
  Analysis	
  of	
  isolates	
  from	
  these	
  
outbreaks	
  revealed	
  a	
  specific	
  strain	
  lineage	
  of	
  C.	
  difficile	
  was	
  associated	
  with	
  the	
  
majority	
  of	
  these	
  cases,	
  typed	
  as	
  NAP1	
  (North	
  American	
  Pulsed-­‐field	
  1),	
  RT	
  027	
  (PCR	
  
ribotype),	
  and	
  BI	
  (restriction	
  endonuclease	
  analysis).	
  	
  In	
  addition,	
  several	
  attributes	
  
of	
  these	
  strains	
  were	
  identified,	
  including	
  production	
  of	
  binary	
  toxin	
  (CDT),	
  
resistance	
  to	
  fluoroquinolones,	
  and	
  a	
  deletion	
  within	
  the	
  tcdC	
  gene,	
  encoding	
  the	
  
negative	
  regulator	
  of	
  TcdA	
  and	
  TcdB	
  (35).	
  	
  Epidemiological	
  studies	
  have	
  reported	
  
that	
  RT	
  027/NAP1	
  strains	
  are	
  associated	
  with	
  up	
  to	
  10-­‐fold	
  higher	
  morbidity	
  rates,	
  
increased	
  cases	
  of	
  severe	
  disease	
  leading	
  to	
  toxic	
  megacolon,	
  and	
  necessity	
  for	
  
colectomy	
  (48).	
  	
  Furthermore,	
  mortality	
  rates	
  have	
  been	
  reported	
  as	
  high	
  as	
  16.7%;	
  a	
  
near	
  10-­‐fold	
  increase	
  from	
  reports	
  previous	
  to	
  the	
  emergence	
  of	
  this	
  strain	
  (48).	
  	
  The	
  
rapid	
  spread	
  of	
  RT	
  027/NAP1	
  strains	
  since	
  these	
  initial	
  outbreaks	
  is	
  remarkable.	
  	
  In	
  
just	
  a	
  few	
  years	
  following,	
  it	
  was	
  reported	
  that	
  57%	
  of	
  478	
  C.	
  difficile	
  clinical	
  isolates	
  
from	
  88	
  hospitals	
  in	
  Quebec,	
  Canada	
  were	
  NAP1	
  strains	
  (49).	
  	
  Moreover,	
  RT	
  027	
  
strains	
  have	
  been	
  shown	
  to	
  be	
  prevalent	
  in	
  many	
  other	
  hospitals	
  and	
  regions	
  (50-­‐53).	
  	
  

	
  

9	
  

In	
  less	
  than	
  a	
  decade,	
  this	
  strain	
  has	
  spread	
  globally	
  and	
  been	
  associated	
  with	
  several	
  
more	
  outbreaks,	
  namely	
  in	
  Europe	
  (54,	
  55).	
  	
  	
  He	
  et	
  al.	
  recently	
  published	
  an	
  extensive	
  
phylogenetic	
  study	
  of	
  RT	
  027	
  C.	
  difficile,	
  using	
  whole	
  genome	
  sequences	
  of	
  a	
  global	
  
set	
  of	
  151	
  isolates	
  (56).	
  	
  	
  Their	
  analyses	
  revealed	
  the	
  global	
  spread	
  patterns	
  of	
  two	
  
distinct	
  lineages	
  within	
  this	
  clade	
  of	
  strains,	
  both	
  originating	
  from	
  North	
  America,	
  
with	
  one	
  undergoing	
  much	
  wider	
  global	
  dissemination.	
  	
  
	
  

Initial	
  reports	
  regarding	
  the	
  emergence	
  of	
  RT	
  027/NAP1	
  strains	
  designated	
  

them	
  as	
  hypervirulent	
  due	
  to	
  associations	
  with	
  increased	
  morbidity	
  and	
  mortality	
  
rates.	
  	
  	
  It	
  has	
  recently	
  come	
  into	
  question,	
  however,	
  if	
  these	
  strains	
  really	
  are	
  
hypervirulent.	
  	
  While	
  several	
  studies	
  have	
  shown	
  increased	
  rates	
  of	
  disease	
  due	
  to	
  
these	
  strains,	
  other	
  reports	
  have	
  demonstrated	
  a	
  lack	
  of	
  correlation	
  between	
  ribotype	
  
and	
  disease	
  severity	
  (57-­‐59).	
  	
  It	
  is	
  possible	
  that	
  outbreak-­‐associated	
  RT	
  027	
  strains	
  
are	
  not	
  representative	
  of	
  the	
  lineage	
  as	
  a	
  whole.	
  	
  Nonetheless,	
  the	
  global	
  spread	
  and	
  
prevalence	
  of	
  this	
  group	
  still	
  warrant	
  investigation.	
  	
  Regardless	
  of	
  the	
  ability	
  of	
  these	
  
strains	
  to	
  cause	
  more	
  severe	
  disease	
  within	
  individuals,	
  this	
  observation	
  suggests	
  
that	
  they	
  have	
  increased	
  transmission	
  rates.	
  	
  In	
  ecological	
  terms,	
  this	
  implies	
  that	
  
they	
  have	
  an	
  increased	
  ecological	
  fitness	
  compared	
  to	
  strains	
  of	
  other,	
  less	
  prevalent,	
  
ribotypes	
  (60).	
  	
  Therefore,	
  it	
  remains	
  to	
  be	
  determined	
  if	
  strain	
  characteristics	
  aside	
  
from	
  archetypical	
  virulence	
  factors	
  contribute	
  to	
  the	
  pervasiveness	
  of	
  the	
  RT	
  027	
  
strains.	
  	
  A	
  significant	
  portion	
  of	
  the	
  work	
  presented	
  in	
  this	
  thesis	
  aims	
  to	
  address	
  this	
  
specific	
  question.	
  	
  
	
  

	
  

10	
  

Current	
  impact	
  on	
  the	
  health	
  care	
  system.	
  	
  According	
  to	
  a	
  recent	
  report	
  from	
  the	
  
Centers	
  for	
  Disease	
  Control	
  and	
  Prevention	
  (CDC),	
  C.	
  difficile	
  infection	
  has	
  surpassed	
  
MRSA	
  and	
  is	
  now	
  the	
  most	
  common	
  hospital-­‐acquired	
  infection	
  in	
  the	
  United	
  States	
  
(61).	
  	
  	
  	
  Based	
  on	
  current	
  surveillance	
  data,	
  C.	
  difficile	
  causes	
  an	
  appalling	
  250,000	
  
infections	
  per	
  year.	
  	
  	
  Moreover,	
  the	
  CDC	
  reported	
  a	
  400%	
  increase	
  in	
  related	
  deaths	
  
between	
  years	
  2000	
  and	
  2007,	
  with	
  a	
  current	
  estimate	
  of	
  14,000	
  deaths	
  per	
  year.	
  	
  In	
  
addition,	
  estimated	
  annual	
  hospital	
  costs	
  associated	
  with	
  CDI	
  are	
  in	
  excess	
  of	
  $3	
  
billion	
  (62).	
  	
  The	
  significance	
  of	
  C.	
  difficile	
  disease	
  is	
  clear,	
  as	
  is	
  the	
  need	
  to	
  better	
  
understand	
  the	
  pathogenesis	
  of	
  this	
  organism.	
  	
  	
  Finding	
  better	
  ways	
  of	
  preventing	
  
and	
  treating	
  this	
  disease	
  is	
  a	
  growing	
  area	
  of	
  current	
  investigation.	
  	
  
	
  
Understanding	
  Disease	
  Development:	
  Colonization	
  Resistance.	
  	
  	
  	
  
Soon	
  after	
  the	
  advent	
  of	
  clinical	
  antibiotic	
  use,	
  it	
  was	
  observed	
  that	
  antibiotic	
  
treatment	
  often	
  resulted	
  in	
  increased	
  susceptibility	
  to	
  enteric	
  pathogens.	
  	
  	
  This	
  led	
  to	
  
the	
  theory	
  that	
  the	
  intestinal	
  microbiota	
  provides	
  a	
  protective	
  function,	
  termed	
  
‘colonization	
  resistance’,	
  by	
  inhibiting	
  growth	
  of	
  pathogens	
  and	
  development	
  of	
  
disease.	
  	
  	
  Research	
  in	
  this	
  area	
  has	
  revealed	
  several	
  different	
  mechanisms	
  at	
  play,	
  
including	
  ones	
  that	
  involve	
  not	
  only	
  microbiota-­‐pathogen	
  interactions,	
  but	
  also	
  
indirect	
  inhibition	
  mediated	
  through	
  microbiota-­‐host	
  interactions	
  (reviewed	
  in	
  (6)).	
  	
  
The	
  intricacies	
  and	
  dynamics	
  involved	
  in	
  colonization	
  resistance	
  are	
  inherently	
  
complex	
  and	
  multi-­‐factorial.	
  	
  Moreover,	
  the	
  key	
  interactions	
  providing	
  resistance	
  are	
  
likely	
  different	
  for	
  different	
  pathogens.	
  	
  	
  Therefore,	
  while	
  there	
  have	
  been	
  some	
  
developments	
  in	
  our	
  understanding	
  of	
  colonization	
  resistance	
  in	
  the	
  case	
  of	
  C.	
  difficile	
  

	
  

11	
  

infection,	
  much	
  more	
  work	
  is	
  needed	
  to	
  be	
  done.	
  	
  There	
  are	
  several	
  mechanisms	
  
proposed	
  to	
  be	
  involved	
  in	
  colonization	
  resistance	
  against	
  C.	
  difficile,	
  and	
  these	
  are	
  
briefly	
  outlined	
  below.	
  	
  	
  	
  
	
  
Bile	
  acids	
  and	
  germination.	
  	
  The	
  capacity	
  of	
  C.	
  difficile	
  to	
  colonize	
  and	
  establish	
  
infection	
  is	
  dependent	
  on	
  the	
  ability	
  of	
  spores	
  to	
  germinate	
  and	
  outgrow.	
  	
  	
  In	
  vivo,	
  
these	
  processes	
  are	
  affected	
  by	
  bile	
  acids;	
  specific	
  primary	
  and	
  secondary	
  bile	
  acids	
  
have	
  been	
  shown	
  to	
  act	
  as	
  germinants,	
  anti-­‐germinants,	
  and	
  growth	
  inhibitors	
  (63-­‐
65).	
  	
  Primary	
  bile	
  acids,	
  both	
  amino	
  acid-­‐conjugated	
  and	
  unconjugated,	
  are	
  secreted	
  
from	
  the	
  gall	
  bladder	
  and	
  function	
  in	
  fat	
  and	
  cholesterol	
  absorption.	
  	
  Some	
  members	
  
of	
  the	
  microbiota	
  can	
  enzymatically	
  deconjugate	
  or	
  convert	
  these	
  primary	
  bile	
  acids	
  
into	
  secondary	
  bile	
  acids.	
  	
  As	
  such,	
  the	
  intestinal	
  track	
  contains	
  a	
  complex	
  mixture	
  of	
  
primary	
  and	
  secondary,	
  conjugated	
  and	
  deconjugated	
  bile	
  acids,	
  and	
  the	
  proportions	
  
of	
  each	
  component	
  within	
  this	
  mixture	
  are	
  affected	
  by	
  the	
  microbiota	
  that	
  are	
  
present.	
  	
  	
  Giel	
  et	
  al.	
  recently	
  demonstrated	
  this	
  using	
  a	
  mouse	
  infection	
  model	
  to	
  show	
  
that	
  the	
  balance	
  of	
  bile	
  acids	
  under	
  normal	
  conditions	
  is	
  such	
  that	
  germination	
  is	
  
inhibited,	
  yet	
  following	
  antibiotic	
  treatment	
  changes	
  to	
  the	
  microbiota	
  results	
  in	
  a	
  
shift	
  in	
  bile	
  acids	
  to	
  a	
  state	
  which	
  allows	
  spore	
  germination	
  and	
  outgrowth	
  (66).	
  	
  	
  To	
  
further	
  support	
  the	
  role	
  of	
  bile	
  acids	
  in	
  colonization	
  resistance	
  and	
  demonstrate	
  that	
  
this	
  mechanism	
  can	
  be	
  exploited	
  as	
  a	
  way	
  to	
  prevent	
  infection,	
  Howerton	
  et	
  al.	
  
showed	
  that	
  a	
  bile	
  acid	
  analog,	
  CamSA,	
  could	
  prevent	
  spore	
  germination	
  and	
  
therefore	
  colonization	
  and	
  development	
  of	
  disease	
  in	
  mice	
  (67).	
  	
  Future	
  work	
  in	
  this	
  

	
  

12	
  

area	
  is	
  needed	
  to	
  better	
  define	
  the	
  role	
  of	
  bile	
  acids	
  in	
  colonization	
  resistance	
  and	
  has	
  
the	
  potential	
  to	
  lead	
  to	
  novel	
  ways	
  of	
  preventing	
  CDI.	
  	
  	
  
	
  
Competitive	
  exclusion.	
  	
  A	
  central	
  mechanism	
  of	
  colonization	
  resistance	
  that	
  can	
  be	
  
applied	
  to	
  many	
  intestinal	
  pathogens	
  involves	
  direct	
  out-­‐competition	
  by	
  the	
  
microbiota	
  for	
  nutrient	
  sources	
  and	
  intestinal	
  binding	
  sites,	
  also	
  referred	
  to	
  as	
  
competitive	
  exclusion	
  or	
  niche	
  exclusion.	
  	
  Indeed,	
  examples	
  for	
  both	
  of	
  these	
  exist	
  in	
  
the	
  case	
  of	
  C.	
  difficile	
  inhibition.	
  	
  	
  Initial	
  evidence	
  that	
  competition	
  occurs	
  between	
  
intestinal	
  microbiota	
  and	
  C.	
  difficile	
  was	
  demonstrated	
  in	
  both	
  an	
  in	
  vitro	
  chemostat	
  
model	
  and	
  an	
  in	
  vivo	
  mouse	
  model	
  (68).	
  	
  	
  Further	
  investigation	
  by	
  Wilson	
  and	
  Perini	
  
identified	
  several	
  compounds	
  for	
  which	
  the	
  microbiota	
  of	
  hamsters	
  can	
  outcompete	
  
C.	
  difficile	
  in	
  vitro	
  (69).	
  	
  These	
  findings	
  were	
  only	
  recently	
  directly	
  tested	
  in	
  an	
  in	
  vivo	
  
mouse	
  model	
  to	
  show	
  that	
  C.	
  difficile	
  metabolism	
  of	
  sialic	
  acids	
  plays	
  a	
  role	
  in	
  gut	
  
colonization	
  (70).	
  	
  	
  Few	
  studies	
  directly	
  testing	
  competition	
  for	
  specific	
  nutrients	
  
exist;	
  however,	
  the	
  general	
  metabolic	
  capabilities	
  of	
  C.	
  difficile	
  will	
  be	
  extensively	
  
reviewed	
  in	
  the	
  next	
  section.	
  	
  	
  	
  
	
  

In	
  addition	
  to	
  nutrient	
  availability,	
  another	
  important	
  element	
  of	
  an	
  

organism’s	
  niche	
  is	
  spatial	
  availability.	
  	
  Regarding	
  intestinal	
  colonization,	
  this	
  
encompasses	
  the	
  ability	
  to	
  bind	
  to	
  mucosal	
  components,	
  such	
  as	
  fibronectin,	
  collagen,	
  
and	
  fibrinogen	
  (71).	
  	
  	
  Several	
  cell	
  surface	
  molecules	
  of	
  C.	
  diffcile	
  have	
  been	
  implicated	
  
to	
  be	
  involved	
  attachment,	
  including	
  fibronectin-­‐binding	
  proteins,	
  cell	
  wall	
  proteins,	
  
and	
  surface	
  layer	
  proteins	
  (reviewed	
  in	
  (43)).	
  	
  The	
  majority	
  of	
  these	
  studies	
  were	
  
conducted	
  using	
  in	
  vitro	
  conditions	
  and	
  therefore	
  the	
  contribution	
  of	
  these	
  adhesins	
  

	
  

13	
  

to	
  in	
  vivo	
  colonization	
  has	
  not	
  been	
  directly	
  tested,	
  nor	
  have	
  the	
  specific	
  competing	
  
members	
  of	
  the	
  microbiota	
  been	
  identified.	
  	
  	
  	
  	
  
	
  

This	
  concept	
  of	
  niche	
  exclusion	
  led	
  to	
  the	
  hypothesis	
  that	
  non-­‐toxigenic	
  C.	
  

difficile	
  could	
  be	
  used	
  to	
  treat	
  or	
  prevent	
  CDI.	
  	
  	
  The	
  premise	
  is	
  that	
  they	
  will	
  either	
  
occupy	
  the	
  specific	
  niche	
  and	
  therefore	
  exclude	
  toxigenic	
  C.	
  difficile,	
  or	
  outcompete	
  
already	
  colonized	
  toxigenic	
  C.	
  difficile,	
  therefore	
  displacing	
  and	
  eliminating	
  them.	
  	
  
Borriello	
  et	
  al.	
  presented	
  evidence	
  supporting	
  this	
  hypothesis	
  when	
  they	
  showed	
  that	
  
pre-­‐colonization	
  of	
  non-­‐toxigenic	
  C.	
  difficile	
  prevented	
  disease	
  development	
  in	
  a	
  
hamster	
  infection	
  model	
  (72).	
  	
  A	
  few	
  years	
  later	
  this	
  was	
  tested	
  in	
  two	
  hospital	
  
patients	
  with	
  recurrent	
  CDI,	
  both	
  of	
  which	
  had	
  no	
  additional	
  bouts	
  of	
  relapsing	
  
disease	
  following	
  administration	
  of	
  the	
  non-­‐toxigenic	
  C.	
  difficile	
  (73).	
  	
  More	
  recently,	
  
additional	
  work	
  investigating	
  non-­‐toxigenic	
  C.	
  difficile	
  prevention	
  of	
  disease	
  has	
  been	
  
conducted	
  in	
  hamsters	
  (74,	
  75).	
  	
  This	
  work	
  supports	
  the	
  theory	
  of	
  niche	
  exclusion	
  in	
  
the	
  intestinal	
  environment	
  and	
  future	
  work	
  to	
  identify	
  specific	
  members	
  of	
  the	
  
microbiota,	
  which	
  compete	
  with	
  C.	
  difficile	
  for	
  its	
  required	
  niche,	
  will	
  contribute	
  to	
  
our	
  understanding	
  of	
  the	
  role	
  of	
  colonization	
  resistance	
  in	
  CDI.	
  
	
  
Direct	
  antagonism.	
  	
  Additional	
  mechanisms	
  of	
  colonization	
  resistance	
  include	
  ones	
  
where	
  there	
  is	
  direct	
  antagonism	
  of	
  C.	
  difficile	
  by	
  production	
  of	
  inhibitory	
  compounds	
  
by	
  the	
  microbiota;	
  these	
  include	
  short	
  chain	
  fatty	
  acids	
  (SCFAs),	
  lactate,	
  and	
  
bacteriocins.	
  	
  Several	
  studies	
  have	
  investigated	
  the	
  effects	
  of	
  SCFAs	
  on	
  C.	
  difficile	
  
colonization	
  in	
  both	
  in	
  vivo	
  and	
  in	
  vitro	
  models	
  (76-­‐78).	
  	
  Much	
  of	
  the	
  work	
  supports	
  
that	
  SCFAs	
  suppress	
  C.	
  difficile	
  colonization,	
  however,	
  some	
  of	
  these	
  results	
  are	
  

	
  

14	
  

conflicting.	
  	
  For	
  example,	
  Su	
  et	
  al.	
  did	
  not	
  see	
  a	
  difference	
  in	
  colonization	
  levels	
  of	
  C.	
  
difficile	
  in	
  mono-­‐associated	
  mice	
  when	
  administered	
  physiologically	
  relevant	
  
concentrations	
  of	
  SCFAs	
  (77).	
  	
  	
  	
  The	
  ability	
  of	
  lactate,	
  another	
  organic	
  acid,	
  to	
  
suppress	
  C.	
  difficile	
  colonization	
  has	
  also	
  been	
  shown	
  (79).	
  	
  	
  In	
  this	
  work,	
  lactate	
  
produced	
  by	
  a	
  potential	
  probiotic	
  strain,	
  Streptococcus	
  thermophilis,	
  was	
  
demonstrated	
  to	
  ameliorate	
  CDI	
  in	
  a	
  mouse	
  infection	
  model.	
  	
  It	
  is	
  known	
  that	
  
intestinal	
  microbiota	
  also	
  produce	
  lactic	
  acid,	
  and	
  therefore,	
  this	
  work	
  supports	
  the	
  
potential	
  role	
  of	
  lactic	
  acid	
  in	
  colonization	
  resistance.	
  	
  Lastly,	
  Rea	
  et	
  al.	
  discovered	
  
that	
  a	
  human	
  fecal	
  isolate,	
  Bacillus	
  theringiensis,	
  produces	
  a	
  bacteriocin	
  with	
  narrow-­‐
spectrum	
  activity	
  against	
  C.	
  difficile	
  (80).	
  	
  	
  To	
  date,	
  this	
  is	
  the	
  only	
  example	
  of	
  a	
  C.	
  
difficile-­‐specific	
  bacteriocin	
  produced	
  by	
  the	
  intestinal	
  microbiota,	
  however,	
  it	
  shows	
  
that	
  there	
  are	
  specific	
  antagonistic	
  mechanisms	
  employed	
  by	
  intestinal	
  microbiota	
  
against	
  C.	
  diffiicle	
  and	
  future	
  work	
  in	
  this	
  area	
  may	
  contribute	
  additional	
  insight.	
  	
  	
  
	
  
In	
  summary,	
  several	
  mechanisms	
  utilized	
  by	
  the	
  intestinal	
  microbiota	
  to	
  either	
  
directly	
  or	
  indirectly	
  suppress	
  C.	
  difficile	
  have	
  been	
  implicated	
  to	
  contribute	
  to	
  
colonization	
  resistance.	
  	
  While	
  some	
  mechanisms	
  have	
  been	
  directly	
  tested	
  in	
  in	
  vivo	
  
models,	
  most	
  have	
  not.	
  	
  Moreover,	
  the	
  specific	
  member/s	
  of	
  the	
  intestinal	
  microbiota	
  
that	
  are	
  responsible	
  for	
  these	
  inhibitory	
  functions	
  have	
  generally	
  not	
  been	
  identified.	
  	
  	
  
Considering	
  the	
  variability	
  in	
  microbial	
  communities	
  among	
  individuals,	
  it	
  is	
  likely	
  
that	
  different	
  strains	
  perform	
  these	
  activities	
  in	
  different	
  people.	
  	
  Therefore,	
  there	
  is	
  
still	
  much	
  to	
  be	
  done	
  in	
  this	
  area	
  to	
  understand	
  colonization	
  resistance	
  and	
  be	
  able	
  to	
  
exploit	
  these	
  mechanisms	
  for	
  prevention	
  or	
  treatment	
  of	
  CDI.	
  	
  	
  	
  

	
  

15	
  

Understanding	
  C.	
  difficile	
  Physiology:	
  Metabolism	
  	
  
The	
  genus	
  Clostridium	
  is	
  very	
  diverse,	
  comprised	
  of	
  both	
  medically	
  and	
  
biotechnologically	
  important	
  members.	
  	
  The	
  metabolic	
  capabilities	
  of	
  many	
  of	
  these	
  
species	
  have	
  been	
  investigated;	
  perhaps	
  the	
  most	
  well	
  known	
  is	
  the	
  industrially	
  
relevant	
  solventogenic	
  clostridia.	
  	
  In	
  contrast	
  to	
  some	
  of	
  these	
  other	
  clostridial	
  
species,	
  little	
  is	
  known	
  about	
  the	
  nutritional	
  requirements	
  and	
  nutrient	
  utilization	
  
capabilities	
  of	
  C.	
  difficile.	
  	
  Since	
  the	
  discovery	
  of	
  CDI,	
  a	
  large	
  portion	
  of	
  research	
  has	
  
been	
  focused	
  on	
  the	
  epidemiology	
  and	
  pathology	
  of	
  C.	
  difficile	
  rather	
  than	
  its	
  
physiology.	
  	
  	
  That	
  approach	
  is	
  changing,	
  however,	
  with	
  the	
  growing	
  body	
  of	
  evidence	
  
supporting	
  that	
  CDI	
  is	
  a	
  microbiome-­‐mediated	
  disease.	
  	
  	
  More	
  attention	
  is	
  being	
  
focused	
  now	
  on	
  the	
  colonization	
  strategies	
  of	
  C.	
  difficile	
  and	
  aspects	
  of	
  its	
  general	
  
biology	
  and	
  physiology,	
  including	
  metabolism,	
  sporulation,	
  and	
  germination.	
  	
  	
  	
  
It	
  is	
  clear	
  that	
  the	
  intestinal	
  microbiota	
  plays	
  a	
  role	
  in	
  resistance	
  to	
  C.	
  difficile	
  
colonization	
  and	
  disease	
  establishment.	
  	
  As	
  reviewed	
  above,	
  a	
  primary	
  mechanism	
  at	
  
play	
  is	
  likely	
  competition	
  for	
  limited	
  nutrients	
  within	
  the	
  gastrointestinal	
  tract.	
  	
  
Understanding	
  better	
  the	
  metabolic	
  strategies	
  that	
  C.	
  difficile	
  employs	
  to	
  grow	
  in	
  the	
  
highly	
  competitive	
  intestinal	
  environment	
  could	
  help	
  to	
  elucidate	
  novel	
  ways	
  to	
  
prevent	
  or	
  control	
  disease	
  development.	
  	
  In	
  the	
  following	
  pages	
  I	
  will	
  review	
  what	
  is	
  
known	
  about	
  C.	
  difficile’s	
  metabolic	
  capabilities	
  based	
  on	
  published	
  literature,	
  with	
  a	
  
focus	
  on	
  nutrients	
  contributing	
  to	
  vegetative	
  growth.	
  	
  	
  The	
  metabolic	
  functions,	
  
associated	
  compounds,	
  experimental	
  approaches	
  used,	
  and	
  relevant	
  references	
  
identified	
  in	
  the	
  literature	
  and	
  included	
  in	
  this	
  review	
  are	
  listed	
  in	
  Table.	
  1.1.	
  	
  
	
  
	
  
	
  

16	
  

Table	
  1.1.	
  	
  C.	
  difficile	
  metabolic	
  functions	
  related	
  to	
  growth.	
  
	
  
Metabolic	
  Function	
  

Sugar	
  fermentation	
  

Stickland	
  
fermentation	
  
Hydrolysis	
  of	
  host	
  
tissue	
  components	
  
Metabolism	
  of	
  
mucin	
  components	
  
Amino	
  acid	
  
utilization	
  
Ethanolamine	
  
utilization	
  

Compounds	
  Metabolized	
  
Glucose,	
  fructose,	
  mannitol,	
  
salicin,	
  xylose,	
  galactose,	
  
maltose,	
  sucrose,	
  lactose,	
  
raffinose,	
  inulin,	
  glycerol	
  
Mannose,	
  melezitose,	
  sorbitol,	
  
ribose,	
  cellobiose,	
  trehalose	
  
raffinose,	
  stachyose	
  
Glucose,	
  N-­‐acetylglucosamine	
  
Amino	
  acids	
  
	
  
Hyaluronic	
  acid	
  
Chondroitin-­‐4-­‐sulfate,	
  
heparin,	
  collagen	
  
Gelatin	
  
Sialic	
  acids	
  (N-­‐
acetyleneuraminic	
  acid)	
  
	
  
Essential	
  amino	
  acids:	
  
cysteine,	
  isoleucine,	
  leucine,	
  
proline,	
  tryptophan,valine	
  
Ethanolamine	
  

Carbon	
  fixation	
  (CO2	
  +	
  H2)	
  
	
  
General	
  metabolism	
  	
   	
  
(In	
  silico	
  analysis)	
  
	
  
General	
  metabolism	
  
(transcriptomics)	
  
	
  
Autotrophy	
  

General	
  metabolism	
  
(proteomics)	
  
General	
  metabolism	
  
(metabolomics)	
  

	
  
Intestinal	
  environment	
  
metabolites	
  

Experimental	
  Approach	
  
In	
  vitro	
  –	
  batch	
  culture	
  

References	
  
(81,	
  82)	
  

In	
  vitro	
  –	
  batch	
  culture	
  
In	
  vitro	
  –	
  batch	
  culture	
  
In	
  vitro	
  gut	
  model	
  
In	
  vitro	
  –	
  batch	
  culture	
  
Proteomics	
  
In	
  vitro	
  -­‐	
  biochemical	
  assay	
  
In	
  vitro	
  -­‐	
  biochemical	
  assay	
  

(82,	
  83)	
  
	
  
(82)	
  
(84)	
  
(85-­‐87)	
  
(88)	
  
(89-­‐92)	
  
(91,	
  92)	
  

In	
  vitro	
  –	
  batch	
  culture	
  
In	
  vitro	
  gut	
  model	
  

(83)	
  

In	
  vivo	
  -­‐	
  mouse	
  model	
  
Defined	
  medium	
  
development	
  

(70)	
  

Structural	
  and	
  biochemical	
  
analysis	
  of	
  eut	
  operon	
  
In	
  vitro	
  –	
  batch	
  culture	
  
	
  
Whole	
  genome	
  sequencing	
  
and	
  annotation	
  
In	
  vitro	
  vs	
  in	
  vivo	
  (pig	
  ileal-­‐
ligated	
  loop	
  model)	
  
In	
  vitro	
  vs	
  in	
  vivo	
  
(monoassociated	
  mice)	
  
In	
  vitro	
  vs	
  in	
  vivo	
  (pig	
  ileal-­‐
ligated	
  loop	
  model)	
  
Analytical	
  chemical	
  analysis	
  
of	
  in	
  vivo	
  (mouse)	
  samples	
  

(84)	
  

(93)	
  
(94)	
  
(95)	
  
	
  
(96)	
  
(97)	
  
(98)	
  
(88)	
  
(82)	
  

	
  
	
  
The	
  nutritional	
  landscape	
  of	
  the	
  gut	
  is	
  remarkably	
  complex,	
  dependent	
  on	
  factors	
  of	
  
the	
  host’s	
  genetics	
  and	
  diet,	
  as	
  well	
  as	
  the	
  intestinal	
  microbial	
  community,	
  which	
  has	
  
been	
  shown	
  to	
  be	
  highly	
  variable	
  from	
  person	
  to	
  person	
  (99).	
  	
  	
  Potential	
  nutrient	
  
sources	
  in	
  the	
  gut	
  come	
  from	
  the	
  host	
  diet,	
  host	
  tissues	
  and	
  secreted	
  compounds,	
  
components	
  of	
  microbial	
  cells,	
  and	
  metabolites	
  of	
  the	
  host	
  and	
  microbiota.	
  	
  While	
  it	
  is	
  

	
  

17	
  

sometimes	
  assumed	
  that	
  the	
  intestinal	
  environment	
  is	
  nutrient-­‐rich,	
  it	
  is	
  in	
  fact	
  a	
  very	
  
nutrient-­‐poor	
  environment,	
  especially	
  in	
  the	
  large	
  intestine.	
  	
  Most	
  nutrients	
  are	
  
depleted	
  from	
  the	
  intestinal	
  contents	
  in	
  transit	
  through	
  the	
  small	
  intestine	
  by	
  the	
  
microbes	
  in	
  this	
  upper	
  region	
  and	
  by	
  host	
  degradation	
  and	
  nutrient	
  uptake.	
  	
  	
  
Moreover,	
  the	
  niche-­‐adapted	
  commensal	
  microbiota	
  present	
  a	
  highly	
  competitive	
  
environment	
  for	
  enteropathogens	
  to	
  obtain	
  the	
  limited	
  nutrients	
  that	
  are	
  available.	
  	
  	
  
The	
  nutrient	
  utilization	
  and	
  competitive	
  strategies	
  of	
  a	
  few	
  enteropathogens,	
  such	
  as	
  
Salmonella	
  enterica	
  and	
  Escherichia	
  coli,	
  have	
  been	
  extensively	
  studied	
  (reviewed	
  in	
  
(100).	
  	
  Some	
  of	
  the	
  key	
  nutritional	
  resources	
  exploited	
  by	
  these	
  pathogens	
  include	
  
oligopeptides,	
  mucus-­‐derived	
  sugars	
  and	
  proteins,	
  ethanolamine,	
  and	
  1,2-­‐propandiol.	
  	
  	
  
While	
  the	
  strategies	
  used	
  by	
  C.	
  difficile	
  are	
  poorly	
  defined	
  to	
  date,	
  it	
  is	
  likely	
  that	
  it	
  
utilizes	
  some	
  of	
  the	
  same	
  resources,	
  and	
  the	
  growing	
  body	
  of	
  literature	
  in	
  this	
  area	
  is	
  
helping	
  to	
  paint	
  an	
  ever-­‐clearer	
  picture	
  of	
  its	
  intestinal	
  niche.	
  	
  	
  
	
  
Early	
  Characterization	
  of	
  C.	
  difficile	
  physiology.	
  	
  In	
  the	
  original	
  paper	
  describing	
  C.	
  
difficile	
  (then	
  named	
  Bacillus	
  difficilis),	
  a	
  classical	
  battery	
  of	
  microbiological	
  tests	
  was	
  
conducted	
  to	
  characterize	
  this	
  newly	
  discovered	
  intestinal	
  bacterium	
  (11).	
  	
  	
  They	
  
noted	
  a	
  difficulty	
  in	
  growing	
  this	
  organism	
  and	
  interpreting	
  the	
  employed	
  assays,	
  
mainly	
  due	
  to	
  the	
  strictly	
  anaerobic	
  conditions	
  required,	
  hence	
  the	
  name	
  “difficilis”.	
  	
  
Nonethethless,	
  a	
  short	
  list	
  of	
  nutrients	
  utilized	
  by	
  C.	
  difficile	
  was	
  reported	
  including	
  
dextrose,	
  fructose,	
  mannitol,	
  salicin,	
  xylose	
  and	
  to	
  a	
  lesser	
  extent	
  galactose,	
  maltose,	
  
sucrose,	
  lactose,	
  raffinose,	
  inulin,	
  and	
  glycerol	
  (Table	
  1.1).	
  	
  	
  Their	
  preliminary	
  
experiments	
  were	
  focused	
  more	
  on	
  characterization	
  of	
  this	
  bacterium	
  in	
  order	
  to	
  

	
  

18	
  

differentiate	
  it	
  from	
  other	
  intestinal	
  isolates	
  and	
  not	
  as	
  an	
  extensive	
  investigation	
  into	
  
its	
  metabolic	
  capabilities.	
  	
  	
  	
  
Over	
  four	
  decades	
  after	
  this	
  initial	
  description	
  of	
  C.	
  difficile,	
  its	
  association	
  with	
  
antibiotic-­‐associated	
  disease,	
  primarily	
  pseudomembranous	
  colitis,	
  was	
  published	
  
(33).	
  	
  At	
  a	
  time	
  when	
  genomic-­‐based	
  diagnostic	
  assays	
  were	
  not	
  available,	
  there	
  was	
  
a	
  need	
  to	
  further	
  develop	
  microbiological	
  assays	
  for	
  clinical	
  identification	
  of	
  this	
  
pathogen.	
  	
  	
  Some	
  additional	
  primary	
  characterization	
  was	
  conducted	
  by	
  Hafiz	
  et	
  al.	
  in	
  
1977,	
  but	
  soon	
  after	
  simplified	
  protocols	
  that	
  were	
  more	
  practical	
  for	
  clinical	
  
application	
  were	
  developed	
  by	
  Nakamura	
  et	
  al.	
  (83,	
  101).	
  	
  In	
  both	
  studies,	
  the	
  
carbohydrate-­‐fermentation	
  properties	
  of	
  large	
  strain	
  sets	
  (30	
  and	
  82,	
  respectively)	
  
were	
  investigated.	
  	
  Nakamura	
  et	
  al.	
  tested	
  for	
  fermentation	
  of	
  28	
  different	
  
carbohydrates,	
  reporting	
  positive	
  results	
  for	
  11	
  of	
  them,	
  many	
  of	
  which	
  overlaped	
  
with	
  the	
  previously	
  published	
  results	
  of	
  Hall	
  and	
  O’Toole.	
  	
  This	
  added	
  mannose,	
  
melezitose,	
  sorbitol,	
  ribose,	
  cellobiose,	
  and	
  trehalose	
  to	
  the	
  list	
  of	
  fermentable	
  
carbohydrates.	
  	
  They	
  also	
  noted	
  that	
  all	
  82	
  strains	
  liquefied	
  2%	
  gelatin,	
  and	
  that	
  the	
  
additional	
  ability	
  to	
  ferment	
  mannitol	
  was	
  a	
  unique	
  characteristic,	
  allowing	
  it	
  to	
  be	
  
distinguished	
  from	
  other	
  subterminaly	
  sporulating	
  clostridia	
  (83).	
  	
  	
  	
  
	
  
Studies	
  of	
  C.	
  difficile-­‐microbiota	
  nutrient	
  competition.	
  	
  The	
  observation	
  that	
  
antibiotic	
  treatment	
  was	
  required	
  for	
  development	
  of	
  CDI	
  sparked	
  much	
  interest	
  into	
  
understanding	
  the	
  role	
  of	
  the	
  intestinal	
  microbiota	
  in	
  resistance	
  to	
  disease.	
  	
  Initial	
  
work	
  replicating	
  this	
  observation	
  in	
  both	
  hamster	
  and	
  mouse	
  models	
  showed	
  that	
  the	
  
intestinal	
  microbiota	
  repressed	
  C.	
  difficile	
  growth	
  since	
  it	
  was	
  only	
  able	
  to	
  colonize	
  

	
  

19	
  

the	
  intestinal	
  tracts	
  of	
  gnotobiotic	
  or	
  antibiotic-­‐treated	
  animals	
  (102,	
  103).	
  	
  Wilson	
  
and	
  Perini	
  went	
  on	
  to	
  further	
  investigate	
  the	
  mechanism	
  of	
  this	
  protective	
  function,	
  
hypothesizing	
  that	
  competition	
  between	
  the	
  intestinal	
  microbiota	
  and	
  C.	
  difficile	
  for	
  
specific	
  limited	
  nutrients	
  was	
  important.	
  	
  A	
  foundation	
  for	
  these	
  studies	
  was	
  
previously	
  established	
  by	
  Freter	
  et	
  al.	
  who	
  studied	
  the	
  interactions	
  and	
  colonization	
  
dynamics	
  between	
  intestinal	
  microbiota	
  and	
  Escherichia	
  coli,	
  and	
  had	
  developed	
  a	
  
continuous-­‐flow	
  culture	
  model	
  of	
  the	
  mouse	
  large	
  intestine	
  (104-­‐107).	
  	
  	
  Indeed,	
  they	
  
had	
  already	
  shown	
  that	
  the	
  intestinal	
  microbiota	
  could	
  displace	
  or	
  inhibit	
  growth	
  of	
  
C.	
  difficile	
  in	
  their	
  in	
  vitro	
  model	
  (68).	
  	
  The	
  fact	
  that	
  these	
  dymanics	
  could	
  be	
  
replicated	
  in	
  the	
  continuous-­‐flow	
  culture	
  system	
  supported	
  the	
  idea	
  that	
  competition	
  
for	
  growth-­‐limiting	
  nutrients	
  was	
  occurring.	
  	
  Wilson	
  and	
  Perini	
  first	
  showed	
  that	
  C.	
  
difficile	
  grew	
  slower	
  than	
  the	
  dilution	
  rate	
  of	
  the	
  in	
  vitro	
  system	
  in	
  spent	
  filtrates	
  of	
  
the	
  model,	
  which	
  had	
  been	
  seeded	
  with	
  mouse	
  microbiota	
  (84).	
  	
  	
  This	
  suggested	
  that	
  
important	
  nutrients	
  had	
  been	
  depleted	
  by	
  the	
  mouse	
  microbiota	
  in	
  the	
  medium,	
  
which	
  was	
  prepared	
  from	
  cecal	
  contents	
  of	
  germ-­‐free	
  mice.	
  	
  Analysis	
  of	
  the	
  
carbohydrates	
  in	
  fresh	
  and	
  spent	
  medium	
  revealed	
  several	
  components	
  that	
  were	
  
significantly	
  depleted	
  in	
  the	
  spent	
  medium.	
  	
  Individually	
  adding	
  these	
  back	
  to	
  the	
  C.	
  
difficile	
  spent	
  medium	
  cultures	
  increased	
  the	
  growth	
  rate	
  of	
  C.	
  difficile,	
  but	
  only	
  for	
  
three	
  of	
  the	
  components	
  tested;	
  glucose,	
  N-­‐acetylglucosamine,	
  and	
  N-­‐
acetylneuraminic	
  acid.	
  	
  	
  It	
  is	
  interesting	
  to	
  note	
  that	
  the	
  latter	
  two	
  of	
  these	
  three	
  are	
  
constituents	
  of	
  mucin.	
  This	
  was	
  the	
  first	
  time	
  that	
  specific	
  nutrients	
  important	
  for	
  C.	
  
difficile	
  growth	
  in	
  the	
  context	
  of	
  the	
  intestinal	
  environment	
  were	
  identified.	
  	
  Another	
  
interesting	
  finding	
  of	
  this	
  work	
  was	
  that	
  C.	
  difficile	
  can	
  only	
  utilize	
  these	
  nutrients	
  in	
  

	
  

20	
  

their	
  free	
  form,	
  since	
  it	
  lacks	
  the	
  ability	
  to	
  cleave	
  them	
  from	
  mucin	
  directly	
  (84).	
  	
  	
  
Therefore,	
  it	
  still	
  requires	
  the	
  degradative	
  functions	
  of	
  other	
  members	
  of	
  the	
  
intestinal	
  microbiota	
  to	
  be	
  able	
  to	
  utilize	
  these	
  compounds	
  in	
  vivo,	
  further	
  supporting	
  
the	
  complexity	
  of	
  these	
  competitive	
  interactions.	
  	
  
Only	
  recently	
  was	
  utilization	
  of	
  mucin-­‐derived	
  nutrients	
  further	
  investigated	
  
in	
  vivo.	
  	
  Ng	
  et	
  al.	
  used	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  invasion	
  to	
  show	
  that	
  sialic	
  acids	
  
(specifically,	
  N-­‐acetylneuraminic	
  acid)	
  impact	
  the	
  levels	
  of	
  C.	
  difficile	
  colonization	
  in	
  
the	
  mouse	
  large	
  intestine	
  (70).	
  	
  Moreover,	
  their	
  work	
  demonstrated	
  that	
  members	
  of	
  
the	
  microbiota	
  (in	
  these	
  experiments,	
  Bacteroides	
  thetaiotaomicron)	
  that	
  can	
  
hydrolyze	
  sialic	
  acids	
  from	
  the	
  mucosa	
  are	
  required	
  for	
  C.	
  difficile	
  to	
  be	
  able	
  to	
  utilize	
  
them,	
  as	
  suggested	
  in	
  the	
  work	
  by	
  Wilson	
  and	
  Perini.	
  	
  	
  C.	
  difficile	
  expansion	
  was	
  lower	
  
in	
  mice	
  when	
  it	
  was	
  co-­‐colonized	
  with	
  a	
  sialidase-­‐deficient	
  strain	
  of	
  B.	
  
thetaiotaomicron	
  rather	
  than	
  the	
  wild-­‐type	
  strain.	
  	
  Furthermore,	
  this	
  decrease	
  in	
  
colonization	
  levels	
  was	
  restored	
  when	
  the	
  mice	
  were	
  fed	
  free	
  sialic	
  acid.	
  	
  Additional	
  
support	
  for	
  the	
  importance	
  of	
  sialic	
  acid	
  availability	
  was	
  demonstrated	
  by	
  showing	
  a	
  
4-­‐fold	
  decrease	
  in	
  colonization	
  between	
  wild-­‐type	
  C.	
  difficile	
  and	
  a	
  mutant	
  unable	
  to	
  
metabolize	
  N-­‐acetylneuraminic	
  acid.	
  	
  	
  However,	
  it	
  is	
  clear	
  that	
  additional	
  nutrients	
  
are	
  important	
  for	
  C.	
  difficile	
  colonization	
  in	
  vivo,	
  since	
  the	
  mutant	
  was	
  still	
  able	
  to	
  
reach	
  relatively	
  high	
  levels	
  (7	
  x	
  107	
  CFU/ml	
  feces).	
  	
  	
  Nonetheless,	
  this	
  work	
  is	
  
important	
  for	
  understanding	
  the	
  metabolic	
  functions	
  of	
  C.	
  difficile	
  playing	
  a	
  role	
  in	
  
intestinal	
  invasion;	
  hopefully,	
  more	
  in	
  vivo	
  work	
  similar	
  to	
  this	
  will	
  be	
  done	
  to	
  
provide	
  insight	
  in	
  this	
  area.	
  	
  	
  
	
  

	
  

21	
  

Hydrolytic	
  enzymes.	
  	
  Hydrolytic	
  enzymes	
  comprise	
  a	
  class	
  of	
  virulence	
  factors	
  
common	
  to	
  bacterial	
  pathogens	
  (108).	
  	
  More	
  specifically,	
  these	
  refer	
  to	
  enzymes	
  that	
  
can	
  hydrolyze	
  different	
  components	
  of	
  host	
  tissues.	
  	
  Some	
  classic	
  examples	
  include	
  
neuraminidase,	
  hyaluronidase,	
  collagenase,	
  gelatinase,	
  and	
  heparinase.	
  	
  These	
  are	
  
important	
  to	
  pathogens	
  for	
  two	
  main	
  reasons.	
  	
  One,	
  they	
  are	
  integral	
  in	
  pathogenesis	
  
because	
  they	
  incite	
  tissue	
  damage,	
  stimulating	
  inflammatory	
  responses	
  and	
  allowing	
  
access	
  to	
  underlying	
  tissues.	
  	
  Secondly,	
  the	
  products	
  of	
  the	
  hydrolysis	
  reactions	
  can	
  
be	
  used	
  as	
  nutrient	
  sources.	
  	
  	
  Furthermore,	
  the	
  activity	
  of	
  hydrolytic	
  enzymes	
  has	
  
been	
  demonstrated	
  for	
  several	
  isolates	
  of	
  the	
  human	
  fecal	
  microbiota,	
  particularity	
  in	
  
reference	
  to	
  the	
  break	
  down	
  of	
  host	
  mucin,	
  supporting	
  that	
  these	
  compounds	
  are	
  an	
  
important	
  source	
  of	
  nutrients	
  for	
  organisms	
  that	
  inhabit	
  the	
  intestinal	
  tract	
  (109).	
  	
  	
  
The	
  hydrolytic	
  activities	
  of	
  C.	
  difficile	
  have	
  not	
  been	
  extensively	
  studied.	
  	
  The	
  
studies	
  that	
  have	
  been	
  done	
  generally	
  include	
  poorly	
  characterized	
  strains,	
  often	
  not	
  
specifying	
  the	
  virulence	
  or	
  toxigenicity	
  of	
  the	
  strains.	
  	
  Drawing	
  conclusions	
  from	
  
these	
  studies	
  is	
  even	
  more	
  difficult	
  in	
  light	
  of	
  the	
  conflicting	
  and	
  inconsistent	
  results.	
  	
  	
  
For	
  example,	
  Hafiz	
  et	
  al.	
  screened	
  21	
  isolates	
  for	
  hyaluronidase	
  activity	
  and	
  reported	
  
them	
  all	
  to	
  be	
  positive,	
  while	
  Popoff	
  and	
  Dodin	
  then	
  screened	
  25	
  isolates	
  finding	
  them	
  
all	
  hyaluronidase	
  negative	
  (89,	
  90).	
  	
  Later,	
  a	
  study	
  was	
  published	
  by	
  Borriello	
  et	
  al.	
  
that	
  screened	
  three	
  groups	
  of	
  strains	
  of	
  varying	
  virulence	
  for	
  the	
  activity	
  of	
  four	
  
different	
  hydrolytic	
  enzymes;	
  hyaluronidase,	
  chondroitin-­‐4-­‐sulfatase,	
  heparinase,	
  and	
  
collagenase	
  (91).	
  	
  They	
  reported	
  nine	
  of	
  the	
  eleven	
  strains	
  to	
  have	
  positive	
  (or	
  weak)	
  
hyaluronidase	
  and	
  chondroitin-­‐4-­‐sulfatse	
  activity,	
  all	
  eleven	
  have	
  positive	
  (or	
  weak)	
  
heparinase	
  activity,	
  and	
  only	
  6	
  to	
  have	
  collagenase	
  activity	
  (of	
  which	
  5	
  were	
  weak).	
  	
  	
  

	
  

22	
  

While	
  there	
  were	
  no	
  strong	
  correlations	
  with	
  virulence,	
  there	
  was	
  a	
  trend	
  of	
  more	
  
hydrolytic	
  activity	
  in	
  the	
  more	
  virulent	
  strains.	
  	
  The	
  same	
  group	
  also	
  published	
  a	
  
larger	
  study	
  of	
  30	
  strains	
  of	
  known	
  virulence	
  or	
  toxin	
  production	
  where	
  they	
  again	
  
screened	
  for	
  the	
  same	
  hydrolytic	
  activities	
  (92).	
  	
  Again,	
  they	
  concluded	
  that	
  there	
  
were	
  no	
  direct	
  correlations	
  between	
  enzymatic	
  activity	
  and	
  virulence	
  or	
  toxigenic	
  
status;	
  however,	
  the	
  most	
  virulent	
  strains	
  in	
  their	
  analyses	
  had	
  the	
  most	
  positive	
  
results,	
  whereas	
  many	
  of	
  the	
  lesser	
  virulent	
  strains	
  had	
  more	
  negative	
  results.	
  	
  It	
  is	
  
unclear	
  whether	
  the	
  hydrolytic	
  capabilities	
  of	
  C.	
  difficile	
  play	
  more	
  of	
  a	
  direct	
  role	
  in	
  
virulence	
  by	
  damaging	
  host	
  tissues	
  and	
  breaking	
  down	
  the	
  mucosal	
  barrier	
  as	
  a	
  
means	
  of	
  allowing	
  the	
  toxins	
  access	
  to	
  the	
  underlying	
  epithelia,	
  or	
  if	
  they	
  serve	
  to	
  play	
  
a	
  more	
  indirect	
  role	
  by	
  providing	
  nutrients	
  for	
  growth	
  and	
  expansion	
  in	
  the	
  intestinal	
  
environment.	
  	
  In	
  essence,	
  these	
  two	
  functions	
  are	
  difficult	
  to	
  separate	
  for	
  C.	
  difficile	
  
since	
  its	
  ability	
  to	
  compete	
  with	
  the	
  intestinal	
  microbiota	
  and	
  colonize	
  is	
  such	
  an	
  
integral	
  part	
  of	
  its	
  pathogenesis.	
  	
  Nonetheless,	
  more	
  investigation	
  in	
  this	
  area	
  is	
  
warranted	
  and	
  would	
  provide	
  much	
  needed	
  insight	
  into	
  important	
  metabolic	
  
strategies	
  of	
  C.	
  difficile.	
  	
  
	
  
Proteolysis,	
  amino	
  acids,	
  and	
  stickland	
  fermentation.	
  	
  In	
  1989,	
  Seddon	
  and	
  
Borriello	
  developed	
  the	
  first	
  chemically	
  defined	
  medium	
  specifically	
  for	
  growth	
  of	
  C.	
  
difficile	
  (110).	
  	
  	
  Components	
  included	
  9	
  amino	
  acids,	
  N-­‐acetylglucosamine,	
  and	
  a	
  mix	
  
of	
  vitamins	
  and	
  minerals.	
  	
  They	
  tested	
  10	
  different	
  strains	
  in	
  this	
  medium,	
  and	
  
growth	
  was	
  adequate,	
  yet	
  poorer	
  than	
  growth	
  in	
  the	
  standard,	
  rich	
  medium	
  typically	
  
used	
  for	
  C.	
  difficile,	
  Brain	
  Heart	
  Infusion	
  (BHI)	
  medium.	
  	
  	
  Several	
  years	
  later,	
  

	
  

23	
  

Karasawa	
  et	
  al.	
  more	
  thoroughly	
  characterized	
  the	
  in	
  vitro	
  growth	
  requirements	
  of	
  C.	
  
difficile	
  by	
  identifying	
  the	
  essential	
  amino	
  acids	
  and	
  vitamins	
  needed	
  for	
  optimum	
  
growth	
  (93).	
  	
  Essential	
  amino	
  acids	
  included	
  cysteine,	
  isoleucine,	
  leucine,	
  proline,	
  
tryptophan,	
  and	
  valine;	
  only	
  three	
  vitamins	
  were	
  found	
  to	
  be	
  essential:	
  biotin,	
  
pantothenate,	
  and	
  pyridoxine.	
  	
  Their	
  optimized	
  defined	
  medium	
  resulted	
  in	
  growth	
  
comparable	
  to	
  that	
  in	
  BHI	
  medium,	
  and	
  was	
  a	
  better	
  alternative	
  for	
  physiological	
  
studies.	
  	
  	
  
Seddon	
  and	
  Borriello	
  observed	
  that	
  C.	
  difficile	
  growth	
  was	
  stimulated	
  when	
  
provided	
  a	
  source	
  of	
  peptides,	
  such	
  as	
  proteose	
  peptone	
  or	
  casein	
  hydrolysate,	
  
hypothesizing	
  that	
  it	
  may	
  prefer	
  to	
  grow	
  on	
  short	
  peptides	
  rather	
  than	
  free	
  amino	
  
acids.	
  	
  To	
  further	
  investigate	
  the	
  proteolytic	
  capabilities	
  of	
  C.	
  difficile,	
  they	
  screened	
  
10	
  strains	
  using	
  various	
  proteolytic	
  assays	
  (111).	
  	
  They	
  used	
  plate	
  assays	
  as	
  well	
  as	
  a	
  
variety	
  of	
  biochemical	
  assays	
  using	
  both	
  whole	
  cell	
  suspensions	
  and	
  supernatant	
  to	
  
separate	
  cell-­‐associated	
  and	
  extracellular	
  enzymatic	
  activity.	
  	
  In	
  agreement	
  with	
  their	
  
previous	
  work	
  investigating	
  the	
  hydrolytic	
  activities	
  of	
  these	
  strains,	
  they	
  found	
  
variable	
  proteolytic	
  activity	
  among	
  the	
  strains	
  with	
  no	
  direct	
  correlation	
  to	
  toxigenic	
  
or	
  virulence	
  status;	
  however,	
  they	
  again	
  observed	
  a	
  trend	
  of	
  the	
  most	
  virulent	
  strains	
  
having	
  the	
  most	
  proteolytic	
  activity.	
  	
  Furthermore,	
  they	
  confirmed	
  the	
  collagenase	
  
activity	
  previously	
  reported,	
  that	
  there	
  was	
  no	
  albuminase	
  activity,	
  and	
  no	
  
chymotrypsin	
  activity.	
  	
  They	
  did	
  observe	
  the	
  presence	
  of	
  an	
  enzyme	
  with	
  trypsin-­‐like	
  
activity	
  and	
  further	
  biochemical	
  characterization	
  of	
  the	
  partially	
  purified	
  enzyme	
  
suggested	
  it	
  was	
  similar	
  to	
  the	
  known	
  proteolytic	
  virulence	
  factor,	
  clostripain,	
  of	
  C.	
  
histolyticum,	
  another	
  Clostridial	
  pathogen.	
  	
  	
  Finally,	
  the	
  authors	
  proposed	
  that	
  C.	
  

	
  

24	
  

difficile’s	
  proteolytic	
  activity	
  allows	
  for	
  generation	
  of	
  amino	
  acid	
  sources	
  for	
  
metabolism	
  and	
  growth	
  and	
  therefore	
  may	
  contribute	
  to	
  the	
  virulence	
  of	
  this	
  
organism.	
  	
  For	
  example,	
  ATP	
  generation	
  through	
  fermentation	
  of	
  amino	
  acids	
  has	
  
been	
  demonstrated	
  and	
  will	
  be	
  discussed	
  next.	
  	
  
First	
  discovered	
  in	
  1935,	
  Stickland	
  observed	
  that	
  extracts	
  from	
  Clostridium	
  
sporogenes	
  were	
  able	
  to	
  ferment	
  amino	
  acids	
  in	
  a	
  pair-­‐wise	
  fashion	
  in	
  which	
  one	
  
amino	
  acid	
  (Stickland	
  donor)	
  was	
  oxidized	
  while	
  another	
  amino	
  acid	
  (Stickland	
  
acceptor)	
  was	
  reduced	
  (112).	
  	
  Subsequent	
  work	
  has	
  shown	
  these	
  reactions	
  to	
  be	
  
utilized	
  by	
  other	
  proteolytic	
  Clostridia.	
  	
  Building	
  upon	
  the	
  previous	
  work	
  showing	
  C.	
  
difficile’s	
  proteoltic	
  activity	
  and	
  ability	
  to	
  utilize	
  free	
  amino	
  acids,	
  Jackson	
  et	
  al.	
  
investigated	
  whether	
  C.	
  difficile	
  was	
  able	
  to	
  also	
  utilize	
  Stickland	
  reactions	
  for	
  growth	
  
(85).	
  	
  Indeed,	
  they	
  showed	
  that	
  C.	
  difficile	
  growth	
  was	
  increased	
  when	
  cells	
  were	
  
grown	
  in	
  minimal	
  medium	
  supplemented	
  with	
  Stickland	
  pairs	
  of	
  amino	
  acids.	
  	
  	
  
Furthermore,	
  this	
  metabolic	
  activity	
  was	
  dependent	
  on	
  the	
  presence	
  of	
  selenium	
  in	
  
the	
  medium.	
  	
  	
  Using	
  the	
  newly	
  sequenced	
  and	
  annotated	
  genome	
  data	
  from	
  the	
  C.	
  
difficile	
  type	
  strain,	
  CD630	
  (96),	
  they	
  identified	
  two	
  selenoenzymes	
  likely	
  to	
  be	
  
involved	
  in	
  these	
  reactions,	
  glycine	
  reductase,	
  and	
  D-­‐proline	
  reductase.	
  	
  	
  Expression	
  
analysis	
  showed	
  both	
  of	
  these	
  enzymes	
  were	
  expressed	
  in	
  the	
  presence	
  of	
  the	
  
appropriate	
  Stickland	
  donor	
  amino	
  acids	
  (glycine,	
  proline,	
  or	
  hydroxyproline).	
  	
  They	
  
were	
  also	
  able	
  to	
  purify	
  and	
  further	
  characterize	
  D-­‐proline	
  reductase,	
  observing	
  that	
  
the	
  biochemistry	
  of	
  this	
  enzyme	
  is	
  slightly	
  different	
  from	
  other	
  known	
  proline	
  
reductases	
  since	
  it	
  is	
  does	
  not	
  require	
  divalent	
  cations	
  and	
  is	
  inhibited	
  by	
  zinc.	
  	
  This	
  
was	
  the	
  first	
  demonstration	
  that	
  Stickland	
  amino	
  acid	
  fermentation	
  is	
  important	
  for	
  

	
  

25	
  

growth	
  of	
  C.	
  difficile	
  and	
  the	
  identification	
  of	
  two	
  enzymes	
  that	
  may	
  play	
  critical	
  roles	
  
in	
  this	
  metabolic	
  function.	
  	
  Moreover,	
  if	
  Stickland	
  fermentations	
  are	
  critical	
  for	
  
optimum	
  growth	
  in	
  the	
  intestinal	
  environment,	
  is	
  suggests	
  that	
  the	
  availability	
  of	
  
selenium	
  could	
  be	
  an	
  important	
  growth	
  factor.	
  	
  More	
  recent	
  work	
  has	
  gone	
  on	
  to	
  
investigate	
  the	
  transcriptional	
  regulation	
  of	
  the	
  glycine	
  reductase	
  and	
  D-­‐proline	
  
reductase	
  gene	
  clusters	
  and	
  has	
  identified	
  that	
  it	
  is	
  proline-­‐dependent	
  and	
  mediated	
  
by	
  the	
  protein	
  PrdR	
  (86).	
  	
  	
  	
  	
  
	
  Based	
  on	
  the	
  evidence	
  suggesting	
  the	
  importance	
  of	
  Stickland	
  fermentation	
  
for	
  C.	
  difficile	
  metabolism	
  and	
  that	
  the	
  organisms	
  utilizing	
  this	
  metabolic	
  strategy	
  are	
  
limited,	
  Wu	
  et	
  al.	
  hypothesized	
  that	
  the	
  enzymes	
  involved	
  could	
  be	
  potential	
  narrow-­‐
spectrum	
  drug	
  targets	
  for	
  development	
  of	
  novel	
  C.	
  difficile	
  treatment	
  (87).	
  	
  	
  
Furthermore,	
  recent	
  transcriptomic	
  and	
  proteomic	
  studies	
  have	
  demonstrated	
  that	
  
these	
  enzymes	
  are	
  expressed	
  in	
  in	
  vivo	
  models	
  (88,	
  98).	
  	
  The	
  activity	
  of	
  D-­‐proline	
  
reductase	
  requires	
  the	
  conversion	
  of	
  L-­‐proline	
  to	
  D-­‐proline,	
  a	
  function	
  carried	
  out	
  
PrdF,	
  D-­‐proline	
  racemase.	
  	
  	
  An	
  insertional	
  knock-­‐out	
  mutant	
  of	
  PrdF	
  showed	
  no	
  
significant	
  reduction	
  in	
  growth	
  rate	
  in	
  early	
  log	
  phase;	
  however,	
  mid	
  and	
  late-­‐log	
  
phase	
  growth	
  was	
  reduced	
  resulting	
  in	
  significantly	
  lower	
  final	
  growth	
  yield	
  of	
  the	
  
mutant.	
  	
  	
  No	
  affect	
  on	
  virulence	
  was	
  reported	
  as	
  there	
  was	
  no	
  difference	
  in	
  survival	
  
time	
  of	
  hamsters	
  infected	
  with	
  the	
  wild	
  type	
  or	
  mutant	
  strains	
  and	
  there	
  was	
  no	
  
significant	
  difference	
  in	
  toxin	
  expression	
  in	
  vitro	
  (87).	
  	
  The	
  generation	
  of	
  knock-­‐out	
  
mutants	
  of	
  several	
  genes	
  in	
  the	
  Stickland	
  fermentation	
  pathways	
  in	
  this	
  and	
  the	
  
Bouillaut	
  et	
  al.	
  study	
  (86)	
  show	
  that	
  these	
  are	
  not	
  essential	
  for	
  growth	
  in	
  vitro.	
  	
  
Additionally,	
  lack	
  of	
  affect	
  of	
  the	
  PrdF	
  knock-­‐out	
  on	
  virulence	
  suggests	
  that	
  Stickland	
  

	
  

26	
  

fermentation	
  is	
  not	
  essential	
  in	
  vivo	
  either.	
  	
  However,	
  more	
  in	
  vivo	
  work	
  in	
  this	
  area	
  
investigating	
  the	
  role	
  of	
  this	
  and	
  other	
  Stickland	
  pathway	
  enzymes	
  on	
  colonization	
  
levels	
  is	
  warranted	
  to	
  determine	
  of	
  Stickland	
  metabolism	
  is	
  important	
  for	
  C.	
  difficile	
  
growth	
  and	
  competition	
  in	
  the	
  intestinal	
  environment.	
  	
  	
  	
  
	
  
“Omics”	
  approaches	
  to	
  investigating	
  metabolism.	
  	
  
Genomics.	
  	
  The	
  first	
  C.	
  difficile	
  whole	
  genome	
  sequence	
  and	
  analysis	
  was	
  published	
  by	
  
Sebaihia	
  et	
  al.	
  in	
  2006	
  for	
  strain	
  CD630,	
  a	
  virulent	
  strain	
  isolated	
  from	
  a	
  patient	
  with	
  
severe	
  pseudomembranous	
  colitis	
  in	
  Switzerland	
  in	
  1982	
  (96).	
  	
  CD630	
  has	
  a	
  4.3	
  Mbp	
  
chromosome	
  (and	
  also	
  a	
  7.9	
  kbp	
  plasmid),	
  with	
  an	
  unusually	
  high	
  (11%)	
  percentage	
  
of	
  mobile	
  elements	
  and	
  a	
  low	
  G+C	
  content	
  (29%).	
  	
  In	
  terms	
  of	
  energy	
  metabolism,	
  
genome	
  annotation	
  identified	
  a	
  large	
  amount	
  of	
  genes	
  involved	
  in	
  carbohydrate	
  
transport	
  and	
  metabolism	
  as	
  well	
  as	
  genes	
  involved	
  in	
  degradation	
  of	
  ethanolamine,	
  a	
  
bacterial	
  carbon	
  and	
  nitrogen	
  source.	
  	
  	
  Ethanolamine	
  is	
  a	
  component	
  of	
  the	
  
phosphatidylethanolamine,	
  a	
  common	
  phospholipid	
  of	
  biological	
  membranes,	
  and	
  
hence	
  is	
  abundant	
  in	
  the	
  gastrointestinal	
  environment	
  due	
  to	
  the	
  multitude	
  of	
  
microbes	
  present	
  as	
  well	
  as	
  the	
  sloughing	
  off	
  of	
  host	
  mucosal	
  epithelia	
  (113).	
  	
  	
  The	
  
ability	
  of	
  several	
  gut-­‐associated	
  bacteria	
  to	
  degrade	
  ethanolamine	
  has	
  been	
  reported	
  
(114),	
  as	
  has	
  its	
  role	
  in	
  virulence	
  and	
  colonization	
  for	
  several	
  pathogens	
  (115).	
  	
  
Recently,	
  a	
  structural	
  and	
  biochemical	
  study	
  of	
  the	
  proteins	
  encoded	
  in	
  the	
  C.	
  difficile	
  
ethanolamine	
  gene	
  cluster	
  and	
  microcompartment	
  formation	
  was	
  reported	
  (94).	
  	
  
However,	
  the	
  direct	
  role	
  of	
  ethanolamine	
  utilization	
  in	
  C.	
  difficile	
  colonization	
  or	
  
virulence	
  has	
  not	
  been	
  investigated.	
  	
  	
  

	
  

27	
  

	
  
Transcriptomics	
  and	
  proteomics.	
  	
  The	
  availability	
  of	
  whole	
  genome	
  sequence	
  data,	
  
especially	
  through	
  web-­‐based	
  microbial	
  genomics	
  sources	
  such	
  as	
  IMG	
  (Integrated	
  
Microbial	
  Genomes)	
  and	
  NCBI	
  (National	
  Center	
  for	
  Biotechnology	
  Information),	
  has	
  
facilitated	
  utilization	
  of	
  “omics”	
  approaches	
  to	
  investigate	
  C.	
  difficile	
  colonization.	
  	
  	
  	
  
Several	
  studies	
  have	
  applied	
  transcriptomics	
  and	
  proteomics	
  to	
  determine	
  which	
  
genes	
  or	
  proteins	
  are	
  differentially	
  expressed	
  under	
  in	
  vivo	
  vs	
  in	
  vitro	
  conditions	
  (88,	
  
97,	
  98,	
  116,	
  117).	
  	
  	
  The	
  focus	
  of	
  many	
  of	
  these	
  studies	
  is	
  on	
  identifying	
  genes	
  related	
  
more	
  directly	
  to	
  virulence	
  (toxin	
  expression)	
  and	
  stress	
  tolerance,	
  yet	
  some	
  data	
  has	
  
emerged	
  providing	
  information	
  about	
  the	
  metabolic	
  strategies	
  employed	
  in	
  vivo.	
  	
  	
  
In	
  an	
  effort	
  to	
  further	
  understand	
  C.	
  difficile	
  adaptation	
  to	
  the	
  host	
  
environment,	
  Scaria	
  et	
  al.	
  developed	
  a	
  pig	
  ileal-­‐ligated	
  loop	
  model	
  and	
  used	
  it	
  to	
  
determine	
  the	
  transcriptional	
  profile	
  of	
  C.	
  difficile	
  4,	
  8,	
  and	
  12	
  hours	
  post-­‐infection	
  
(97).	
  	
  	
  	
  Comparison	
  of	
  microarray	
  data	
  between	
  in	
  vivo	
  (pig	
  ligated	
  loop	
  model)	
  and	
  in	
  
vitro	
  (BHI	
  medium)	
  conditions	
  revealed	
  upregulation	
  of	
  several	
  pathways	
  involved	
  in	
  
transport	
  and	
  metabolism	
  of	
  amino	
  acids	
  and	
  various	
  carbohydrates,	
  anaerobic	
  
respiration,	
  and	
  lipid	
  degradation.	
  	
  In	
  particular,	
  the	
  data	
  suggested	
  metabolism	
  of	
  
xylose,	
  mannose,	
  and	
  glycogen	
  to	
  be	
  important	
  in	
  vivo.	
  	
  	
  This	
  study	
  was	
  one	
  of	
  the	
  
first	
  transcriptional	
  profiling	
  studies	
  of	
  C.	
  difficile	
  gene	
  expression	
  in	
  an	
  in	
  vivo	
  model.	
  	
  
The	
  same	
  group	
  then	
  used	
  a	
  proteomics	
  approach	
  to	
  look	
  at	
  C.	
  difficile	
  protein	
  
expression	
  in	
  a	
  similar	
  pig	
  ligated	
  loop	
  model	
  (88).	
  	
  Again,	
  they	
  compared	
  samples	
  
from	
  in	
  vitro	
  and	
  in	
  vivo	
  conditions	
  4,	
  8,	
  and	
  12	
  hours	
  post	
  infection;	
  of	
  the	
  705	
  
quantifiable	
  proteins	
  identified,	
  109	
  were	
  differentially	
  expressed.	
  	
  	
  Many	
  proteins	
  

	
  

28	
  

involved	
  in	
  metabolism	
  and	
  energy	
  production	
  were	
  identified,	
  including	
  those	
  
important	
  for	
  amino	
  acid	
  metabolism,	
  carbohydrate	
  metabolism,	
  and	
  Stickland	
  
fermentation.	
  	
  Many	
  of	
  the	
  pathways	
  upregulated	
  in	
  this	
  study	
  overlap	
  with	
  their	
  
previous	
  transcriptomic	
  study;	
  however,	
  some	
  did	
  not	
  and	
  this	
  is	
  probably	
  a	
  
reflection	
  of	
  the	
  differences	
  in	
  the	
  technical	
  challenges	
  of	
  these	
  two	
  techniques,	
  and	
  
inherent	
  differences	
  in	
  the	
  targets	
  of	
  the	
  analyses.	
  	
  	
  Regardless,	
  these	
  studies	
  provide	
  
useful	
  data	
  suggesting	
  important	
  metabolic	
  adaptations	
  of	
  C.	
  difficile	
  to	
  its	
  host	
  
environment	
  and	
  can	
  be	
  mined	
  for	
  potential	
  functions	
  to	
  more	
  directly	
  test	
  through	
  
genetic	
  manipulation	
  and	
  in	
  vivo	
  colonization	
  and	
  virulence	
  experiments.	
  	
  
Another	
  in	
  vivo	
  transcriptomic	
  study	
  recently	
  published	
  by	
  Janoir	
  et	
  al.	
  used	
  
this	
  comprehensive	
  approach	
  by	
  first	
  identifying	
  genes	
  upregulated	
  in	
  vivo	
  and	
  then	
  
testing	
  knock-­‐out	
  mutants	
  of	
  those	
  genes	
  for	
  defects	
  in	
  colonization	
  (98).	
  	
  They	
  
compared	
  temporal	
  gene	
  expression	
  in	
  vitro	
  (TY	
  medium)	
  vs	
  in	
  vivo,	
  using	
  C.	
  difficile-­‐
monoassociated	
  mice;	
  549	
  differentially	
  expressed	
  genes	
  were	
  identified,	
  20%	
  of	
  
which	
  were	
  ascribed	
  to	
  metabolic	
  pathways.	
  	
  Like	
  the	
  other	
  in	
  vivo	
  transcriptomic	
  
and	
  proteomic	
  studies	
  in	
  the	
  pig	
  ileal-­‐ligated	
  loop	
  model,	
  the	
  pathways	
  included	
  lipid	
  
and	
  carbohydrate	
  transport	
  and	
  metabolism,	
  and	
  fermentation	
  of	
  carbohydrates	
  and	
  
amino	
  acids.	
  	
  In	
  addition,	
  genes	
  involved	
  in	
  glycogen	
  synthesis	
  and	
  degradation	
  of	
  
polysaccharides	
  were	
  upregulated	
  in	
  vivo.	
  	
  Evidence	
  for	
  a	
  preference	
  for	
  the	
  
leucine/proline	
  Stickland	
  fermentation	
  pair	
  was	
  also	
  apparent	
  in	
  the	
  data.	
  	
  
Interestingly,	
  one-­‐sixth	
  of	
  the	
  upregulated	
  genes	
  were	
  of	
  unknown	
  function,	
  two	
  of	
  
which	
  were	
  selected	
  for	
  further	
  investigation.	
  	
  Genetic	
  knock-­‐out	
  mutants	
  of	
  these	
  
two	
  genes	
  were	
  tested	
  against	
  the	
  wild	
  type	
  strain	
  for	
  decreased	
  colonization	
  fitness.	
  	
  

	
  

29	
  

One	
  of	
  them,	
  CD1581,	
  had	
  significantly	
  lower	
  colonization	
  in	
  the	
  co-­‐infection	
  
experiment,	
  providing	
  direct	
  evidence	
  for	
  the	
  importance	
  of	
  the	
  product	
  of	
  this	
  gene	
  
for	
  colonization	
  in	
  vivo.	
  	
  The	
  observation	
  that	
  the	
  other	
  hypothetical	
  protein	
  mutant	
  
did	
  not	
  show	
  a	
  colonization	
  defect	
  even	
  though	
  the	
  gene	
  was	
  differentially	
  expressed	
  
supports	
  that	
  colonization	
  factors	
  identified	
  in	
  these	
  initial	
  genetic	
  studies	
  must	
  be	
  
directly	
  tested	
  in	
  follow-­‐up	
  experiments.	
  	
  	
  	
  	
  	
  
	
  
Metabolomics.	
  	
  Yet	
  another	
  approach	
  to	
  understanding	
  the	
  metabolic	
  strategies	
  used	
  
by	
  C.	
  difficile	
  for	
  colonization	
  was	
  recently	
  published	
  by	
  Theriot	
  et	
  al.	
  (82).	
  	
  They	
  used	
  
metabolomics	
  to	
  first	
  identify	
  changes	
  in	
  the	
  available	
  nutrients	
  between	
  caecal	
  
contents	
  of	
  C.	
  difficile	
  resistant	
  and	
  susceptible	
  (antibiotic-­‐treated)	
  mice,	
  identifying	
  
significant	
  changes	
  in	
  levels	
  of	
  some	
  amino	
  acids,	
  carbohydrates,	
  lipids,	
  peptides	
  and	
  
xenobiotics.	
  	
  In	
  vitro	
  growth	
  experiments	
  confirmed	
  the	
  ability	
  of	
  C.	
  difficile	
  to	
  utilize	
  
some	
  of	
  the	
  carbohydrates	
  shown	
  to	
  increase	
  in	
  the	
  C.	
  difficile-­‐susceptible	
  metabolic	
  
environment,	
  namely	
  glucose,	
  fructose,	
  mannitol,	
  sorbitol,	
  raffinose,	
  and	
  stachyose.	
  	
  
Additionally,	
  they	
  identified	
  two	
  different	
  C.	
  difficile-­‐resistant	
  states,	
  one	
  before	
  
antibiotic	
  treatment,	
  and	
  one	
  six	
  weeks	
  after	
  antibiotic	
  treatment.	
  	
  Interestingly,	
  the	
  
bacterial	
  communities	
  between	
  these	
  states	
  were	
  very	
  different,	
  however,	
  their	
  
metabolomic	
  profiles	
  were	
  strikingly	
  similar,	
  especially	
  with	
  regard	
  to	
  compounds	
  
found	
  to	
  be	
  important	
  for	
  C.	
  difficile	
  germination	
  and	
  growth.	
  	
  This	
  supports	
  that	
  the	
  
functions	
  of	
  the	
  intestinal	
  microbiota,	
  which	
  shape	
  the	
  metabolic	
  environment,	
  rather	
  
than	
  the	
  specific	
  community	
  member	
  are	
  more	
  important	
  for	
  resistance	
  to	
  C.	
  difficile.	
  	
  	
  
This	
  study	
  took	
  a	
  novel	
  approach	
  to	
  understanding	
  C.	
  difficile	
  colonization	
  by	
  first	
  

	
  

30	
  

characterizing	
  the	
  nutrients	
  available	
  to	
  C.	
  difficile,	
  then	
  testing	
  its	
  ability	
  to	
  utilize	
  
these	
  compounds.	
  	
  In	
  contrast,	
  previous	
  studies	
  first	
  identified	
  C.	
  difficile	
  compound	
  
utilization	
  by	
  genomic	
  and	
  phenotypic	
  analyses,	
  then	
  tested	
  if	
  these	
  affected	
  
colonization.	
  	
  Additional	
  studies	
  combining	
  data	
  from	
  these	
  two	
  complementary	
  
approaches	
  will	
  undoubtedly	
  provide	
  even	
  more	
  insight	
  into	
  understanding	
  C.	
  difficile	
  
intestinal	
  colonization	
  and	
  bring	
  us	
  closer	
  to	
  development	
  of	
  novel	
  ways	
  to	
  prevent	
  
and	
  treat	
  CDI.	
  	
  	
  
	
  
Autotrophy.	
  	
  A	
  completely	
  novel	
  growth	
  strategy	
  used	
  by	
  C.	
  difficile	
  was	
  recently	
  
discovered	
  by	
  Kopke	
  et	
  al.	
  (95).	
  	
  This	
  was	
  an	
  exciting	
  discovery	
  as	
  it	
  was	
  the	
  first	
  time	
  
a	
  bacterial	
  pathogen	
  was	
  demonstrated	
  to	
  be	
  able	
  to	
  grow	
  autotrophically.	
  	
  	
  Metabolic	
  
byproducts	
  of	
  the	
  microbial	
  breakdown	
  of	
  some	
  intestinal	
  saccharide	
  and	
  protein	
  
sources	
  include	
  carbon	
  dioxide	
  and	
  hydrogen	
  gases.	
  	
  	
  C.	
  difficile	
  can	
  in	
  turn	
  fix	
  the	
  CO2	
  
and	
  form	
  acetate	
  by	
  way	
  of	
  the	
  Wood-­‐Ljungdahl	
  pathway;	
  several	
  genes	
  belonging	
  to	
  
this	
  pathway	
  are	
  conserved	
  among	
  C.	
  difficile	
  isolates.	
  	
  Indeed,	
  it	
  was	
  demonstrated	
  
that	
  C.	
  difficile	
  can	
  grow	
  on	
  CO2	
  and	
  H2	
  as	
  the	
  sole	
  source	
  of	
  carbon	
  and	
  energy	
  in	
  a	
  
defined	
  medium	
  (95).	
  	
  Experiments	
  designed	
  to	
  directly	
  test	
  the	
  contribution	
  of	
  
autotrophic	
  growth	
  to	
  in	
  vivo	
  colonization	
  are	
  needed;	
  however,	
  the	
  ability	
  of	
  C.	
  
difficile	
  to	
  grow	
  autotrophically	
  adds	
  another	
  growth	
  strategy	
  to	
  C.	
  difficile’s	
  
metabolic	
  repertoire.	
  	
  	
  	
  	
  	
  	
  
Summary	
  	
  
Disease	
  caused	
  by	
  C.	
  difficile	
  is	
  a	
  common	
  complication	
  in	
  antibiotic	
  treatment.	
  	
  Until	
  
recently,	
  it	
  was	
  not	
  considered	
  a	
  significant	
  nosocomial	
  threat.	
  	
  However,	
  increases	
  in	
  

	
  

31	
  

disease	
  prevalence	
  and	
  severity	
  have	
  generated	
  an	
  urgency	
  to	
  understand	
  the	
  
epidemiology	
  and	
  etiology	
  of	
  this	
  disease.	
  	
  	
  A	
  likely	
  contributing	
  factor	
  to	
  the	
  
increased	
  disease	
  incidence	
  is	
  the	
  emergence	
  of	
  epidemic	
  strains	
  that	
  appear	
  to	
  
spread	
  more	
  readily	
  and	
  potentially	
  cause	
  more	
  severe	
  disease.	
  	
  C.	
  difficile	
  infection	
  is	
  
distinctive	
  in	
  that	
  there	
  is	
  a	
  clear	
  role	
  of	
  the	
  intestinal	
  microbiota	
  in	
  the	
  development	
  
of	
  disease.	
  	
  However,	
  the	
  specific	
  mechanisms	
  that	
  govern	
  outgrowth	
  of	
  C.	
  difficile	
  in	
  
the	
  colon	
  are	
  poorly	
  understood.	
  	
  One	
  hypothesis	
  regarding	
  inhibition	
  of	
  C.	
  difficile	
  in	
  
the	
  absence	
  of	
  antibiotic	
  treatment	
  suggests	
  that	
  members	
  of	
  the	
  microbiota	
  
outcompete	
  C.	
  difficile	
  for	
  key	
  nutrient	
  sources.	
  	
  Several	
  studies	
  have	
  been	
  published	
  
providing	
  evidence	
  that	
  supports	
  this	
  hypothesis.	
  	
  There	
  is	
  some	
  understanding	
  of	
  the	
  
general	
  metabolic	
  capabilities	
  of	
  C.	
  difficile,	
  however,	
  which	
  specific	
  nutrients	
  are	
  key	
  
players	
  in	
  colonization	
  resistance	
  and	
  how	
  much	
  of	
  an	
  impact	
  they	
  have	
  on	
  the	
  ability	
  
of	
  C.	
  difficile	
  to	
  outgrow	
  in	
  the	
  colon	
  is	
  poorly	
  understood.	
  	
  Understanding	
  all	
  of	
  the	
  
strategies	
  C.	
  difficile	
  uses	
  to	
  compete	
  for	
  the	
  limited	
  resources	
  available	
  in	
  the	
  
gastrointestinal	
  tract	
  could	
  provide	
  insight	
  into	
  how	
  the	
  microbiota	
  impacts	
  disease	
  
progression,	
  and	
  ultimately	
  lead	
  to	
  new	
  infection	
  control	
  strategies.	
  	
  Furthermore,	
  it	
  
is	
  unclear	
  if	
  characteristics	
  of	
  the	
  newly	
  emerged	
  epidemic	
  strains	
  related	
  to	
  
colonization	
  and	
  competitive	
  dynamics	
  within	
  the	
  intestinal	
  environment	
  contribute	
  
to	
  their	
  increased	
  prevalence	
  and	
  ability	
  to	
  cause	
  disease.	
  	
  	
  The	
  work	
  presented	
  in	
  
this	
  thesis	
  aims	
  to	
  address	
  some	
  of	
  these	
  issues.	
  	
  	
  Chapter	
  2	
  presents	
  the	
  development	
  
of	
  an	
  in	
  vitro	
  model	
  to	
  study	
  fecal	
  bacterial	
  communities,	
  which	
  was	
  adapted	
  in	
  
chapter	
  3	
  as	
  a	
  C.	
  difficile	
  infection	
  model.	
  	
  We	
  used	
  the	
  in	
  vitro	
  C.	
  difficile	
  infection	
  
model	
  to	
  show	
  that	
  in	
  the	
  presence	
  of	
  complex	
  fecal	
  microbiota,	
  epidemic	
  ribotype	
  

	
  

32	
  

027	
  strains	
  outcompete	
  strains	
  of	
  other	
  ribotypes.	
  	
  Also	
  presented	
  are	
  in	
  vivo	
  data	
  
supporting	
  this	
  competitive	
  advantage	
  of	
  the	
  RT	
  027	
  strains	
  in	
  a	
  mouse	
  infection	
  
model.	
  	
  Chapter	
  4	
  focuses	
  on	
  the	
  growth	
  metabolism	
  of	
  C.	
  difficile	
  and	
  shows	
  that	
  
epidemic	
  strains	
  are	
  able	
  to	
  grow	
  more	
  efficiently	
  on	
  the	
  disaccharide	
  trehalose.	
  	
  	
  
This	
  metabolic	
  characteristic	
  of	
  epidemic	
  strains	
  may	
  contribute	
  to	
  the	
  competitive	
  
advantage	
  we	
  observed	
  in	
  our	
  in	
  vitro	
  and	
  in	
  vivo	
  models.	
  	
  	
  	
  Moreover,	
  if	
  the	
  
importance	
  of	
  trehalose	
  metabolism	
  translates	
  into	
  the	
  human	
  intestinal	
  
environment,	
  this	
  work	
  provides	
  insight	
  into	
  defining	
  the	
  intestinal	
  niche	
  of	
  C.	
  diffcile,	
  
and	
  understanding	
  the	
  evolution	
  and	
  emergence	
  of	
  epidemic	
  strains.	
  	
  	
  
	
  

	
  	
  

	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

33	
  

REFERENCES	
  

	
  

34	
  

	
  
	
  
1.	
  

REFERENCES	
  
	
  
Aminov	
  RI.	
  2010.	
  A	
  brief	
  history	
  of	
  the	
  antibiotic	
  era:	
  lessons	
  learned	
  and	
  
challenges	
  for	
  the	
  future.	
  Front	
  Microbiol	
  1:134.	
  

2.	
  

Bartlett	
  JG.	
  2002.	
  Antibiotic-­‐associated	
  diarrhea.	
  New	
  England	
  Journal	
  of	
  
Medicine	
  346:334–339.	
  

3.	
  

Högenauer	
  C,	
  Hammer	
  HF,	
  Krejs	
  GJ,	
  Reisinger	
  C.	
  1998.	
  Mechanisms	
  and	
  
management	
  of	
  antibiotic-­‐associated	
  diarrhea.	
  Clin.	
  Infect.	
  Dis.	
  27:702–710.	
  

4.	
  

Qin	
  J,	
  Li	
  R,	
  Raes	
  J,	
  Arumugam	
  M,	
  Burgdorf	
  KS,	
  Manichanh	
  C,	
  Nielsen	
  T,	
  
Pons	
  N,	
  Levenez	
  F,	
  Yamada	
  T,	
  Mende	
  DR,	
  Li	
  J,	
  Xu	
  J,	
  Li	
  S,	
  Li	
  D,	
  Cao	
  J,	
  Wang	
  
B,	
  Liang	
  H,	
  Zheng	
  H,	
  Xie	
  Y,	
  Tap	
  J,	
  Lepage	
  P,	
  Bertalan	
  M,	
  Batto	
  J-­‐M,	
  Hansen	
  
T,	
  Le	
  Paslier	
  D,	
  Linneberg	
  A,	
  Nielsen	
  HB,	
  Pelletier	
  E,	
  Renault	
  P,	
  Sicheritz-­‐
Ponten	
  T,	
  Turner	
  K,	
  Zhu	
  H,	
  Yu	
  C,	
  Li	
  S,	
  Jian	
  M,	
  Zhou	
  Y,	
  Li	
  Y,	
  Zhang	
  X,	
  Li	
  S,	
  
Qin	
  N,	
  Yang	
  H,	
  Wang	
  J,	
  Brunak	
  S,	
  Doré	
  J,	
  Guarner	
  F,	
  Kristiansen	
  K,	
  
Pedersen	
  O,	
  Parkhill	
  J,	
  Weissenbach	
  J,	
  MetaHIT	
  Consortium,	
  Bork	
  P,	
  
Ehrlich	
  SD,	
  Wang	
  J.	
  2010.	
  A	
  human	
  gut	
  microbial	
  gene	
  catalogue	
  established	
  
by	
  metagenomic	
  sequencing.	
  Nature	
  464:59–65.	
  

5.	
  

de	
  Vos	
  WM,	
  de	
  Vos	
  EA.	
  2012.	
  Role	
  of	
  the	
  intestinal	
  microbiome	
  in	
  health	
  and	
  
disease:	
  from	
  correlation	
  to	
  causation.	
  Nutrition	
  Reviews	
  70:S45–S56.	
  

6.	
  

Stecher	
  B,	
  Hardt	
  W-­‐D.	
  2008.	
  The	
  role	
  of	
  microbiota	
  in	
  infectious	
  disease.	
  
Trends	
  in	
  Microbiology	
  16:107–114.	
  

7.	
  

Paterson	
  DL.	
  2004.	
  “Collateral	
  Damage”	
  from	
  Cephalosporin	
  or	
  Quinolone	
  
Antibiotic	
  Therapy.	
  CLIN	
  INFECT	
  DIS	
  38:S341–S345.	
  

8.	
  

Ayyagari	
  A,	
  Agarwal	
  J,	
  Garg	
  A.	
  2003.	
  Antibiotic	
  associated	
  diarrhoea:	
  
infectious	
  causes.	
  Indian	
  J	
  Med	
  Microbiol	
  21:6–11.	
  

9.	
  

Yassin	
  SF,	
  Young-­‐Fadok	
  TM,	
  Zein	
  NN,	
  Pardi	
  DS.	
  2001.	
  Clostridium	
  difficile-­‐
associated	
  diarrhea	
  and	
  colitis.	
  Mayo	
  Clin.	
  Proc.	
  76:725–730.	
  

10.	
  

Kelly	
  CP,	
  Pothoulakis	
  C,	
  LaMont	
  JT.	
  1994.	
  Clostridium	
  difficile	
  Colitis.	
  N.	
  
Engl.	
  J.	
  Med.	
  330:257–262.	
  

11.	
  

HALL	
  IC.	
  1935.	
  Intestinal	
  flora	
  in	
  new-­‐born	
  infants.	
  Am	
  J	
  Dis	
  Child	
  49:390–
402.	
  

12.	
  

Wilson	
  KH.	
  1993.	
  The	
  microecology	
  of	
  Clostridium	
  difficile.	
  Clin.	
  Infect.	
  Dis.	
  
16	
  Suppl	
  4:S214–8.	
  

	
  

35	
  

13.	
  

Rotimi	
  VO,	
  Duerden	
  BI.	
  1981.	
  The	
  development	
  of	
  the	
  bacterial	
  flora	
  in	
  
normal	
  neonates.	
  Journal	
  of	
  Medical	
  Microbiology	
  14:51–62.	
  

14.	
  

SCHAEDLER	
  RW,	
  DUBOS	
  R,	
  COSTELLO	
  R.	
  1965.	
  The	
  Development	
  Of	
  The	
  
Bacterial	
  Flora	
  In	
  The	
  Gastrointestinal	
  Tract	
  Of	
  Mice.	
  J.	
  Exp.	
  Med.	
  122:59–66.	
  

15.	
  

Favier	
  CF,	
  Vaughan	
  EE,	
  de	
  Vos	
  WM,	
  Akkermans	
  ADL.	
  2002.	
  Molecular	
  
monitoring	
  of	
  succession	
  of	
  bacterial	
  communities	
  in	
  human	
  neonates.	
  Appl	
  
Microbiol	
  68:219–226.	
  

16.	
  

Koenig	
  JE,	
  Spor	
  A,	
  Scalfone	
  N,	
  Fricker	
  AD,	
  Stombaugh	
  J,	
  Knight	
  R,	
  
Angenent	
  LT,	
  Ley	
  RE.	
  2011.	
  Succession	
  of	
  microbial	
  consortia	
  in	
  the	
  
developing	
  infant	
  gut	
  microbiome.	
  Proceedings	
  of	
  the	
  National	
  Academy	
  of	
  
Sciences	
  108	
  Suppl	
  1:4578–4585.	
  

17.	
  

Rea	
  MC,	
  O'Sullivan	
  O,	
  Shanahan	
  F,	
  O'Toole	
  PW,	
  Stanton	
  C,	
  Ross	
  RP,	
  Hill	
  C.	
  
2012.	
  Clostridium	
  difficile	
  Carriage	
  in	
  Elderly	
  Subjects	
  and	
  Associated	
  
Changes	
  in	
  the	
  Intestinal	
  Microbiota.	
  Journal	
  of	
  Clinical	
  Microbiology	
  50:867–
875.	
  

18.	
  

Roberts	
  K,	
  Smith	
  CF,	
  Snelling	
  AM,	
  Kerr	
  KG,	
  Banfield	
  KR,	
  Sleigh	
  PA,	
  Beggs	
  
CB.	
  2008.	
  Aerial	
  Dissemination	
  of	
  Clostridium	
  difficile	
  spores.	
  BMC	
  Infect	
  Dis	
  
8:7.	
  

19.	
  

Hellickson	
  LA,	
  Owens	
  KL.	
  2007.	
  Cross-­‐Contamination	
  of	
  Clostridium	
  difficile	
  
Spores	
  on	
  Bed	
  Linen	
  During	
  Laundering.	
  American	
  Journal	
  of	
  Infection	
  
Control	
  35:E32–E33.	
  

20.	
  

Jabbar	
  U	
  ,	
  BA,	
  Leischner	
  J	
  ,	
  MD,	
  Kasper	
  D	
  ,	
  MD,	
  Gerber	
  R	
  ,	
  MD,	
  Sambol	
  SP	
  ,	
  
BS,	
  MT(A,	
  Parada	
  JP	
  ,	
  MD,	
  MPH,	
  Johnson	
  S	
  ,	
  MD,	
  Gerding	
  DN	
  ,	
  MD.	
  2010.	
  
Effectiveness	
  of	
  Alcohol-­‐Based	
  Hand	
  Rubs	
  for	
  Removal	
  of	
  Clostridium	
  
difficileSpores	
  from	
  Hands.	
  Infection	
  Control	
  and	
  Hospital	
  Epidemiology	
  
31:565–570.	
  

21.	
  

Gerding	
  DN,	
  Muto	
  CA,	
  Owens	
  RC	
  Jr.	
  2008.	
  Measures	
  to	
  Control	
  and	
  Prevent	
  
Clostridium	
  difficileInfection.	
  Clin	
  Infect	
  Dis	
  46:S43–S49.	
  

22.	
  

Rupnik	
  M.	
  2007.	
  Is	
  Clostridium	
  difficile-­‐associated	
  infection	
  a	
  potentially	
  
zoonotic	
  and	
  foodborne	
  disease?	
  Clin	
  Microbiol	
  Infect	
  13:457–459.	
  

23.	
  

Rousseau	
  C,	
  Poilane	
  I,	
  De	
  Pontual	
  L,	
  Maherault	
  A-­‐C,	
  Le	
  Monnier	
  A,	
  
Collignon	
  A.	
  2012.	
  Clostridium	
  difficile	
  carriage	
  in	
  healthy	
  infants	
  in	
  the	
  
community:	
  a	
  potential	
  reservoir	
  for	
  pathogenic	
  strains.	
  Clin	
  Infect	
  Dis	
  
55:1209–1215.	
  

24.	
  

Alam	
  MJ,	
  Anu	
  A,	
  Walk	
  ST,	
  Garey	
  KW.	
  2014.	
  Investigation	
  of	
  potentially	
  
pathogenic	
  Clostridium	
  difficile	
  contamination	
  in	
  household	
  environs.	
  

	
  

36	
  

Anaerobe	
  27C:31–33.	
  
25.	
  

Johnson	
  S.	
  2009.	
  Recurrent	
  Clostridium	
  difficile	
  infection:	
  A	
  review	
  of	
  risk	
  
factors,	
  treatments,	
  and	
  outcomes.	
  Journal	
  of	
  Infection	
  58:403–410.	
  

26.	
  

Wilcox	
  MH,	
  Fawley	
  WN,	
  Settle	
  CD,	
  Davidson	
  A.	
  1998.	
  Recurrence	
  of	
  
symptoms	
  in	
  Clostridium	
  difficile	
  infection—relapse	
  or	
  reinfection?	
  Journal	
  of	
  
Hospital	
  Infection	
  38:93–100.	
  

27.	
  

Figueroa	
  I,	
  Johnson	
  S,	
  Sambol	
  SP,	
  Goldstein	
  EJ,	
  Citron	
  DM,	
  Gerding	
  DN.	
  
2012.	
  Relapse	
  versus	
  reinfection:	
  recurrent	
  Clostridium	
  difficile	
  infection	
  
following	
  treatment	
  with	
  fidaxomicin	
  or	
  vancomycin.	
  Clin.	
  Infect.	
  Dis.	
  
55:S104–S109.	
  

28.	
  

Chang	
  JY,	
  Antonopoulos	
  DA,	
  Kalra	
  A,	
  Tonelli	
  A,	
  Khalife	
  WT,	
  Schmidt	
  TM,	
  
Young	
  VB.	
  2008.	
  Decreased	
  Diversity	
  of	
  the	
  Fecal	
  Microbiome	
  in	
  Recurrent	
  
Clostridium	
  difficile–Associated	
  Diarrhea.	
  J	
  Infect	
  Dis	
  197:435–438.	
  

29.	
  

Peterfreund	
  GL,	
  Vandivier	
  LE,	
  Sinha	
  R,	
  Marozsan	
  AJ,	
  Olson	
  WC,	
  Zhu	
  J,	
  
Bushman	
  FD.	
  2012.	
  Succession	
  in	
  the	
  gut	
  microbiome	
  following	
  antibiotic	
  
and	
  antibody	
  therapies	
  for	
  Clostridium	
  difficile.	
  PLoS	
  One	
  7:e46966.	
  

30.	
  

Jakobsson	
  HE,	
  Jernberg	
  C,	
  Andersson	
  AF,	
  Sjölund-­‐Karlsson	
  M,	
  Jansson	
  JK,	
  
Engstrand	
  L.	
  2010.	
  Short-­‐term	
  antibiotic	
  treatment	
  has	
  differing	
  long-­‐term	
  
impacts	
  on	
  the	
  human	
  throat	
  and	
  gut	
  microbiome.	
  PLoS	
  One	
  5:e9836.	
  

31.	
  

Schubert	
  AM,	
  Rogers	
  MAM,	
  Ring	
  C,	
  Mogle	
  J,	
  Petrosino	
  JP,	
  Young	
  VB,	
  
Aronoff	
  DM,	
  Schloss	
  PD.	
  2014.	
  Microbiome	
  Data	
  Distinguish	
  Patients	
  with	
  
Clostridium	
  difficile	
  Infection	
  and	
  Non-­‐C.	
  difficile-­‐Associated	
  Diarrhea	
  from	
  
Healthy	
  Controls.	
  MBio	
  5:e01021–14–e01021–14.	
  

32.	
  

Jernberg	
  C,	
  Löfmark	
  S,	
  Edlund	
  C,	
  Jansson	
  JK.	
  2007.	
  Long-­‐term	
  ecological	
  
impacts	
  of	
  antibiotic	
  administration	
  on	
  the	
  human	
  intestinal	
  microbiota.	
  ISME	
  
J	
  1:56–66.	
  

33.	
  

Bartlett	
  JG,	
  Chang	
  TW,	
  Gurwith	
  M,	
  Gorbach	
  SL,	
  Onderdonk	
  AB.	
  1978.	
  
Antibiotic-­‐associated	
  pseudomembranous	
  colitis	
  due	
  to	
  toxin-­‐producing	
  
clostridia.	
  N.	
  Engl.	
  J.	
  Med.	
  298:531–534.	
  

34.	
  

McDonald	
  LC,	
  Owings	
  M,	
  Jernigan	
  DB.	
  2006.	
  Clostridium	
  difficile	
  infection	
  
in	
  patients	
  discharged	
  from	
  US	
  short-­‐stay	
  hospitals,	
  1996–2003.	
  Emerg.	
  
Infect.	
  Dis.	
  12:409.	
  

35.	
  

Loo	
  VG,	
  Poirier	
  L,	
  Miller	
  MA,	
  Oughton	
  M,	
  Libman	
  MD,	
  Michaud	
  S,	
  
Bourgault	
  A-­‐M,	
  Nguyen	
  T,	
  Frenette	
  C,	
  Kelly	
  M,	
  Vibien	
  A,	
  Brassard	
  P,	
  Fenn	
  
S,	
  Dewar	
  K,	
  Hudson	
  TJ,	
  Horn	
  R,	
  René	
  P,	
  Monczak	
  Y,	
  Dascal	
  A.	
  2005.	
  A	
  
predominantly	
  clonal	
  multi-­‐institutional	
  outbreak	
  of	
  Clostridium	
  difficile-­‐

	
  

37	
  

associated	
  diarrhea	
  with	
  high	
  morbidity	
  and	
  mortality.	
  N.	
  Engl.	
  J.	
  Med.	
  
353:2442–2449.	
  
36.	
  

Barbut	
  F,	
  Petit	
  JC.	
  2001.	
  Epidemiology	
  of	
  Clostridium	
  difficile-­‐associated	
  
infections.	
  Clin	
  Microbiol	
  Infect	
  7:405–410.	
  

37.	
  

Kuijper	
  EJ,	
  Barbut	
  F,	
  Brazier	
  JS,	
  Kleinkauf	
  N,	
  Eckmanns	
  T,	
  Lambert	
  ML,	
  
Drudy	
  D,	
  Fitzpatrick	
  F,	
  Wiuff	
  C,	
  Brown	
  DJ,	
  Coia	
  JE,	
  Pituch	
  H,	
  Reichert	
  P,	
  
Even	
  J,	
  Mossong	
  J,	
  Widmer	
  AF,	
  Olsen	
  KE,	
  Allerberger	
  F,	
  Notermans	
  DW,	
  
Delmee	
  M,	
  Coignard	
  B,	
  Wilcox	
  M,	
  Patel	
  B,	
  Frei	
  R,	
  Nagy	
  E,	
  Bouza	
  E,	
  Marin	
  
M,	
  Akerlund	
  T,	
  Virolainen-­‐Julkunen	
  A,	
  Lyytikäinen	
  O,	
  Kotila	
  S,	
  
Ingebretsen	
  A,	
  Smyth	
  B,	
  Rooney	
  P,	
  Poxton	
  IR,	
  Monnet	
  DL.	
  2008.	
  Update	
  of	
  
Clostridium	
  difficile	
  infection	
  due	
  to	
  PCR	
  ribotype	
  027	
  in	
  Europe,	
  2008.	
  Euro	
  
Surveill.	
  13.	
  

38.	
  

Poutanen	
  SM.	
  2004.	
  Clostridium	
  difficile-­‐associated	
  diarrhea	
  in	
  adults.	
  
Canadian	
  Medical	
  Association	
  Journal	
  171:51–58.	
  

39.	
  

McFarland	
  LV.	
  1998.	
  Epidemiology,	
  Risk	
  Factors	
  and	
  Treatments	
  for	
  
Antibiotic‐Associated	
  Diarrhea.	
  Dig	
  Dis	
  16:292–307.	
  

40.	
  

Voth	
  DE,	
  Ballard	
  JD.	
  2005.	
  Clostridium	
  difficile	
  toxins:	
  mechanism	
  of	
  action	
  
and	
  role	
  in	
  disease.	
  Clin.	
  Microbiol.	
  Rev.	
  18:247–263.	
  

41.	
  

Shen	
  A.	
  2012.	
  Clostridium	
  difficile	
  toxins:	
  mediators	
  of	
  inflammation.	
  J	
  Innate	
  
Immun	
  4:149–158.	
  

42.	
  

Gerding	
  DN,	
  Johnson	
  S,	
  Rupnik	
  M,	
  Aktories	
  K.	
  2013.	
  Clostridium	
  difficile	
  
binary	
  toxin	
  CDT:	
  Mechanism,	
  epidemiology,	
  and	
  potential	
  clinical	
  
importance.	
  gutmicrobes	
  5:6–18.	
  

43.	
  

Vedantam	
  G,	
  Clark	
  A,	
  Chu	
  M,	
  McQuade	
  R,	
  Mallozzi	
  M,	
  Viswanathan	
  VK.	
  
2012.	
  Clostridium	
  difficile	
  infection:	
  toxins	
  and	
  non-­‐toxin	
  virulence	
  factors,	
  
and	
  their	
  contributions	
  to	
  disease	
  establishment	
  and	
  host	
  response.	
  
gutmicrobes	
  3:121–134.	
  

44.	
  

Tasteyre	
  A,	
  Barc	
  MC,	
  Collignon	
  A,	
  Boureau	
  H,	
  Karjalainen	
  T.	
  2001.	
  Role	
  of	
  
FliC	
  and	
  FliD	
  flagellar	
  proteins	
  of	
  Clostridium	
  difficile	
  in	
  adherence	
  and	
  gut	
  
colonization.	
  Infection	
  and	
  Immunity	
  69:7937–7940.	
  

45.	
  

Killgore	
  G,	
  Thompson	
  A,	
  Johnson	
  S,	
  Brazier	
  J,	
  Kuijper	
  E,	
  Pepin	
  J,	
  Frost	
  EH,	
  
Savelkoul	
  P,	
  Nicholson	
  B,	
  van	
  den	
  Berg	
  RJ,	
  KATO	
  H,	
  Sambol	
  SP,	
  Zukowski	
  
W,	
  Woods	
  C,	
  Limbago	
  B,	
  Gerding	
  DN,	
  McDonald	
  LC.	
  2008.	
  Comparison	
  of	
  
Seven	
  Techniques	
  for	
  Typing	
  International	
  Epidemic	
  Strains	
  of	
  Clostridium	
  
difficile:	
  Restriction	
  Endonuclease	
  Analysis,	
  Pulsed-­‐Field	
  Gel	
  Electrophoresis,	
  
PCR-­‐Ribotyping,	
  Multilocus	
  Sequence	
  Typing,	
  Multilocus	
  Variable-­‐Number	
  
Tandem-­‐Repeat	
  Analysis,	
  Amplified	
  Fragment	
  Length	
  Polymorphism,	
  and	
  

	
  

38	
  

Surface	
  Layer	
  Protein	
  A	
  Gene	
  Sequence	
  Typing.	
  Journal	
  of	
  Clinical	
  
Microbiology	
  46:431–437.	
  
46.	
  

Manzo	
  CE,	
  Merrigan	
  MM,	
  Johnson	
  S,	
  Gerding	
  DN,	
  Riley	
  TV,	
  Silva	
  J,	
  Brazier	
  
JS,	
  the	
  International	
  Clostridium	
  difficile	
  Study	
  Group.	
  2014.	
  International	
  
typing	
  study	
  of	
  Clostridium	
  difficile.	
  Anaerobe	
  28C:4–7.	
  

47.	
  

McDonald	
  LC,	
  Killgore	
  GE,	
  Thompson	
  A,	
  Owens	
  RC,	
  Kazakova	
  SV,	
  Sambol	
  
SP,	
  Johnson	
  S,	
  Gerding	
  DN.	
  2005.	
  An	
  epidemic,	
  toxin	
  gene-­‐variant	
  strain	
  of	
  
Clostridium	
  difficile.	
  N.	
  Engl.	
  J.	
  Med.	
  353:2433–2441.	
  

48.	
  

Bartlett	
  JG.	
  2006.	
  Narrative	
  review:	
  the	
  new	
  epidemic	
  of	
  Clostridium	
  difficile-­‐
associated	
  enteric	
  disease.	
  Ann.	
  Intern.	
  Med.	
  145:758–764.	
  

49.	
  

Hubert	
  B,	
  Loo	
  VG,	
  Bourgault	
  A-­‐M,	
  Poirier	
  L,	
  Dascal	
  A,	
  Fortin	
  E,	
  Dionne	
  M,	
  
Lorange	
  M.	
  2007.	
  A	
  portrait	
  of	
  the	
  geographic	
  dissemination	
  of	
  the	
  
Clostridium	
  difficile	
  North	
  American	
  pulsed-­‐field	
  type	
  1	
  strain	
  and	
  the	
  
epidemiology	
  of	
  C.	
  difficile-­‐associated	
  disease	
  in	
  Québec.	
  Clin	
  Infect	
  Dis	
  
44:238–244.	
  

50.	
  

Waslawski	
  S,	
  Lo	
  ES,	
  Ewing	
  SA,	
  Young	
  VB,	
  Aronoff	
  DM,	
  Sharp	
  SE,	
  Novak-­‐
Weekley	
  SM,	
  Crist	
  AE,	
  Dunne	
  WM,	
  Hoppe-­‐Bauer	
  J,	
  Johnson	
  M,	
  Brecher	
  
SM,	
  Newton	
  DW,	
  Walk	
  ST.	
  2013.	
  Clostridium	
  difficile	
  ribotype	
  diversity	
  at	
  
six	
  health	
  care	
  institutions	
  in	
  the	
  United	
  States.	
  Journal	
  of	
  Clinical	
  
Microbiology	
  51:1938–1941.	
  

51.	
  

Goorhuis	
  A,	
  Van	
  der	
  Kooi	
  T,	
  Vaessen	
  N,	
  Dekker	
  FW,	
  Van	
  den	
  Berg	
  R,	
  
Harmanus	
  C,	
  van	
  den	
  Hof	
  S,	
  Notermans	
  DW,	
  Kuijper	
  EJ.	
  2007.	
  Spread	
  and	
  
Epidemiology	
  of	
  Clostridium	
  difficile	
  Polymerase	
  Chain	
  Reaction	
  Ribotype	
  
027/Toxinotype	
  III	
  in	
  The	
  Netherlands.	
  CLIN	
  INFECT	
  DIS	
  45:695–703.	
  

52.	
  

Wilcox	
  MH,	
  Shetty	
  N,	
  Fawley	
  WN,	
  Shemko	
  M,	
  Coen	
  P,	
  Birtles	
  A,	
  Cairns	
  M,	
  
Curran	
  MD,	
  Dodgson	
  KJ,	
  Green	
  SM,	
  Hardy	
  KJ,	
  Hawkey	
  PM,	
  Magee	
  JG,	
  Sails	
  
AD,	
  Wren	
  MWD.	
  2012.	
  Changing	
  epidemiology	
  of	
  Clostridium	
  difficile	
  
infection	
  following	
  the	
  introduction	
  of	
  a	
  national	
  ribotyping-­‐based	
  
surveillance	
  scheme	
  in	
  England.	
  CLIN	
  INFECT	
  DIS	
  55:1056–1063.	
  

53.	
  

Wultańska	
  D,	
  Młynarczyk	
  G,	
  Harmanus	
  C.	
  2014.	
  Occurrence	
  of	
  Clostridium	
  
difficile	
  PCR-­‐ribotype	
  027	
  and	
  it's	
  closely	
  related	
  PCR-­‐ribotype	
  176	
  in	
  
hospitals	
  in	
  Poland	
  in	
  2008-­‐2010.	
  Anaerobe.	
  

54.	
  

Kuijper	
  EJ,	
  van	
  Dissel	
  JT.	
  2008.	
  Spectrum	
  of	
  Clostridium	
  difficile	
  infections	
  
outside	
  health	
  care	
  facilities.	
  Canadian	
  Medical	
  Association	
  Journal	
  179:747–
748.	
  

55.	
  

Gerding	
  DN	
  ,	
  MD.	
  2010.	
  Global	
  Epidemiology	
  of	
  Clostridium	
  difficileInfection	
  
in	
  2010.	
  Infection	
  Control	
  and	
  Hospital	
  Epidemiology	
  31:S32–S34.	
  

	
  

39	
  

56.	
  

He	
  M,	
  Miyajima	
  F,	
  Roberts	
  P,	
  Ellison	
  L,	
  Pickard	
  DJ,	
  Martin	
  MJ,	
  Connor	
  TR,	
  
Harris	
  SR,	
  Fairley	
  D,	
  Bamford	
  KB,	
  D'Arc	
  S,	
  Brazier	
  J,	
  Brown	
  D,	
  Coia	
  JE,	
  
Douce	
  G,	
  Gerding	
  D,	
  Kim	
  HJ,	
  Koh	
  TH,	
  Kato	
  H,	
  Senoh	
  M,	
  Louie	
  T,	
  Michell	
  S,	
  
Butt	
  E,	
  Peacock	
  SJ,	
  Brown	
  NM,	
  Riley	
  T,	
  Songer	
  G,	
  Wilcox	
  M,	
  Pirmohamed	
  
M,	
  Kuijper	
  E,	
  Hawkey	
  P,	
  Wren	
  BW,	
  Dougan	
  G,	
  Parkhill	
  J,	
  Lawley	
  TD.	
  2013.	
  
Emergence	
  and	
  global	
  spread	
  of	
  epidemic	
  healthcare-­‐associated	
  Clostridium	
  
difficile.	
  Nat	
  Genet	
  45:109–113.	
  

57.	
  

Sirard	
  S,	
  Valiquette	
  L,	
  Fortier	
  LC.	
  2011.	
  Lack	
  of	
  Association	
  between	
  Clinical	
  
Outcome	
  of	
  Clostridium	
  difficile	
  Infections,	
  Strain	
  Type,	
  and	
  Virulence-­‐
Associated	
  Phenotypes.	
  Journal	
  of	
  Clinical	
  Microbiology	
  49:4040–4046.	
  

58.	
  

Carlson	
  PE,	
  Walk	
  ST,	
  Bourgis	
  AET,	
  Liu	
  MW,	
  Kopliku	
  F,	
  Lo	
  E,	
  Young	
  VB,	
  
Aronoff	
  DM,	
  Hanna	
  PC.	
  2013.	
  The	
  relationship	
  between	
  phenotype,	
  ribotype,	
  
and	
  clinical	
  disease	
  in	
  human	
  Clostridium	
  difficile	
  isolates.	
  Anaerobe	
  24:109–
116.	
  

59.	
  

Walk	
  ST,	
  Micic	
  D,	
  Jain	
  R,	
  Lo	
  ES,	
  Trivedi	
  I,	
  Liu	
  EW,	
  Almassalha	
  LM,	
  Ewing	
  
SA,	
  Ring	
  C,	
  Galecki	
  AT.	
  2012.	
  Clostridium	
  difficile	
  ribotype	
  does	
  not	
  predict	
  
severe	
  infection.	
  Clin.	
  Infect.	
  Dis.	
  55:1661–1668.	
  

60.	
  

Smits	
  WK.	
  2013.	
  Hype	
  or	
  hypervirulence:	
  a	
  reflection	
  on	
  problematic	
  C.	
  
difficile	
  strains.	
  Virulence	
  4:592–596.	
  

61.	
  

US	
  CFDCAP.	
  2013.	
  Antibiotic	
  resistance	
  threats	
  in	
  the	
  United	
  States,	
  2013.	
  

62.	
  

Dubberke	
  E.	
  2012.	
  Clostridium	
  difficile	
  infection:	
  The	
  scope	
  of	
  the	
  problem.	
  J.	
  
Hosp.	
  Med.	
  7:S1–S4.	
  

63.	
  

Sorg	
  JA,	
  Sonenshein	
  AL.	
  2008.	
  Bile	
  Salts	
  and	
  Glycine	
  as	
  Cogerminants	
  for	
  
Clostridium	
  difficile	
  Spores.	
  Journal	
  of	
  Bacteriology	
  190:2505–2512.	
  

64.	
  

Sorg	
  JA,	
  Sonenshein	
  AL.	
  2009.	
  Chenodeoxycholate	
  is	
  an	
  inhibitor	
  of	
  
Clostridium	
  difficile	
  spore	
  germination.	
  Journal	
  of	
  Bacteriology	
  191:1115–
1117.	
  

65.	
  

Wilson	
  KH.	
  1983.	
  Efficiency	
  of	
  various	
  bile	
  salt	
  preparations	
  for	
  stimulation	
  
of	
  Clostridium	
  difficile	
  spore	
  germination.	
  Journal	
  of	
  Clinical	
  Microbiology	
  
18:1017–1019.	
  

66.	
  

Giel	
  JL,	
  Sorg	
  JA,	
  Sonenshein	
  AL,	
  Zhu	
  J.	
  2010.	
  Metabolism	
  of	
  Bile	
  Salts	
  in	
  Mice	
  
Influences	
  Spore	
  Germination	
  in	
  Clostridium	
  difficile.	
  PLoS	
  ONE	
  5:e8740.	
  

67.	
  

Howerton	
  A,	
  Patra	
  M,	
  Abel-­‐Santos	
  E.	
  2013.	
  A	
  new	
  strategy	
  for	
  the	
  
prevention	
  of	
  Clostridium	
  difficile	
  infection.	
  J	
  Infect	
  Dis	
  207:1498–1504.	
  

68.	
  

Wilson	
  KH,	
  Freter	
  R.	
  1986.	
  Interaction	
  of	
  Clostridium	
  difficile	
  and	
  

	
  

40	
  

Escherichia	
  coli	
  with	
  microfloras	
  in	
  continuous-­‐flow	
  cultures	
  and	
  gnotobiotic	
  
mice.	
  Infection	
  and	
  Immunity	
  54:354–358.	
  
69.	
  

Wilson	
  KH,	
  Perini	
  F.	
  1988.	
  Role	
  of	
  competition	
  for	
  nutrients	
  in	
  suppression	
  
of	
  Clostridium	
  difficile	
  by	
  the	
  colonic	
  microflora.	
  Infection	
  and	
  Immunity	
  
56:2610–2614.	
  

70.	
  

Ng	
  KM,	
  Ferreyra	
  JA,	
  Higginbottom	
  SK,	
  Lynch	
  JB,	
  Kashyap	
  PC,	
  Gopinath	
  S,	
  
Naidu	
  N,	
  Choudhury	
  B,	
  Weimer	
  BC,	
  Monack	
  DM,	
  Sonnenburg	
  JL.	
  2013.	
  
Microbiota-­‐liberated	
  host	
  sugars	
  facilitate	
  post-­‐antibiotic	
  expansion	
  of	
  enteric	
  
pathogens.	
  Nature	
  502:96–99.	
  

71.	
  

Vengadesan	
  K,	
  Narayana	
  SVL.	
  2011.	
  Structural	
  biology	
  of	
  Gram-­‐positive	
  
bacterial	
  adhesins.	
  Protein	
  Sci.	
  20:759–772.	
  

72.	
  

Borriello	
  SP,	
  BARCLAY	
  FE.	
  1985.	
  Protection	
  of	
  hamsters	
  against	
  Clostridium	
  
difficile	
  ileocaecitis	
  by	
  prior	
  colonisation	
  with	
  non-­‐pathogenic	
  strains.	
  Journal	
  
of	
  Medical	
  Microbiology	
  19:339–350.	
  

73.	
  

Seal	
  D,	
  Borriello	
  SP,	
  Barclay	
  F,	
  Welch	
  A,	
  Piper	
  M,	
  Bonnycastle	
  M.	
  1987.	
  
Treatment	
  of	
  relapsing	
  Clostridium	
  difficile	
  diarrhoea	
  by	
  administration	
  of	
  a	
  
non-­‐toxigenic	
  strain.	
  Eur.	
  J.	
  Clin.	
  Microbiol.	
  6:51–53.	
  

74.	
  

Merrigan	
  MM,	
  Sambol	
  SP,	
  Johnson	
  S,	
  Gerding	
  DN.	
  2009.	
  New	
  approach	
  to	
  
the	
  management	
  of	
  Clostridium	
  difficile	
  infection:	
  colonisation	
  with	
  non-­‐
toxigenic	
  C.	
  difficile	
  during	
  daily	
  ampicillin	
  or	
  ceftriaxone	
  administration.	
  Int.	
  
J.	
  Antimicrob.	
  Agents	
  33:S46–S50.	
  

75.	
  

Nagaro	
  KJ,	
  Phillips	
  ST,	
  Cheknis	
  AK,	
  Sambol	
  SP,	
  Zukowski	
  WE,	
  Johnson	
  S,	
  
Gerding	
  DN.	
  2013.	
  Nontoxigenic	
  Clostridium	
  difficile	
  protects	
  hamsters	
  
against	
  challenge	
  with	
  historic	
  and	
  epidemic	
  strains	
  of	
  toxigenic	
  
BI/NAP1/027	
  C.	
  difficile.	
  Antimicrobial	
  Agents	
  and	
  Chemotherapy	
  57:5266–
5270.	
  

76.	
  

Rolfe	
  RD.	
  1984.	
  Role	
  of	
  volatile	
  fatty	
  acids	
  in	
  colonization	
  resistance	
  to	
  
Clostridium	
  difficile.	
  Infection	
  and	
  Immunity	
  45:185–191.	
  

77.	
  

Su	
  WJ,	
  Waechter	
  MJ,	
  Bourlioux	
  P,	
  Dolegeal	
  M,	
  Fourniat	
  J,	
  Mahuzier	
  G.	
  
1987.	
  Role	
  of	
  volatile	
  fatty	
  acids	
  in	
  colonization	
  resistance	
  to	
  Clostridium	
  
difficile	
  in	
  gnotobiotic	
  mice.	
  Infection	
  and	
  Immunity	
  55:1686–1691.	
  

78.	
  

May	
  T,	
  Mackie	
  RI,	
  Fahey	
  GC,	
  Cremin	
  JC,	
  Garleb	
  KA.	
  1994.	
  Effect	
  of	
  fiber	
  
source	
  on	
  short-­‐chain	
  fatty	
  acid	
  production	
  and	
  on	
  the	
  growth	
  and	
  toxin	
  
production	
  by	
  Clostridium	
  difficile.	
  Scand	
  J	
  Gastroenterol	
  29:916–922.	
  

79.	
  

Kolling	
  GL,	
  Wu	
  M,	
  Warren	
  CA,	
  Durmaz	
  E,	
  Klaenhammer	
  TR,	
  Guerrant	
  RL.	
  
2012.	
  Lactic	
  acid	
  production	
  by	
  Streptococcus	
  thermophilus	
  alters	
  

	
  

41	
  

Clostridium	
  difficile	
  infection	
  and	
  in	
  vitro	
  Toxin	
  A	
  production.	
  gutmicrobes	
  
3:0–6.	
  
80.	
  

Rea	
  MC,	
  Sit	
  CS,	
  Clayton	
  E,	
  O'Connor	
  PM,	
  Whittal	
  RM,	
  Zheng	
  J,	
  Vederas	
  JC,	
  
Ross	
  RP,	
  Hill	
  C.	
  2010.	
  Thuricin	
  CD,	
  a	
  posttranslationally	
  modified	
  bacteriocin	
  
with	
  a	
  narrow	
  spectrum	
  of	
  activity	
  against	
  Clostridium	
  difficile.	
  Proceedings	
  
of	
  the	
  National	
  Academy	
  of	
  Sciences	
  107:9352–9357.	
  

81.	
  

HALL	
  IC,	
  O'TOOLE	
  E.	
  1935.	
  Intestinal	
  flora	
  in	
  new-­‐born	
  infantswith	
  a	
  
description	
  of	
  a	
  new	
  pathogenic	
  anaerobe,	
  Bacillus	
  difficilis.	
  Am	
  J	
  Dis	
  Child	
  
49:390–402.	
  

82.	
  

Theriot	
  CM,	
  Koenigsknecht	
  MJ,	
  Carlson	
  PE,	
  Hatton	
  GE,	
  Nelson	
  AM,	
  Li	
  B,	
  
Huffnagle	
  GB,	
  Li	
  JZ,	
  Young	
  VB.	
  2014.	
  ncomms4114.	
  Nature	
  Communications	
  
5:1–10.	
  

83.	
  

Nakamura	
  S,	
  Nakashio	
  S,	
  Yamakawa	
  K,	
  Tanabe	
  N,	
  Nishida	
  S.	
  2013.	
  
Carbohydrate	
  Fermentation	
  by	
  Clostridium	
  difficile.	
  Microbiol.	
  Immunol.	
  
26:107–111.	
  

84.	
  

Wilson	
  KH,	
  Perini	
  F.	
  1988.	
  Role	
  of	
  competition	
  for	
  nutrients	
  in	
  suppression	
  
of	
  Clostridium	
  difficile	
  by	
  the	
  colonic	
  microflora.	
  Infection	
  and	
  Immunity	
  
56:2610–2614.	
  

85.	
  

Jackson	
  S,	
  Calos	
  M,	
  Myers	
  A,	
  Self	
  WT.	
  2006.	
  Analysis	
  of	
  Proline	
  Reduction	
  in	
  
the	
  Nosocomial	
  Pathogen	
  Clostridium	
  difficile.	
  Journal	
  of	
  Bacteriology	
  
188:8487–8495.	
  

86.	
  

Bouillaut	
  L,	
  Self	
  WT,	
  Sonenshein	
  AL.	
  2013.	
  Proline-­‐Dependent	
  Regulation	
  of	
  
Clostridium	
  difficile	
  Stickland	
  Metabolism.	
  Journal	
  of	
  Bacteriology	
  195:844–
854.	
  

87.	
  

Wu	
  X,	
  Hurdle	
  JG.	
  2014.	
  The	
  Clostridium	
  difficileproline	
  racemase	
  is	
  not	
  
essential	
  for	
  early	
  logarithmic	
  growth	
  and	
  infection.	
  Can.	
  J.	
  Microbiol.	
  60:251–
254.	
  

88.	
  

Janvilisri	
  T,	
  Scaria	
  J,	
  Teng	
  C-­‐H,	
  McDonough	
  SP,	
  Gleed	
  RD,	
  Fubini	
  SL,	
  Zhang	
  
S,	
  Akey	
  B,	
  Chang	
  Y-­‐F.	
  2012.	
  Temporal	
  differential	
  proteomes	
  of	
  Clostridium	
  
difficile	
  in	
  the	
  pig	
  ileal-­‐ligated	
  loop	
  model.	
  PLoS	
  ONE	
  7:e45608.	
  

89.	
  

Hafiz	
  S.	
  1974.	
  Clostridium	
  difficile	
  and	
  its	
  toxins.	
  

90.	
  

Popoff	
  MR,	
  Dodin	
  A.	
  1985.	
  Survey	
  of	
  neuraminidase	
  production	
  by	
  
Clostridium	
  butyricum,	
  Clostridium	
  beijerinckii,	
  and	
  Clostridium	
  difficile	
  
strains	
  from	
  clinical	
  and	
  nonclinical	
  sources.	
  Journal	
  of	
  Clinical	
  Microbiology	
  
22:873–876.	
  

	
  

42	
  

91.	
  

Borriello	
  SP,	
  Davies	
  HA,	
  Kamiya	
  S,	
  Reed	
  PJ,	
  Seddon	
  S.	
  1990.	
  Virulence	
  
factors	
  of	
  Clostridium	
  difficile.	
  Clin.	
  Infect.	
  Dis.	
  12:S185–S191.	
  

92.	
  

Seddon	
  SV,	
  HEMINGWAY	
  I,	
  Borriello	
  SP.	
  1990.	
  Hydrolytic	
  enzyme	
  
production	
  by	
  Clostridium	
  difficile	
  and	
  its	
  relationship	
  to	
  toxin	
  production	
  
and	
  virulence	
  in	
  the	
  hamster	
  model.	
  Journal	
  of	
  Medical	
  Microbiology	
  31:169–
174.	
  

93.	
  

KARASAWA	
  T,	
  Ikoma	
  S,	
  Yamakawa	
  K,	
  Nakamura	
  S.	
  1995.	
  A	
  defined	
  growth	
  
medium	
  for	
  Clostridium	
  difficile.	
  Microbiology	
  (Reading,	
  Engl.)	
  141:371–375.	
  

94.	
  

Pitts	
  AC,	
  Tuck	
  LR,	
  Faulds-­‐Pain	
  A,	
  Lewis	
  RJ,	
  Marles-­‐Wright	
  J.	
  2012.	
  
Structural	
  insight	
  into	
  the	
  Clostridium	
  difficile	
  ethanolamine	
  utilisation	
  
microcompartment.	
  PLoS	
  ONE	
  7:e48360.	
  

95.	
  

Köpke	
  M,	
  Straub	
  M,	
  Dürre	
  P.	
  2013.	
  Clostridium	
  difficile	
  is	
  an	
  autotrophic	
  
bacterial	
  pathogen.	
  PLoS	
  ONE	
  8:e62157.	
  

96.	
  

Sebaihia	
  M,	
  Wren	
  BW,	
  Mullany	
  P,	
  Fairweather	
  NF,	
  Minton	
  N,	
  Stabler	
  R,	
  
Thomson	
  NR,	
  Roberts	
  AP,	
  Cerdeño-­‐Tárraga	
  AM,	
  Wang	
  H,	
  Holden	
  MT,	
  
Wright	
  A,	
  Churcher	
  C,	
  Quail	
  MA,	
  Baker	
  S,	
  Bason	
  N,	
  Brooks	
  K,	
  
Chillingworth	
  T,	
  Cronin	
  A,	
  Davis	
  P,	
  Dowd	
  L,	
  Fraser	
  A,	
  Feltwell	
  T,	
  Hance	
  Z,	
  
Holroyd	
  S,	
  Jagels	
  K,	
  Moule	
  S,	
  Mungall	
  K,	
  Price	
  C,	
  Rabbinowitsch	
  E,	
  Sharp	
  
S,	
  Simmonds	
  M,	
  Stevens	
  K,	
  Unwin	
  L,	
  Whithead	
  S,	
  Dupuy	
  B,	
  Dougan	
  G,	
  
Barrell	
  B,	
  Parkhill	
  J.	
  2006.	
  The	
  multidrug-­‐resistant	
  human	
  pathogen	
  
Clostridium	
  difficile	
  has	
  a	
  highly	
  mobile,	
  mosaic	
  genome.	
  Nat	
  Genet	
  38:779–
786.	
  

97.	
  

Scaria	
  J,	
  Janvilisri	
  T,	
  Fubini	
  S,	
  Gleed	
  RD,	
  McDonough	
  SP,	
  Chang	
  YF.	
  2011.	
  
Clostridium	
  difficile	
  Transcriptome	
  Analysis	
  Using	
  Pig	
  Ligated	
  Loop	
  Model	
  
Reveals	
  Modulation	
  of	
  Pathways	
  Not	
  Modulated	
  In	
  Vitro.	
  J	
  Infect	
  Dis	
  
203:1613–1620.	
  

98.	
  

Janoir	
  C,	
  Denève	
  C,	
  Bouttier	
  S,	
  Barbut	
  F,	
  Hoys	
  S,	
  Caleechum	
  L,	
  Chapetón-­‐
Montes	
  D,	
  Pereira	
  FC,	
  Henriques	
  AO,	
  Collignon	
  A,	
  Monot	
  M,	
  Dupuy	
  B.	
  
2013.	
  Adaptive	
  strategies	
  and	
  pathogenesis	
  of	
  Clostridium	
  difficile	
  from	
  in	
  
vivo	
  transcriptomics.	
  Infection	
  and	
  Immunity	
  81:3757–3769.	
  

99.	
  

Blottiere	
  HM,	
  de	
  Vos	
  WM,	
  Ehrlich	
  SD,	
  Doré	
  J.	
  2013.	
  Human	
  intestinal	
  
metagenomics:	
  state	
  of	
  the	
  art	
  and	
  future.	
  Current	
  Opinion	
  in	
  Microbiology	
  
16:232–239.	
  

100.	
  

Staib	
  L,	
  Fuchs	
  TM.	
  2014.	
  From	
  food	
  to	
  cell:	
  nutrient	
  exploitation	
  strategies	
  of	
  
enteropathogens.	
  Microbiology.	
  

101.	
  

HAFIZ	
  S,	
  OAKLEY	
  CL.	
  1976.	
  Clostridium	
  difficile:	
  isolation	
  and	
  characteristics	
  
(Plate	
  VIII).	
  Journal	
  of	
  Medical	
  Microbiology	
  9:129–136.	
  

	
  

43	
  

102.	
  

Wilson	
  KH,	
  Sheagren	
  JN,	
  Freter	
  R.	
  1985.	
  Population	
  dynamics	
  of	
  ingested	
  
Clostridium	
  difficile	
  in	
  the	
  gastrointestinal	
  tract	
  of	
  the	
  Syrian	
  hamster.	
  J	
  
INFECT	
  DIS	
  151:355–361.	
  

103.	
  

Wilson	
  KH,	
  Sheagren	
  JN,	
  Freter	
  R,	
  Weatherbee	
  L,	
  Lyerly	
  D.	
  1986.	
  
Gnotobiotic	
  models	
  for	
  study	
  of	
  the	
  microbial	
  ecology	
  of	
  Clostridium	
  difficile	
  
and	
  Escherichia	
  coli.	
  J	
  INFECT	
  DIS	
  153:547–551.	
  

104.	
  

Freter	
  R,	
  Stauffer	
  E,	
  Cleven	
  D,	
  Holdeman	
  LV,	
  Moore	
  WE.	
  1983.	
  Continuous-­‐
flow	
  cultures	
  as	
  in	
  vitro	
  models	
  of	
  the	
  ecology	
  of	
  large	
  intestinal	
  flora.	
  
Infection	
  and	
  Immunity	
  39:666–675.	
  

105.	
  

Freter	
  R,	
  Brickner	
  H,	
  Botney	
  M,	
  Cleven	
  D,	
  Aranki	
  A.	
  1983.	
  Mechanisms	
  that	
  
control	
  bacterial	
  populations	
  in	
  continuous-­‐flow	
  culture	
  models	
  of	
  mouse	
  
large	
  intestinal	
  flora.	
  Infection	
  and	
  Immunity	
  39:676–685.	
  

106.	
  

Freter	
  R,	
  Brickner	
  H,	
  Fekete	
  J,	
  Vickerman	
  MM,	
  Carey	
  KE.	
  1983.	
  Survival	
  
and	
  implantation	
  of	
  Escherichia	
  coli	
  in	
  the	
  intestinal	
  tract.	
  Infection	
  and	
  
Immunity	
  39:686–703.	
  

107.	
  

Ozawa	
  A,	
  Freter	
  R.	
  1964.	
  Ecological	
  Mechanism	
  Controlling	
  Growth	
  Of	
  
Escherichia	
  Coli	
  In	
  Continuous	
  Flow	
  Cultures	
  And	
  In	
  The	
  Mouse	
  Intestine.	
  J	
  
Infect	
  Dis	
  114:235–242.	
  

108.	
  

Duerden	
  BI.	
  1994.	
  Virulence	
  factors	
  in	
  anaerobes.	
  Clin.	
  Infect.	
  Dis.	
  18	
  Suppl	
  
4:S253–9.	
  

109.	
  

Corfield	
  AP,	
  Wagner	
  SA,	
  Clamp	
  JR,	
  Kriaris	
  MS,	
  Hoskins	
  LC.	
  1992.	
  Mucin	
  
degradation	
  in	
  the	
  human	
  colon:	
  production	
  of	
  sialidase,	
  sialate	
  O-­‐
acetylesterase,	
  N-­‐acetylneuraminate	
  lyase,	
  arylesterase,	
  and	
  glycosulfatase	
  
activities	
  by	
  strains	
  of	
  fecal	
  bacteria.	
  Infection	
  and	
  Immunity	
  60:3971–3978.	
  

110.	
  

Seddon	
  SV,	
  Borriello	
  SP.	
  1989.	
  A	
  chemically	
  defined	
  and	
  minimal	
  medium	
  
for	
  Clostridium	
  difficile.	
  Letters	
  in	
  applied	
  microbiology	
  9:237–239.	
  

111.	
  

Seddon	
  SV,	
  Borriello	
  SP.	
  1992.	
  Proteolytic	
  activity	
  of	
  Clostridium	
  difficile.	
  
Journal	
  of	
  Medical	
  Microbiology	
  36:307–311.	
  

112.	
  

Stickland	
  LH.	
  1934.	
  Studies	
  in	
  the	
  metabolism	
  of	
  the	
  strict	
  anaerobes	
  (genus	
  
Clostridium):	
  The	
  chemical	
  reactions	
  by	
  which	
  Cl.	
  sporogenes	
  obtains	
  its	
  
energy.	
  Biochemical	
  Journal	
  28:1746.	
  

113.	
  

Garsin	
  DA.	
  2010.	
  Ethanolamine	
  utilization	
  in	
  bacterial	
  pathogens:	
  roles	
  and	
  
regulation.	
  Nat	
  Rev	
  Micro	
  8:290–295.	
  

114.	
  

Tsoy	
  O,	
  Ravcheev	
  D,	
  Mushegian	
  A.	
  2009.	
  Comparative	
  genomics	
  of	
  
ethanolamine	
  utilization.	
  Journal	
  of	
  Bacteriology	
  191:7157–7164.	
  

	
  

44	
  

115.	
  

Garsin	
  DA.	
  2012.	
  Ethanolamine:	
  a	
  signal	
  to	
  commence	
  a	
  host-­‐associated	
  
lifestyle?	
  MBio	
  3:e00172–12.	
  

116.	
  

Janvilisri	
  T,	
  Scaria	
  J,	
  Chang	
  Y-­‐F.	
  2010.	
  Transcriptional	
  profiling	
  of	
  
Clostridium	
  difficile	
  and	
  Caco-­‐2	
  cells	
  during	
  infection.	
  J	
  Infect	
  Dis	
  202:282–
290.	
  

117.	
  

Emerson	
  JE,	
  Stabler	
  RA,	
  Wren	
  BW,	
  Fairweather	
  NF.	
  2008.	
  Microarray	
  
analysis	
  of	
  the	
  transcriptional	
  responses	
  of	
  Clostridium	
  difficile	
  to	
  
environmental	
  and	
  antibiotic	
  stress.	
  Journal	
  of	
  Medical	
  Microbiology	
  57:757–
764.	
  

	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

45	
  

CHAPTER	
  2	
  
	
  
Development	
  of	
  Single	
  Chamber	
  Human	
  Gut	
  Microbiota	
  Mini-­‐bioreactor	
  Arrays	
  
(MBRA)	
  	
  
	
  
	
  
	
  
	
  
	
  

The	
  material	
  presented	
  in	
  this	
  chapter	
  was	
  an	
  equal	
  collaboration	
  with	
  Jennifer	
  M.	
  
Auchtung	
  (Britton	
  Lab,	
  Department	
  of	
  Microbiology	
  and	
  Molecular	
  Genetics	
  MSU)	
  
and	
  Robert	
  D.	
  Stedtfeld	
  (Hashsham	
  Lab,	
  Department	
  of	
  Civil	
  and	
  Environmental	
  
Engineering,	
  MSU).	
  	
  	
  Robert	
  Stedtfeld	
  drafted	
  the	
  bioreactor	
  design	
  for	
  fabrication,	
  
and	
  performed	
  the	
  HPLC	
  analysis	
  for	
  SCFA	
  quantification.	
  	
  Jennifer	
  Auchtung	
  
conducted	
  the	
  bioreactor	
  community	
  analysis	
  and	
  microbial	
  ecology.	
  

	
  

46	
  

Abstract	
  	
  
We	
  developed	
  single	
  chamber	
  continuous-­‐flow	
  mini-­‐bioreactor	
  arrays	
  (MBRA)	
  for	
  
in	
  vitro	
  human	
  fecal	
  microbiome	
  studies.	
  The	
  small	
  size	
  (15	
  ml	
  volume)	
  and	
  setup	
  
allowed	
  many	
  individual	
  reactors	
  to	
  be	
  operated	
  in	
  parallel,	
  providing	
  flexibility	
  in	
  
terms	
  of	
  testing	
  multiple	
  conditions	
  and	
  replicates	
  simultaneously.	
  Initially,	
  we	
  
tested	
  four	
  variations	
  of	
  media	
  to	
  determine	
  how	
  well	
  these	
  media	
  would	
  be	
  able	
  to	
  
reproduce	
  the	
  microbial	
  diversity	
  and	
  functional	
  stability	
  of	
  a	
  human	
  fecal	
  
community.	
  Triplicate	
  reactors	
  were	
  inoculated	
  with	
  each	
  medium	
  and	
  their	
  
stability	
  was	
  monitored	
  over	
  time	
  (28	
  days).	
  	
  We	
  identified	
  one	
  medium	
  that	
  best	
  
supported	
  growth	
  of	
  a	
  diverse	
  fecal	
  microbial	
  community	
  with	
  a	
  stable	
  short	
  chain	
  
fatty	
  acid	
  (SCFA)	
  profile	
  most	
  similar	
  to	
  the	
  starting	
  fecal	
  inoculum.	
  Detailed	
  
analyses	
  of	
  communities	
  in	
  this	
  medium	
  demonstrated	
  that	
  the	
  microbial	
  
community	
  in	
  each	
  reactor	
  was	
  dynamic	
  throughout	
  the	
  culture	
  period.	
  The	
  largest	
  
changes	
  in	
  the	
  communities	
  were	
  observed	
  during	
  the	
  first	
  week	
  in	
  culture,	
  after	
  
which	
  the	
  communities	
  stabilized	
  to	
  a	
  constant	
  low	
  rate	
  of	
  change.	
  Over	
  time,	
  the	
  
communities	
  present	
  in	
  each	
  replicate	
  reactor	
  diverged	
  both	
  from	
  the	
  starting	
  
community	
  and	
  from	
  the	
  other	
  replicate	
  reactors,	
  but	
  still	
  shared	
  some	
  similarity	
  
with	
  the	
  starting	
  fecal	
  inoculum.	
  This	
  study	
  demonstrates	
  the	
  utility	
  of	
  multiplexed,	
  
single	
  chamber	
  MBRA	
  to	
  support	
  growth	
  of	
  complex,	
  dynamic,	
  functionally-­‐stable	
  
fecal	
  microbial	
  communities.	
  	
  
	
  

	
  

47	
  

Introduction	
  
The	
   trillions	
   of	
   microorganisms	
   forming	
   the	
   commensal	
   microbiota	
   of	
   the	
   human	
  
gastrointestinal	
   tract	
   contribute	
   to	
   many	
   different	
   aspects	
   of	
   host	
   physiology,	
  
including	
   metabolism	
   and	
   energy	
   stasis,	
   maturation	
   of	
   the	
   immune	
   system,	
  
metabolism	
  of	
  xenobiotics,	
  and	
  protection	
  from	
  pathogen	
  invasion	
  (reviewed	
  in	
  (1-­‐
3)).	
   Although	
   many	
   insights	
   about	
   the	
   role	
   of	
   the	
   microbiota	
   in	
   host	
   health	
   and	
  
disease	
  have	
  been	
  gained	
  by	
  studying	
  the	
  microbiota	
  in	
  association	
  with	
  its	
  human	
  
host	
   (e.g.,	
   (4-­‐6)),	
   the	
   availability	
   of	
   less	
   complex	
   models	
   of	
   the	
   microbiota	
   (i.e.,	
  
conventional	
   and	
   humanized	
   animal	
   (7-­‐10)	
   and	
   in	
  vitro	
   models	
   (11-­‐16)	
   have	
   also	
  
played	
   an	
   important	
   role	
   in	
   elucidating	
   the	
   roles	
   of	
   the	
   microbiota.	
   The	
   primary	
  
advantage	
  of	
  using	
  these	
  models	
  are	
  the	
  ability	
  to	
  perform	
  controlled	
  experiments	
  
at	
   a	
   higher	
   throughput	
   than	
   can	
   be	
   achieved	
   with	
   human	
   studies,	
   whereas	
   the	
  
primary	
  disadvantage	
  is	
  the	
  lack	
  of	
  interactions	
  with	
  the	
  human	
  host.	
  
Several	
   different	
   in	
   vitro	
   models	
   of	
   the	
   human	
   intestinal	
   tract	
   have	
   been	
  
developed,	
   ranging	
   from	
   simple	
   batch	
   culture	
   models	
   (e.g.,	
   (17,	
   18))	
   to	
   more	
  
complex	
   continuous	
   flow	
   culture	
   models	
   (e.g.,	
   (11-­‐16)).	
   Continuous-­‐flow	
   culture	
  
models	
   are	
   beneficial	
   for	
   studying	
   the	
   complex	
   interactions	
   between	
   members	
   of	
  
the	
  host	
  microbiota	
  in	
  vitro	
  because	
  they	
  allow	
  for	
  studies	
  to	
  be	
  completed	
  during	
  
an	
   extended	
   period	
   of	
   time	
   under	
   conditions	
   where	
   pH,	
   nutrient	
   availability	
   and	
  
washout	
  of	
  waste	
  products	
  and	
  dead	
  cells	
  can	
  be	
  better	
  controlled	
  (reviewed	
  in	
  (19-­‐
21)).	
  	
  	
  
Three	
   multi-­‐stage	
   continuous-­‐culture	
   models	
   that	
   have	
   been	
   well	
   studied	
  
and	
   validated	
   for	
   their	
   use	
   as	
   in	
   vitro	
   models	
   of	
   the	
   human	
   microbiota	
   are	
   the	
  

	
  

48	
  

Simulator	
  of	
  Human	
  Intestinal	
  Microbial	
  Ecosystem	
  (SHIME,	
  (14,	
  22,	
  23)),	
  the	
  TIM-­‐2	
  
in	
   vitro	
   model	
   of	
   the	
   human	
   intestine	
   (12,	
   24),	
   and	
   the	
   three-­‐stage	
   compound	
  
continuous	
   culture	
   system	
   (13,	
   16).	
   Although	
   they	
   share	
   several	
   features,	
   each	
  
model	
   has	
   its	
   own	
   unique	
   properties.	
   The	
   SHIME	
   and	
   three-­‐stage	
   compound	
  
continuous	
   culture	
   system	
   simulate	
   multiple	
   colon	
   compartments	
   (ascending,	
  
transverse,	
   and	
   descending	
   colon)	
   with	
   fresh	
   nutrients	
   coming	
   into	
   the	
   system	
   at	
  
the	
  ascending	
  colon.	
  In	
  contrast,	
  the	
  TIM-­‐2	
  model	
  has	
  four	
  chambers	
  that	
  simulate	
  
the	
  proximal	
  colon	
  and	
  are	
  connected	
  in	
  a	
  loop	
  with	
  contents	
  circulated	
  between	
  the	
  
four	
   chambers	
   and	
   fresh	
   medium	
   brought	
   into	
   one	
   of	
   the	
   chambers	
   through	
   a	
  
peristaltic	
  valve.	
  All	
  three	
  models	
  are	
  inoculated	
  with	
  human	
  feces,	
  maintained	
  at	
  a	
  
constant	
   temperature	
   of	
   37°C,	
   kept	
   anaerobic	
   by	
   continuous	
   flushing	
   of	
   anaerobic	
  
gas	
   (N2	
   (SHIME,	
   TIM-­‐2)	
   or	
   CO2	
   (three-­‐stage),	
   and	
   incorporate	
   dynamic	
   pH	
  
monitoring	
  and	
  adjustment.	
  	
  
Another	
   common	
   feature	
   of	
   all	
   three	
   model	
   systems	
   described	
   above	
   is	
   their	
  
size	
   and	
   complexity,	
   which	
   makes	
   high	
   numbers	
   of	
   biological	
   replication	
   of	
  
experiments	
   challenging.	
   The	
   goal	
   of	
   our	
   study	
   was	
   to	
   develop	
   a	
   simpler,	
  
continuous-­‐flow	
  in	
  vitro	
  model	
  that	
  would	
  support	
  growth	
  of	
  complex	
  human	
  fecal	
  
microbial	
  communities	
  but	
  also	
  facilitate	
  higher	
  throughput	
  studies.	
  We	
  developed	
  
mini-­‐bioreactor	
  arrays	
  (MBRAs)	
  that	
  allow	
  for	
  simultaneous	
  operation	
  of	
  up	
  to	
  24	
  
independent	
   reactors	
   in	
   a	
   single	
   anaerobic	
   chamber.	
   After	
   testing	
   four	
   different	
  
media,	
   we	
   identified	
   a	
   single	
   medium	
   that	
   best	
   supported	
   growth	
   of	
   microbial	
  
communities	
   that	
   were	
   functionally	
   and	
   ecologically	
   similar	
   to	
   the	
   initial	
   fecal	
  

	
  

49	
  

community.	
  Based	
  upon	
  these	
  results,	
  we	
  anticipate	
  that	
  our	
  MBRA	
  model	
  will	
  be	
  a	
  
useful	
  tool	
  for	
  moderate	
  throughput	
  studies	
  of	
  human	
  fecal	
  microbial	
  communities.	
  
	
  
Materials	
  and	
  Methods	
  
Design	
  of	
  mini-­‐bioreactor	
  arrays	
  (MBRA).	
  MBRAs	
  were	
  designed	
  using	
  CAD	
  
software	
  (Argon,	
  Asheller-­‐Vellum),	
  and	
  fabricated	
  with	
  DSM	
  Somos	
  Watershed	
  XC	
  
11122	
  via	
  stereolithography	
  (FineLine	
  prototyping).	
  	
  In	
  detail,	
  six	
  reactors	
  with	
  an	
  
internal	
  volume	
  of	
  25	
  mL	
  were	
  designed	
  into	
  a	
  single	
  strip	
  with	
  200mm	
  x	
  47	
  mm	
  x	
  
36	
  mm	
  dimensions	
  (Fig.	
  1A).	
  	
  Reactors	
  were	
  drawn	
  with	
  25	
  x	
  25	
  x	
  40	
  mm	
  
dimensions,	
  including	
  10	
  mm	
  radial	
  blends	
  on	
  the	
  bottom	
  corners	
  (to	
  prevent	
  
buildup	
  of	
  cells	
  and	
  other	
  insoluble	
  materials),	
  and	
  5	
  mm	
  radial	
  blends	
  on	
  the	
  top.	
  
Reactors	
  were	
  spaced	
  32.5	
  mm	
  center	
  to	
  center	
  to	
  match	
  dimensions	
  on	
  the	
  
selected	
  stir	
  plate	
  (described	
  below).	
  	
  Three	
  5.55	
  mm	
  diameter	
  holes	
  (influent,	
  
effluent,	
  inoculation/sampling)	
  were	
  placed	
  into	
  the	
  top	
  of	
  each	
  reactor,	
  and	
  spaced	
  
16	
  mm	
  apart	
  center	
  to	
  center.	
  	
  Holes	
  were	
  tapped	
  with	
  threads	
  to	
  fit	
  with	
  
conventional	
  leur	
  connectors.	
  	
  Inner	
  walls	
  of	
  the	
  reactors	
  were	
  placed	
  2	
  mm	
  from	
  
the	
  bottom	
  of	
  the	
  strip,	
  5	
  mm	
  from	
  the	
  top	
  of	
  the	
  strip,	
  5	
  mm	
  from	
  the	
  side	
  of	
  the	
  
strip,	
  and	
  3.25	
  mm	
  into	
  the	
  strip.	
  A	
  1.25	
  x	
  16	
  x	
  31	
  mm	
  intrusion	
  was	
  designed	
  into	
  
the	
  bottom	
  corners	
  of	
  the	
  MBRA	
  for	
  fixing	
  into	
  a	
  custom	
  built	
  acrylic	
  block	
  that	
  held	
  
the	
  reactors	
  upright	
  and	
  properly	
  aligned	
  with	
  the	
  stir-­‐plate.	
  	
  A	
  CAD	
  file	
  (.stl)	
  that	
  
can	
  be	
  used	
  directly	
  for	
  fabrication	
  via	
  stereolithography	
  is	
  available	
  in	
  the	
  
supplemental	
  material.	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
	
  

	
  

50	
  

Collection	
  of	
  fecal	
  samples.	
  Fecal	
  samples	
  were	
  donated	
  by	
  twelve	
  healthy,	
  
anonymous	
  donors	
  that	
  were	
  between	
  the	
  ages	
  of	
  25	
  and	
  64,	
  had	
  not	
  taken	
  
antibiotics	
  for	
  at	
  least	
  two	
  months	
  and	
  had	
  refrained	
  from	
  consuming	
  products	
  
containing	
  intentionally-­‐added	
  live	
  microbes	
  for	
  at	
  least	
  two	
  days	
  prior	
  to	
  donation.	
  
Fresh	
  samples	
  were	
  collected	
  into	
  sterile	
  containers,	
  which	
  were	
  then	
  packed	
  in	
  wet	
  
ice	
  in	
  a	
  sealed	
  (8.1	
  quart,	
  Sterilite	
  Ultraseal)	
  container	
  with	
  two	
  anaerobic	
  gaspaks	
  
(BD	
  biosciences)	
  and	
  transported	
  to	
  the	
  laboratory	
  within	
  24	
  hours.	
  Protocols	
  for	
  
participation	
  of	
  human	
  subjects	
  were	
  reviewed	
  and	
  approved	
  by	
  the	
  Institutional	
  
Review	
  Board	
  of	
  Michigan	
  State	
  University.	
  	
  
	
  
Preparation	
  of	
  fecal	
  samples.	
  Once	
  received	
  in	
  the	
  laboratory,	
  samples	
  were	
  
transferred	
  to	
  an	
  anaerobic	
  chamber	
  and	
  manually	
  mixed	
  with	
  sterile	
  equipment	
  
(spatula	
  and/or	
  pestle).	
  Aliquots	
  were	
  transferred	
  to	
  sterile	
  cryogenic	
  vials	
  and	
  
stored	
  at	
  -­‐80°C	
  until	
  use.	
  Prior	
  to	
  inoculation	
  into	
  the	
  bioreactors,	
  aliquots	
  from	
  
each	
  donor	
  were	
  thawed	
  in	
  the	
  anaerobic	
  chamber	
  and	
  resupsended	
  in	
  sterile,	
  
phosphate-­‐buffered	
  saline	
  with	
  0.1%	
  cysteine	
  at	
  a	
  concentration	
  of	
  20%	
  w/v.	
  
Samples	
  were	
  vortexed	
  vigorously	
  for	
  5	
  min,	
  then	
  allowed	
  to	
  settle	
  for	
  5	
  min	
  prior	
  
to	
  removal	
  of	
  the	
  supernatant,	
  which	
  was	
  used	
  for	
  inoculation	
  into	
  the	
  reactors.	
  
Inoculum	
  was	
  introduced	
  into	
  each	
  reactor	
  through	
  an	
  ethanol-­‐sterilized	
  rubber	
  
septum	
  via	
  a	
  sterile	
  syringe	
  and	
  needle.	
  The	
  final	
  concentration	
  of	
  inoculum	
  in	
  each	
  
reactor	
  was	
  1%.	
  Excess	
  fecal	
  slurry	
  was	
  diluted	
  30%	
  with	
  sterile	
  glycerol	
  to	
  a	
  final	
  
concentration	
  of	
  15%	
  glycerol,	
  aliquoted	
  into	
  1	
  ml	
  volumes	
  in	
  sterile	
  cryovials,	
  flash	
  
frozen	
  in	
  liquid	
  nitrogen,	
  and	
  stored	
  at	
  -­‐80°C.	
  Two	
  aliquots	
  of	
  excess	
  fecal	
  slurry	
  

	
  

51	
  

were	
  used	
  for	
  analyzing	
  the	
  SCFA	
  and	
  microbial	
  community	
  composition	
  of	
  the	
  fecal	
  
inoculum.	
  
	
  
Media.	
  One	
  liter	
  of	
  BRMW	
  was	
  composed	
  of	
  tryptone	
  (1g),	
  proteose	
  peptone	
  #3	
  
(2g),	
  yeast	
  extract	
  (2g),	
  arabinogalactan	
  (1g),	
  maltose	
  (1.5g),	
  D-­‐cellobiose	
  (1.5g),	
  
sodium	
  chloride	
  (0.4g),	
  hemin	
  (5mg),	
  magnesium	
  sulfate	
  (1mg),	
  calclium	
  chloride	
  
(1mg),	
  tween	
  80	
  (2ml),	
  taurocholate	
  (1g),	
  D-­‐glucose	
  (400mg),	
  inulin	
  (2g),	
  sodium	
  
bicarbonate	
  (2g),	
  potassium	
  phosphate	
  monobasic	
  (6.1g),	
  potassium	
  phosphate	
  
dibasic	
  (7.6g),	
  and	
  vitamin	
  K3	
  (1mg).	
  The	
  composition	
  of	
  BRMW10	
  was	
  the	
  same	
  as	
  
BRMW	
  with	
  the	
  exception	
  of	
  10-­‐fold	
  reductions	
  in	
  the	
  concentration	
  of	
  
arabinogalactan,	
  maltose,	
  D-­‐cellobiose,	
  D-­‐glucose,	
  and	
  inulin.	
  One	
  liter	
  of	
  BRMG	
  was	
  
composed	
  of	
  trypticase	
  peptone	
  (2g),	
  yeast	
  extract	
  (1g),	
  D-­‐glucose	
  (0.4g),	
  cellobiose	
  
(1g),	
  maltose	
  (1g),	
  fructose	
  (1g),	
  beef	
  extract	
  (5g),	
  magnesium	
  sulfate	
  heptahydrate	
  
(32	
  mg),	
  sodium	
  chloride	
  (80mg),	
  calcium	
  chloride	
  (0.8mg),	
  iron	
  sulfate	
  (2.5mg),	
  
hematin	
  (1.2mg),	
  histidine	
  (31mg),	
  tween	
  80	
  (0.5ml),	
  ATCC	
  vitamin	
  mix	
  (10%	
  v/v),	
  
boric	
  acid	
  (30µg),	
  manganese	
  chloride	
  (600µg),	
  cobalt	
  (II)	
  chloride	
  (190µg),	
  nickel	
  
(II)	
  chloride	
  (124µg),	
  copper	
  (II)	
  chloride	
  (102µg),	
  zinc	
  sulfate	
  (144µg),	
  sodium	
  
molybdate	
  (36µg),	
  sodium	
  metavanadate	
  (25µg),	
  sodium	
  tungstate	
  (25µg),	
  sodium	
  
selenite	
  (6µg),	
  isovaleric	
  acid	
  (0.1ml),	
  propionic	
  acid	
  (2ml),	
  butyric	
  acid	
  (2ml),	
  
taurocholate	
  (1g),	
  sodium	
  bicarbonate	
  (2g),	
  potassium	
  phosphate	
  monobasic	
  (6.1g),	
  
potassium	
  phosphate	
  dibasic	
  (7.6g),	
  and	
  vitamin	
  K3	
  (1mg).	
  The	
  composition	
  of	
  
BRMG10	
  was	
  the	
  same	
  as	
  BRMG	
  with	
  the	
  exception	
  of	
  10-­‐fold	
  reductions	
  in	
  the	
  
concentration	
  of	
  maltose,	
  D-­‐cellobiose,	
  D-­‐glucose,	
  and	
  fructose.	
  We	
  adjusted	
  the	
  pH	
  

	
  

52	
  

of	
  all	
  four	
  media	
  to	
  6.8	
  and	
  sterilized	
  by	
  a	
  combination	
  of	
  autoclaving	
  and	
  filter	
  
sterilization	
  of	
  heat	
  labile	
  reagents.	
  
	
  
MBRA	
  operating	
  conditions	
  and	
  sampling.	
  Media	
  were	
  transferred	
  from	
  the	
  
source	
  bottles	
  through	
  a	
  combination	
  of	
  1/8	
  in	
  inner	
  diameter	
  (ID)	
  C-­‐flex	
  tubing	
  
(6424-­‐67,	
  Cole-­‐Parmer)	
  and	
  0.89	
  mm	
  ID	
  2-­‐stop	
  Tygon	
  lab	
  tubing	
  supplied	
  to	
  the	
  
reactors	
  via	
  a	
  24-­‐channel	
  peristaltic	
  pump	
  (IPC-­‐24,	
  Ismatec).	
  The	
  waste	
  line	
  in	
  each	
  
MBRA	
  was	
  set	
  to	
  maintain	
  a	
  working	
  volume	
  of	
  15	
  ml	
  and	
  the	
  pumps	
  were	
  
calibrated	
  for	
  a	
  flow	
  rate	
  of	
  0.625	
  ml/hr	
  (24	
  hr	
  retention	
  time).	
  Waste	
  was	
  removed	
  
from	
  the	
  reactors	
  through	
  a	
  combination	
  of	
  1/8	
  in	
  ID	
  C-­‐flex	
  tubing	
  and	
  1.14	
  mm	
  ID	
  
2-­‐stop	
  Tygon	
  lab	
  tubing	
  drawn	
  from	
  the	
  reactors	
  via	
  the	
  same	
  24-­‐channel	
  peristaltic	
  
pump.	
  Omnifit	
  caps	
  were	
  used	
  to	
  connect	
  media	
  and	
  waste	
  bottles	
  to	
  tubing.	
  	
  
Reactors	
  were	
  stirred	
  using	
  magnetic	
  stir	
  bars	
  driven	
  by	
  independent	
  magnets	
  on	
  a	
  
60-­‐spot	
  magnetic	
  stir	
  plate	
  (VarioMAG	
  HP	
  60,	
  Vario-­‐MAG	
  USA).	
  Assembled	
  MBRA	
  
were	
  sterilized	
  by	
  autoclaving	
  and	
  were	
  operated	
  in	
  an	
  anaerobic	
  chamber	
  (5%	
  H2,	
  
5%	
  CO2,	
  90%	
  N2)	
  maintained	
  at	
  37°C.	
  MBRA	
  and	
  source	
  media	
  were	
  allowed	
  to	
  
equilibrate	
  to	
  the	
  anaerobic	
  environment	
  of	
  the	
  chamber	
  for	
  at	
  least	
  72	
  hours	
  prior	
  
to	
  use.	
  Media	
  was	
  pumped	
  through	
  tubing	
  into	
  the	
  reactors	
  and	
  allowed	
  to	
  
equilibrate	
  for	
  48	
  hrs	
  in	
  the	
  reactors	
  prior	
  to	
  inoculation	
  to	
  ensure	
  sterility.	
  1	
  ml	
  
samples	
  were	
  removed	
  for	
  SCFA	
  and	
  microbial	
  ecological	
  analyses	
  every	
  two	
  days	
  
by	
  sterile	
  needle	
  and	
  syringe	
  through	
  ethanol-­‐sterilized	
  septa.	
  	
  (Removal	
  of	
  1	
  ml	
  of	
  
sample	
  decreases	
  the	
  total	
  volume	
  of	
  the	
  culture	
  in	
  the	
  reactor	
  by	
  6.7%.	
  	
  Reactor	
  
volume	
  returns	
  to	
  the	
  pre-­‐sampling	
  volume	
  of	
  15	
  ml	
  within	
  1.6	
  hours	
  of	
  sampling	
  

	
  

53	
  

through	
  the	
  addition	
  of	
  fresh	
  medium.	
  During	
  these	
  1.6	
  hours,	
  waste	
  is	
  unlikely	
  to	
  
be	
  removed	
  since	
  the	
  volume	
  of	
  the	
  reactor	
  is	
  below	
  the	
  threshold	
  volume	
  for	
  waste	
  
removal.)	
  Samples	
  were	
  centrifuged	
  for	
  1	
  min	
  at	
  maximum	
  speed,	
  supernatants	
  
were	
  removed	
  from	
  cell	
  pellets,	
  and	
  both	
  pellets	
  and	
  supernatants	
  were	
  stored	
  at	
  -­‐
80°C	
  until	
  further	
  processing.	
  The	
  pH	
  of	
  thawed	
  supernatant	
  samples	
  was	
  tested	
  
with	
  pH	
  strips	
  (pHyrdrion	
  papers,	
  pH	
  4.5-­‐7.5,	
  MicroEssential	
  Laboratories).	
  
	
  
HPLC	
  of	
  SCFA.	
  To	
  ensure	
  cells	
  were	
  removed,	
  thawed	
  supernatant	
  samples	
  were	
  
centrifuged	
  at	
  maximum	
  speed	
  for	
  an	
  additional	
  15	
  min.	
  	
  Centrifuged	
  reactor	
  
samples	
  were	
  filtered	
  through	
  0.22-­‐μm-­‐pore-­‐size	
  filters	
  (SLGS033SS,	
  Millipore)	
  and	
  
acidified	
  with	
  0.1	
  M	
  H2SO4.	
  	
  SCFAs	
  were	
  analyzed	
  with	
  a	
  high-­‐performance	
  liquid	
  
chromatograph	
  (HPLC)	
  equipped	
  with	
  a	
  250	
  x	
  4.6	
  mm	
  Discovery	
  C8	
  column	
  (59354-­‐
U,	
  Supelco	
  Analytical)	
  connected	
  to	
  a	
  UV/Vis	
  absorbance	
  detector	
  (Series	
  200,	
  
Perken	
  Elmer)	
  set	
  at	
  210	
  nm.	
  The	
  mobile	
  phase	
  was	
  25	
  mM	
  potassium	
  phosphate	
  
adjusted	
  to	
  a	
  pH	
  of	
  2.8	
  with	
  phosphoric	
  acid,	
  at	
  a	
  flow	
  rate	
  of	
  1	
  ml/min,	
  and	
  100	
  μl	
  
was	
  injected	
  into	
  the	
  HPLC	
  using	
  an	
  autosampler	
  (Series	
  200,	
  Perkin	
  Elmer).	
  All	
  
compounds	
  detectable	
  with	
  the	
  UV/Vis	
  absorbance	
  detector	
  within	
  50	
  min	
  were	
  
monitored.	
  Lactate,	
  acetate,	
  butyrate,	
  propionate,	
  and	
  isobutyrate	
  were	
  identified	
  by	
  
their	
  retention	
  times	
  compared	
  to	
  those	
  of	
  standards.	
  	
  
	
  
SCFA	
  Analysis.	
  The	
  area	
  under	
  peaks	
  were	
  calculated	
  using	
  TotalChrom	
  Navigator	
  
(Perkin	
  Elmer),	
  and	
  the	
  concentration	
  of	
  SCFA	
  was	
  derived	
  via	
  standard	
  dilutions	
  of	
  
lactate,	
  acetate,	
  butyrate,	
  propionate,	
  and	
  isobutyrate	
  run	
  in	
  parallel.	
  	
  Pearson	
  

	
  

54	
  

Product	
  Moment	
  Correlation	
  Coefficients	
  were	
  calculated	
  using	
  the	
  Pearson	
  
function	
  in	
  Microsoft	
  Excel.	
  Additional	
  transformations	
  of	
  the	
  data	
  are	
  reported	
  in	
  
the	
  figure	
  or	
  table	
  legend.	
  
	
  
DNA	
  extraction.	
  We	
  extracted	
  DNA	
  from	
  samples	
  using	
  bead	
  beating	
  followed	
  by	
  
cleanup	
  with	
  a	
  Qiagen	
  DNEasy	
  Tissue	
  Kit.	
  Samples	
  were	
  resuspended	
  in	
  360	
  µl	
  
buffer	
  ATL	
  (Qiagen),	
  transferred	
  to	
  a	
  MoBio	
  fecal	
  bead	
  tube,	
  and	
  homogenized	
  on	
  
full	
  speed	
  for	
  1	
  min.	
  Homogenates	
  were	
  incubated	
  with	
  proteinase	
  K	
  (40	
  µl	
  of	
  >600	
  
mAU/ml,	
  Qiagen)	
  for	
  1	
  hour	
  at	
  55°C,	
  followed	
  by	
  incubation	
  with	
  200	
  µl	
  of	
  buffer	
  AL	
  
for	
  10-­‐30	
  minutes	
  at	
  70°C.	
  200	
  µl	
  of	
  ethanol	
  was	
  added	
  prior	
  to	
  loading	
  on	
  the	
  
column.	
  Column	
  washing	
  was	
  as	
  described	
  in	
  the	
  Qiagen	
  protocol	
  and	
  samples	
  were	
  
eluted	
  in	
  100	
  µl	
  of	
  Buffer	
  AE.	
  DNA	
  concentrations	
  were	
  determined	
  with	
  Quant-­‐IT	
  
(Invitrogen)	
  according	
  to	
  the	
  manufacturer’s	
  protocol.	
  
	
  
Preparation	
  of	
  16S	
  rRNA	
  amplicons	
  for	
  sequencing.	
  We	
  used	
  40	
  ng	
  of	
  each	
  DNA	
  
sample	
  as	
  template	
  in	
  PCR	
  with	
  the	
  following	
  final	
  concentrations	
  of	
  reagents:	
  200	
  
nM	
   357F	
   primer,	
   200	
   nM	
   926R	
   primer,	
   1X	
   AccuPrime	
   PCR	
   Buffer	
   II	
   (Invitrogen),	
  
0.75	
  U	
  of	
  AccuPrime	
  Taq	
  DNA	
  High	
  Fidelity	
  (Invitrogen).	
  357F/962R	
  were	
  designed	
  
by	
   the	
   Human	
   Microbiome	
   Project,	
   amplify	
   the	
   V3-­‐V5	
   variable	
   regions	
   of	
   the	
   16S	
  
rRNA	
  gene,	
  and	
  contain	
  unique	
  barcodes	
  that	
  can	
  be	
  used	
  to	
  multiplex	
  sequencing	
  
reactions	
   (25).	
   Each	
   reaction	
   was	
   set	
   up	
   in	
   triplicate	
   and	
   amplified	
   using	
   the	
  
following	
  cycle:	
  95°C	
  for	
  2	
  min,	
  followed	
  by	
  30	
  cycles	
  of	
  95°C	
  for	
  20	
  sec,	
  50°C	
  for	
  30	
  
sec,	
   and	
   72°C	
   for	
   5	
   min,	
   with	
   a	
   final	
   extension	
   at	
   72°C	
   for	
   5	
   min.	
   Successful	
   PCR	
  
	
  

55	
  

amplification	
   products	
   from	
   triplicate	
   reactions	
   were	
   pooled	
   and	
   cleaned	
   with	
  
Agencourt	
   AMPure	
   XP	
   beads	
   essentially	
   according	
   to	
   protocol	
   with	
   minor	
  
modifications	
   (Beckman-­‐Coulter).	
   Briefly,	
   products	
   were	
   resuspended	
   with	
   a	
   0.7X	
  
volume	
   of	
   beads,	
   washed	
   twice	
   with	
   70%	
   ethanol,	
   and	
   eluted	
   with	
   25	
   µl	
   of	
   low	
  
EDTA	
   TE	
   Buffer	
   (10	
   mM	
   Tris,	
   0.1	
   mM	
   EDTA).	
   	
   Concentrations	
   of	
   purified	
   DNA	
  
samples	
   were	
   determined	
   using	
   Quant-­‐IT	
   (Invitrogen)	
   according	
   to	
   the	
  
manufacturer’s	
   protocol	
   and	
   were	
   pooled	
   in	
   equimolar	
   amounts.	
   Nucleotide	
  
sequencing	
  was	
  performed	
  on	
  a	
  454	
  GS	
  Junior	
  (Roche	
  Diagnostics)	
  at	
  Michigan	
  State	
  
University	
   according	
   to	
   the	
   manufacturer’s	
   protocols.	
   Four	
   sequencing	
   runs	
   were	
  
performed.	
  Run	
  1	
  included	
  BRMW	
  &	
  BRMW10	
  Reactors	
  1,	
  2	
  &	
  3	
  on	
  days	
  2,	
  4,	
  8,	
  14,	
  
and	
   28	
   and	
   BRMG	
   Reactors	
   1,	
   2,	
   &	
   3	
   on	
   days	
   2,	
   4,	
   8,	
   14,	
   22,	
   and	
   28	
   in	
   culture;	
   Run	
   2	
  
included	
  BRMW	
  &	
  BRMW10	
  Reactors	
  1,	
  2,	
  &	
  3	
  on	
  day	
  20	
  in	
  culture;	
  Run	
  3	
  included	
  
BRMW10	
  Reactors	
  1,	
  2,	
  &	
  3	
  on	
  days	
  2,	
  4,	
  6,	
  8,	
  10,	
  12,	
  14,	
  16,	
  18,	
  and	
  20	
  in	
  culture	
  as	
  
well	
  as	
  two	
  independently	
  prepared	
  replicates	
  of	
  the	
  fecal	
  slurry;	
  Run	
  4	
  was	
  an	
  in-­‐
depth	
  analysis	
  of	
  the	
  two	
  replicates	
  of	
  the	
  fecal	
  slurry.	
  
	
  
Analysis	
  of	
  amplicon	
  data.	
  All	
  sequence	
  data	
  were	
  analyzed	
  using	
  mothur	
  (26)	
  
Version	
  1.27.0	
  (August	
  2012).	
  The	
  sequences	
  from	
  the	
  four	
  sequencing	
  runs	
  
described	
  above	
  were	
  initially	
  processed	
  independently	
  and	
  were	
  quality	
  trimmed	
  
and	
  filtered	
  to	
  remove	
  those	
  sequences	
  that	
  had	
  any	
  ambiguous	
  bases,	
  mismatches	
  
to	
  the	
  reverse	
  primer	
  or	
  barcode,	
  homopolymeric	
  stretches	
  longer	
  than	
  8	
  nt,	
  were	
  
shorter	
  than	
  200	
  nt,	
  and/or	
  had	
  an	
  average	
  quality	
  score	
  over	
  a	
  50	
  nt	
  window	
  less	
  
than	
  30	
  (27).	
  Sequences	
  from	
  all	
  four	
  runs	
  were	
  then	
  compiled	
  into	
  a	
  single	
  fasta	
  file	
  

	
  

56	
  

and	
  aligned	
  to	
  the	
  SILVA	
  reference	
  alignment	
  using	
  the	
  NAST-­‐based	
  aligner	
  in	
  
mothur,	
  trimmed	
  to	
  ensure	
  that	
  sequences	
  overlapped,	
  and	
  pre-­‐clustered,	
  allowing	
  
a	
  difference	
  between	
  sequences	
  of	
  2	
  bp	
  or	
  less	
  (27).	
  Potentially	
  chimeric	
  sequences	
  
were	
  removed	
  using	
  the	
  mothur-­‐implementation	
  of	
  UChime	
  (28);	
  remaining	
  
sequences	
  were	
  classified	
  using	
  RDP	
  training	
  set	
  version	
  9	
  (March	
  2012)	
  and	
  
mothur’s	
  implementation	
  of	
  the	
  kmer-­‐based	
  Bayesian	
  classifier.	
  Sequences	
  that	
  
classified	
  as	
  Mitochondria,	
  Chloroplasts,	
  Eukarya	
  or	
  unknown	
  were	
  removed	
  prior	
  
to	
  further	
  analysis.	
  Sequences	
  were	
  binned	
  into	
  Operational	
  Taxonomic	
  Units	
  
(OTUs)	
  with	
  ≤3%	
  sequence	
  dissimilarity	
  using	
  the	
  average	
  neighbor	
  algorithm	
  of	
  
mothur.	
  Taxonomy	
  was	
  assigned	
  to	
  each	
  OTU	
  based	
  upon	
  the	
  majority	
  sequence	
  
consensus	
  within	
  that	
  OTU	
  (29).	
  	
  
Comparison	
  of	
  cultures	
  grown	
  in	
  BRMW,	
  BRMW10	
  and	
  BRMG	
  to	
  the	
  fecal	
  slurry.	
  In	
  
order	
  to	
  determine	
  the	
  differences	
  in	
  microbial	
  community	
  structure	
  and	
  
composition	
  within	
  and	
  between	
  cultures	
  grown	
  in	
  BRMW,	
  BRMW10,	
  BRMG	
  and	
  the	
  
fecal	
  slurry,	
  we	
  extracted	
  these	
  samples	
  (from	
  Runs	
  1	
  (all	
  samples	
  included),	
  2	
  (all	
  
samples	
  included),	
  and	
  3	
  (fecal	
  slurry	
  only))	
  from	
  the	
  larger	
  pool	
  of	
  samples	
  prior	
  to	
  
further	
  analysis.	
  For	
  OTU-­‐based	
  analyses,	
  we	
  then	
  removed	
  those	
  sequences	
  that	
  
were	
  represented	
  by	
  only	
  a	
  single	
  sequence	
  across	
  the	
  samples	
  studied	
  (singletons).	
  
Before	
  determining	
  alpha-­‐diversity	
  measures	
  (Inverse	
  Simpson,	
  Chao	
  Richness,	
  
Simpson	
  Evenness)	
  for	
  the	
  communities	
  using	
  the	
  calculators	
  present	
  in	
  mothur,	
  we	
  
randomly	
  subsampled	
  our	
  data	
  to	
  the	
  maximum	
  number	
  of	
  sequences	
  present	
  in	
  the	
  
smallest	
  group	
  (nseqs=554)	
  ten	
  times	
  and	
  presented	
  the	
  mean	
  ±	
  standard	
  
deviations	
  of	
  these	
  calculations.	
  	
  We	
  also	
  used	
  one	
  iteration	
  of	
  our	
  randomly	
  

	
  

57	
  

subsampled	
  data	
  to	
  calculate	
  the	
  theta	
  dissimilarity	
  measure	
  described	
  by	
  Yue	
  and	
  
Clayton	
  (θYC	
  (30))	
  using	
  the	
  calculators	
  present	
  in	
  mothur.	
  	
  
For	
  taxonomic-­‐based	
  analyses,	
  we	
  classified	
  sequences	
  using	
  the	
  RDP	
  
training	
  set	
  and	
  the	
  mothur-­‐implementation	
  of	
  the	
  RDP-­‐classifier	
  as	
  described	
  
above.	
  Sequences	
  that	
  were	
  classified	
  to	
  the	
  same	
  taxonomic-­‐level	
  (Genus,	
  Family,	
  
Order,	
  Class,	
  or	
  Phylum)	
  were	
  binned	
  together	
  into	
  phylotypes.	
  Prior	
  to	
  calculation	
  
of	
  the	
  θYC	
  dissimilarity	
  index,	
  those	
  phylotypes	
  that	
  were	
  represented	
  by	
  a	
  single	
  
sequence	
  across	
  the	
  samples	
  were	
  removed	
  and	
  the	
  data	
  was	
  randomly	
  subsampled	
  
to	
  the	
  largest	
  number	
  of	
  sequences	
  present	
  in	
  the	
  smallest	
  sample	
  (nseqs=557).	
  
	
  	
  	
  	
  	
  For	
  analysis	
  of	
  phylogenetic	
  diversity,	
  we	
  randomly	
  subsampled	
  our	
  data	
  to	
  the	
  
maximum	
  number	
  of	
  sequences	
  present	
  in	
  the	
  smallest	
  group	
  (nseqs=557),	
  
calculated	
  an	
  uncorrected	
  pairwise	
  distance	
  matrix	
  between	
  the	
  aligned	
  sequences,	
  
and	
  then	
  built	
  a	
  relaxed	
  neighbor	
  joining	
  phylogenetic	
  tree	
  of	
  the	
  sequence	
  data	
  
using	
  the	
  mothur-­‐implementation	
  of	
  clearcut	
  (31).	
  We	
  calculated	
  the	
  phylogenetic	
  
distance	
  between	
  samples	
  using	
  the	
  mothur-­‐implementation	
  of	
  weighted	
  unifrac	
  
(32).	
  	
  We	
  used	
  the	
  mothur-­‐implementation	
  of	
  principal	
  coordinates	
  analysis	
  
(PCOA),	
  ANOSIM,	
  and	
  AMOVA	
  for	
  analysis	
  of	
  θYC	
  Dissimilarity	
  Indices.	
  PCOA	
  data	
  
was	
  plotted	
  in	
  R	
  whereas	
  other	
  distance	
  measures	
  were	
  plotted	
  in	
  Microsoft	
  Excel.	
  	
  
	
  	
  	
  	
  	
  The	
  percent	
  abundance	
  of	
  each	
  phylotype	
  at	
  the	
  class-­‐level	
  was	
  based	
  upon	
  
analysis	
  of	
  total	
  sequences	
  present	
  in	
  the	
  indicated	
  samples	
  (not	
  subsampled).	
  The	
  
percent	
  abundance	
  in	
  the	
  fecal	
  slurry	
  samples	
  was	
  based	
  upon	
  the	
  in-­‐depth	
  
sequencing	
  (nseqs	
  =	
  38,169	
  and	
  44,792)	
  of	
  duplicate	
  fecal	
  slurry	
  samples	
  
(Sequencing	
  Run	
  4).	
  

	
  

58	
  

	
  
Analysis	
  of	
  variation	
  in	
  cultures	
  grown	
  in	
  BRMW10	
  and	
  comparison	
  to	
  the	
  starting	
  
fecal	
  inoculum.	
  In	
  order	
  to	
  calculate	
  the	
  differences	
  that	
  occurred	
  between	
  the	
  three	
  
reactors	
  grown	
  in	
  BRMW10	
  and	
  between	
  BRMW10	
  and	
  the	
  fecal	
  inoculum,	
  we	
  
extracted	
  sequence	
  data	
  from	
  all	
  three	
  reactors	
  on	
  Day	
  2,	
  4,	
  6,	
  8,	
  10,	
  12,	
  14,	
  16,	
  18	
  
and	
  20	
  and	
  the	
  two	
  replicates	
  of	
  the	
  fecal	
  slurry	
  samples	
  (Sequencing	
  Run	
  3).	
  For	
  
OTU-­‐based	
  analyses,	
  we	
  removed	
  sequences	
  represented	
  by	
  only	
  a	
  single	
  sequence	
  
across	
  the	
  data	
  points	
  and	
  randomly	
  subsampled	
  all	
  groups	
  to	
  the	
  largest	
  number	
  
present	
  in	
  the	
  smallest	
  sample	
  (nseqs=1814	
  for	
  comparisons	
  without	
  fecal	
  slurry	
  
samples;	
  nseqs=1582	
  for	
  comparisons	
  with	
  fecal	
  slurry	
  samples).	
  For	
  taxonomic-­‐
based	
  analyses,	
  we	
  classified	
  sequences	
  using	
  the	
  RDP-­‐classifier	
  and	
  binned	
  into	
  
phylotypes,	
  removed	
  singletons,	
  and	
  subsampled	
  as	
  described	
  above.	
  We	
  calculated	
  
the	
  θYC	
  dissimilarity	
  index,	
  and	
  ANOSIM	
  and	
  AMOVA	
  of	
  the	
  θYC	
  distance	
  matrix	
  as	
  
described	
  above.	
  We	
  calculated	
  the	
  Pearson	
  Product	
  Moment	
  Correlation	
  Coefficient	
  
using	
  the	
  Pearson	
  function	
  in	
  Microsoft	
  Excel.	
  Data	
  was	
  plotted	
  in	
  Microsoft	
  Excel.	
  	
  
	
  
Results	
  
Mini-­‐bioreactor	
   design.	
   After	
   evaluating	
   different	
   materials	
   for	
   use	
   in	
   the	
   MBRA,	
  
we	
   determined	
   that	
   DSM	
   Somos	
   Watershed	
   XC	
   11122	
   was	
   the	
   most	
   appropriate	
  
material	
  due	
  to	
  its	
  transparency,	
  resistance	
  to	
  water	
  and	
  humidity,	
  durability	
  after	
  
autoclaving,	
   previous	
   use	
   in	
   microbiological	
   studies	
   (33),	
   and	
   its	
   ability	
   for	
   use	
   with	
  
enclosed	
   3D	
   fabrication.	
   Stereolithography	
   was	
   used	
   to	
   create	
   blocks	
   of	
   six	
  
bioreactor	
   chambers	
   within	
   each	
   mini-­‐bioreactor	
   array	
   (Fig.	
   2.1A).	
   We	
   chose	
  

	
  

59	
  

stereolithography	
   for	
   synthesis	
   in	
   order	
   to	
   have	
   a	
   completely	
   enclosed	
   reactor	
  
design,	
  thereby	
  minimizing	
  points	
  at	
  which	
  contaminants	
  could	
  be	
  introduced	
  into	
  
the	
   reactors.	
   The	
   interior	
   volume	
   of	
   each	
   reactor	
   was	
   designed	
   to	
   allow	
   a	
   15	
   ml	
  
medium	
   volume	
   and	
   a	
   10	
   ml	
   headspace	
   volume.	
   This	
   reactor	
   volume	
   was	
   chosen	
   to	
  
allow	
  for	
  a	
  less	
  drastic	
  change	
  in	
  volume	
  of	
  medium	
  upon	
  sampling	
  (<10%	
  for	
  a	
  1	
  ml	
  
sample)	
  while	
  still	
  maintaining	
  sufficient	
  headspace	
  within	
  the	
  reactor	
  to	
  reduce	
  the	
  
risk	
   of	
   contamination	
   of	
   the	
   source	
   medium.	
   	
   Further,	
   the	
   small	
   medium	
   volume	
  
(compared	
   to	
   liter	
   scale	
   reactors)	
   reduced	
   medium	
   consumption	
   and	
   waste	
  
production,	
  thereby	
  limiting	
  expense	
  of	
  operation.	
  
	
  
	
  
	
  
	
  

	
  

60	
  

	
  
Figure	
  2.1.	
  Schematic	
  of	
  MBRA	
  design	
  and	
  set-­‐up.	
  (A)	
  Picture	
  of	
  a	
  single	
  mini-­‐
bioreactor	
  (MBRA)	
  strip	
  with	
  dimensions	
  indicated.	
  (B)	
  Schematic	
  of	
  bioreactor	
  
setup.	
  Each	
  mini-­‐bioreactor	
  was	
  operated	
  with	
  a	
  total	
  volume	
  of	
  15	
  ml.	
  We	
  used	
  
peristaltic	
  pump	
  tubing	
  with	
  0.89	
  mm	
  inner	
  diameter	
  (ID)	
  and	
  the	
  lowest	
  peristaltic	
  
pump	
  setting	
  to	
  produce	
  a	
  flow	
  rate	
  from	
  the	
  feed	
  of	
  0.625	
  ml/hr	
  (24	
  hr	
  retention	
  
time).	
  We	
  used	
  1.14	
  ID	
  peristaltic	
  tubing	
  on	
  the	
  waste	
  lines	
  to	
  prevent	
  clogging.	
  
Each	
  reactor	
  was	
  constantly	
  stirred	
  using	
  a	
  magnetic	
  stir	
  bar	
  driven	
  by	
  a	
  stir	
  plate.	
  
Reactors	
  were	
  maintained	
  in	
  an	
  anaerobic	
  chamber	
  (5%	
  H2,	
  5%	
  CO2,	
  90%	
  N2)	
  at	
  
37°C.	
  
	
  
	
  
MBRA	
  operation.	
  We	
  operated	
  the	
  MBRA	
  as	
  continuously-­‐stirred	
  tank	
  bioreactors	
  
with	
  a	
  retention	
  time	
  of	
  24	
  hours	
  (Fig.	
  2.1B).	
  Similar	
  operating	
  conditions	
  have	
  been	
  
used	
   in	
   other	
   in	
  vitro	
   colonic	
   models	
   of	
   the	
   distal	
   colon	
   that	
   supported	
   growth	
   of	
  
diverse	
   microbial	
   communities	
   (14,	
   34).	
   These	
   operating	
   conditions	
   allowed	
   us	
   to	
  
monitor	
  growth	
  of	
  our	
  fecal	
  communities	
  in	
  culture	
  for	
  28	
  days.	
  In	
  order	
  to	
  facilitate	
  

	
  

61	
  

growth	
  of	
  strict	
  anaerobes,	
  we	
  maintained	
  the	
  MBRA	
  in	
  a	
  37°C	
  anaerobic	
  chamber	
  
with	
  5%	
  CO2,	
  5%	
  H2	
  and	
  90%	
  N2	
  atmosphere.	
  We	
  evaluated	
  four	
  different	
  types	
  of	
  
media,	
  which	
  were	
  variations	
  of	
  media	
  developed	
  in	
  the	
  Wilcox	
  (13)	
  and	
  Gordon	
  (8)	
  
laboratories	
   for	
   culturing	
   diverse	
   fecal	
   communities.	
   We	
   named	
   the	
   media	
  
variations	
  BRMW	
  (Bioreactor	
  Medium	
  Wilcox),	
  BRMG	
  (Bioreactor	
  Medium	
  Gordon),	
  
BRMW10	
  and	
  BRMG10.	
  BRMW10	
  and	
  BRMG10	
  are	
  modifications	
  of	
  BRMW	
  and	
  BRMG	
  
that	
   contain	
   10%	
   of	
   the	
   carbohydrates	
   found	
   in	
   the	
   original	
   media,	
   which	
   we	
  
evaluated	
   because	
   previous	
   work	
   with	
   the	
   three-­‐stage	
   compound	
   continuous	
  
culture	
   system	
   had	
   shown	
   that	
   the	
   majority	
   of	
   carbohydrates	
   in	
   the	
   medium	
   are	
  
depleted	
   prior	
   to	
   reaching	
   the	
   third	
   chamber	
   of	
   the	
   system	
   that	
   mimics	
   the	
   distal	
  
colon	
   (34).	
   The	
   media	
   were	
   buffered	
   with	
   both	
   phosphate	
   buffer	
   and	
   bicarbonate,	
  
which	
  was	
  sufficient	
  to	
  maintain	
  a	
  relatively	
  constant	
  pH	
  without	
  addition	
  of	
  acid	
  or	
  
base	
   (pH	
   was	
   between	
   6.5	
   and	
   7.0	
   across	
   all	
   twelve	
   reactors,	
   when	
   measured	
   every	
  
2	
  days).	
  
	
  
Short	
  chain	
  fatty	
  acid	
  profiles	
  of	
  cultures	
  grown	
  in	
  each	
  medium	
  revealed	
  that	
  
BRMW10	
   communities	
   were	
   highly	
   stable	
   and	
   most	
   similar	
   to	
   the	
   fecal	
  
inoculum.	
   	
   One	
   key	
   function	
   the	
   gut	
   microbiota	
   provides	
   to	
   its	
   host	
   is	
   the	
  
fermentation	
   of	
   otherwise	
   indigestible	
   polysaccharides	
   to	
   short	
   chain	
   fatty	
   acids	
  
(SCFAs),	
  which	
  can	
  both	
  be	
  used	
  by	
  the	
  host	
  for	
  energy	
  metabolism	
  and	
  influence	
  a	
  
wide	
   array	
   of	
   host	
   functions	
   (reviewed	
   in	
   (3)).	
   Changes	
   in	
   the	
   concentration	
   and	
  
proportion	
  of	
  individual	
  SCFAs	
  correlate	
  with	
  changes	
  in	
  bacterial	
  groups	
  (35-­‐37),	
  
thus	
   monitoring	
   SCFAs	
   can	
   be	
   a	
   good	
   measure	
   of	
   community	
   functional	
   stability.	
  

	
  

62	
  

Using	
   HPLC,	
   we	
   monitored	
   the	
   concentrations	
   of	
   five	
   SCFAs	
   (acetate,	
   butyrate,	
  
isobutryate,	
  lactate	
  and	
  propionate)	
  in	
  our	
  reactors	
  over	
  time	
  and	
  compared	
  these	
  
concentrations	
   to	
   the	
   starting	
   fecal	
   inoculum.	
   While	
   the	
   low	
   concentration	
   of	
  
isobutyrate,	
   an	
   isoform	
   of	
   butyrate,	
   is	
   typically	
   neglected,	
   it	
   was	
   monitored	
   in	
   this	
  
study	
   because	
   it	
   has	
   been	
   described,	
   along	
   with	
   butyrate,	
   as	
   the	
   most	
   rapid	
  
indicators	
  for	
  monitoring	
  process	
  instability	
  (38,	
  39).	
  	
  	
  
We	
   found	
   considerable	
   differences	
   in	
   SCFA	
   profiles	
   and	
   concentrations	
  
across	
  the	
  four	
  different	
  media	
  (Figure	
  2.2,	
  Figure	
  S2.1).	
  For	
  an	
  individual	
  reactor	
  in	
  
a	
  given	
  medium,	
  we	
  generally	
  observed	
  that	
  stabilization	
  in	
  the	
  SCFA	
  profile,	
  which	
  
we	
   defined	
   as	
   a	
   Pearson	
   Product	
   Moment	
   Correlation	
   Coefficient	
   >	
   0.8	
   from	
   one	
  
sample	
  day	
  to	
  the	
  next	
  (as	
  had	
  been	
  previously	
  reported	
  to	
  represent	
  stability	
  (23,	
  
40)),	
   occurred	
   by	
   day	
   two	
   of	
   growth	
   in	
   culture	
   and	
   persisted	
   throughout	
   the	
  
remainder	
   of	
   time	
   in	
   culture	
   (Fig	
   2.3A-­‐D).	
   The	
   two	
   exceptions	
   to	
   this	
   trend	
   were	
  
from	
  a	
  single	
  reactor	
  grown	
  in	
  BRMG	
  and	
  another	
  reactor	
  grown	
  in	
  BRMG10	
  where	
  
the	
   correlation	
   coefficient	
   dropped	
   to	
   0.73	
   and	
   0.77,	
   respectively	
   during	
   single	
   two-­‐
day	
  windows.	
  In	
  general,	
  the	
  correlation	
  coefficient	
  was	
  above	
  0.9	
  for	
  most	
  points	
  in	
  
culture.	
   A	
   notable	
   exception	
   occurred	
   between	
   day	
   18	
   and	
   20	
   in	
   culture	
   in	
   BRMG	
  
and	
   BRMG10	
   where	
   the	
   correlation	
   coefficient	
   dropped	
   quickly	
   then	
   returned	
   to	
  
stability.	
   This	
   time	
   frame	
   corresponded	
   to	
   a	
   change	
   from	
   one	
   batch	
   of	
   source	
  
medium	
  to	
  a	
  new	
  batch	
  of	
  source	
  medium.	
  	
  
	
  

	
  

63	
  

	
  
Figure	
  2.2.	
  	
  Short	
  Chain	
  Fatty	
  Acid	
  analysis	
  of	
  bioreactor	
  cutures.	
  	
  Comparison	
  
of	
  the	
  Short	
  Chain	
  Fatty	
  Acid	
  Profiles	
  of	
  the	
  fecal	
  inoculum	
  to	
  bioreactors	
  cultured	
  in	
  
BRMW,	
  BRMG,	
  BRMW10,	
  or	
  BRMG10.	
  (A-­‐E)	
  Mean	
  Percent	
  Abundance	
  of	
  SCFA	
  in	
  
duplicate	
  samples	
  of	
  original	
  fecal	
  inoculum	
  (A)	
  and	
  from	
  triplicate	
  reactors	
  
cultured	
  in	
  BRMW(B),	
  BRMG(C),	
  BRMW10(D),	
  or	
  BRMG10	
  (E)	
  sampled	
  on	
  every	
  two	
  
days.	
  (F-­‐G)	
  	
  We	
  calculated	
  Pearson	
  Product	
  Moment	
  Correlation	
  Coefficients	
  of	
  the	
  
SCFA	
  composition	
  (percent	
  abundance	
  of	
  acetate,	
  butyrate,	
  isobutyrate,	
  lactate,	
  and	
  
propionate)	
  for	
  each	
  day	
  in	
  culture	
  compared	
  to	
  the	
  initial	
  fecal	
  inoculum.	
  Data	
  
presented	
  are	
  the	
  mean	
  of	
  triplicate	
  reactors	
  (error	
  bars=SD	
  of	
  mean).	
  Symbols	
  
represent	
  cultures	
  grown	
  in	
  BRMW	
  (n),	
  BRMW10	
  (o),	
  BRMG(l)	
  or	
  BRMG10	
  (¡).	
  
	
  
	
  

	
  

64	
  

Pearson Correlation Coefficient!

A!

B!

1!

0.95!

0.85!

0.9!

0.7!
0!

C!

1!

4!

8! 12! 16! 20! 24! 28!

D!

1!

0.9875!

0!

4!

8! 12! 16! 20! 24! 28!

0!

4!

8! 12! 16! 20! 24! 28!
Days in Culture!

1!

0.85!

0.975!
0! 4! 8! 12! 16! 20! 24! 28!
Days in Culture!

0.7!

	
  
Figure	
   2.3.	
   Pearson	
   Product	
   Moment	
   Correlation	
   Coefficient	
   of	
   the	
   SCFA	
  
composition	
   in	
   bioreactor	
   cultures.	
   For	
   each	
   individual	
   reactor	
   grown	
   in	
  
BRMW(A),	
  BRMG(B),	
  BRMW10	
  (C),	
  or	
  BRMG10(D),	
  we	
  calculated	
  the	
  Pearson	
  Product	
  
Moment	
  Correlation	
  Coefficient	
  of	
  the	
  SCFA	
  composition	
  from	
  day	
  X	
  to	
  day	
  X+2	
  in	
  
culture	
  (X=2-­‐28	
  at	
  two	
  day	
  intervals)	
  and	
  plotted	
  this	
  is	
  a	
  function	
  of	
  day	
  X.	
  Black	
  
symbols	
   represent	
   the	
   data	
   from	
   reactor	
   1,	
   gray	
   symbols	
   represent	
   the	
   data	
   from	
  
reactor	
  2,	
  and	
  white	
  symbols	
  represent	
  the	
  data	
  for	
  reactor	
  3	
  in	
  each	
  panel.	
  
	
  
The	
  similarity	
  in	
  SCFA	
  profiles	
  that	
  we	
  observed	
  from	
  one	
  replicate	
  reactor	
  
to	
   another	
   varied	
   by	
   the	
   type	
   of	
   medium	
   evaluated.	
   Bioreactor	
   communities	
  
established	
  in	
  BRMW10	
  were	
  highly	
  similar	
  to	
  one	
  another,	
  with	
  a	
  mean	
  correlation	
  
coefficient	
  from	
  one	
  replicate	
  to	
  another	
  on	
  the	
  same	
  day	
  in	
  culture	
  of	
  >0.99	
  (Figure	
  
S2.2B).	
   In	
   contrast,	
   one	
   of	
   the	
   replicate	
   reactors	
   grown	
   in	
   BRMW	
   diverged	
  
drastically	
  from	
  the	
  other	
  two	
  replicates	
  by	
  day	
  4	
  in	
  culture	
  (Figure	
  S2.2A).	
  In	
  this	
  
case,	
   the	
   SCFA	
   profiles	
   of	
   two	
   reactors	
   were	
   highly	
   similar	
   (mean	
   correlation	
  
coefficient	
   across	
   all	
   days	
   sampled	
   >0.99),	
   whereas	
   the	
   third	
   reactor	
   had	
   a	
   SCFA	
  
profile	
   dissimilar	
   to	
   either	
   of	
   the	
   reactors	
   (mean	
   correlation	
   coefficient	
   across	
   all	
  
days	
  sampled	
  of	
  0.46	
  and	
  0.41).	
  The	
  replicate	
  cultures	
  grown	
  in	
  BRMG	
  and	
  BRMG10	
  

	
  

65	
  

were	
  intermediate	
  between	
  these	
  two	
  extremes,	
  with	
  mean	
  correlation	
  coefficients	
  
amongst	
  reactors	
  ranging	
  from	
  0.82	
  to	
  0.98	
  (Figure	
  S2.2C	
  and	
  S2.2D)	
  
We	
   also	
   compared	
   the	
   similarity	
   of	
   the	
   SCFA	
   profiles	
   of	
   our	
   different	
  
bioreactor	
  cultures	
  to	
  the	
  starting	
  fecal	
  slurry.	
  The	
  SCFA	
  profile	
  of	
  our	
  fecal	
  slurry	
  
(87%	
  acetate,	
  11%	
  butyrate,	
  1%	
  propionate,	
  and	
  1	
  %	
  lactate,	
  Table	
  2.1)	
  was	
  similar	
  
to	
  previously	
  reported	
  fecal	
  SCFA	
  profiles	
  (35,	
  41).	
  Cultures	
  grown	
  in	
  BRMW10	
  had	
  a	
  
SCFA	
   profile	
   most	
   similar	
   to	
   the	
   fecal	
   inoculum	
   (Fig.	
   2.2F),	
   although	
   acetate	
  
concentrations	
   were	
   15%	
   lower	
   than	
   in	
   the	
   fecal	
   inoculum	
   with	
   increased	
  
concentrations	
   of	
   the	
   remaining	
   four	
   SCFAs	
   (Table	
   2.1).	
   Cultures	
   grown	
   in	
   BRMW	
  
had	
   a	
   SCFA	
   profile	
   least	
   similar	
   to	
   the	
   fecal	
   inoculum	
   (Fig.	
   2.2F).	
   Acetate	
  
concentrations	
   dropped	
   to	
   24%	
   of	
   the	
   total	
   SCFA	
   pool	
   with	
   large	
   increases	
   in	
  
butyrate	
   (50%)	
   and	
   propionate	
   (20%,	
   Table	
   2.1).	
   Cultures	
   grown	
   in	
   BRMG	
   and	
  
BRMG10	
  had	
   SCFA	
   profiles	
   more	
   similar	
   to	
   the	
   fecal	
   inoculum	
   than	
   those	
   grown	
   in	
  
BRMW,	
   but	
   less	
   similar	
   than	
   cultures	
   grown	
   in	
   BRMW10	
   (Fig.	
   2.2F	
   &	
   2.2G).	
   As	
  
expected,	
   concentrations	
   of	
   acetate,	
   butyrate,	
   and	
   propionate	
   were	
   higher	
   in	
  
cultures	
   grown	
   in	
   BRMG	
   or	
   BRMG10	
  (Fig.	
   S2.1),	
   which	
   contain	
   added	
   SCFAs	
   (acetate,	
  
29.6	
  mM;	
  butyrate,	
  21.6	
  mM;	
  propionate,	
  26.5	
  mM).	
  However,	
  even	
  after	
  correcting	
  
for	
  the	
  calculated	
  levels	
  of	
  SCFAs	
  in	
  the	
  starting	
  media,	
  cultures	
  grown	
  in	
  BRMG	
  and	
  
BRMG10	
  still	
   had	
   reduced	
   acetate	
   concentrations	
   (47%	
   and	
   55%,	
   respectively,	
   Table	
  
2.1)	
  and	
  higher	
  concentrations	
  of	
  butyrate	
  (41%	
  and	
  33%,	
  respectively,	
  Table	
  2.1)	
  
than	
  the	
  fecal	
  inoculum.	
  	
  
	
  
	
  

	
  

66	
  

Table	
  2.1.	
  SCFA	
  profiles	
  for	
  fecal	
  inoculum	
  and	
  MBRA	
  cultures.	
  
	
  
Fecal	
  
Inoculum1	
  
BRMW2	
  

Acetate	
  
Butyrate	
  	
  
Propionate	
  	
   Lactate	
  
Isobutyrate	
  	
  
87%	
  ±	
  
10%	
  ±	
  1.5%	
   1.0%	
  ±	
  0.4%	
   1.0	
  ±	
  0.0%	
   ND	
  
1.1%	
  
24.1	
  ±	
  
ND	
  
2.6%	
  
50.7	
  ±	
  2.9%	
   19.9	
  ±	
  2.5%	
   5.3	
  ±	
  1.2%	
  
	
  
	
  
	
  
	
  
2	
  
BRMW10
70.5±	
  2.9%	
   16.5	
  ±	
  1.9%	
   6.1	
  ±	
  1.3%	
  
4.7	
  ±	
  0.2%	
   1.9±0.09%	
  
	
  
	
  
	
  
	
  
	
  
2,3
BRMG 	
  
43.2	
  ±	
  
2.6%	
  
33.9	
  ±	
  6.1%	
   21	
  ±	
  4.5%	
  
1.7	
  ±	
  0.6%	
   0.09	
  ±	
  0.1%	
  
	
  
	
  
	
  
	
  
	
  
BRMG102,3	
   47.3	
  ±	
  
5.5%	
  
30.8	
  ±	
  6.6%	
   20.4	
  ±	
  1.9%	
   1.2	
  ±	
  0.2%	
   0.2	
  ±	
  0.2%	
  
	
  
	
  
	
  
	
  
	
  
1	
  Values	
  reported	
  are	
  the	
  mean	
  of	
  duplicate	
  slurry	
  samples	
  for	
  fecal	
  inoculum	
  ±	
  SD	
  	
  
2	
  Values	
  reported	
  are	
  the	
  mean	
  of	
  samples	
  collected	
  every	
  two	
  days	
  for	
  28	
  days	
  
across	
  triplicate	
  bioreactors	
  ±	
  SD	
  
3	
  Background	
  concentrations	
  of	
  acetate,	
  butyrate	
  and	
  propionate	
  (29.6	
  mM,	
  21.6	
  
mM,	
  and	
  26.5	
  mM,	
  respectively)	
  in	
  BRMG	
  and	
  BRMG10	
  were	
  subtracted	
  prior	
  to	
  
determining	
  the	
  mean	
  percent	
  abundance	
  of	
  each	
  SCFA	
  species.	
  
	
  
	
  
Comparison	
  of	
  MBRA	
  microbial	
  composition	
  and	
  structure	
  demonstrates	
  that	
  
communities	
   cultured	
   in	
   BRMW10	
   are	
   most	
   similar	
   to	
   the	
   fecal	
   inoclum.	
   In	
  
order	
   to	
   study	
   how	
   culturing	
   in	
   the	
   bioreactors	
   influenced	
   the	
   structure	
   of	
   the	
  
microbial	
   communities,	
   we	
   analyzed	
   16S	
   rRNA	
  gene	
   diversity	
   of	
   samples	
   collected	
  
from	
   triplicate	
   reactors	
   grown	
   in	
   BRMW,	
   BRMG,	
   and	
   BRMW10	
   at	
   several	
   points	
  
during	
   growth	
   as	
   well	
   as	
   from	
   duplicate	
   samples	
   of	
   the	
   initial	
   fecal	
   slurry.	
   We	
   did	
  
not	
   include	
   cultures	
   grown	
   in	
   BRMG10	
   in	
   these	
   analyses	
   because	
   preliminary	
   16S	
  
rRNA	
   gene	
   sequence	
   analyses	
   of	
   BRMG10	
   communities	
   revealed	
   that	
   the	
   cultures	
  
were	
  dissimilar	
  from	
  the	
  fecal	
  inoculum	
  (data	
  not	
  shown).	
  

	
  

67	
  

Overall,	
  culturing	
  in	
  any	
  of	
  the	
  three	
  media	
  resulted	
  in	
  significant	
  decreases	
  
in	
   microbial	
   diversity	
   and	
   species	
   richness	
   and	
   an	
   increase	
   in	
   species	
   evenness	
  
compared	
   to	
   the	
   fecal	
   inoculum	
   (Fig	
   2.4A-­‐C).	
   The	
   highest	
   levels	
   of	
   diversity	
   and	
  
richness	
   were	
   seen	
   in	
   the	
   communities	
   cultured	
   in	
   BRMW10,	
   whereas	
   the	
   lowest	
  
levels	
  of	
  diversity	
  and	
  richness	
  were	
  seen	
  in	
  the	
  communities	
  cultured	
  in	
  BRMG.	
  As	
  
might	
   be	
   expected	
   based	
   upon	
   the	
   changes	
   in	
   diversity,	
   richness	
   and	
   evenness,	
  
culturing	
  also	
  resulted	
  in	
  changes	
  in	
  overall	
  community	
  structure	
  (Fig.	
  2.4D-­‐F).	
  We	
  
examined	
  the	
  changes	
  in	
  community	
  structure	
  at	
  multiple	
  taxonomic	
  levels	
  (Order,	
  
Genus,	
   and	
   operational	
   taxonomic	
   units	
   (OTUs)	
   with	
   sequence	
   dissimilarity	
   ≤	
   3%	
   in	
  
Fig.	
  2.4D,	
  2.4E,	
  and	
  2.4F,	
  respectively)	
  using	
  the	
  dissimilarity	
  measure	
  described	
  by	
  
Yue	
   and	
   Clayton	
   (θYC	
   (30)),	
   which	
   compares	
   community	
   structures	
   based	
   upon	
  
shared	
   community	
   membership	
   and	
   abundance.	
   We	
   found	
   that	
   the	
   communities	
  
cultured	
  in	
  BRMG	
  were	
  least	
  similar	
  to	
  the	
  starting	
  fecal	
  inoculum	
  at	
  all	
  taxonomic	
  
levels	
  (Fig.	
  2.4D-­‐F,	
  scaled	
  from	
  0	
  to	
  1,	
  with	
  1	
  being	
  least	
  similar).	
  When	
  examined	
  at	
  
the	
   Order	
   level,	
   communities	
   cultured	
   in	
   BRMW	
   appeared	
   slightly	
   more	
   similar	
   to	
  
the	
  starting	
  fecal	
  inoculum	
  than	
  communities	
  cultured	
  in	
  BRMW10,	
  throughout	
  their	
  
time	
   in	
   culture,	
   although	
   these	
   differences	
   were	
   not	
   statistically	
   significant	
  
(student’s	
  two-­‐tailed	
  t-­‐test,	
  p>0.07	
  for	
  all	
  time	
  points).	
  	
  As	
  expected,	
  comparing	
  the	
  
BRMW	
   and	
   BRMW10	
   communities	
   at	
   finer	
   taxonomic	
   resolution	
   (Fig.	
   2.4E	
   and	
   2.4F)	
  
revealed	
   more	
   dissimilarity	
   to	
   the	
   starting	
   fecal	
   inoculum	
   after	
   culture	
   in	
   either	
  
medium.	
   Culturing	
   in	
   either	
   medium	
   produced	
   similar	
   levels	
   of	
   change	
   in	
   the	
  
community	
  structure	
  relative	
  to	
  the	
  fecal	
  inoculum,	
  with	
  exception	
  of	
  the	
  bioreactor	
  
communities	
   in	
   BRMW	
   reactors	
   on	
   day	
   4,	
   which	
   were	
   much	
   more	
   similar	
   to	
   the	
  

	
  

68	
  

starting	
  fecal	
  inoculum.	
  Similar	
  differences	
  in	
  community	
  structure	
  among	
  the	
  three	
  
media	
  were	
  also	
  observed	
  when	
  we	
  compared	
  their	
  16S	
  rRNA	
  gene	
  diversity	
  using	
  
shared	
  phylogenetic	
  distance	
  with	
  weighted	
  Unifrac	
  ((32),	
  data	
  not	
  shown).	
  
	
  
A!

B!

C!

10!
5!

100!

0!

0!

1!
0.8!
0.6!
0.4!
0.2!
0!

Order!

0! 4! 8! 12! 16! 20! 24! 28!
Days in Culture!

E!
θYC Dissimilarity!

D!

0.2!

0!
0! 4! 8! 12! 16! 20! 24! 28!
Days in Culture!
Genus!

1!
0.9!
0.8!
0.7!
0.6!
0.5!
0! 4! 8! 12!16!20!24!28!
Days in Culture!

0! 4! 8! 12! 16! 20! 24! 28!
Days in Culture!

F!
θYC Dissimilarity!

0! 4! 8! 12! 16! 20! 24! 28!
Days in Culture!

θYC Dissimilarity!

0.4!

Evenness!

200!

Richness!

Diversity!

15!

3% OTU!
1!
0.9!
0.8!
0.7!
0.6!
0.5!
0! 4! 8! 12!16!20!24!28!
Days in Culture!

	
  
Figure	
  2.4.	
  Microbial	
  ecology	
  of	
  bioreactors	
  compared	
  to	
  the	
  fecal	
  inoculum.	
  
We	
  compared	
  the	
  composition	
  and	
  structures	
  of	
  microbial	
  communities	
  cultured	
  in	
  
BRMW,	
  BRMW10,	
  or	
  BRMG	
  medium	
  to	
  the	
  fecal	
  inoculum	
  by	
  analysis	
  of	
  16S	
  rRNA	
  
gene	
  diversity.	
  We	
  partitioned	
  sequences	
  with	
  ≤ 3%	
  sequence	
  dissimilarity	
  into	
  
operational	
  taxonomic	
  units	
  (OTUs,	
  A-­‐C,	
  G)	
  as	
  well	
  as	
  examining	
  shared	
  community	
  
composition	
  at	
  the	
  Order	
  (E)	
  and	
  Genus	
  (F)	
  levels.	
  Triplicate	
  BRMW	
  or	
  BRMW10	
  
bioreactor	
  communities	
  were	
  analyzed	
  on	
  days	
  2,	
  4,	
  8,	
  14,	
  20,	
  and	
  28	
  in	
  culture;	
  
triplicate	
  BRMG	
  bioreactor	
  communities	
  were	
  analyzed	
  on	
  days	
  2,	
  4,	
  8,	
  14,	
  22,	
  and	
  
28.	
  In	
  (A-­‐C),	
  we	
  determined	
  the	
  mean	
  (A)	
  microbial	
  diversity	
  (Inverse	
  Simpson	
  
(1/D)),	
  (B)	
  species	
  richness	
  (Chao	
  Richness	
  Estimate),	
  and	
  (C)	
  species	
  evenness	
  
(Simpson	
  Evenness)	
  of	
  triplicate	
  bioreactor	
  communities	
  after	
  the	
  indicated	
  times	
  
in	
  culture	
  in	
  BRMW	
  (n),	
  BRMW10	
  (l),	
  and	
  BRMG	
  (r)	
  to	
  the	
  fecal	
  inoculum	
  (Í).	
  In	
  
(D-­‐F),	
  we	
  compared	
  the	
  differences	
  in	
  the	
  overall	
  community	
  structure	
  (Yue	
  and	
  
Clayton	
  θ	
  Dissimilarity	
  Index	
  (θYC)	
  at	
  the	
  Order	
  (E),	
  Genus	
  (F),	
  and	
  3%	
  OTU	
  (G)	
  
levels	
  between	
  each	
  day	
  in	
  culture	
  in	
  BRMW	
  (n),	
  BRMW10	
  (l),	
  or	
  BRMG	
  (r)	
  to	
  the	
  
fecal	
  inoculum.	
  We	
  plotted	
  the	
  mean	
  of	
  the	
  dissimilarity	
  index	
  for	
  each	
  triplicate	
  set	
  
of	
  reactors	
  as	
  a	
  function	
  of	
  the	
  day	
  in	
  culture.	
  Error	
  bars	
  represent	
  the	
  standard	
  
deviation	
  of	
  the	
  mean.	
  
	
  
	
  

69	
  

Examining	
  how	
  the	
  community	
  composition	
  varied	
  amongst	
  media,	
  we	
  found	
  
that	
  each	
  medium	
  supported	
  a	
  microbial	
  community	
  that	
  was	
  distinct	
  in	
  structure	
  
from	
   the	
   other	
   media	
   tested.	
   We	
   found	
   that	
   both	
   analysis	
   of	
   similarities	
   (ANOSIM	
  
(42),	
   Fig.	
   S2.3)	
   as	
   well	
   as	
   analysis	
   of	
   molecular	
   variance	
   (AMOVA	
   (43)),	
   indicated	
  
that	
   the	
   composition	
   and	
   structure	
   of	
   each	
   community	
   was	
   distinct	
   (P	
   <	
   0.001).	
  
These	
   observations	
   held	
   true	
   whether	
   we	
   examined	
   shared	
   community	
   structure	
   as	
  
measured	
  by	
  common	
  OTUs	
  (Fig.	
  S2.3)	
  or	
  phylogenetic	
  distance	
  (data	
  not	
  shown).	
  	
  
	
  
Comparison	
   of	
   class-­‐level	
   differences	
   among	
   reactor	
   communities	
   revealed	
  
the	
   extent	
   of	
   reorganization	
   of	
   the	
   microbial	
   community	
   during	
   culture.	
  
Examining	
   the	
   class-­‐level	
   distribution	
   of	
   organisms	
   in	
   the	
   different	
   media	
   after	
   4	
  
weeks	
  in	
  culture	
  (Day	
  28),	
  we	
  found	
  that	
  in	
  all	
  three	
  media	
  there	
  was	
  a	
  significant	
  
decrease	
   in	
   the	
   relative	
   abundance	
   of	
   Clostridia	
   (Fig.	
   2.5).	
   This	
   decrease	
   in	
   the	
  
abundance	
   of	
   Clostridia	
   was	
   observed	
   by	
   the	
   second	
   day	
   in	
   culture	
   (Fig	
   S2.4).	
   In	
  
cultures	
   grown	
   with	
   BRMW	
   or	
   BRMW10,	
   this	
   decrease	
   in	
   Clostridia	
   correlated	
  
primarily	
  with	
  increased	
  representation	
  of	
  Bacteroides	
  (Fig.	
  2.5).	
  In	
  two	
  of	
  the	
  three	
  
reactors	
   cultured	
   in	
   BRMG,	
   the	
   loss	
   of	
   Clostridia	
   correlated	
   with	
   a	
   much	
   higher	
  
population	
   of	
   γ-­‐Proteobacteria	
   (~50%	
   of	
   all	
   organisms	
   present),	
   whereas	
   in	
   the	
  
third	
   reactor	
   there	
   was	
   a	
   bloom	
   of	
   Bacilli.	
   Two	
   other	
   classes	
   of	
   Bacteria	
   that	
  
significantly	
   expanded	
   in	
   culture	
   were	
   the	
   Synergistia	
   and	
   Fusobacteria,	
   with	
   the	
  
largest	
  expansions	
  observed	
  in	
  reactors	
  cultured	
  with	
  BRMW10	
  and	
  BRMG.	
  
	
  
	
  

	
  

70	
  

I!

100%!

100%!

80%!

80%!

60%!

60%!

I! MW! MW10! MG!

40%!

40%!

20%!

20%!
0%!

0%!

1 2 3 1 2 3 1 2 3!
Reactor Number!

R1!
Other!
1.2! 0.17!
Other!
Other!
UC
Bacteria!
0.34! 0.17!
Unclassified
Unclassified
Bacteria! Bacteria!
Synergistia!
Synergistia!
Synergistia! 0.0024! 1.3!
Fusobacteria!
Fusobacteria!
Fusobacteria! 0.0035! 5.4!
Gammaproteobacteria!
γ-Proteobacteria!
0.088! 11!
Gammaproteobacteria!
Deltaproteobacteria!
Δ-Proteobacteria!
0.087! 0.34!
Deltaproteobacteria!
Bacteroidia!
Bacteroidia!
20! 45!
Bacteroidia!
Unclassified
Firmicutes!

MW!
R2!
0.3!
2.7!
2.5!
2.3!
9.4!
0.3!
39!

R3!
0.6!
1.2!
0.66!
0.83!
18!
0.25!
46!

R1!
0.56!
0!
3.4!
18!
14!
0.37!
39!

MW10!
R2! R3! R1!
0.93! 0.59! 0.2!
0! 0.074! 0!
10! 7.3! 4.4!
13! 7.2! 2.9!
0!
10! 51!
1.2! 2.9! 0.10!
39! 49! 5.4!

Negativicutes!
Unclassified Firmicutes!
UC Firmicutes!
3.0!
0!
0! 0.25! 0.28! 1.4! 0.74! 0!
Erysipelotrichia!
Negativicutes! 3.5! 5.4! 1.2! 0.66! 2.7! 0.86! 2.4! 8.4!
Negativicutes!
Clostridia!
Erysipelotrichia! 1.5! 0.085! 0! 3.1! 0! 0! 0.07! 0!
Erysipelotrichia!
Bacilli!

Clostridia!
Clostridia!
Bacilli!
Bacilli!

70!
0.11!

31!
0!

42! 28! 21!
0! 0.91! 0!

33!
0!

20!
0!

16!
11!

MG!
R2!
0!
0!
2.1!
8.0!
66!
0.13!
7.6!

R3!
0.48!
0!
7.9!
8.9!
2.2!
3.6!
15!

0!
0!
9.2! 8.1!
0!
0!
6.7! 8.0!
0!
46!

	
  
Figure	
  2.5.	
  Comparison	
  of	
  the	
  class-­‐level	
  distribution	
  of	
  microbes	
  in	
  
bioreactors	
  and	
  the	
  fecal	
  inoculum.	
  	
  Comparison	
  of	
  the	
  class-­‐level	
  distribution	
  of	
  
microbes	
  on	
  day	
  28	
  in	
  culture	
  in	
  BR(MW),	
  BR(MW10),	
  	
  or	
  BR(MG)	
  to	
  the	
  fecal	
  
(I)noculum.	
  The	
  percent	
  abundance	
  of	
  each	
  class	
  of	
  Bacteria	
  in	
  the	
  individual	
  
bioreactor	
  samples	
  is	
  shown	
  as	
  is	
  the	
  mean	
  percent	
  abundance	
  of	
  each	
  class	
  in	
  
duplicate	
  fecal	
  inoculum	
  samples.	
  Reactor	
  numbers	
  correspond	
  to	
  those	
  given	
  in	
  
Figures	
  3	
  and	
  4.	
  UC	
  Bacteria	
  and	
  UC	
  Firmicutes	
  indicate	
  phylotypes	
  that	
  could	
  not	
  
be	
  classified	
  with	
  greater	
  than	
  80%	
  confidence	
  beyond	
  the	
  Domain	
  (Bacteria)	
  and	
  
Phylum	
  (Firmicutes)	
  levels	
  by	
  the	
  ribosomal	
  database	
  project	
  classifier	
  release	
  9.	
  
Phylotypes	
  classified	
  as	
  “Other”	
  were	
  present	
  in	
  <1%	
  abundance	
  in	
  any	
  of	
  the	
  
samples	
  and	
  include	
  Actinobacteria,	
  Uncalssified	
  Bacteroidetes,	
  Lentisphaeria,	
  
Methanobacteria,	
  α-­‐Proteobacteria,	
  β-­‐Proteobacteria,	
  Unclassified	
  Proteobacteria,	
  
and	
  Verrucomicrobiae.	
  
	
  
	
  
Examining	
  the	
  microbial	
  community	
  structure	
  in	
  BRMW10	
  communities	
  at	
  
more	
  frequent	
  time	
  intervals	
  revealed	
  how	
  communities	
  diverge	
  from	
  day-­‐to-­‐
day	
  and	
  reactor-­‐to-­‐reactor.	
  After	
  examining	
  the	
  SCFA	
  profiles	
  and	
  community	
  
composition,	
  it	
  was	
  clear	
  that	
  the	
  cultures	
  grown	
  in	
  BRMG	
  were	
  least	
  similar	
  to	
  the	
  
starting	
  fecal	
  inoculum	
  and	
  were	
  not	
  further	
  studied.	
  Because	
  cultures	
  grown	
  in	
  
BRMW10	
  had	
  a	
  SCFA	
  profile	
  that	
  was	
  more	
  similar	
  to	
  the	
  fecal	
  inoculum	
  (Fig.	
  2.2),	
  
showed	
  lower	
  inter-­‐reactor	
  variation	
  in	
  SCFA	
  profiles	
  across	
  replicates	
  (Fig.	
  2.3,	
  Fig.	
  
S2.2),	
  and	
  maintained	
  levels	
  of	
  microbial	
  diversity,	
  species	
  richness,	
  evenness,	
  and	
  

	
  

71	
  

community	
  composition	
  with	
  as	
  much	
  similarity	
  to	
  the	
  fecal	
  inoculum	
  as	
  cultures	
  
grown	
  in	
  BRMW	
  (Fig.	
  2.5),	
  we	
  focused	
  on	
  the	
  microbial	
  communities	
  that	
  were	
  
established	
  in	
  these	
  reactors	
  in	
  more	
  detail.	
  We	
  examined	
  changes	
  in	
  microbial	
  
community	
  composition	
  in	
  the	
  triplicate	
  reactors	
  grown	
  in	
  this	
  medium	
  every	
  two	
  
days	
  for	
  the	
  first	
  twenty	
  days	
  in	
  culture	
  in	
  order	
  to	
  determine	
  the	
  extent	
  of	
  intra-­‐	
  
and	
  inter-­‐reactor	
  variation	
  that	
  occurred	
  over	
  time	
  in	
  culture.	
  	
  
Two	
  overall	
  trends	
  were	
  apparent	
  from	
  these	
  analyses:	
  1)	
  the	
  communities	
  
present	
  in	
  each	
  individual	
  reactor	
  changed	
  over	
  time	
  in	
  culture	
  with	
  larger	
  changes	
  
in	
  community	
  structure	
  at	
  earlier	
  times	
  in	
  culture	
  (Fig.	
  2.6),	
  and	
  2)	
  the	
  structures	
  of	
  
the	
  microbial	
  communities	
  in	
  replicate	
  reactors	
  diverged	
  from	
  each	
  other	
  over	
  time	
  
(Fig.	
  2.7).	
  We	
  determined	
  the	
  average	
  θYC	
  dissimilarity	
  of	
  each	
  day	
  in	
  culture	
  from	
  all	
  
other	
  days	
  in	
  culture	
  in	
  each	
  replicate	
  reactor	
  (a	
  similar	
  measure	
  of	
  community	
  
stability	
  based	
  upon	
  average	
  Unifrac	
  distances	
  was	
  described	
  by	
  Werner	
  (44)).	
  	
  
Upon	
  plotting	
  these	
  dissimilarities	
  as	
  a	
  function	
  of	
  the	
  day	
  in	
  culture,	
  we	
  found	
  that	
  
the	
  communities	
  present	
  early	
  during	
  cultivation	
  (Days	
  2-­‐6)	
  were	
  less	
  similar	
  to	
  the	
  
overall	
  communities	
  present	
  later	
  in	
  culture	
  (Days	
  8-­‐20,	
  Fig.	
  2.6A).	
  The	
  statistical	
  
significance	
  of	
  the	
  differences	
  was	
  supported	
  by	
  both	
  ANOSIM	
  and	
  AMOVA	
  (Fig.	
  
2.6B).	
  
We	
  also	
  examined	
  the	
  variation	
  in	
  community	
  structure	
  between	
  replicate	
  
reactors	
  over	
  time	
  in	
  culture.	
  We	
  found	
  that	
  the	
  variation	
  between	
  replicate	
  
reactors	
  was	
  smallest	
  during	
  the	
  first	
  week	
  in	
  culture	
  and	
  increased	
  between	
  the	
  
replicate	
  reactors	
  in	
  subsequent	
  weeks	
  (Fig.	
  2.7A).	
  This	
  pattern	
  is	
  most	
  obvious	
  
when	
  comparing	
  reactors	
  1	
  and	
  3,	
  where	
  both	
  ANOSIM	
  and	
  AMOVA	
  revealed	
  

	
  

72	
  

statistically	
  significant	
  differences	
  in	
  community	
  structure	
  during	
  the	
  last	
  two	
  
weeks	
  in	
  culture	
  that	
  were	
  absent	
  during	
  the	
  first	
  week	
  of	
  culture	
  (Fig.	
  2.7B).	
  In	
  
contrast,	
  the	
  community	
  present	
  in	
  reactor	
  2	
  had	
  statistically	
  significant	
  differences	
  
in	
  structure	
  compared	
  to	
  reactors	
  1	
  and	
  3	
  throughout	
  its	
  time	
  in	
  culture	
  (Fig.	
  2.7A	
  
and	
  2.7B),	
  although	
  the	
  differences	
  in	
  community	
  structure	
  between	
  these	
  reactors	
  
early	
  in	
  culture	
  indicated	
  by	
  the	
  ANOSIM	
  values	
  (R=0.78	
  between	
  reactors	
  1	
  &	
  2	
  and	
  
reactors	
  2	
  &	
  3),	
  were	
  not	
  statistically	
  significant.	
  	
  
	
  

θYC Dissimilarity!

A!
Early!

0.6!

Late!

R1!
R2!
R3!

0.45!
0.3!
0.15!
0!
0!

4!

8!
12!
Days in Culture!

16!

20!

B!
ANOSIM!

AMOVA!

Reactor 1!

R=0.73, P=0.008!

0.004!

Reactor 2!

R=0.95, P=0.012!

<0.001!

Reactor 3!

R=0.99, P=0.01!

<0.001!

	
  
	
  
Figure	
  2.6.	
  	
  Analysis	
  of	
  changes	
  in	
  community	
  structure	
  across	
  time	
  in	
  
replicate	
  reactors.	
  	
  We	
  examined	
  the	
  changes	
  that	
  occurred	
  in	
  the	
  structure	
  of	
  
microbial	
  communities	
  based	
  upon	
  changes	
  in	
  shared	
  OTUs	
  (≤	
  3%	
  sequence	
  
dissimilarity)	
  every	
  two	
  days	
  from	
  Day	
  2	
  to	
  Day	
  20	
  during	
  culturing	
  in	
  triplicate	
  
reactors	
  grown	
  in	
  BRMW10.	
  In	
  (A),	
  we	
  compared	
  the	
  mean	
  distance	
  from	
  each	
  day	
  in	
  
culture	
  from	
  all	
  other	
  days	
  in	
  culture	
  in	
  Reactor	
  1	
  (n),	
  Reactor	
  2	
  (l),	
  or	
  Reactor	
  3	
  
(r)	
  using	
  the	
  θYC	
  (A)	
  Dissimilarity	
  Indices.	
  Error	
  bars	
  reflect	
  the	
  standard	
  errors	
  of	
  
the	
  mean.	
  In	
  (B),	
  we	
  calculated	
  ANOSIM	
  and	
  AMOVA	
  between	
  early	
  (Days	
  2-­‐6)	
  and	
  
late	
  (Days	
  8-­‐20)	
  days	
  in	
  culture	
  for	
  each	
  individual	
  reactor	
  based	
  upon	
  θYC	
  
Dissimilarity	
  Indices.	
  

	
  

73	
  

	
  

θYC Dissimilarity!

A!
Early!
R1!
R2!
R3!

0.6!
0.45!
0.3!

Late!

0.15!
0!
0!

B!

4!

8!
12!
Days in Culture!

ANOSIM!

16!

20!

AMOVA!

Early!

Late!

Early!

Late!

Reactor !
1 & 2!

R=0.78,
P=0.079!

R=0.93,
P<0.001!

<0.001!

<0.001!

Reactor !
1 & 3!

R=-0.26
P=0.709!

R=0.51,
P=0.001!

0.597!

0.002!

Reactor !
2 & 3!

R=0.78,
P=0.132!

R=0.98,
P<0.001!

<0.001!

0.001!

	
  
Figure	
  2.7.	
  Pairwise	
  analyses	
  of	
  changes	
  in	
  comminty	
  structure	
  across	
  time	
  
between	
  reactors.	
  	
  We	
  compared	
  the	
  changes	
  that	
  occurred	
  in	
  the	
  structure	
  of	
  
microbial	
  communities	
  between	
  triplicate	
  reactors	
  grown	
  in	
  BRMW10	
  based	
  upon	
  
changes	
  in	
  shared	
  OTUs	
  (≤	
  3%	
  sequence	
  dissimilarity)	
  every	
  two	
  days	
  from	
  Day	
  2	
  to	
  
Day	
  20.	
  In	
  (A),	
  we	
  plotted	
  the	
  the	
  θYC	
  (C)	
  Dissimiarity	
  Indices	
  between	
  Reactors	
  1	
  
and	
  2	
  (n),	
  Reactors	
  1	
  and	
  3	
  (l),	
  or	
  Reactors	
  2	
  and	
  3(r)	
  on	
  each	
  day	
  as	
  a	
  function	
  of	
  
that	
  day	
  in	
  culture.	
  In	
  (B),	
  we	
  calculated	
  ANOSIM	
  and	
  AMOVA	
  between	
  reactors	
  
during	
  early	
  (Days	
  2-­‐6)	
  and	
  late	
  (Days	
  8-­‐20)	
  days	
  in	
  culture	
  based	
  upon	
  θYC	
  (C)	
  or	
  
Jaccard	
  (D)	
  Dissimilarity	
  Indices.	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

74	
  

Discussion	
  
We	
  developed	
  relatively	
  simple,	
  single-­‐stage	
  bioreactors	
  that	
  allow	
  for	
  higher	
  
throughput	
  in	
  vitro	
  studies	
  of	
  fecal	
  microbial	
  communities.	
  We	
  tested	
  four	
  different	
  
media	
  (BRMW,	
  BRMW10,	
  BRMG,	
  and	
  BRMG10)	
  and	
  identified	
  a	
  medium,	
  BRMW10,	
  
that	
  best	
  supported	
  cultivation	
  of	
  microbially-­‐diverse,	
  functionally-­‐stable	
  human	
  
fecal	
  communities.	
  BRMW10	
  communities	
  produced	
  the	
  short	
  chain	
  fatty	
  acids	
  
acetate,	
  butyrate,	
  isobutyrate,	
  latctate,	
  and	
  propionate	
  in	
  highly	
  stable	
  proportions	
  
throughout	
  their	
  time	
  in	
  culture	
  and	
  at	
  levels	
  that	
  were	
  most	
  similar	
  to	
  the	
  starting	
  
human	
  fecal	
  inoculum.	
  Further,	
  16S	
  rRNA	
  gene	
  sequence	
  analysis	
  revealed	
  that	
  
cultures	
  grown	
  in	
  BRMW10	
  were	
  most	
  similar	
  to	
  the	
  fecal	
  inoculum,	
  having	
  the	
  
highest	
  levels	
  of	
  microbial	
  diversity	
  and	
  richness	
  of	
  the	
  media	
  tested,	
  as	
  well	
  as	
  
overall	
  community	
  composition	
  and	
  structure	
  that	
  was	
  as	
  or	
  more	
  similar	
  to	
  the	
  
fecal	
  inoculum	
  than	
  the	
  other	
  media	
  tested.	
  
From	
  our	
  comparisons	
  of	
  microbial	
  community	
  structure	
  between	
  the	
  initial	
  
fecal	
  inoculum	
  and	
  the	
  communities	
  cultured	
  in	
  the	
  BRMW,	
  BRMW10	
  and	
  BRMG	
  
(Fig.	
  2.4),	
  it	
  was	
  clear	
  that	
  the	
  communities	
  we	
  cultivated	
  in	
  all	
  three	
  media	
  were	
  
significantly	
  different	
  from	
  the	
  initial	
  fecal	
  inoculum,	
  with	
  decreases	
  in	
  microbial	
  
diversity	
  and	
  richness	
  and	
  increases	
  in	
  evenness,	
  and	
  that	
  these	
  changes	
  occurred	
  
by	
  day	
  two	
  in	
  culture.	
  Many	
  factors	
  might	
  have	
  contributed	
  to	
  our	
  inability	
  to	
  
cultivate	
  organisms	
  from	
  the	
  original	
  fecal	
  sample.	
  	
  One	
  of	
  these	
  factors	
  is	
  likely	
  the	
  
absence	
  of	
  specific	
  nutrients	
  in	
  our	
  media	
  that	
  are	
  required	
  for	
  growth	
  of	
  these	
  fecal	
  
organisms.	
  Because	
  our	
  assessment	
  of	
  the	
  composition	
  of	
  the	
  initial	
  fecal	
  inoculum	
  
is	
  based	
  upon	
  16S	
  rRNA	
  gene	
  analysis,	
  we	
  also	
  do	
  not	
  know	
  what	
  fraction	
  of	
  the	
  

	
  

75	
  

community	
  we	
  sequenced	
  represents	
  non-­‐viable	
  cells	
  that	
  could	
  not	
  be	
  cultivated	
  
under	
  any	
  condition.	
  Although	
  our	
  protocol	
  for	
  sample	
  collection	
  and	
  processing	
  
was	
  intended	
  to	
  minimize	
  damage	
  to	
  sensitive	
  anaerobic	
  bacteria,	
  it	
  is	
  possible	
  that	
  
immediate	
  cultivation	
  from	
  freshly	
  voided	
  fecal	
  samples	
  could	
  produce	
  better	
  
results.	
  In	
  addition,	
  since	
  our	
  initial	
  fecal	
  inoculum	
  was	
  pooled	
  from	
  twelve	
  distinct	
  
fecal	
  donors,	
  each	
  with	
  a	
  different	
  microbial	
  community	
  (data	
  not	
  shown),	
  it	
  is	
  
possible	
  that	
  competitive	
  forces	
  that	
  shaped	
  each	
  individual	
  microbiota	
  would	
  make	
  
it	
  unlikely	
  that	
  all	
  of	
  these	
  microbes	
  would	
  coexist	
  in	
  a	
  single	
  community	
  under	
  in	
  
vitro	
  or	
  in	
  vivo	
  conditions.	
  Comparing	
  the	
  impact	
  of	
  cultivation	
  on	
  inocula	
  from	
  an	
  
individual	
  donor	
  to	
  pooled	
  communities	
  could	
  provide	
  some	
  insight	
  into	
  this	
  
question.	
  Finally,	
  because	
  our	
  MBRA	
  were	
  designed	
  as	
  continuously-­‐stirred	
  tank	
  
bioreactors	
  that	
  are	
  operated	
  anaerobically,	
  they	
  provide	
  fewer	
  niches	
  than	
  would	
  
be	
  available	
  within	
  the	
  host.	
  The	
  increased	
  evenness	
  of	
  our	
  communities	
  in	
  culture	
  
could	
  indicate	
  that	
  there	
  has	
  been	
  a	
  reduction	
  in	
  available	
  niches	
  that	
  promote	
  
growth	
  in	
  different	
  proportions.	
  It	
  is	
  possible	
  that	
  coexistence	
  of	
  this	
  community	
  
requires	
  further	
  niches	
  to	
  be	
  present,	
  such	
  as	
  those	
  that	
  could	
  be	
  provided	
  by	
  
including	
  beads	
  coated	
  with	
  mucin	
  or	
  glycopeptides	
  typically	
  found	
  in	
  the	
  distal	
  
colon.	
  	
  
Based	
  upon	
  in-­‐depth	
  analysis	
  of	
  the	
  communities	
  cultivated	
  in	
  BRMW10,	
  it	
  
was	
  clear	
  that	
  adaptation	
  to	
  growth	
  in	
  culture	
  caused	
  the	
  biggest	
  change	
  in	
  the	
  
structure	
  of	
  the	
  microbial	
  community,	
  with	
  the	
  largest	
  changes	
  observed	
  between	
  
the	
  initial	
  fecal	
  inoculum	
  and	
  Day	
  2	
  in	
  culture	
  (Fig.	
  2.4).	
  However,	
  it	
  was	
  also	
  
evident	
  that	
  there	
  was	
  continuous	
  change	
  in	
  the	
  community	
  composition	
  and	
  

	
  

76	
  

structure	
  throughout	
  cultivation,	
  with	
  the	
  largest	
  changes	
  observed	
  during	
  the	
  first	
  
week	
  in	
  culture	
  (Fig.	
  2.6).	
  Further,	
  these	
  changes	
  appeared	
  to	
  follow	
  an	
  independent	
  
path	
  in	
  each	
  replicate	
  reactor,	
  resulting	
  in	
  communities	
  in	
  each	
  reactor	
  at	
  the	
  end	
  of	
  
cultivation	
  that	
  were	
  more	
  dissimilar	
  than	
  at	
  the	
  beginning	
  of	
  cultivation	
  (Fig.	
  2.6,	
  
2.7).	
  In	
  spite	
  of	
  these	
  changes	
  in	
  community	
  composition,	
  SCFA	
  production	
  
remained	
  highly	
  stable	
  throughout	
  the	
  time	
  in	
  culture,	
  indicating	
  that	
  in	
  these	
  
communities,	
  functional	
  stability	
  can	
  be	
  obtained	
  in	
  the	
  presence	
  of	
  ongoing	
  
changes	
  in	
  community	
  structure	
  at	
  the	
  species-­‐level.	
  
Other	
  significant	
  changes	
  that	
  were	
  apparent	
  in	
  the	
  BMRW10	
  reactor	
  
communities	
  relative	
  to	
  the	
  fecal	
  inoculum	
  were	
  the	
  increases	
  in	
  the	
  abundance	
  of	
  γ-­‐
Proteobacteria,	
  Fusobacteria	
  and	
  Synergistetes	
  (Fig.	
  2.5).	
  The	
  γ-­‐Proteobacteria	
  
detected	
  were	
  primarily	
  classified	
  with	
  ≥	
  80%	
  confidence	
  as	
  members	
  of	
  the	
  
Enterobacteriaceae	
  family.	
  The	
  Enterobacteriaceae	
  are	
  facultative	
  anaerobes	
  that	
  
have	
  a	
  large	
  number	
  of	
  readily	
  cultivable	
  representatives	
  (e.g.,	
  Escherichia,	
  
Salmonella,	
  Shigella).	
  It	
  is	
  unclear	
  what	
  niche	
  they	
  may	
  be	
  occupying	
  in	
  these	
  
cultures	
  that	
  allow	
  for	
  their	
  expansion.	
  In	
  vivo,	
  high	
  levels	
  of	
  Enterobacteriaceae	
  
correlate	
  with	
  inflammation,	
  colitis,	
  and	
  colorectal	
  cancer	
  (e.g.,	
  (45)).	
  Similarly,	
  we	
  
cannot	
  predict	
  what	
  niche	
  Fusobacteria	
  may	
  be	
  filling	
  within	
  the	
  reactor	
  
communities,	
  but	
  their	
  abundance	
  has	
  also	
  been	
  correlated	
  with	
  colorectal	
  cancer	
  in	
  
vivo,	
  although	
  it	
  is	
  unclear	
  whether	
  these	
  organisms	
  may	
  play	
  a	
  causative	
  role	
  or	
  be	
  
a	
  secondary	
  invader	
  (46,	
  47).	
  All	
  of	
  the	
  Synergistetes	
  isolates	
  that	
  were	
  identified	
  
classified	
  as	
  members	
  of	
  the	
  genus	
  Cloacibacillus.	
  C.	
  evryensis,	
  a	
  cultured	
  
representative	
  of	
  the	
  genus	
  Cloacibacillus	
  was	
  isolated	
  from	
  an	
  anaerobic	
  sludge	
  

	
  

77	
  

digestor	
  of	
  a	
  wastewater	
  treatment	
  plant	
  and	
  characterized	
  by	
  its	
  ability	
  to	
  ferment	
  
amino	
  acids	
  to	
  acetate,	
  butyrate,	
  H2	
  and	
  CO2	
  (48).	
  Subsequently,	
  Synergistetes,	
  
including	
  organisms	
  classified	
  as	
  Cloacibacillus,	
  have	
  been	
  found	
  as	
  naturally	
  
occurring	
  members	
  of	
  the	
  human	
  microflora,	
  but	
  have	
  also	
  been	
  implicated	
  as	
  
opportunistic	
  pathogens,	
  being	
  cultured	
  from	
  infections	
  of	
  the	
  peritoneal	
  fluid,	
  a	
  
sacral	
  pressure	
  ulcer,	
  and	
  a	
  blood	
  culture	
  ((49),	
  and	
  references	
  therein).	
  
The	
  immediate	
  stability	
  that	
  we	
  observed	
  in	
  the	
  SCFA	
  profiles	
  (between	
  the	
  
first	
  two	
  time	
  points	
  tested,	
  Day	
  2	
  and	
  Day	
  4)	
  in	
  the	
  MBRAs	
  differs	
  from	
  other	
  
studies,	
  in	
  which	
  functional	
  stability	
  was	
  not	
  reached	
  until	
  two	
  to	
  three	
  weeks	
  after	
  
inoculation	
  (23,	
  40).	
  	
  We	
  suspect	
  that	
  this	
  difference	
  may	
  be	
  due	
  to	
  the	
  fact	
  that	
  
there	
  are	
  initially	
  large	
  changes	
  seen	
  as	
  the	
  inoculum	
  acclimates	
  to	
  growth	
  in	
  
culture	
  (differences	
  between	
  the	
  inoculum	
  and	
  day	
  2	
  in	
  culture)	
  and	
  that	
  once	
  
established,	
  the	
  simplistic	
  continuous-­‐flow	
  setup	
  of	
  the	
  reactors	
  causes	
  less	
  
perturbation	
  to	
  the	
  functional	
  activity	
  of	
  the	
  communities.	
  	
  	
  	
  	
  	
  	
  	
  
	
  
Comparison	
   of	
   community	
   ecology	
   in	
   BRMW10	
   reactors	
   to	
   previously	
  
published	
   models	
   reveals	
   similar	
   trends	
   among	
   the	
   in	
   vitro	
   models.	
   It	
   is	
  
somewhat	
   difficult	
   to	
   compare	
   the	
   changes	
   observed	
   between	
   the	
   communities	
  
cultured	
  in	
  BRMW10	
  and	
  the	
  fecal	
  inoculum	
  with	
  those	
  previously	
  reported	
  for	
  other	
  
gut	
   models.	
   In	
   the	
   three-­‐stage	
   compound	
   continuous	
   culture	
   system	
   developed	
   by	
  
Macfarlane	
   et	
   al.	
   (1998),	
   community	
   composition	
   is	
   assessed	
   by	
   plating	
   on	
   selective	
  
media,	
  which	
  does	
  not	
  resolve	
  differences	
  in	
  community	
  structure	
  at	
  the	
  genus	
  and	
  
species	
   level.	
   Both	
   Van	
   den	
   Abeele	
   (2010)	
   and	
   Rajilic-­‐Stojanovic	
   (2010)	
   used	
   a	
  

	
  

78	
  

phylogenetic	
   microarray	
   (HITChip)	
   to	
   compare	
   the	
   composition	
   of	
   their	
   reactors	
  
(SHIME	
  and	
  TIM-­‐2,	
  respectively)	
  to	
  their	
  fecal	
  inoculum.	
  HITChip	
  has	
  a	
  much	
  finer	
  
phylogenetic	
  resolution	
  than	
  selective	
  plating,	
  but	
  still	
  produces	
  signals	
  at	
  a	
  variety	
  
of	
   phylogenetic	
   distances	
   (60%	
   of	
   signals	
   at	
   98%	
   identity	
   phylotype	
   level,	
   29%	
   of	
  
signals	
  at	
  genus-­‐level	
  identity,	
  and	
  9%	
  of	
  signals	
  at	
  order-­‐level	
  identity	
  (50)).	
  	
  
Van	
   den	
   Abeele	
   reported	
   the	
   similarity	
   in	
   composition	
   between	
   their	
   twin	
  
SHIME	
  reactors	
  at	
  the	
  Phylum	
  and	
  Group	
  levels	
  (23).	
  They	
  found	
  differences	
  at	
  both	
  
the	
   Phylum	
   and	
   Group	
   levels	
   after	
   19	
   and	
   26	
   days	
   in	
   culture,	
   with	
   a	
   ~1.5-­‐fold	
  
increase	
   in	
   Bacteroidetes	
   and	
   an	
   ~3-­‐fold	
   decrease	
   in	
   Firmicutes.	
   Similarly,	
   we	
  
observed	
  a	
  ~2-­‐fold	
  increase	
  in	
  Bacteroidetes	
  and	
  a	
  ~2-­‐3-­‐fold	
  decrease	
  in	
   Firmicutes	
  
on	
  Day	
  28	
  in	
  culture	
  for	
  BRMW10	
  reactors	
  compared	
  to	
  the	
  fecal	
  inoculum	
  (Fig.	
  2.5).	
  
Rajilic-­‐Stojanovic	
   (2010)	
   reported	
   the	
   overall	
   similarity	
   of	
   their	
   fecal	
   inoculum	
   to	
  
culturing	
  in	
  the	
  TIM-­‐2	
  model	
  for	
  up	
  to	
  72	
  hours	
  and	
  found	
  that	
  the	
  mean	
  similarity	
  
was	
   50%	
   based	
   upon	
   Pearson	
   Product	
   Moment	
   Correlation	
   Coefficients	
   of	
   their	
  
HITChip	
   profiles	
   of	
   cultured	
   populations	
   and	
   the	
   fecal	
   inoculum.	
   When	
   we	
  
calculated	
   the	
   Pearson	
   Product	
   Moment	
   Correlation	
   Coefficient	
   for	
   our	
   data	
   at	
  
different	
  taxonomic	
  levels,	
  we	
  found	
  that	
  it	
  increased	
  at	
  increasing	
  taxonomic	
  levels	
  
(mean	
  r2	
  =	
  29.7%	
  at	
  the	
  Genus-­‐level;	
  mean	
  r2	
  =	
  31.0%	
  at	
  the	
  Family-­‐level;	
  mean	
  r2	
  =	
  
66%	
   at	
   the	
   Order-­‐level).	
   Rajilic-­‐Stojanovic	
   also	
   observed	
   significant	
   increases	
   in	
  
Bacteroidetes	
   in	
   culture,	
   whereas	
   the	
   differences	
   amongst	
   Firmicutes	
   were	
   more	
  
obvious	
   at	
   the	
   Class-­‐level,	
   with	
   increases	
   in	
   Bacilli	
   and	
   decreases	
   in	
   Clostridia.	
   We	
  
did	
   not	
   observe	
   significant	
   increases	
   in	
   Bacilli	
   in	
   reactors	
   cultured	
   in	
   BRMW10.	
  
These	
   differences	
   may	
   be	
   due	
   to	
   differences	
   in	
   media	
   or	
   inoculum	
   preparation	
   or	
  

	
  

79	
  

due	
   to	
   the	
   operation	
   of	
   our	
   reactors	
   in	
   an	
   anaerobic	
   chamber	
   as	
   opposed	
   to	
  
maintaining	
  anaerobicity	
  through	
  streaming	
  of	
  O2-­‐free	
  gas.	
  With	
  our	
  reactor	
  setup,	
  
we	
   found	
   that	
   streaming	
   of	
   O2-­‐free	
   gas	
   was	
   not	
   sufficient	
   to	
   maintain	
   a	
   strict	
  
anaerobic	
   environment	
   due	
   to	
   the	
   multiple	
   points	
   at	
   which	
   O2	
   could	
   diffuse	
   into	
   the	
  
system	
  (data	
  not	
  shown).	
  
	
  
Conclusions.	
   Although	
   the	
   MBRA	
   lack	
   the	
   more	
   complex	
   interactions	
   that	
   are	
  
possible	
   through	
   in	
   vivo	
   models	
   (e.g.,	
   lack	
   of	
   immune	
   interaction,	
  
absorption/excretion	
   of	
   water	
   and	
   SCFA	
   between	
   the	
   host	
   and	
   microbiota),	
   they	
  
represent	
   an	
   excellent	
   opportunity	
   to	
   study	
   microbial	
   interactions	
   in	
   a	
   more	
  
structured,	
   high-­‐throughput	
   environment.	
   When	
   MBRA	
   were	
   operated	
   with	
  
BRMW10	
   medium,	
   we	
   found	
   that	
   they	
   developed	
   dynamic	
   human	
   fecal	
   microbial	
  
communities	
  that	
  produced	
  SCFA	
  profiles	
  very	
  similar	
  to	
  the	
  starting	
  fecal	
  inoculum.	
  
Thus,	
   the	
   BRMW10	
   medium	
   and	
   MBRA	
   represent	
   an	
   exciting	
   avenue	
   for	
   future	
  
studies	
  of	
  human	
  fecal	
  microbiota	
  community	
  dynamics.	
  	
  	
  	
  	
  	
  
	
  
Acknowledgements	
  
The	
  authors	
  would	
  like	
  to	
  acknowledge	
  Byron	
  Smith,	
  Alec	
  Bonifer,	
  and	
  Lilian	
  Jensen	
  
for	
  technical	
  assistance,	
  as	
  well	
  as	
  Terrence	
  Marsh	
  and	
  Thomas	
  Schmidt	
  for	
  
constructive	
  comments	
  on	
  the	
  manuscript.	
  This	
  research	
  was	
  funded	
  as	
  part	
  of	
  the	
  
Michigan	
  State	
  University	
  Enteric	
  Research	
  Investigation	
  Network	
  award	
  from	
  that	
  
National	
  Institutes	
  of	
  Health	
  to	
  Robert	
  Britton	
  (Project	
  Number	
  5U19AI090872-­‐02).	
  
	
  

	
  

80	
  

APPENDIX	
  

	
  

81	
  

	
  
Appendix	
  

Figure	
  S2.1.	
  SCFA	
  concentrations	
  in	
  bioreactor	
  cultures.	
  	
  SCFA	
  concentrations	
  
measured	
  in	
  reactors	
  cultured	
  with	
  BRMW	
  (▲)/	
  BRMW10	
  (◊;	
  A,	
  C,	
  E,	
  G)	
  or	
  BRMG	
  
(▲)/BRMG10	
  (◊;	
  B,	
  D,	
  F,	
  H)	
  media.	
  	
  Similarity	
  indices	
  based	
  on	
  Pearson	
  product-­‐
moment	
  correlation	
  coefficients	
  generated	
  between	
  day	
  X	
  and	
  day	
  X-­‐2	
  (K,	
  L)	
  for	
  
samples	
  taken	
  at	
  2,	
  4,	
  6,	
  8,	
  10,	
  12,	
  14,	
  16,	
  18,	
  20,	
  22,	
  24,	
  26,	
  and	
  28	
  days.	
  	
  

	
  

82	
  

	
  

Pearson Correlation Coefficient!

	
  
	
  
1!

1!

0.5!

0.975!

0!

0.95!
0!

4!

8! 12! 16! 20! 24! 28!

C!

D!

1!

0.7!

0!

4!

8! 12! 16! 20! 24! 28!

0!

4! 8! 12! 16! 20! 24! 28!
Days in Culture!

1!
0.7!

0.4!

0.4!
0!

4!

8! 12! 16! 20! 24! 28!
Days in Culture!

	
  
	
  	
  
Figure	
  S2.2.	
  	
  Pearson	
  Product	
  Moment	
  Correlation	
  Coefficient	
  of	
  the	
  SCFA	
  
profiles.	
  We	
  calculated	
  the	
  Pearson	
  Product	
  Moment	
  Correlation	
  Coefficient	
  of	
  the	
  
SCFA	
  profiles	
  (percent	
  abundance	
  of	
  acetate,	
  butyrate,	
  isobutyrate,	
  lactate,	
  and	
  
propionate)	
  on	
  each	
  day	
  across	
  the	
  three	
  reactors	
  grown	
  in	
  in	
  BRMW(A),	
  BRMW10	
  
(B),	
  BRMG(C),	
  or	
  BRMG10(D)	
  and	
  plotted	
  this	
  is	
  a	
  function	
  of	
  day	
  in	
  culture.	
  Black	
  
symbols	
  represent	
  the	
  comparison	
  between	
  reactors	
  1	
  and	
  2,	
  gray	
  symbols	
  
represent	
  the	
  comparison	
  between	
  reactors	
  2	
  and	
  3,	
  and	
  white	
  symbols	
  represent	
  
the	
  comparison	
  between	
  reactors	
  1	
  and	
  3.	
  Reactor	
  numbering	
  is	
  the	
  same	
  as	
  in	
  
Figure	
  2.3.	
  
	
  
	
  

	
  

83	
  

3&2$$[LV
ï



A"

θYC Dissimilarity"
BRMG!

BRMW!
BRMW10! ● ●
●

● ●

ï

●
●

● ●●
●● ● ●
● ●

ï
ï

3&2$$[LV



B"
BRMW"
BRMW10"

R = 0.34!
P < 0.001 !

BRMG"

R = 0.76!
P < 0.001!

BRMW10"

R = 0.79!
P < 0.001!

	
  	
  
Figure	
  S2.3.	
  	
  Shared	
  community	
  structure	
  of	
  bioreactor	
  cultures.	
  	
  Analysis	
  of	
  
shared	
  community	
  structure	
  of	
  cultures	
  grown	
  in	
  BRMW,	
  BRMW10,	
  and	
  BRMG	
  
evaluated	
  using	
  OTUs	
  with	
  ≤	
  0.03%	
  sequence	
  dissimilarity	
  in	
  the	
  16S	
  rRNA	
  gene.	
  (A)	
  
Principle	
  Coordinates	
  Analysis	
  of	
  the	
  θYC	
  Dissimilarity	
  measures	
  among	
  triplicate	
  
bioreactor	
  communities	
  cultured	
  in	
  BRMW	
  (n),	
  BRMW10	
  (l),	
  or	
  BRMG	
  (r)	
  and	
  
sampled	
  on	
  Day	
  2,	
  4,	
  8,	
  14,	
  20/22,	
  and	
  28	
  (Day	
  20	
  for	
  BRMW	
  and	
  BRMW10;	
  Day	
  22	
  
for	
  BRMG	
  reactors).	
  (B)	
  ANOSIM	
  θYC	
  Dissimilarity	
  (D)	
  of	
  communities	
  plotted	
  in	
  A.	
  
	
  

	
  

84	
  

	
  
	
  
A!
Percent Abundance!

100%!

(!!"#

BRMW!

80%!
60%!
40%!
20%!

0%! 2 4 8 14 20 28 2 4 8 14 20 28 2 4 8 14 20 28!
BRMW10!
100%!

'!"#

&!"#

%!"#

Percent Abundance!

B!

$!"#

Percent Abundance!

!"#

Deltaproteobacteria!
Bacteroidia!

40%!

Unclassified Firmicutes!

20%!

100%!

Synergistia!
Gammaproteobacteria!

60%!

C!

Unclassified Bacteria!
Fusobacteria!

80%!

0%!

Other!

Negativicutes!
2 4 8 14 20 28 2 4 8 14 20 28 2 4 8 14 20 28!
BRMG!

Erysipelotrichia!
Clostridia!
Bacilli!

80%!
60%!
40%!
20%!
0%!

2 4 8 14 22 28 2 4 8 14 22 28 2 4 8 14 22 28!
Reactor 1!
Reactor 2!
Reactor 3!

	
  
Figure	
  S2.4.	
  	
  Class-­‐level	
  community	
  analysis	
  of	
  replicate	
  reactors	
  in	
  different	
  
media.	
  	
  We	
  compared	
  the	
  class-­‐level	
  distribution	
  of	
  community	
  members	
  on	
  Day	
  2,	
  
4,	
  8,	
  14,	
  20	
  or	
  22	
  (BRMW/BRMW10	
  and	
  BRMG,	
  respectively),	
  and	
  28	
  in	
  culture	
  
across	
  triplicate	
  reactors.	
  The	
  percent	
  abundance	
  of	
  each	
  class	
  of	
  Bacteria	
  in	
  the	
  
individual	
  bioreactor	
  samples	
  is	
  shown.	
  Reactor	
  numbers	
  correspond	
  to	
  those	
  given	
  
in	
  Figures	
  2.3,	
  2.4,	
  and	
  2.6.	
  UC	
  Bacteria	
  and	
  UC	
  Firmicutes	
  indicate	
  phylotypes	
  that	
  
could	
  not	
  be	
  classified	
  with	
  greater	
  than	
  80%	
  confidence	
  beyond	
  the	
  Domain	
  
(Bacteria)	
  and	
  Phylum	
  (Firmicutes)	
  levels	
  by	
  the	
  ribosomal	
  database	
  project	
  	
  
	
  

	
  

85	
  

Figure	
  S2.4	
  (cont’d)	
  
classifier	
  release	
  9.	
  Phylotypes	
  classified	
  as	
  “Other”	
  were	
  present	
  in	
  <1%	
  abundance	
  
in	
  any	
  of	
  the	
  samples	
  and	
  include	
  Actinobacteria,	
  Uncalssified	
  Bacteroidetes,	
  	
  
Lentisphaeria,	
  Methanobacteria,	
  α-­‐Proteobacteria,	
  β-­‐Proteobacteria,	
  Unclassified	
  
Proteobacteria,	
  and	
  Verrucomicrobiae.	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

86	
  

REFERENCES	
  

	
  

87	
  

REFERENCES	
  
	
  
1.	
  

Blumberg	
  R,	
  Powrie	
  F.	
  2012.	
  Microbiota,	
  Disease,	
  and	
  Back	
  to	
  Health:	
  A	
  
Metastable	
  Journey.	
  Science	
  Translational	
  Medicine	
  4:137rv7–137rv7.	
  

2.	
  

Hooper	
  LV,	
  Littman	
  DR,	
  Macpherson	
  AJ.	
  2012.	
  Interactions	
  Between	
  the	
  
Microbiota	
  and	
  the	
  Immune	
  System.	
  Science	
  336:1268–1273.	
  

3.	
  

Nicholson	
  JK,	
  Holmes	
  E,	
  Kinross	
  J,	
  Burcelin	
  R,	
  Gibson	
  G,	
  Jia	
  W,	
  Pettersson	
  
S.	
  2012.	
  Host-­‐Gut	
  Microbiota	
  Metabolic	
  Interactions.	
  Science	
  336:1262–1267.	
  

4.	
  

Wu	
   GD,	
  Chen	
   J,	
  Hoffmann	
   C,	
  Bittinger	
   K,	
  Chen	
   YY,	
  Keilbaugh	
   SA,	
  Bewtra	
   M,	
  
Knights	
  D,	
  Walters	
  WA,	
  Knight	
  R,	
  Sinha	
  R,	
  Gilroy	
  E,	
  Gupta	
  K,	
  Baldassano	
  R,	
  
Nessel	
  L,	
  Li	
  H,	
  Bushman	
  FD,	
  Lewis	
  JD.	
  2011.	
  Linking	
  Long-­‐Term	
  Dietary	
  
Patterns	
  with	
  Gut	
  Microbial	
  Enterotypes.	
  Science	
  334:105–108.	
  

5.	
  

Sokol	
  H,	
  Pigneur	
  B,	
  Watterlot	
  L,	
  Lakhdari	
  O,	
  Bermúdez-­‐Humarán	
  LG,	
  
Gratadoux	
  J-­‐J,	
  Blugeon	
  S,	
  Bridonneau	
  C,	
  Furet	
  J-­‐P,	
  Corthier	
  G,	
  Grangette	
  C,	
  
Vasquez	
  N,	
  Pochart	
  P,	
  Trugnan	
  G,	
  Thomas	
  G,	
  Blottiere	
  HM,	
  Doré	
  J,	
  
Marteau	
  P,	
  Seksik	
  P,	
  Langella	
  P.	
  2008.	
  Faecalibacterium	
  prausnitzii	
  is	
  an	
  
anti-­‐inflammatory	
  commensal	
  bacterium	
  identified	
  by	
  gut	
  microbiota	
  analysis	
  
of	
  Crohn	
  disease	
  patients.	
  Proc.	
  Natl.	
  Acad.	
  Sci.	
  U.S.A.	
  105:16731–16736.	
  

6.	
  

Chang	
  JY,	
  Antonopoulos	
  DA,	
  Kalra	
  A,	
  Tonelli	
  A,	
  Khalife	
  WT,	
  Schmidt	
  TM,	
  
Young	
  VB.	
  2008.	
  Decreased	
  Diversity	
  of	
  the	
  Fecal	
  Microbiome	
  in	
  Recurrent	
  
Clostridium	
  difficile–Associated	
  Diarrhea.	
  J	
  Infect	
  Dis	
  197:435–438.	
  

7.	
  

Wang	
  Z,	
  Klipfell	
  E,	
  Bennett	
  BJ,	
  Koeth	
  R,	
  Levison	
  BS,	
  DuGar	
  B,	
  Feldstein	
  AE,	
  
Britt	
  EB,	
  Fu	
  X,	
  Chung	
  Y-­‐M,	
  Wu	
  Y,	
  Schauer	
  P,	
  Smith	
  JD,	
  Allayee	
  H,	
  Tang	
  
WHW,	
  DiDonato	
  JA,	
  Lusis	
  AJ,	
  Hazen	
  SL.	
  2011.	
  Gut	
  flora	
  metabolism	
  of	
  
phosphatidylcholine	
  promotes	
  cardiovascular	
  disease.	
  Nature	
  472:57–63.	
  

8.	
  

Goodman	
  AL,	
  Kallstrom	
  G,	
  Faith	
  JJ,	
  Reyes	
  A,	
  Moore	
  A,	
  Dantas	
  G,	
  Gordon	
  JI.	
  
2011.	
  Extensive	
  personal	
  human	
  gut	
  microbiota	
  culture	
  collections	
  
characterized	
  and	
  manipulated	
  in	
  gnotobiotic	
  mice.	
  Proceedings	
  of	
  the	
  
National	
  Academy	
  of	
  Sciences	
  108:6252–6257.	
  

9.	
  

Bäckhed	
  F,	
  Ding	
  H,	
  Wang	
  T,	
  Hooper	
  LV,	
  Koh	
  GY,	
  Nagy	
  A,	
  Semenkovich	
  CF,	
  
Gordon	
  JI.	
  2004.	
  The	
  gut	
  microbiota	
  as	
  an	
  environmental	
  factor	
  that	
  
regulates	
  fat	
  storage.	
  Proc.	
  Natl.	
  Acad.	
  Sci.	
  U.S.A.	
  101:15718–15723.	
  

10.	
  

Gootenberg	
  DB,	
  Turnbaugh	
  PJ.	
  2011.	
  Companion	
  animals	
  symposium:	
  
humanized	
  animal	
  models	
  of	
  the	
  microbiome.	
  J.	
  Anim.	
  Sci.	
  89:1531–1537.	
  

11.	
  

Allison	
  C,	
  McFarlan	
  C,	
  Macfarlane	
  GT.	
  1989.	
  Studies	
  on	
  mixed	
  populations	
  

	
  

88	
  

of	
  human	
  intestinal	
  bacteria	
  grown	
  in	
  single-­‐stage	
  and	
  multistage	
  continuous	
  
culture	
  systems.	
  Appl	
  Microbiol	
  55:672–678.	
  
12.	
  

Minekus	
  M,	
  Smeets-­‐Peeters	
  M,	
  Bernalier	
  A,	
  Marol-­‐Bonnin	
  S,	
  Havenaar	
  R,	
  
Marteau	
  P,	
  Alric	
  M,	
  Fonty	
  G,	
  Huis	
  in't	
  Veld	
  JH.	
  1999.	
  A	
  computer-­‐controlled	
  
system	
  to	
  simulate	
  conditions	
  of	
  the	
  large	
  intestine	
  with	
  peristaltic	
  mixing,	
  
water	
  absorption	
  and	
  absorption	
  of	
  fermentation	
  products.	
  Appl	
  Microbiol	
  
Biotechnol	
  53:108–114.	
  

13.	
  

Freeman	
  J.	
  2003.	
  Effects	
  of	
  cefotaxime	
  and	
  desacetylcefotaxime	
  upon	
  
Clostridium	
  difficile	
  proliferation	
  and	
  toxin	
  production	
  in	
  a	
  triple-­‐stage	
  
chemostat	
  model	
  of	
  the	
  human	
  gut.	
  Journal	
  of	
  Antimicrobial	
  Chemotherapy	
  
52:96–102.	
  

14.	
  

Molly	
  K,	
  Vande	
  Woestyne	
  M,	
  Verstraete	
  W.	
  1993.	
  Development	
  of	
  a	
  5-­‐step	
  
multi-­‐chamber	
  reactor	
  as	
  a	
  simulation	
  of	
  the	
  human	
  intestinal	
  microbial	
  
ecosystem.	
  Appl	
  Microbiol	
  Biotechnol	
  39:254–258.	
  

15.	
  

Honda	
  H,	
  Gibson	
  GR,	
  Farmer	
  S,	
  Keller	
  D,	
  McCartney	
  AL.	
  2011.	
  Use	
  of	
  a	
  
continuous	
  culture	
  fermentation	
  system	
  to	
  investigate	
  the	
  effect	
  of	
  
GanedenBC30	
  (Bacillus	
  coagulans	
  GBI-­‐30,	
  6086)	
  supplementation	
  on	
  
pathogen	
  survival	
  in	
  the	
  human	
  gut	
  microbiota.	
  Anaerobe	
  17:36–42.	
  

16.	
  

Macfarlane	
  G,	
  Macfarlane	
  S,	
  Gibson	
  G.	
  1998.	
  Validation	
  of	
  a	
  Three-­‐Stage	
  
Compound	
  Continuous	
  Culture	
  System	
  for	
  Investigating	
  the	
  Effect	
  of	
  
Retention	
  Time	
  on	
  the	
  Ecology	
  and	
  Metabolism	
  of	
  Bacteria	
  in	
  the	
  Human	
  
Colon.	
  Microbial	
  Ecology	
  35:180–187.	
  

17.	
  

WANG	
  X,	
  Gibson	
  GR.	
  1993.	
  Effects	
  of	
  the	
  in	
  vitro	
  fermentation	
  of	
  
oligofructose	
  and	
  inulin	
  by	
  bacteria	
  growing	
  in	
  the	
  human	
  large	
  intestine.	
  J.	
  
Appl.	
  Bacteriol.	
  75:373–380.	
  

18.	
  

Kim	
  B-­‐S,	
  Kim	
  JN,	
  Cerniglia	
  CE.	
  2011.	
  In	
  Vitro	
  Culture	
  Conditions	
  for	
  
Maintaining	
  a	
  Complex	
  Population	
  of	
  Human	
  Gastrointestinal	
  Tract	
  
Microbiota.	
  J.	
  Biomed.	
  Biotechnol.	
  2011:1–10.	
  

19.	
  

Gibson	
  GR,	
  Fuller	
  R.	
  2000.	
  Aspects	
  of	
  in	
  vitro	
  and	
  in	
  vivo	
  research	
  
approaches	
  directed	
  toward	
  identifying	
  probiotics	
  and	
  prebiotics	
  for	
  human	
  
use.	
  The	
  Journal	
  of	
  nutrition	
  130:391S–395S.	
  

20.	
  

Macfarlane	
  GT,	
  Macfarlane	
  S.	
  2007.	
  Models	
  for	
  intestinal	
  fermentation:	
  
association	
  between	
  food	
  components,	
  delivery	
  systems,	
  bioavailability	
  and	
  
functional	
  interactions	
  in	
  the	
  gut.	
  Current	
  Opinion	
  in	
  Biotechnology	
  18:156–
162.	
  

21.	
  

Sousa	
  T,	
  Paterson	
  R,	
  Moore	
  V,	
  Carlsson	
  A,	
  Abrahamsson	
  B,	
  Basit	
  AW.	
  
2008.	
  The	
  gastrointestinal	
  microbiota	
  as	
  a	
  site	
  for	
  the	
  biotransformation	
  of	
  

	
  

89	
  

drugs.	
  International	
  Journal	
  of	
  Pharmaceutics	
  363:1–25.	
  
22.	
  

Molly	
  K,	
  Woestyne	
  MV,	
  Smet	
  ID,	
  Verstraete	
  W.	
  1994.	
  Validation	
  of	
  the	
  
simulator	
  of	
  the	
  human	
  intestinal	
  microbial	
  ecosystem	
  (SHIME)	
  reactor	
  using	
  
microorganism-­‐associated	
  activities.	
  Microb	
  Ecol	
  Health	
  Dis	
  7:191–200.	
  

23.	
   Van	
  den	
  Abbeele	
  P,	
  Grootaert	
  C,	
  Marzorati	
  M,	
  Possemiers	
  S,	
  Verstraete	
  W,	
  
Gerard	
  P,	
  Rabot	
  S,	
  Bruneau	
  A,	
  Aidy	
  El	
  S,	
  Derrien	
  M,	
  Zoetendal	
  E,	
  
Kleerebezem	
  M,	
  Smidt	
  H,	
  Van	
  de	
  Wiele	
  T.	
  2010.	
  Microbial	
  Community	
  
Development	
  in	
  a	
  Dynamic	
  Gut	
  Model	
  Is	
  Reproducible,	
  Colon	
  Region	
  Specific,	
  
and	
  Selective	
  for	
  Bacteroidetes	
  and	
  Clostridium	
  Cluster	
  IX.	
  Applied	
  and	
  
Environmental	
  Microbiology	
  76:5237–5246.	
  
24.	
  

Rajilic-­‐Stojanovic	
  M,	
  Maathuis	
  A,	
  Heilig	
  HGHJ,	
  Venema	
  K,	
  de	
  Vos	
  WM,	
  
Smidt	
  H.	
  2010.	
  Evaluating	
  the	
  microbial	
  diversity	
  of	
  an	
  in	
  vitro	
  model	
  of	
  the	
  
human	
  large	
  intestine	
  by	
  phylogenetic	
  microarray	
  analysis.	
  Microbiology	
  
156:3270–3281.	
  

25.	
  

Consortium	
  THMP.	
  2012.	
  Structure,	
  function	
  and	
  diversity	
  of	
  the	
  healthy	
  
human	
  microbiome.	
  Nature	
  486:207–214.	
  

26.	
  

Schloss	
  PD.	
  2009.	
  A	
  High-­‐Throughput	
  DNA	
  Sequence	
  Aligner	
  for	
  Microbial	
  
Ecology	
  Studies.	
  PLoS	
  ONE	
  4:e8230.	
  

27.	
  

Schloss	
  PD,	
  Gevers	
  D,	
  Westcott	
  SL.	
  2011.	
  Reducing	
  the	
  Effects	
  of	
  PCR	
  
Amplification	
  and	
  Sequencing	
  Artifacts	
  on	
  16S	
  rRNA-­‐Based	
  Studies.	
  PLoS	
  ONE	
  
6:e27310.	
  

28.	
  

Edgar	
  RC,	
  Haas	
  BJ,	
  Clemente	
  JC,	
  Quince	
  C,	
  Knight	
  R.	
  2011.	
  UCHIME	
  
improves	
  sensitivity	
  and	
  speed	
  of	
  chimera	
  detection.	
  Bioinformatics	
  
27:2194–2200.	
  

29.	
  

Schloss	
  PD,	
  Westcott	
  SL.	
  2011.	
  Assessing	
  and	
  improving	
  methods	
  used	
  in	
  
operational	
  taxonomic	
  unit-­‐based	
  approaches	
  for	
  16S	
  rRNA	
  gene	
  sequence	
  
analysis.	
  Applied	
  and	
  Environmental	
  Microbiology	
  77:3219–3226.	
  

30.	
  

Yue	
  J,	
  Clayton	
  M.	
  2005.	
  A	
  Similarity	
  Measure	
  Based	
  on	
  Species	
  Proportions.	
  
Communications	
  in	
  Statistics:	
  Theory	
  and	
  Methods	
  34:2123–2131.	
  

31.	
  

Evans	
  J,	
  Sheneman	
  L,	
  Foster	
  J.	
  2006.	
  Relaxed	
  Neighbor	
  Joining:	
  A	
  Fast	
  
Distance-­‐Based	
  Phylogenetic	
  Tree	
  Construction	
  Method.	
  Journal	
  of	
  Molecular	
  
Evolution	
  62:785–792.	
  

32.	
  

Lozupone	
  CA,	
  Hamady	
  M,	
  Kelley	
  ST,	
  Knight	
  R.	
  2007.	
  Quantitative	
  and	
  
qualitative	
  beta	
  diversity	
  measures	
  lead	
  to	
  different	
  insights	
  into	
  factors	
  that	
  
structure	
  microbial	
  communities.	
  Applied	
  and	
  Environmental	
  Microbiology	
  
73:1576–1585.	
  

	
  

90	
  

33.	
  

Stoner	
  DL,	
  Watson	
  SM,	
  Stedtfeld	
  RD,	
  Meakin	
  P,	
  Griffel	
  LK,	
  Tyler	
  TL,	
  
Pegram	
  LM,	
  Barnes	
  JM,	
  Deason	
  VA.	
  2005.	
  Application	
  of	
  Stereolithographic	
  
Custom	
  Models	
  for	
  Studying	
  the	
  Impact	
  of	
  Biofilms	
  and	
  Mineral	
  Precipitation	
  
on	
  Fluid	
  Flow.	
  Applied	
  and	
  Environmental	
  Microbiology	
  71:8721–8728.	
  

34.	
  

2009.	
  Validation	
  of	
  a	
  Three-­‐Stage	
  Compound	
  Continuous	
  Culture	
  System	
  for	
  
Investigating	
  the	
  Effect	
  of	
  Retention	
  Time	
  on	
  the	
  Ecology	
  and	
  Metabolism	
  of	
  
Bacteria	
  in	
  the	
  Human	
  Colon	
  1–8.	
  

35.	
   Schwiertz	
   A,	
  Taras	
   D,	
  Schäfer	
   K,	
  Beijer	
   S,	
  Bos	
   NA,	
  Donus	
   C,	
  Hardt	
   PD.	
  2009.	
  
Microbiota	
  and	
  SCFA	
  in	
  Lean	
  and	
  Overweight	
  Healthy	
  Subjects.	
  Obesity	
  
18:190–195.	
  
36.	
  

Flint	
  HJ,	
  Bayer	
  EA,	
  Rincon	
  MT,	
  Lamed	
  R,	
  White	
  BA.	
  2008.	
  Polysaccharide	
  
utilization	
  by	
  gut	
  bacteria:	
  potential	
  for	
  new	
  insights	
  from	
  genomic	
  analysis.	
  
Nat	
  Rev	
  Micro	
  6:121–131.	
  

37.	
  

Duncan	
  SH,	
  Belenguer	
  A,	
  Holtrop	
  G,	
  Johnstone	
  AM,	
  Flint	
  HJ,	
  Lobley	
  GE.	
  
2007.	
  Reduced	
  Dietary	
  Intake	
  of	
  Carbohydrates	
  by	
  Obese	
  Subjects	
  Results	
  in	
  
Decreased	
  Concentrations	
  of	
  Butyrate	
  and	
  Butyrate-­‐Producing	
  Bacteria	
  in	
  
Feces.	
  Applied	
  and	
  Environmental	
  Microbiology	
  73:1073–1078.	
  

38.	
  

Ahring	
  BK,	
  Sandberg	
  M,	
  Angelidaki	
  I.	
  1995.	
  Volatile	
  fatty	
  acids	
  as	
  indicators	
  
of	
  process	
  imbalance	
  in	
  anaerobic	
  digestors.	
  Appl	
  Microbiol	
  Biotechnol	
  
43:559–565.	
  

39.	
  

Ito	
  Y,	
  Moriwaki	
  H,	
  Muto	
  Y,	
  Kato	
  N,	
  Watanabe	
  K,	
  Ueno	
  K.	
  1997.	
  Effect	
  of	
  
lactulose	
  on	
  short-­‐chain	
  fatty	
  acids	
  and	
  lactate	
  production	
  and	
  on	
  the	
  growth	
  
of	
  faecal	
  flora,	
  with	
  special	
  reference	
  to	
  Clostridium	
  difficile.	
  Journal	
  of	
  
Medical	
  Microbiology	
  46:80–84.	
  

40.	
  

Possemiers	
  S,	
  VerthÃ	
  K,	
  Uyttendaele	
  S,	
  Verstraete	
  W.	
  PCR-­‐DGGE-­‐based	
  
quantification	
  of	
  stability	
  of	
  the	
  microbial	
  community	
  in	
  a	
  simulator	
  of	
  the	
  
human	
  intestinal	
  microbial	
  ecosystem.	
  FEMS	
  Microbiology	
  Ecology	
  49:495–
507.	
  

41.	
  

Kotani	
  A,	
  Miyaguchi	
  Y,	
  Kohama	
  M,	
  Ohtsuka	
  T,	
  Shiratori	
  T,	
  Kusu	
  F.	
  2009.	
  
Determination	
  of	
  short-­‐chain	
  fatty	
  acids	
  in	
  rat	
  and	
  human	
  feces	
  by	
  high-­‐
performance	
  liquid	
  chromatography	
  with	
  electrochemical	
  detection.	
  
Analytical	
  sciences	
  :	
  the	
  international	
  journal	
  of	
  the	
  Japan	
  Society	
  for	
  
Analytical	
  Chemistry	
  25:1007–1011.	
  

42.	
  

CLARKE	
  KR.	
  1993.	
  Non-­‐parametric	
  multivariate	
  analyses	
  of	
  changes	
  in	
  
community	
  structure	
  18:117–143.	
  

43.	
  

Excoffier	
  L,	
  Smouse	
  PE,	
  Quattro	
  JM.	
  1992.	
  Analysis	
  of	
  molecular	
  variance	
  
inferred	
  from	
  metric	
  distances	
  among	
  DNA	
  haplotypes:	
  application	
  to	
  human	
  

	
  

91	
  

mitochondrial	
  DNA	
  restriction	
  data.	
  Genetics	
  131:479–491.	
  
44.	
  

Werner	
  JJ,	
  Knights	
  D,	
  Garcia	
  ML,	
  Scalfone	
  NB,	
  Smith	
  S,	
  Yarasheski	
  K,	
  
Cummings	
  TA,	
  Beers	
  AR,	
  Knight	
  R,	
  Angenent	
  LT.	
  2011.	
  Bacterial	
  
community	
  structures	
  are	
  unique	
  and	
  resilient	
  in	
  full-­‐scale	
  bioenergy	
  systems.	
  
Proceedings	
  of	
  the	
  National	
  Academy	
  of	
  Sciences	
  108:4158–4163.	
  

45.	
  

Arthur	
  JC,	
  Perez-­‐Chanona	
  E,	
  Mühlbauer	
  M,	
  Tomkovich	
  S,	
  Uronis	
  JM,	
  Fan	
  
T-­‐J,	
  Campbell	
  BJ,	
  Abujamel	
  T,	
  Dogan	
  B,	
  Rogers	
  AB,	
  Rhodes	
  JM,	
  Stintzi	
  A,	
  
Simpson	
  KW,	
  Hansen	
  JJ,	
  Keku	
  TO,	
  Fodor	
  AA,	
  Jobin	
  C.	
  2012.	
  Intestinal	
  
inflammation	
  targets	
  cancer-­‐inducing	
  activity	
  of	
  the	
  microbiota.	
  Science	
  
338:120–123.	
  

46.	
  

Kostic	
  AD,	
  Gevers	
  D,	
  Pedamallu	
  CS,	
  Michaud	
  M,	
  Duke	
  F,	
  Earl	
  AM,	
  Ojesina	
  
AI,	
  Jung	
  J,	
  Bass	
  AJ,	
  Tabernero	
  J,	
  Baselga	
  J,	
  Liu	
  C,	
  Shivdasani	
  RA,	
  Ogino	
  S,	
  
Birren	
  BW,	
  Huttenhower	
  C,	
  Garrett	
  WS,	
  Meyerson	
  M.	
  2012.	
  Genomic	
  
analysis	
  identifies	
  association	
  of	
  Fusobacterium	
  with	
  colorectal	
  carcinoma.	
  
Genome	
  Research	
  22:292–298.	
  

47.	
  

Castellarin	
  M,	
  Warren	
  RL,	
  Freeman	
  JD,	
  Dreolini	
  L,	
  Krzywinski	
  M,	
  Strauss	
  
J,	
  Barnes	
  R,	
  Watson	
  P,	
  Allen-­‐Vercoe	
  E,	
  Moore	
  RA,	
  Holt	
  RA.	
  2012.	
  
Fusobacterium	
  nucleatum	
  infection	
  is	
  prevalent	
  in	
  human	
  colorectal	
  
carcinoma.	
  Genome	
  Research	
  22:299–306.	
  

48.	
  

Ganesan	
  A,	
  Chaussonnerie	
  S,	
  Tarrade	
  A,	
  Dauga	
  C,	
  Bouchez	
  T,	
  Pelletier	
  E,	
  
Le	
  Paslier	
  D,	
  Sghir	
  A.	
  2008.	
  Cloacibacillus	
  evryensis	
  gen.	
  nov.,	
  sp.	
  nov.,	
  a	
  
novel	
  asaccharolytic,	
  mesophilic,	
  amino-­‐acid-­‐degrading	
  bacterium	
  within	
  the	
  
phylum	
  “Synergistetes,”	
  isolated	
  from	
  an	
  anaerobic	
  sludge	
  digester.	
  Int.	
  J.	
  Syst.	
  
Evol.	
  Microbiol.	
  58:2003–2012.	
  

49.	
  

Marchandin	
  H,	
  Damay	
  A,	
  RoudiEre	
  L,	
  Teyssier	
  C,	
  Zorgniotti	
  I,	
  Dechaud	
  H,	
  
Jean-­‐Pierre	
  H,	
  Jumas-­‐Bilak	
  E.	
  2010.	
  Phylogeny,	
  diversity	
  and	
  host	
  
specialization	
  in	
  the	
  phylum	
  Synergistetes	
  with	
  emphasis	
  on	
  strains	
  and	
  
clones	
  of	
  human	
  origin.	
  Research	
  in	
  Microbiologoy	
  161:91–100.	
  

50.	
  

Rajilić-­‐Stojanović	
  M,	
  Heilig	
  HGHJ,	
  Molenaar	
  D,	
  Kajander	
  K,	
  Surakka	
  A,	
  
Smidt	
  H,	
  de	
  Vos	
  WM.	
  2009.	
  Development	
  and	
  application	
  of	
  the	
  human	
  
intestinal	
  tract	
  chip,	
  a	
  phylogenetic	
  microarray:	
  analysis	
  of	
  universally	
  
conserved	
  phylotypes	
  in	
  the	
  abundant	
  microbiota	
  of	
  young	
  and	
  elderly	
  adults.	
  
Environ	
  Microbiol	
  11:1736–1751.	
  

	
  	
  

	
  

92	
  

CHAPTER	
  3	
  
	
  
Epidemic	
  Clostridium	
  difficile	
  strains	
  demonstrate	
  increased	
  competitive	
  
fitness	
  over	
  non-­‐epidemic	
  isolates	
  
	
  

	
  

	
  
	
  
	
  
These	
  results	
  have	
  been	
  published	
  in	
  the	
  article:	
  	
  Robinson*,	
  C.D,	
  J.	
  M.	
  Auchtung*,	
  J.	
  
Collins,	
  and	
  R.	
  A.	
  Britton.	
  	
  Epidemic	
  Clostridium	
  difficile	
  strains	
  demonstrate	
  
increased	
  competitive	
  fitness	
  over	
  non-­‐epidemic	
  isolates.	
  	
  	
  Infection	
  and	
  Immunity.	
  	
  
Published	
  ahead	
  of	
  print	
  April	
  14,	
  2014.	
  	
  
*Co-­‐first	
  authors	
  	
  
Copyright	
  ©	
  American	
  Society	
  for	
  Microbiology,	
  Infection	
  and	
  Immunity,	
  82:7,	
  2014,	
  
2815-­‐2825,	
  DOI:	
  10.1128/IAI.01524-­‐14	
  
	
  
	
  
	
  
Development	
  of	
  the	
  in	
  vitro	
  C.	
  difficile	
  infection	
  model	
  was	
  an	
  equal	
  collaboration	
  
with	
  Jennifer	
  M.	
  Auchtung.	
  	
  Jennifer	
  Auchtung	
  conducted	
  the	
  community	
  analysis	
  
and	
  microbial	
  ecology	
  of	
  the	
  bioreactors.	
  	
  James	
  Collins	
  developed	
  the	
  humanized	
  
mouse	
  infection	
  model	
  and	
  facilitated	
  the	
  mouse	
  competition	
  experiments.	
  	
  
	
  
	
  
	
  
	
  
	
  

93	
  

Abstract	
  	
  
Clostridium	
  difficile	
  infection	
  (CDI)	
  is	
  the	
  most	
  common	
  cause	
  of	
  severe	
  cases	
  of	
  
antibiotic	
  associated	
  diarrhea	
  (AAD)	
  and	
  is	
  a	
  significant	
  health	
  burden.	
  Recent	
  
increases	
  in	
  the	
  rate	
  of	
  CDI	
  have	
  paralleled	
  the	
  emergence	
  of	
  a	
  specific	
  phylogenetic	
  
clade	
  of	
  C.	
  difficile	
  strains	
  (ribotype	
  027,	
  North	
  American	
  Pulsed-­‐Field	
  
Electrophoresis	
  (NAP)	
  1,	
  Restriction	
  Endonuclease	
  Analysis	
  (REA)	
  Group	
  BI).	
  Initial	
  
reports	
  indicated	
  that	
  ribotype	
  027	
  strains	
  were	
  associated	
  with	
  increased	
  
morbidity	
  and	
  mortality	
  and	
  may	
  be	
  hypervirulent.	
  Although	
  subsequent	
  work	
  has	
  
cast	
  some	
  doubt	
  as	
  to	
  whether	
  ribotype	
  027	
  strains	
  are	
  hypervirulent,	
  these	
  strains	
  
are	
  considered	
  epidemic	
  isolates	
  that	
  have	
  caused	
  severe	
  outbreaks	
  across	
  the	
  
globe.	
  We	
  hypothesized	
  one	
  factor	
  that	
  could	
  lead	
  to	
  the	
  increased	
  prevalence	
  of	
  
ribotype	
  027	
  strains	
  would	
  be	
  if	
  these	
  strains	
  had	
  increased	
  competitive	
  fitness	
  
compared	
  to	
  strains	
  of	
  other	
  ribotypes.	
  We	
  developed	
  a	
  moderate	
  throughput	
  in	
  
vitro	
  model	
  of	
  C.	
  difficile	
  infection	
  and	
  used	
  it	
  to	
  test	
  competition	
  between	
  four	
  
ribotype	
  027	
  clinical	
  isolates	
  and	
  clinical	
  isolates	
  of	
  four	
  other	
  ribotypes	
  (001,	
  002,	
  
014,	
  and	
  053).	
  We	
  found	
  that	
  ribotype	
  027	
  strains	
  out-­‐competed	
  strains	
  of	
  other	
  
ribotypes.	
  A	
  similar	
  competitive	
  advantage	
  was	
  observed	
  when	
  two	
  ribotype	
  pairs	
  
were	
  competed	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection.	
  Based	
  upon	
  these	
  results	
  
we	
  conclude	
  that	
  one	
  possible	
  mechanism	
  that	
  ribotype	
  027	
  strains	
  have	
  caused	
  
outbreaks	
  worldwide	
  is	
  due	
  to	
  their	
  increased	
  ability	
  to	
  compete	
  in	
  the	
  presence	
  of	
  
a	
  complex	
  microbiota.	
  
	
  
	
  

	
  

94	
  

Introduction	
  
Diarrhea	
  and	
  colitis	
  are	
  some	
  of	
  the	
  most	
  common	
  side	
  effects	
  of	
  antibiotic	
  
treatment	
  (1).The	
  generally	
  accepted	
  paradigm	
  of	
  antibiotic-­‐associated	
  diarrhea	
  
(AAD)	
  is	
  that	
  antibiotics	
  cause	
  a	
  perturbation	
  of	
  the	
  intestinal	
  microbiota,	
  
presenting	
  conditions	
  that	
  allow	
  for	
  the	
  growth	
  of	
  toxigenic	
  bacteria	
  and	
  viruses.	
  It	
  
was	
  not	
  until	
  the	
  late	
  1970’s	
  that	
  toxigenic	
  C.	
  difficile	
  was	
  identified	
  as	
  a	
  common	
  
causative	
  agent	
  of	
  AAD	
  and	
  colitis	
  (2).	
  It	
  is	
  now	
  estimated	
  that	
  30%	
  of	
  antibiotic-­‐
associated	
  diarrhea	
  cases	
  are	
  attributable	
  to	
  C.	
  difficile	
  (3)	
  and	
  that	
  healthy	
  GI	
  
microbial	
  communities	
  play	
  an	
  essential	
  role	
  in	
  providing	
  colonization	
  resistance	
  to	
  
C.	
  difficile	
  infection	
  (1,	
  4,	
  5).	
  	
  
The	
  incidence	
  of	
  C.	
  difficile	
  infection	
  (CDI)	
  has	
  been	
  steadily	
  rising	
  over	
  the	
  
past	
  decade,	
  with	
  CDI	
  recently	
  becoming	
  the	
  most	
  common	
  nosocomial	
  infection	
  in	
  
the	
  United	
  States	
  (2,	
  6).	
  This	
  rise	
  in	
  the	
  rate	
  of	
  CDI	
  has	
  co-­‐occurred	
  with	
  an	
  
increased	
  prevalence	
  of	
  infection	
  caused	
  by	
  a	
  specific	
  phylogenetic	
  clade	
  of	
  strains	
  
characterized	
  as	
  ribotype	
  027	
  (3,	
  7).	
  Initial	
  clinical	
  and	
  epidemiological	
  studies	
  
reported	
  ribotype	
  027	
  strains	
  as	
  being	
  associated	
  with	
  increased	
  rates	
  of	
  morbidity	
  
and	
  mortality,	
  leading	
  to	
  the	
  hypothesis	
  that	
  these	
  strains	
  are	
  hypervirulent	
  (7-­‐13).	
  
However,	
  subsequent	
  analyses	
  comparing	
  the	
  outcomes	
  of	
  endemic	
  CDI	
  caused	
  by	
  
ribotype	
  027	
  strains	
  to	
  CDI	
  caused	
  by	
  strains	
  of	
  other	
  C.	
  difficile	
  ribotypes	
  have	
  
yielded	
  conflicting	
  results.	
  Several	
  studies	
  have	
  found	
  that	
  infection	
  with	
  ribotype	
  
027	
  strains	
  did	
  not	
  result	
  in	
  more	
  severe	
  clinical	
  outcomes	
  across	
  a	
  number	
  of	
  
institutions	
  (14-­‐17);	
  whereas	
  other	
  studies	
  have	
  seen	
  higher	
  mortality	
  caused	
  by	
  
ribotype	
  027	
  strains	
  compared	
  to	
  strains	
  from	
  some	
  of	
  the	
  other	
  C.	
  difficile	
  

	
  

95	
  

ribotypes	
  (18,	
  19).	
  What	
  is	
  clear,	
  however,	
  is	
  that	
  ribotype	
  027	
  strains	
  have	
  been	
  
associated	
  with	
  several	
  large	
  outbreaks,	
  have	
  undergone	
  rapid,	
  global	
  spread	
  since	
  
their	
  emergence,	
  and	
  have	
  become	
  a	
  prevalent	
  ribotype	
  in	
  many	
  hospitals	
  and	
  
regions	
  (20,	
  21),	
  and	
  references	
  therein).	
  Therefore,	
  rather	
  than	
  ribotype	
  027	
  
strains	
  being	
  hypervirulent	
  (capable	
  of	
  causing	
  more	
  severe	
  C.	
  difficile	
  disease),	
  we	
  
hypothesized	
  that	
  ribotype	
  027	
  strains	
  may	
  instead	
  have	
  increased	
  ecological	
  
fitness	
  over	
  strains	
  of	
  other	
  C.	
  difficile	
  ribotypes.	
  	
  
In	
  order	
  to	
  test	
  this	
  hypothesis,	
  we	
  examined	
  competition	
  between	
  several	
  
ribotype	
  027	
  and	
  non-­‐027	
  strains	
  in	
  human	
  fecal	
  bioreactors.	
  Fecal	
  bioreactors	
  have	
  
previously	
  been	
  used	
  to	
  study	
  C.	
  difficile	
  invasion	
  of	
  complex	
  microbial	
  communities	
  
in	
  vitro,	
  as	
  well	
  as	
  the	
  effects	
  of	
  potential	
  antibiotic	
  and	
  probiotic	
  treatments	
  (e.g.,	
  (4,	
  
22-­‐24)).	
  When	
  developing	
  our	
  fecal	
  bioreactors,	
  we	
  modified	
  parameters	
  of	
  design	
  
and	
  operation	
  used	
  in	
  other	
  established	
  models	
  to	
  allow	
  for	
  simpler,	
  higher-­‐
throughput	
  fecal	
  mini-­‐bioreactor	
  arrays	
  (MBRA).	
  Using	
  these	
  fecal	
  MBRA,	
  we	
  
examined	
  competition	
  between	
  four	
  different	
  pairs	
  of	
  ribotype	
  027	
  and	
  non-­‐027	
  
clinical	
  isolates.	
  We	
  found	
  that	
  in	
  all	
  competitions	
  studied,	
  ribotype	
  027	
  strains	
  
demonstrated	
  a	
  clear	
  competitive	
  advantage	
  over	
  non-­‐027	
  strains,	
  often	
  increasing	
  
in	
  abundance	
  more	
  than	
  two	
  to	
  three	
  orders	
  of	
  magnitude	
  by	
  the	
  end	
  of	
  the	
  
experiment.	
  We	
  then	
  performed	
  similar	
  competitions	
  between	
  ribotype	
  027	
  and	
  
non-­‐027	
  strains	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection	
  and	
  saw	
  similar	
  increased	
  
competitive	
  advantage	
  of	
  the	
  ribotype	
  027	
  strains.	
  These	
  results	
  support	
  our	
  
hypothesis	
  that	
  ribotype	
  027	
  strains	
  have	
  become	
  more	
  prevalent	
  due	
  to	
  increased	
  
ecological	
  fitness	
  compared	
  to	
  strains	
  of	
  other	
  C.	
  difficile	
  ribotypes.	
  

	
  

96	
  

Materials	
  And	
  Methods	
  
Mini-­‐bioreactor	
  array	
  (MBRA)	
  design	
  and	
  operation.	
  	
  MBRAs	
  were	
  designed	
  
using	
  CAD	
  software	
  (Argon,	
  Asheller-­‐Vellum),	
  and	
  fabricated	
  with	
  DSM	
  Somos	
  
Watershed	
  XC	
  11122	
  via	
  stereolithography	
  (FineLine	
  prototyping,	
  Fig.	
  3.1).	
  Each	
  
MBRA	
  consisted	
  of	
  six	
  reactors	
  with	
  an	
  internal	
  volume	
  of	
  25	
  ml	
  and	
  a	
  working	
  
volume	
  of	
  15	
  ml.	
  MBRA	
  were	
  operated	
  under	
  an	
  atmosphere	
  of	
  5%	
  CO2/5%	
  
H2/90%	
  N2	
  at	
  37°C	
  in	
  a	
  heated	
  anaerobic	
  chamber.	
  Media	
  was	
  continuously	
  
replenished	
  and	
  waste	
  removed	
  at	
  a	
  flow	
  rate	
  of	
  1.875	
  ml/hr.	
  Prior	
  to	
  use,	
  MBRA	
  
and	
  media	
  were	
  sterilized	
  by	
  autoclaving	
  and	
  allowed	
  to	
  equilibrate	
  to	
  the	
  
anaerobic	
  environment	
  for	
  ≥72	
  hrs.	
  Reactor	
  contents	
  were	
  continuously	
  stirred.	
  
Additional	
  details	
  regarding	
  MBRA	
  design	
  and	
  operation	
  are	
  available	
  in	
  the	
  
Supplementary	
  Methods.	
  
	
  
Strains,	
  media,	
  and	
  growth	
  conditions.	
  	
  All	
  C.	
  difficile	
  strains	
  used	
  in	
  this	
  study	
  are	
  
clinical	
  isolates	
  obtained	
  from	
  the	
  Michigan	
  Department	
  of	
  Community	
  Health	
  
(MDCH,	
  Table	
  3.1).	
  They	
  were	
  collected	
  from	
  Michigan	
  hospitals	
  between	
  December	
  
2007	
  and	
  May	
  2008.	
  MDCH	
  determined	
  strain	
  toxinotype	
  and	
  NAP	
  status.	
  
Ribotyping	
  was	
  determined	
  by	
  Seth	
  Walk	
  (University	
  of	
  Michigan).	
  All	
  growth	
  
studies	
  were	
  carried	
  out	
  in	
  a	
  37°	
  C	
  anaerobic	
  chamber	
  (Coy,	
  Grass	
  Lake,	
  MI)	
  under	
  
5%	
  CO2/5%	
  H2/90%	
  N2	
  atmosphere	
  using	
  pre-­‐equilibrated	
  media.	
  	
  
	
  
	
  
	
  
	
  

	
  

97	
  

Table	
  3.1.	
  Characterization	
  of	
  Strains	
  Used	
  in	
  this	
  Study.	
  
	
  
PFGE	
  Type	
  
Strain	
  
Toxinotype	
  
(NAP	
  status)	
  

Ribotype	
  

CD1014	
  

0	
  

MI-­‐NAP4	
  

014	
  

CD2015	
  

III	
  

MI-­‐NAP1	
  

027	
  

CD2048	
  

0	
  

MI-­‐NAP3	
  

053	
  

CD3014	
  

0	
  

MI-­‐NAP2	
  

001	
  

CD3017	
  

III	
  

MI-­‐NAP1	
  

027	
  

CD4004	
  

0	
  

MI-­‐NAP6	
  

002	
  

CD4010	
  

III	
  

MI-­‐UN13	
  

027	
  

CD4015	
  

III	
  

MI-­‐NAP1	
  

027	
  

	
  
	
  
BHIS	
  and	
  TCCFA	
  were	
  made	
  as	
  previously	
  described	
  (25),	
  except	
  that	
  
cysteine	
  was	
  excluded	
  from	
  both	
  media.	
  One	
  liter	
  of	
  bioreactor	
  medium	
  (BRM)	
  
contained:	
  1	
  g	
  tryptone,	
  2	
  g	
  proteose	
  peptone	
  #3,	
  2	
  g	
  yeast	
  extract,	
  0.1	
  g	
  
arabinogalactan,	
  0.15	
  g	
  maltose,	
  0.15	
  g	
  D-­‐cellobiose,	
  0.4	
  g	
  sodium	
  chloride,	
  5	
  mg	
  
hemin,	
  0.1	
  g	
  magnesium	
  sulfate,	
  0.1	
  g	
  calclium	
  chloride,	
  0.4	
  g	
  potassium	
  phosphate	
  
monobasic,	
  0.4	
  g	
  potassium	
  phosphate	
  dibasic,	
  and	
  2	
  ml	
  tween	
  80;	
  which	
  were	
  
adjusted	
  to	
  pH	
  6.8	
  and	
  autoclaved	
  at	
  121°C	
  for	
  30	
  min.	
  Following	
  autoclaving,	
  a	
  
filter-­‐sterilized	
  mix	
  of	
  1	
  g	
  taurocholic	
  acid,	
  sodium	
  salt,	
  40	
  mg	
  D-­‐glucose,	
  0.2	
  g	
  
inulin,	
  2	
  g	
  sodium	
  bicarbonate,	
  and	
  1	
  mg	
  vitamin	
  K3	
  was	
  added.	
  When	
  needed	
  to	
  
solidify	
  media,	
  Bacto	
  agar	
  was	
  added	
  to	
  a	
  final	
  concentration	
  of	
  1.5	
  %	
  w/v.	
  	
  
	
  
Collection	
  and	
  preparation	
  of	
  fecal	
  samples	
  for	
  fecal	
  MBRA	
  experiments.	
  	
  Fecal	
  
	
  

98	
  

samples	
  were	
  donated	
  by	
  twelve	
  healthy,	
  anonymous	
  donors	
  that	
  were	
  between	
  the	
  
ages	
  of	
  25	
  and	
  64,	
  had	
  not	
  taken	
  antibiotics	
  for	
  at	
  least	
  two	
  months	
  and	
  had	
  not	
  
consumed	
  probiotic	
  products	
  for	
  at	
  least	
  two	
  days	
  prior	
  to	
  donation.	
  Fresh	
  samples	
  
were	
  collected	
  into	
  sterile	
  containers,	
  which	
  were	
  then	
  packed	
  in	
  wet	
  ice	
  in	
  a	
  sealed	
  
(8.1	
  quart,	
  Sterilite	
  Ultraseal)	
  container	
  with	
  two	
  anaerobic	
  gaspaks	
  (BD	
  
biosciences)	
  and	
  transported	
  to	
  the	
  laboratory	
  within	
  24	
  hours.	
  Samples	
  were	
  then	
  
transferred	
  to	
  an	
  anaerobic	
  chamber	
  and	
  manually	
  mixed	
  with	
  sterile	
  equipment.	
  
Aliquots	
  were	
  transferred	
  to	
  sterile	
  cryogenic	
  vials	
  and	
  stored	
  at	
  -­‐80°C	
  until	
  use.	
  
Prior	
  to	
  inoculation,	
  aliquots	
  were	
  resupsended	
  in	
  sterile,	
  anaerobic	
  phosphate-­‐
buffered	
  saline	
  at	
  a	
  concentration	
  of	
  25%	
  w/v.	
  Samples	
  (pooled	
  at	
  equal	
  volumes)	
  
were	
  vortexed	
  vigorously	
  for	
  5	
  min,	
  large	
  particulates	
  were	
  removed	
  by	
  
centrifugation	
  at	
  201	
  X	
  g	
  for	
  5	
  min,	
  and	
  supernatants	
  were	
  used	
  for	
  inoculation	
  of	
  
the	
  reactors.	
  	
  
	
  
C.	
  difficile	
  invasion	
  and	
  competition	
  growth	
  studies	
  in	
  fecal	
  community	
  MBRA.	
  
Reactors	
  were	
  inoculated	
  with	
  4	
  ml	
  of	
  25%	
  fecal	
  slurry	
  inoculum	
  and	
  allowed	
  to	
  
grow	
  in	
  batch	
  culture	
  for	
  16-­‐18	
  hr.	
  After	
  16-­‐18	
  hrs,	
  fresh	
  media	
  flow	
  and	
  waste	
  
removal	
  was	
  initiated.	
  36	
  hrs	
  later,	
  we	
  began	
  dosing	
  clindamycin	
  (500	
  µg/ml	
  final	
  
concentration)	
  or	
  an	
  equivalent	
  volume	
  of	
  water	
  (solvent	
  for	
  clindamycin;	
  both	
  
stored	
  aerobically	
  at	
  4°	
  C	
  until	
  use)	
  twice	
  daily	
  for	
  4	
  days.	
  1	
  ml	
  samples	
  were	
  
removed	
  from	
  the	
  excess	
  fecal	
  slurry	
  sample,	
  and	
  from	
  reactors	
  prior	
  to	
  the	
  
initiation	
  of	
  dosing,	
  and	
  daily	
  thereafter	
  throughout	
  the	
  experiment.	
  Samples	
  were	
  

	
  

99	
  

centrifuged	
  at	
  21000	
  X	
  g	
  for	
  1	
  min,	
  supernatants	
  discarded,	
  and	
  cell	
  pellets	
  were	
  
stored	
  at	
  -­‐80°C	
  until	
  subjected	
  to	
  further	
  analyses.	
  	
  
For	
  invasion	
  studies	
  with	
  CD2015,	
  CD2015	
  was	
  grown	
  in	
  BRM	
  broth	
  batch	
  
culture	
  overnight	
  and	
  reactors	
  were	
  inoculated	
  with	
  either	
  a	
  1:100	
  dilution	
  of	
  the	
  
overnight	
  culture	
  (Fig.	
  3.2)	
  or	
  from	
  dilutions	
  of	
  the	
  exponentially	
  growing	
  
subculture	
  in	
  BRM	
  broth	
  (Fig.	
  3.3;	
  CD2015	
  concentrations	
  specified	
  in	
  figure)	
  on	
  day	
  
7	
  of	
  operation.	
  Prior	
  to	
  inoculation,	
  reactors	
  were	
  tested	
  for	
  C.	
  difficile	
  
contamination	
  by	
  selective	
  plating	
  of	
  an	
  aliquot	
  of	
  each	
  reactors’	
  contents	
  on	
  TCCFA	
  
supplemented	
  with	
  rifampicin	
  (rif,	
  50	
  µg/ml)	
  and	
  erythromycin	
  (erm,	
  20	
  µg/ml)	
  
and	
  by	
  qPCR	
  with	
  C.	
  difficile	
  specific	
  primers	
  (methods	
  described	
  below).	
  Additional	
  
200	
  µl	
  samples	
  were	
  removed	
  from	
  each	
  reactor	
  either	
  15	
  min	
  (Fig.	
  3.3)	
  or	
  3	
  hrs	
  
(Fig.	
  3.2)	
  post-­‐inoculation	
  and	
  CD2015	
  levels	
  were	
  determined	
  by	
  selective	
  plating	
  
on	
  TCCFA	
  rif	
  erm.	
  On	
  subsequent	
  days	
  of	
  MBRA	
  operation,	
  C.	
  difficile	
  levels	
  were	
  
determined	
  via	
  selective	
  plating	
  of	
  an	
  appropriately	
  diluted	
  100	
  µl	
  aliquot	
  from	
  the	
  
1	
  ml	
  daily	
  sample	
  as	
  well	
  as	
  by	
  C.	
  difficile	
  specific	
  qPCR	
  at	
  the	
  times	
  indicated.	
  	
  
For	
  competition	
  studies,	
  bioreactors	
  were	
  set-­‐up	
  following	
  the	
  C.	
  difficile	
  in	
  
vitro	
  invasion	
  model	
  described	
  above	
  with	
  the	
  following	
  modifications.	
  Strains	
  were	
  
inoculated	
  into	
  5	
  ml	
  BRM	
  overnights,	
  which	
  were	
  subcultured	
  into	
  10-­‐30	
  ml	
  BRM	
  
media	
  and	
  allowed	
  to	
  grow	
  at	
  37°	
  C	
  for	
  4	
  hr	
  before	
  inoculation	
  into	
  the	
  reactors	
  to	
  
ensure	
  active	
  growth	
  at	
  the	
  time	
  of	
  inoculation.	
  Subcultures	
  were	
  mixed	
  at	
  various	
  
ratios	
  (1:1,	
  1:2,	
  or	
  1:5	
  027:non-­‐027	
  as	
  indicated)	
  and	
  inoculated	
  into	
  the	
  reactors.	
  
For	
  replicates	
  indicated	
  in	
  Table	
  S1	
  (three	
  replicates	
  of	
  two	
  competition	
  pairs,	
  
CD3017/CD3014	
  and	
  CD4010/CD4004),	
  C.	
  difficile	
  was	
  inoculated	
  into	
  reactors	
  on	
  

	
  

100	
  

day	
  8	
  of	
  operation	
  instead	
  of	
  day	
  7.	
  0.5	
  ml	
  samples	
  were	
  removed	
  from	
  the	
  reactors	
  
immediately	
  prior	
  to	
  C.	
  difficile	
  inoculation	
  and	
  2	
  hr	
  after	
  inoculation	
  and	
  processed	
  
as	
  described	
  above.	
  0.25	
  ml	
  of	
  the	
  aliquots	
  collected	
  prior	
  to	
  C.	
  difficile	
  inoculation	
  
were	
  used	
  in	
  qPCR	
  reactions	
  (described	
  below)	
  to	
  detect	
  possible	
  C.	
  difficile	
  
contamination.	
  MBRA	
  were	
  run	
  for	
  an	
  additional	
  10-­‐12	
  days	
  and	
  sampled	
  daily	
  as	
  
described	
  above.	
  	
  
	
  
Quantitative	
  PCR	
  of	
  tcdA	
  gene	
  to	
  quantify	
  C.	
  difficile	
  invasion.	
  	
  Frozen	
  culture	
  
cell	
  pellets	
  were	
  resuspended	
  in	
  0.5	
  ml	
  sterile	
  water	
  and	
  transferred	
  to	
  2	
  ml	
  screw-­‐
top	
  tubes	
  containing	
  ~200	
  µl	
  0.1	
  mm	
  silica	
  beads	
  (Biospec	
  Products).	
  The	
  samples	
  
were	
  homogenized	
  by	
  bead-­‐beating	
  (BioSpec	
  Products)	
  on	
  the	
  homogenize	
  setting	
  
for	
  1.5	
  min,	
  centrifuged	
  for	
  1	
  min	
  at	
  21000	
  X	
  g,	
  and	
  the	
  supernatant	
  was	
  transferred	
  
to	
  a	
  new	
  tube.	
  When	
  not	
  in	
  use,	
  processed	
  supernatants	
  were	
  stored	
  at	
  -­‐20°C.	
  C.	
  
difficile	
  levels	
  were	
  determined	
  by	
  qPCR	
  with	
  primers	
  specific	
  to	
  the	
  C.	
  difficile	
  Toxin	
  
A	
  gene	
  (tcdA,	
  Table	
  S3.2).	
  
We	
  calibrated	
  the	
  tcdA	
  signal	
  observed	
  in	
  our	
  reactors	
  to	
  a	
  known	
  
concentration	
  of	
  C.	
  difficile	
  cells	
  grown	
  under	
  fecal	
  bioreactor	
  conditions	
  in	
  the	
  
MBRA	
  and	
  enumerated	
  by	
  plate	
  counting.	
  We	
  processed	
  the	
  samples	
  as	
  described	
  
above	
  and	
  spiked	
  them	
  into	
  pooled	
  supernatant	
  samples	
  prepared	
  from	
  bioreactors	
  
prior	
  to	
  C.	
  difficile	
  inoculation.	
  We	
  generated	
  10-­‐fold	
  dilutions	
  of	
  C.	
  difficile	
  in	
  this	
  
background	
  community	
  DNA	
  and	
  used	
  them	
  to	
  generate	
  a	
  standard	
  curve	
  for	
  
determining	
  the	
  absolute	
  amounts	
  of	
  C.	
  difficile	
  in	
  our	
  bioreactor	
  samples.	
  We	
  also	
  
used	
  10-­‐fold	
  dilutions	
  of	
  community	
  DNA	
  alone	
  to	
  generate	
  a	
  standard	
  curve	
  for	
  

	
  

101	
  

assessing	
  the	
  total	
  bacterial	
  signal	
  from	
  our	
  reactors	
  using	
  universal	
  16S	
  primers	
  
(Table	
  S3.2).	
  
Real-­‐time	
  PCR	
  reactions	
  were	
  performed	
  in	
  triplicate	
  and	
  contained	
  the	
  
following	
  components:	
  4	
  µl	
  supernatant	
  (undiluted	
  (tcdA)	
  or	
  1:500	
  dilution	
  in	
  
sterile	
  water	
  (universal	
  16S	
  rRNA)),	
  12.5	
  µL	
  Power	
  SYBR	
  Green	
  PCR	
  Master	
  Mix	
  
(ABI,	
  Carlsbad,	
  CA),	
  0.25	
  µL	
  each	
  primer	
  (5	
  µM)	
  (Table	
  S3.2),	
  and	
  8	
  µL	
  Milli-­‐Q	
  water.	
  
Real-­‐time	
  PCR	
  was	
  performed	
  using	
  an	
  Eppendorf	
  Mastercycler	
  PCR	
  machine	
  under	
  
the	
  following	
  conditions:	
  95°C	
  10	
  min,	
  40	
  cycles	
  of	
  95°C	
  for	
  15	
  sec	
  followed	
  by	
  60°C	
  
for	
  1	
  min.	
  A	
  20	
  min	
  melting	
  curve	
  was	
  also	
  performed	
  from	
  60°C	
  to	
  95°C.	
  We	
  
calculated	
  the	
  tcdA	
  copies/ml	
  from	
  our	
  experimental	
  samples	
  using	
  the	
  CT	
  values	
  
and	
  concentrations	
  from	
  our	
  standard	
  curve	
  described	
  above.	
  If	
  a	
  sample’s	
  CT	
  value	
  
fell	
  below	
  the	
  lowest	
  concentration	
  from	
  our	
  standard	
  curve,	
  it	
  was	
  designated	
  
below	
  the	
  limit	
  of	
  detection,	
  which	
  was	
  1000	
  tcdA	
  copies/ml.	
  We	
  also	
  determined	
  
the	
  total	
  bacterial	
  load	
  per	
  sample	
  based	
  upon	
  CT	
  value	
  with	
  broad-­‐host	
  range	
  16S	
  
rRNA	
  primers	
  (Table	
  S3.2).	
  We	
  used	
  these	
  CT	
  values,	
  which	
  varied	
  by	
  less	
  than	
  3	
  
cycles	
  across	
  all	
  samples	
  (CT=18.75-­‐21.97,	
  Fig.	
  S3.1),	
  to	
  normalize	
  the	
  tcdA	
  copy	
  
numbers	
  that	
  are	
  reported	
  in	
  Fig.	
  3.2.	
  	
  
	
  
Preparation	
  of	
  16S	
  rRNA	
  amplicon	
  sequencing.	
  	
  We	
  extracted	
  DNA	
  from	
  samples	
  
using	
  bead	
  beating	
  followed	
  by	
  modified	
  cleanup	
  with	
  a	
  Qiagen	
  DNEasy	
  Tissue	
  Kit	
  
as	
  described	
  (26).	
  DNA	
  concentrations	
  were	
  determined	
  by	
  spectrophotometry	
  at	
  
260	
  and	
  280	
  nm	
  (Nanodrop).	
  We	
  used	
  40	
  ng	
  of	
  each	
  DNA	
  as	
  template	
  in	
  PCR	
  with	
  
the	
  following	
  final	
  concentrations	
  of	
  reagents:	
  200	
  nM	
  357F	
  primer,	
  200	
  nM	
  926R	
  

	
  

102	
  

primer,	
  1X	
  AccuPrime	
  PCR	
  Buffer	
  II	
  (Invitrogen),	
  0.75	
  U	
  of	
  AccuPrime	
  Taq	
  DNA	
  High	
  
Fidelity	
  (Invitrogen).	
  357F/962R	
  were	
  designed	
  by	
  the	
  Human	
  Microbiome	
  Project,	
  
amplify	
  the	
  V3-­‐V5	
  variable	
  regions	
  of	
  the	
  16S	
  rRNA	
  gene,	
  and	
  contain	
  unique	
  
barcodes	
  that	
  can	
  be	
  used	
  to	
  multiplex	
  sequencing	
  reactions	
  (27).	
  Each	
  reaction	
  was	
  
set	
  up	
  in	
  triplicate	
  and	
  amplified	
  using	
  the	
  following	
  cycle:	
  95°C	
  for	
  2	
  min,	
  followed	
  
by	
  30	
  cycles	
  of	
  95°C	
  for	
  20	
  sec,	
  50°C	
  for	
  30	
  sec,	
  and	
  72°C	
  for	
  5	
  min,	
  with	
  a	
  final	
  
extension	
  at	
  72°C	
  for	
  5	
  min.	
  Successful	
  PCR	
  amplification	
  products	
  from	
  triplicate	
  
reactions	
  were	
  pooled	
  and	
  cleaned	
  with	
  Agencourt	
  AMPure	
  XP	
  (Beckman-­‐Coulter).	
  
Products	
  were	
  resuspended	
  with	
  a	
  0.7X	
  volume	
  of	
  beads,	
  washed	
  twice	
  with	
  70%	
  
ethanol,	
  and	
  eluted	
  with	
  25	
  µl	
  of	
  low	
  EDTA	
  TE	
  Buffer	
  (10	
  mM	
  Tris,	
  0.1	
  mM	
  EDTA).	
  
Concentrations	
  of	
  purified	
  DNA	
  were	
  determined	
  using	
  Quant-­‐IT	
  (Invitrogen)	
  
according	
  to	
  the	
  manufacturer’s	
  protocol	
  and	
  were	
  pooled	
  in	
  equimolar	
  amounts.	
  
Nucleotide	
  sequencing	
  was	
  performed	
  on	
  a	
  454	
  GS	
  Junior	
  (Roche	
  Diagnostics)	
  at	
  
MSU	
  according	
  to	
  the	
  manufacturer’s	
  protocols.	
  	
  
	
  
Processing	
  and	
  analysis	
  of	
  Sequencing	
  Data.	
  	
  All	
  sequence	
  data	
  was	
  processed	
  
using	
  mothur	
  (28)	
  Version	
  1.29.1	
  (January	
  2013).	
  Sequences	
  were	
  initially	
  quality	
  
filtered	
  using	
  the	
  mothur-­‐implementation	
  of	
  PyroNoise	
  to	
  remove	
  low	
  quality	
  
sequences	
  as	
  well	
  as	
  trimmed	
  and	
  filtered	
  to	
  remove	
  those	
  sequences	
  that	
  had	
  any	
  
ambiguous	
  bases,	
  mismatches	
  to	
  the	
  reverse	
  primer	
  or	
  barcode,	
  homopolymeric	
  
stretches	
  longer	
  than	
  8	
  nt,	
  and	
  were	
  shorter	
  than	
  200	
  nt	
  (29).	
  Sequences	
  from	
  
independent	
  sequencing	
  runs	
  were	
  then	
  compiled	
  into	
  a	
  single	
  fasta	
  file	
  and	
  aligned	
  
to	
  the	
  SILVA	
  reference	
  alignment	
  using	
  the	
  NAST-­‐based	
  aligner	
  in	
  mothur,	
  trimmed	
  

	
  

103	
  

to	
  ensure	
  that	
  sequences	
  overlapped,	
  and	
  pre-­‐clustered,	
  allowing	
  a	
  difference	
  
between	
  sequences	
  of	
  2	
  bp	
  or	
  less	
  (29).	
  Potentially	
  chimeric	
  sequences	
  were	
  
removed	
  using	
  the	
  mothur-­‐implementation	
  of	
  UChime	
  (30);	
  remaining	
  sequences	
  
were	
  classified	
  using	
  RDP	
  training	
  set	
  version	
  9	
  (March	
  2012)	
  and	
  mothur’s	
  
implementation	
  of	
  the	
  kmer-­‐based	
  Bayesian	
  classifier.	
  Sequences	
  were	
  binned	
  into	
  
Operational	
  Taxonomic	
  Units	
  (OTUs)	
  with	
  ≤3%	
  sequence	
  dissimilarity	
  using	
  the	
  
average	
  neighbor	
  algorithm	
  of	
  mothur.	
  Taxonomy	
  was	
  assigned	
  to	
  each	
  OTU	
  based	
  
upon	
  the	
  majority	
  sequence	
  consensus	
  within	
  that	
  OTU	
  (31).	
  Number	
  of	
  OTUs,	
  
evenness	
  (Simpson	
  Evenness)	
  and	
  diversity	
  (Inverse	
  Simpson)	
  were	
  calculated	
  
using	
  mothur.	
  Differences	
  in	
  OTU	
  abundances	
  between	
  treated	
  and	
  untreated	
  
bioreactors	
  was	
  determined	
  using	
  the	
  mothur-­‐implementation	
  of	
  metastats	
  (32).	
  
Bray-­‐Curtis	
  dissimilarities	
  were	
  calculated	
  from	
  the	
  OTU	
  distributions	
  of	
  each	
  
sample,	
  which	
  were	
  log10	
  transformed	
  and	
  normalized	
  by	
  dividing	
  the	
  abundance	
  of	
  
each	
  OTU	
  in	
  a	
  sample	
  by	
  the	
  maximum	
  abundance	
  observed	
  for	
  that	
  OTU	
  followed	
  
by	
  normalizing	
  the	
  total	
  abundance	
  of	
  OTUs	
  across	
  each	
  sample	
  to	
  the	
  same	
  number	
  
using	
  the	
  vegan	
  package	
  in	
  R	
  (33).	
  The	
  metaMDS	
  function	
  of	
  vegan	
  was	
  used	
  to	
  
determine	
  the	
  optimal	
  ordination	
  distances	
  for	
  the	
  Bray-­‐Curtis	
  dissimilarities,	
  
which	
  were	
  also	
  plotted	
  in	
  R.	
  The	
  significance	
  of	
  community	
  differences	
  was	
  
determined	
  by	
  Analysis	
  of	
  Similarities	
  (ANOSIM),	
  which	
  was	
  calculated	
  in	
  R.	
  
	
  
Quantitative	
  PCR	
  Analysis	
  of	
  Competition	
  Cultures	
  and	
  Calculations	
  of	
  
Competitive	
  Index.	
  	
  Strain-­‐specific	
  genes	
  thyA	
  (027)	
  and	
  thyX	
  (non-­‐027)	
  were	
  used	
  
to	
  differentiate	
  strains	
  in	
  competitions.	
  We	
  first	
  identified	
  the	
  strain	
  specificity	
  of	
  

	
  

104	
  

these	
  thymidylate	
  synthase	
  genes	
  while	
  doing	
  in	
  silico	
  genomic	
  comparisons	
  of	
  C.	
  
difficile	
  genomes.	
  We	
  then	
  screened	
  a	
  collection	
  of	
  88	
  strains	
  belonging	
  to	
  several	
  
NAP	
  groups,	
  including	
  all	
  of	
  the	
  strains	
  used	
  in	
  this	
  study,	
  for	
  the	
  presence	
  of	
  thyA	
  or	
  
thyX	
  (Fig.	
  S3.2),	
  and	
  verified	
  that	
  thyA	
  was	
  unique	
  to	
  NAP1	
  (ribotype	
  027)	
  strains	
  by	
  
screening.	
  This	
  correlation	
  of	
  thyA	
  with	
  ribotype	
  027	
  strains	
  has	
  been	
  noted	
  earlier.	
  
(34).	
  	
  
Frozen	
  culture	
  cell	
  pellets	
  (from	
  0.5	
  or	
  1	
  ml	
  cells)	
  were	
  washed	
  in	
  the	
  same	
  
volume	
  of	
  sterile	
  water,	
  resuspended	
  in	
  the	
  same	
  volume	
  of	
  sterile	
  water,	
  and	
  
transferred	
  to	
  2	
  ml	
  screw-­‐top	
  tubes	
  containing	
  ~200	
  µL	
  0.1	
  mm	
  silica	
  beads.	
  The	
  
samples	
  were	
  then	
  placed	
  in	
  a	
  BeadBeater	
  cell	
  disruptor	
  (BioSpec	
  Products,	
  
Bartlesville,	
  OK)	
  on	
  the	
  homogenize	
  setting	
  for	
  1	
  min,	
  centrifuged	
  at	
  21000	
  X	
  g	
  for	
  1	
  
min,	
  and	
  diluted	
  1:10	
  in	
  sterile	
  water.	
  Real-­‐time	
  PCR	
  reactions	
  were	
  set-­‐up	
  by	
  
combining	
  the	
  following	
  components:	
  12.5	
  µL	
  Power	
  Sybr	
  Green	
  PCR	
  Master	
  Mix	
  
(ABI,	
  Carlsbad,	
  CA),	
  0.25	
  µL	
  each	
  primer	
  (100	
  µM),	
  11	
  µL	
  Mili-­‐Q	
  water,	
  1	
  µL	
  diluted	
  
culture	
  supernatant.	
  Primers	
  used	
  are	
  described	
  in	
  Table	
  S3.2.	
  PCR	
  was	
  performed	
  
using	
  the	
  conditions	
  described	
  above.	
  All	
  PCR	
  reactions	
  were	
  performed	
  in	
  technical	
  
triplicate	
  and	
  the	
  CT	
  values	
  are	
  an	
  average	
  of	
  the	
  triplicate	
  data	
  points.	
  The	
  
amplification	
  efficiency	
  (E)	
  of	
  each	
  primer	
  set	
  was	
  determined	
  by	
  plotting	
  the	
  CT	
  
values	
  of	
  a	
  standard	
  curve	
  generated	
  by	
  serial	
  4-­‐log	
  dilutions	
  of	
  C.	
  difficile	
  template;	
  
the	
  sample	
  with	
  the	
  highest	
  signal	
  was	
  diluted	
  into	
  sample	
  with	
  no	
  C.	
  difficile	
  
inoculated	
  (community	
  background	
  DNA).	
  Primer	
  efficiencies	
  were	
  calculated	
  using	
  
the	
  method	
  described	
  by	
  Pfaffl	
  et	
  al.;	
  E=	
  10(-­‐1/slope)	
  (35).	
  Competitive	
  Indices	
  (CI)	
  
were	
  calculated	
  by	
  dividing	
  the	
  end	
  point	
  ribotype	
  027:non-­‐027	
  ratio	
  by	
  the	
  ratio	
  at	
  

	
  

105	
  

T0	
  (Ratio=	
  2CT	
  (non-­‐027)-­‐CT	
  (027)).	
  Primer	
  efficiencies	
  were	
  not	
  factored	
  into	
  the	
  CI	
  
calculations;	
  however,	
  they	
  differ	
  by	
  <5%	
  (EthyA=2.04;	
  EthyX=1.95)	
  when	
  calculated	
  
from	
  reactions	
  using	
  sample	
  containing	
  C.	
  difficile	
  diluted	
  into	
  fecal	
  community	
  
background	
  DNA.	
  	
  
	
  
C.	
  difficile	
  competition	
  experiments	
  in	
  humanized	
  microbiota	
  mice	
  (hmmice).	
  	
  
Germ	
  free	
  C57/B6	
  mice	
  were	
  gavaged	
  with	
  fecal	
  slurry	
  pooled	
  from	
  the	
  twelve	
  
human	
  fecal	
  donors	
  described	
  above.	
  Following	
  initial	
  establishment	
  a	
  stable	
  
humanized	
  microbiota	
  was	
  passed	
  from	
  hmmice	
  to	
  their	
  progeny.	
  Descendants	
  of	
  
these	
  original	
  hmmice	
  were	
  maintained	
  under	
  specific	
  pathogen	
  free	
  conditions	
  and	
  
used	
  for	
  all	
  experiments.	
  To	
  induce	
  susceptibility	
  to	
  C.	
  difficile	
  infection	
  (36),	
  an	
  
antibiotic	
  mixture	
  of	
  kanamycin	
  (0.4	
  mg/ml),	
  gentamicin	
  (0.035	
  mg/ml),	
  colistin	
  
(850	
  U/ml),	
  metronidazole	
  (0.215	
  mg/ml),	
  and	
  vancomycin	
  (0.045	
  mg/ml)	
  was	
  
administered	
  in	
  drinking	
  water	
  ad	
  libitum	
  for	
  3	
  days	
  and	
  then	
  replaced	
  with	
  fresh	
  
drinking	
  water.	
  After	
  24	
  hours	
  of	
  plain	
  drinking	
  water,	
  mice	
  were	
  treated	
  
intraperitoneally	
  with	
  clindamycin	
  (10	
  mg/kg)	
  and	
  24	
  hours	
  post	
  injection	
  
challenged	
  with	
  either	
  104	
  pure	
  or	
  mixed	
  027/non-­‐027	
  C.	
  difficile	
  spores.	
  Spores	
  
were	
  cultivated	
  by	
  spread	
  plating	
  overnight	
  BHIS	
  cultures	
  of	
  C.	
  difficile	
  on	
  BHIS	
  
medium	
  and	
  incubating	
  anaerobically	
  at	
  37°C	
  for	
  5	
  days.	
  	
  Cells	
  were	
  scraped	
  from	
  
the	
  plates	
  and	
  resuspended	
  in	
  sterile	
  water,	
  heat-­‐treated	
  at	
  60°C	
  for	
  30	
  min	
  to	
  kill	
  
vegetative	
  cells,	
  and	
  the	
  number	
  of	
  viable	
  spores	
  were	
  enumerated	
  by	
  plating	
  
appropriate	
  serial	
  dilutions	
  on	
  BHIS	
  supplemented	
  with	
  0.1%	
  taurocholic	
  acid.	
  
Spore	
  preparations	
  were	
  diluted	
  in	
  sterile	
  water	
  to	
  yield	
  the	
  desired	
  concentrations	
  

	
  

106	
  

(~104	
  or	
  ~105	
  spores/ml),	
  then	
  mixed,	
  when	
  appropriate,	
  prior	
  to	
  gavaging	
  a	
  total	
  
of	
  ~104	
  spores/mouse.	
  For	
  competition	
  experiments,	
  mice	
  were	
  gavaged	
  with	
  the	
  
following	
  ratios	
  of	
  ribotype	
  027:non-­‐027	
  spores:	
  1:14	
  for	
  CD3017	
  +	
  CD1014	
  and	
  
1:50	
  for	
  CD4015	
  +	
  CD2048.	
  Mice	
  were	
  observed	
  daily	
  for	
  disease	
  symptoms	
  and	
  
morbidity.	
  Fecal	
  samples	
  were	
  collected	
  daily	
  and	
  frozen	
  until	
  analyzed.	
  	
  
C.	
  difficile	
  levels	
  were	
  quantified	
  in	
  fecal	
  samples	
  by	
  plating.	
  Fecal	
  samples	
  
were	
  weighed,	
  diluted	
  in	
  500	
  µl	
  sterile	
  water,	
  and	
  heat-­‐treated	
  at	
  65°C	
  for	
  30	
  min	
  to	
  
reduce	
  background	
  growth	
  of	
  mouse	
  fecal	
  microbiota.	
  Total	
  heat-­‐resistant	
  CFU/g	
  of	
  
feces	
  were	
  determined	
  by	
  spotting	
  appropriate	
  serial	
  dilutions	
  on	
  BHIS	
  plates	
  
supplemented	
  with	
  0.1%	
  taurocholic	
  acid.	
  Ribotype	
  027-­‐specific	
  heat-­‐resistant	
  
CFU/g	
  of	
  feces	
  were	
  determined	
  by	
  spotting	
  appropriate	
  serial	
  dilutions	
  on	
  BHIS	
  
plates	
  supplemented	
  with	
  0.1%	
  taurocholic	
  acid	
  and	
  either	
  50	
  µg/ml	
  rifampicin	
  and	
  
10	
  µg/ml	
  erythromycin	
  (CD4015	
  +	
  CD2048	
  competition	
  samples)	
  or	
  10	
  µg/ml	
  
erythromycin	
  only	
  (CD3017	
  +CD	
  1014	
  competition	
  samples).	
  	
  Plates	
  were	
  incubated	
  
anaerobically	
  for	
  24-­‐48	
  hrs.	
  Colonies	
  formed	
  on	
  antibiotic-­‐supplemented	
  medium	
  
represented	
  levels	
  of	
  ribotype	
  027	
  strains	
  (CD3017	
  and	
  CD4015).	
  Non-­‐027	
  ribotype	
  
strain	
  levels	
  were	
  determined	
  by	
  subtracting	
  the	
  number	
  of	
  colonies	
  formed	
  on	
  
selective	
  plates	
  from	
  the	
  number	
  on	
  non-­‐selective	
  plates	
  (total	
  C.	
  difficile).	
  The	
  fecal	
  
sample	
  weights	
  were	
  then	
  used	
  to	
  determine	
  CFU/g	
  feces.	
  CI’s	
  were	
  calculated	
  by	
  
dividing	
  the	
  ribotype	
  027:non-­‐027	
  strain	
  ratios	
  at	
  day	
  4	
  by	
  the	
  ratios	
  in	
  the	
  gavaged	
  
spore	
  mixtures.	
  In	
  cases	
  where	
  the	
  number	
  of	
  colonies	
  on	
  selective	
  plates	
  were	
  the	
  
same	
  as	
  the	
  number	
  on	
  non-­‐selective	
  plates,	
  the	
  ribotype	
  027:non-­‐027	
  ratio	
  was	
  set	
  
to	
  10.	
  This	
  was	
  determined	
  as	
  a	
  reasonable	
  level	
  for	
  the	
  non-­‐027	
  limit	
  of	
  detection	
  

	
  

107	
  

based	
  on	
  this	
  type	
  of	
  subtractive	
  analysis	
  taking	
  into	
  account	
  plating	
  error	
  and	
  
would	
  most	
  likely	
  result	
  in	
  an	
  underestimate	
  of	
  the	
  ribotype	
  027	
  CI.	
  	
  
Because	
  we	
  used	
  subtractive	
  plating	
  to	
  measure	
  the	
  levels	
  of	
  ribotype	
  
027:non-­‐027	
  spores	
  and	
  our	
  detection	
  limit	
  set	
  the	
  maximum	
  observable	
  ratio	
  
between	
  ribotype	
  027:non-­‐ribotype	
  027	
  strains	
  at	
  10:1,	
  we	
  found	
  that	
  starting	
  with	
  
the	
  ribotype	
  027	
  strain	
  in	
  the	
  minority	
  provided	
  a	
  larger	
  dynamic	
  range	
  that	
  
allowed	
  us	
  to	
  more	
  readily	
  detect	
  an	
  increase	
  in	
  the	
  ribotype	
  027	
  strain	
  levels	
  
compared	
  to	
  non-­‐027	
  levels.	
  Although	
  the	
  ratios	
  of	
  ribotype	
  027:non-­‐ribotype	
  027	
  
spores	
  gavaged	
  were	
  lower	
  than	
  originally	
  intended,	
  placing	
  the	
  ribotype	
  027	
  
strains	
  in	
  the	
  minority	
  ensures	
  that	
  the	
  027	
  strains	
  do	
  not	
  have	
  an	
  advantage	
  by	
  
being	
  even	
  at	
  a	
  slightly	
  higher	
  proportion	
  in	
  the	
  mixed	
  spore	
  preparations,	
  since	
  
there	
  is	
  always	
  a	
  low	
  level	
  of	
  error	
  in	
  diluting	
  and	
  plating	
  
	
  
Results	
  	
  
Fecal	
  mini-­‐bioreactors	
  (MBRA)	
  provide	
  an	
  in	
  vitro	
  model	
  to	
  study	
  C.	
  difficile	
  
invasion	
  in	
  complex	
  microbial	
  communities.	
  	
  Our	
  objective	
  was	
  to	
  design	
  human	
  
fecal	
  bioreactors	
  that	
  recapitulated	
  antibiotic-­‐	
  induced	
  C.	
  difficile	
  invasion	
  of	
  a	
  
resistant	
  community	
  that	
  also	
  allowed	
  for	
  testing	
  of	
  multiple	
  experimental	
  
parameters	
  in	
  replicate	
  reactors	
  simultaneously.	
  Therefore,	
  we	
  pursued	
  a	
  relatively	
  
simple	
  bioreactor	
  design,	
  an	
  array	
  of	
  six	
  single	
  vessel	
  chambers	
  (mini-­‐bioreactor	
  
array,	
  MBRA)	
  with	
  modest	
  operating	
  volume	
  (15	
  ml)	
  that	
  would	
  allow	
  us	
  to	
  operate	
  
up	
  to	
  24	
  continuous-­‐flow	
  fecal	
  bioreactors	
  simultaneously	
  in	
  the	
  same	
  anaerobic	
  
chamber	
  (Fig.	
  3.1).	
  The	
  reactors	
  were	
  fabricated	
  with	
  DSM	
  Somos	
  Watershed	
  XC	
  

	
  

108	
  

11122	
  resin,	
  which	
  allowed	
  for	
  direct	
  observation	
  of	
  the	
  reactor	
  contents	
  and	
  the	
  
ability	
  to	
  autoclave	
  and	
  reuse	
  the	
  reactors.	
  	
  

Influent
Effluent
Sample
port
47 mm
32.5 mm
200 mm
Inner Dimensions:
25 X 25 X 40 mm (25 ml)

	
  
Figure	
  3.1.	
  An	
  example	
  of	
  a	
  minibioreactor	
  array	
  (MBRA)	
  used	
  for	
  cultivation	
  
of	
  fecal	
  microbial	
  communities.	
  The	
  placement	
  of	
  the	
  influent,	
  effluent	
  and	
  sample	
  
port	
  for	
  one	
  of	
  the	
  six	
  bioreactor	
  chambers	
  is	
  indicated	
  as	
  are	
  some	
  of	
  the	
  key	
  
dimensions.	
  
	
  
Continuous-­‐flow	
  MBRAs	
  were	
  inoculated	
  with	
  fecal	
  samples	
  pooled	
  from	
  
twelve	
  healthy,	
  C.	
  difficile-­‐negative	
  donors	
  and	
  bacterial	
  communities	
  were	
  allowed	
  
to	
  adapt	
  to	
  growth	
  in	
  culture	
  before	
  challenging	
  the	
  communities	
  with	
  clindamycin,	
  
an	
  antibiotic	
  known	
  to	
  support	
  C.	
  difficile	
  invasion	
  (37).	
  Initial	
  studies	
  compared	
  the	
  
ability	
  of	
  three	
  clindamycin-­‐treated	
  and	
  three	
  mock-­‐treated	
  communities	
  to	
  resist	
  
invasion	
  by	
  C.	
  difficile.	
  We	
  found	
  that	
  in	
  mock-­‐treated	
  communities	
  challenged	
  with	
  
106	
  vegetative	
  cells	
  of	
  C.	
  difficile,	
  C.	
  difficile	
  levels	
  were	
  reduced	
  to	
  below	
  the	
  level	
  of	
  

	
  

109	
  

detection	
  in	
  replicate	
  reactors	
  within	
  one	
  to	
  four	
  days	
  following	
  inoculation	
  (Fig.	
  
3.2,	
  open	
  symbols).	
  In	
  contrast,	
  when	
  MBRAs	
  were	
  treated	
  twice	
  daily	
  with	
  
clindamycin	
  for	
  four	
  days	
  prior	
  to	
  C.	
  difficile	
  inoculation,	
  C.	
  difficile	
  levels	
  were	
  
maintained	
  at	
  the	
  same	
  high	
  levels	
  at	
  which	
  they	
  were	
  inoculated	
  for	
  eight	
  days	
  
following	
  inoculation	
  (Fig.	
  3.2,	
  closed	
  symbols).	
  	
  
	
  

C.  difficile  CFU/ml

A.  

tcdA  copies/+l

B.  

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  

107
106
105
104
103
102
101

107
106
105
104
103
102
101

7

9
11
13
Days  in  Culture

15

7

9
11
13
Days  in  Culture

15
	
  

Figure	
  3.2.	
  Fecal	
  bioreactor	
  communities	
  prevent	
  invasion	
  by	
  C.	
  difficile	
  unless	
  
disrupted	
  by	
  treatment	
  with	
  clindamycin.	
  We	
  monitored	
  C.	
  difficile	
  proliferation	
  
in	
  three	
  independent	
  fecal	
  bioreactors	
  that	
  were	
  either	
  clindamycin-­‐treated	
  
(reactors	
  1-­‐3;	
  closed	
  squares,	
  circles	
  and	
  diamonds,	
  respectively)	
  or	
  mock-­‐treated	
  
(reactors	
  4-­‐6;	
  open	
  squares,	
  circles	
  and	
  diamonds,	
  respectively)	
  through	
  selective	
  
plating	
  (A)	
  or	
  quantitative	
  PCR	
  (B).	
  The	
  gray	
  dashed	
  line	
  in	
  panel	
  A	
  represents	
  the	
  
theoretical	
  washout	
  rate	
  of	
  non-­‐proliferating	
  C.	
  difficile	
  cells.	
  The	
  black	
  dashed	
  line	
  
represents	
  the	
  limit	
  of	
  detection	
  for	
  selective	
  plating	
  (A)	
  or	
  qPCR	
  (B)	
  in	
  our	
  
experiments.	
  
	
  

	
  

110	
  

Because	
  we	
  used	
  such	
  a	
  high	
  inoculum	
  in	
  this	
  initial	
  experiment,	
  we	
  were	
  
interested	
  in	
  establishing	
  the	
  minimal	
  number	
  of	
  C.	
  difficile	
  cells	
  required	
  for	
  
invasion.	
  We	
  found	
  that	
  inoculation	
  with	
  104	
  or	
  150	
  cells	
  was	
  sufficient	
  to	
  allow	
  
clindamycin-­‐induced	
  invasion	
  by	
  C.	
  difficile,	
  and	
  that	
  the	
  final	
  cfu/ml	
  reached	
  105-­‐
106	
  (Fig.	
  3.3A).	
  These	
  levels	
  are	
  only	
  ~10-­‐100-­‐fold	
  lower	
  than	
  the	
  107	
  cfu/ml	
  that	
  
pure	
  C.	
  difficile	
  reaches	
  in	
  reactors	
  operating	
  under	
  these	
  continuous-­‐culture	
  
conditions	
  (Fig.	
  3.3B).	
  	
  
107

C.  difficile  CFU/ml

A.

106
105
104
103
102

7

9
11
Days  in  Culture

13

2
3
4
5
Days  in  Culture

6

C.  difficile  CFU/ml

B.

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  

108
107
106
105
104

0

1

	
  

Figure	
  3.3.	
  C.	
  difficile	
  proliferation	
  was	
  assayed	
  in	
  fecal	
  bioreactors	
  with	
  
different	
  levels	
  of	
  inoculum	
  and	
  in	
  pure	
  culture	
  under	
  the	
  continuous-­‐culture	
  
conditions	
  used	
  for	
  bioreactors.	
  In	
  (A),	
  we	
  monitored	
  C.	
  difficile	
  proliferation	
  in	
  
four	
  independent	
  fecal	
  bioreactors	
  that	
  were	
  clindamycin-­‐treated	
  and	
  inoculated	
  at	
  
the	
  indicated	
  densities.	
  In	
  (B),	
  we	
  measured	
  C.	
  difficile	
  proliferation	
  in	
  pure	
  culture	
  
in	
  three	
  replicate	
  continuous-­‐culture	
  bioreactors	
  operated	
  under	
  flow	
  conditions	
  
used	
  for	
  fecal	
  bioreactors.	
  	
  
	
  

111	
  

	
  
In	
  addition	
  to	
  the	
  invasion	
  that	
  we	
  observed	
  for	
  the	
  single	
  ribotype	
  027	
  
strain	
  (CD2015)	
  shown	
  in	
  Fig.	
  3.2,	
  we	
  found	
  similar	
  invasion	
  dynamics	
  for	
  other	
  
ribotype	
  027	
  strains	
  (CD3017,	
  CD4010,	
  and	
  CD4015)	
  as	
  well	
  as	
  strains	
  from	
  other	
  
ribotypes	
  (CD1014,	
  CD3014,	
  and	
  CD4004,	
  data	
  not	
  shown).	
  Based	
  upon	
  these	
  
results,	
  we	
  concluded	
  that	
  the	
  communities	
  established	
  in	
  our	
  fecal	
  MBRA	
  
demonstrated	
  the	
  key	
  attribute	
  of	
  an	
  in	
  vitro	
  C.	
  difficile	
  invasion	
  model	
  that	
  we	
  
intended	
  to	
  achieve	
  -­‐	
  the	
  ability	
  to	
  resist	
  invasion	
  by	
  C.	
  difficile	
  until	
  disrupted	
  by	
  
antibiotic	
  treatment.	
  	
  
	
  
MBRAs	
  support	
  complex	
  fecal	
  microbial	
  communities.	
  	
  In	
  order	
  to	
  investigate	
  
whether	
  the	
  resistance	
  to	
  invasion	
  that	
  we	
  observed	
  in	
  our	
  unperturbed	
  MBRA	
  
communities	
  was	
  due	
  to	
  the	
  presence	
  of	
  complex	
  microbial	
  communities	
  or	
  simple	
  
communities	
  composed	
  of	
  a	
  few	
  strains	
  that	
  were	
  inhibitory	
  to	
  C.	
  difficile	
  growth,	
  
we	
  sequenced	
  the	
  V3-­‐V5	
  hypervariable	
  region	
  of	
  the	
  16S	
  rRNA	
  gene	
  from	
  samples	
  
collected	
  from	
  the	
  triplicate	
  clindamycin	
  and	
  mock-­‐treated	
  reactors	
  through	
  
pyrosequencing.	
  Samples	
  were	
  collected	
  prior	
  to	
  antibiotic	
  treatment	
  (day	
  2)	
  and	
  
every	
  two	
  days	
  after	
  the	
  initiation	
  of	
  treatment	
  (days	
  4,	
  6,	
  8,	
  10	
  and	
  12;	
  C.	
  difficile	
  
was	
  added	
  to	
  all	
  reactors	
  on	
  day	
  7)	
  and	
  sequencing	
  data	
  from	
  these	
  samples	
  were	
  
compared	
  to	
  duplicate	
  samples	
  from	
  the	
  initial	
  fecal	
  inoculum.	
  	
  
At	
  the	
  sequence	
  depth	
  examined	
  (1053	
  sequences/sample),	
  we	
  detected	
  a	
  
mean	
  of	
  69	
  operational	
  taxonomic	
  units	
  (OTUs;	
  97%	
  sequence	
  similarity)	
  in	
  the	
  
untreated	
  (day	
  2,	
  all	
  reactors)	
  and	
  mock-­‐treated	
  (days	
  4,	
  6,	
  8,	
  10	
  and	
  12)	
  bioreactor	
  

	
  

112	
  

communities	
  (Fig.	
  3.4A).	
  The	
  number	
  of	
  OTUs	
  observed	
  in	
  untreated	
  or	
  mock-­‐
treated	
  reactors	
  was	
  ~2.4-­‐fold	
  lower	
  than	
  that	
  observed	
  in	
  the	
  original	
  fecal	
  
inoculum	
  (mean=168	
  OTUs,	
  Fig.	
  3.4A).	
  

#  of  OTUs  (3%  ID)

A.   200
150
100

Relative  Abundance

B.  

50
0

0

2

4
6
8
Days  in  Culture

10

12

1.0

0.5

0.0

	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  

1          2          1          2          3          4          5          6
Fecal
Day  2  Bioreactors
Inoculum
Actinobacteria
Firmicutes
Bacteriodes
Verrucomicrobia
Proteobacteria
Synergistetes
Fusobacteria
Unclassified  Bacteria 	
  

Figure	
  3.4.	
  Comparison	
  of	
  the	
  community	
  structure	
  between	
  fecal	
  samples,	
  
mock-­‐treated	
  and	
  clindamycin-­‐treated	
  reactors.	
  We	
  analyzed	
  the	
  16S	
  rRNA	
  gene	
  
abundances	
  from	
  the	
  three	
  mock-­‐treated	
  and	
  clindamycin-­‐treated	
  communities	
  
described	
  in	
  Fig.	
  3.2	
  on	
  days	
  2,	
  4,	
  6,	
  8,	
  10	
  and	
  12	
  in	
  MBRAs	
  as	
  well	
  as	
  duplicate	
  
samples	
  from	
  the	
  initial	
  fecal	
  inoculum.	
  In	
  (A),	
  we	
  plotted	
  the	
  mean	
  number	
  of	
  OTUs	
  
in	
  the	
  fecal	
  inoculum	
  (asterisks),	
  and	
  clindamycin	
  (closed	
  circles)	
  and	
  mock-­‐treated	
  
(open	
  squares)	
  communities	
  ±	
  the	
  standard	
  deviations.	
  In	
  (B),	
  we	
  classified	
  each	
  
sequence	
  to	
  the	
  phylum	
  level	
  with	
  at	
  least	
  80%	
  confidence	
  (sequences	
  <80%	
  were	
  
designated	
  “Unclassified	
  Bacteria”).	
  We	
  then	
  plotted	
  the	
  relative	
  abundance	
  of	
  each	
  
phylum	
  in	
  the	
  duplicate	
  samples	
  of	
  initial	
  fecal	
  inoculum	
  and	
  the	
  six	
  replicate	
  
bioreactor	
  samples	
  from	
  day	
  2	
  prior	
  to	
  the	
  initiation	
  of	
  treatment.	
  Reactors	
  are	
  
numbered	
  as	
  in	
  Fig	
  3.2.	
  	
  
	
  

	
  

113	
  

When	
  we	
  compared	
  the	
  composition	
  of	
  the	
  original	
  fecal	
  inoculum	
  
community	
  to	
  the	
  bioreactor	
  communities,	
  we	
  found	
  a	
  significant	
  shift	
  in	
  
composition	
  upon	
  culturing	
  in	
  the	
  bioreactors,	
  even	
  by	
  day	
  2	
  (Fig.	
  3.4B).	
  The	
  fecal	
  
inoculum	
  was	
  dominated	
  by	
  members	
  of	
  the	
  Firmicutes	
  phylum,	
  which	
  comprised	
  
74%	
  ±	
  3%	
  of	
  the	
  sequences.	
  In	
  contrast,	
  members	
  of	
  the	
  Bacteroides	
  phylum	
  were	
  
dominant	
  members	
  of	
  the	
  bioreactor	
  communities	
  on	
  day	
  2	
  in	
  culture	
  (Fig.	
  3.4B),	
  
comprising	
  67%	
  ±	
  3%	
  of	
  the	
  sequences	
  in	
  all	
  six	
  replicate	
  reactors	
  studied.	
  	
  
	
  

	
  

Bioreactor	
  community	
  composition	
  changes	
  in	
  response	
  to	
  clindamycin	
  
treatment.	
  	
  Because	
  clindamycin-­‐treated	
  bioreactor	
  communities	
  become	
  
susceptible	
  to	
  C.	
  difficile	
  invasion	
  (Fig.	
  3.2),	
  we	
  anticipated	
  that	
  we	
  would	
  observe	
  
changes	
  in	
  the	
  microbial	
  composition	
  of	
  these	
  communities	
  compared	
  to	
  the	
  mock-­‐
treated	
  communities,	
  and	
  that	
  these	
  changes	
  would	
  be	
  consistent	
  with	
  previously	
  
published	
  models	
  of	
  C.	
  difficile	
  invasion.	
  One	
  significant	
  change	
  we	
  observed	
  in	
  our	
  
clindamycin-­‐treated	
  communities	
  was	
  a	
  significant	
  reduction	
  in	
  the	
  number	
  of	
  OTUs	
  
compared	
  to	
  mock-­‐treated	
  reactors	
  (Fig.	
  3.4A,	
  p<0.01	
  for	
  days	
  4-­‐12	
  with	
  student’s	
  t-­‐
test).	
  The	
  ability	
  of	
  antibiotic	
  treatment	
  to	
  significantly	
  reduce	
  species	
  complexity	
  
has	
  been	
  previously	
  reported	
  (26,	
  38).	
  Although	
  the	
  species	
  richness	
  declined,	
  
quantitative	
  PCR	
  with	
  broad-­‐range	
  16S	
  rRNA	
  gene	
  primers	
  indicated	
  the	
  total	
  
amount	
  of	
  bacteria	
  in	
  the	
  reactors	
  was	
  equivalent	
  to	
  untreated	
  reactors	
  after	
  
clindamycin	
  treatment	
  (Fig.	
  S3.1),	
  indicating	
  that	
  C.	
  difficile	
  invasion	
  was	
  not	
  
dependent	
  upon	
  a	
  decreased	
  bacterial	
  load	
  in	
  the	
  bioreactors.	
  	
  

	
  

114	
  

This	
  change	
  in	
  microbial	
  composition	
  was	
  also	
  evident	
  when	
  comparing	
  the	
  
composition	
  of	
  the	
  communities	
  using	
  the	
  Bray-­‐Curtis	
  dissimilarity	
  measure,	
  which	
  
compares	
  the	
  relative	
  abundances	
  of	
  shared	
  OTUs	
  between	
  communities.	
  When	
  we	
  
plotted	
  these	
  data	
  using	
  nonmetric	
  multi-­‐dimensional	
  scaling	
  (NMDS,	
  Fig.	
  3.5),	
  we	
  
found	
  that	
  all	
  six	
  bioreactor	
  communities	
  were	
  highly	
  similar	
  prior	
  to	
  treatment	
  on	
  
day	
  2	
  and	
  that	
  bioreactor	
  communities	
  diverged	
  in	
  response	
  to	
  clindamycin-­‐
treatment	
  as	
  well	
  as	
  time	
  in	
  culture.	
  Analysis	
  of	
  similarities	
  (ANOSIM,	
  (39))	
  found	
  
strong	
  statistical	
  support	
  (p<0.05)	
  for	
  the	
  distinct	
  partitioning	
  of	
  the	
  communities	
  
into	
  pre-­‐treatment,	
  mock-­‐treatment	
  and	
  clindamycin-­‐treatment	
  groups,	
  with	
  
clindamycin-­‐treatment	
  causing	
  a	
  more	
  significant	
  shift	
  in	
  community	
  structure	
  than	
  
time	
  in	
  culture.	
  Using	
  metastats	
  (32),	
  we	
  identified	
  several	
  specific	
  OTUs	
  that	
  were	
  
significantly	
  different	
  between	
  treated	
  and	
  untreated	
  communities.	
  We	
  observed	
  
decreases	
  in	
  specific	
  members	
  of	
  Ruminococcaceae,	
  Lachnospiraceae,	
  and	
  
Clostridiaceae	
  families	
  that	
  were	
  consistent	
  with	
  changes	
  observed	
  in	
  previous	
  
animal	
  and	
  human	
  studies	
  (40,	
  41),	
  providing	
  further	
  support	
  for	
  the	
  relevance	
  of	
  
this	
  model	
  for	
  studying	
  aspects	
  of	
  C.	
  difficile	
  invasion	
  in	
  vitro.	
  
	
  
	
  
	
  
	
  

	
  

115	
  

	
  

Figure	
  3.5.	
  Community	
  structure	
  changes	
  in	
  response	
  to	
  clindamycin-­‐
treatment.	
  We	
  plotted	
  the	
  Bray-­‐Curtis	
  dissimilarity	
  (3%	
  OTUs)	
  between	
  samples	
  
using	
  nonmetric	
  multidimensional	
  scaling	
  (NMDS).	
  Samples	
  were	
  plotted	
  from	
  
three	
  clindamycin-­‐treated	
  replicates	
  (1-­‐3;	
  closed	
  symbols)	
  and	
  three	
  mock-­‐treated	
  
replicates	
  (4-­‐6;	
  open	
  symbols)	
  every	
  two	
  days	
  from	
  day	
  2	
  (pre-­‐treatment)	
  through	
  
day	
  12.	
  Data	
  was	
  normalized	
  prior	
  to	
  analysis	
  as	
  described	
  in	
  the	
  methods.	
  The	
  
numbers	
  in	
  boxes	
  indicate	
  the	
  points	
  for	
  the	
  indicated	
  reactors	
  on	
  day	
  2	
  and	
  day	
  12	
  
and	
  correspond	
  to	
  the	
  reactor	
  numbers	
  indicated	
  in	
  Fig.	
  3.4.	
  Intervening	
  time	
  points	
  
are	
  represented	
  by	
  symbols	
  and	
  are	
  connected	
  in	
  sequential	
  order	
  by	
  lines.	
  The	
  
ellipses	
  indicate	
  the	
  95%	
  confidence	
  intervals	
  for	
  the	
  indicated	
  groups	
  (pre-­‐
treatment;	
  mock-­‐treatment;	
  clindamycin-­‐treatment.)	
  Distinct	
  distributions	
  between	
  
clindamycin	
  and	
  early/mock-­‐treated	
  samples	
  were	
  also	
  supported	
  by	
  ANOSIM,	
  with	
  
p-­‐values	
  less	
  than	
  0.001.	
  The	
  plot	
  stress	
  was	
  0.171.	
  
	
  
	
  	
  
Ribotype	
  027	
  strains	
  exhibit	
  a	
  competitive	
  advantage	
  over	
  non-­‐027	
  strains	
  in	
  
the	
  presence	
  of	
  a	
  complex	
  microbiota.	
  	
  Having	
  developed	
  the	
  MBRA	
  C.	
  difficile	
  
invasion	
  model,	
  we	
  investigated	
  if	
  ribotype	
  027	
  strains	
  were	
  able	
  to	
  better	
  compete	
  
than	
  non-­‐027	
  ribotype	
  strains	
  for	
  their	
  available	
  niche	
  after	
  antibiotic	
  treatment	
  in	
  
the	
  presence	
  of	
  the	
  complex	
  MBRA	
  communities.	
  We	
  chose	
  to	
  study	
  recent	
  clinical	
  
isolates	
  of	
  ribotype	
  027	
  and	
  non-­‐027	
  C.	
  difficile	
  strains	
  collected	
  by	
  the	
  Michigan	
  
	
  

116	
  

Department	
  of	
  Community	
  Health	
  in	
  order	
  to	
  avoid	
  confounding	
  effects	
  of	
  strain	
  
adaptation	
  to	
  laboratory	
  conditions.	
  Eighty-­‐eight	
  isolates	
  were	
  characterized	
  by	
  
North	
  American	
  Pulsed-­‐Field	
  Electrophoresis	
  (NAP)	
  fingerprint,	
  toxinotype,	
  and	
  
ribotype;	
  we	
  selected	
  eight	
  strains	
  for	
  further	
  study	
  (Table	
  3.1).	
  Four	
  different	
  
ribotype	
  027	
  strains	
  were	
  competed	
  against	
  four	
  different	
  non-­‐027	
  ribotype	
  strains	
  
in	
  order	
  to	
  avoid	
  selecting	
  strains	
  with	
  unrepresentatively	
  high	
  or	
  low	
  competitive	
  
fitness	
  for	
  either	
  ribotype	
  group.	
  For	
  the	
  non-­‐027	
  ribotypes,	
  we	
  selected	
  strains	
  that	
  
were	
  different	
  ribotypes	
  and	
  had	
  different	
  NAP	
  designations	
  to	
  broaden	
  the	
  
phylogenetic	
  breadth	
  of	
  strains	
  tested.	
  
	
  

Exponentially	
  growing	
  pure	
  cultures	
  of	
  ribotype	
  027	
  and	
  non-­‐027	
  C.	
  difficile	
  

strains	
  were	
  mixed	
  together	
  and	
  inoculated	
  into	
  clindamycin-­‐treated	
  MBRAs.	
  At	
  
days	
  3,	
  7,	
  and	
  11	
  post-­‐inoculation	
  samples	
  were	
  taken	
  and	
  quantitative	
  PCR	
  was	
  
conducted	
  to	
  determine	
  the	
  relative	
  ratios	
  of	
  the	
  competing	
  strains.	
  Plotted	
  in	
  Fig.	
  
3.6	
  are	
  the	
  competitive	
  indices	
  (CI)	
  of	
  ribotype	
  027	
  strains	
  at	
  day	
  7	
  for	
  all	
  of	
  the	
  
replicates	
  in	
  each	
  competition	
  pair,	
  calculated	
  as	
  the	
  ratio	
  of	
  027:non-­‐027	
  at	
  day	
  7	
  
divided	
  by	
  the	
  ratio	
  at	
  day	
  0.	
  The	
  mean	
  competitive	
  indices	
  (range)	
  for	
  these	
  
competitions	
  are	
  7.8	
  (0.5	
  to	
  22.4)	
  for	
  CD2015,	
  832.4	
  (21.9	
  to	
  3993.2)	
  for	
  CD3017,	
  
131.4	
  (1.4	
  to	
  593.4)	
  for	
  CD4015,	
  and	
  327.9	
  (30.3	
  to	
  764.9)	
  for	
  CD4010.	
  Fig.	
  S3.3	
  
shows	
  the	
  027:non-­‐027	
  ratios	
  plotted	
  across	
  time	
  for	
  each	
  individual	
  reactor	
  of	
  all	
  
competition	
  pairs.	
  The	
  competition	
  dynamics	
  vary	
  between	
  replicates	
  both	
  within	
  
and	
  across	
  competition	
  pairs.	
  However,	
  there	
  is	
  a	
  strong	
  trend	
  of	
  increasing	
  ratios	
  
over	
  time	
  for	
  the	
  ribotype	
  027	
  strains,	
  even	
  when	
  started	
  at	
  different	
  initial	
  input	
  
ratios,	
  further	
  supporting	
  their	
  competitive	
  advantage.	
  The	
  competitive	
  indices	
  

	
  

117	
  

calculated	
  from	
  these	
  ratios	
  across	
  all	
  time	
  points	
  (days	
  3,	
  7,	
  and	
  11;	
  some	
  days	
  
vary,	
  see	
  figure	
  legend)	
  are	
  reported	
  in	
  Table	
  S3.1.	
  Across	
  all	
  22	
  competition	
  
replicates,	
  only	
  two	
  CIs	
  were	
  <1.0	
  at	
  day	
  7;	
  these	
  account	
  for	
  two	
  of	
  the	
  six	
  
replicates	
  of	
  the	
  CD2015	
  (ribotype	
  027)	
  +	
  CD3014	
  (ribotype	
  001)	
  competition	
  pair.	
  
This	
  competition	
  pair	
  was	
  particularly	
  interesting	
  in	
  that	
  this	
  ribotype	
  027	
  strain	
  
(CD2015)	
  displayed	
  an	
  initial	
  drop	
  in	
  strain	
  ratio	
  in	
  the	
  majority	
  of	
  the	
  competition	
  
replicates,	
  sometimes	
  emerging	
  as	
  low	
  as	
  2%	
  of	
  the	
  total	
  C.	
  difficile	
  population	
  at	
  
day	
  3	
  (Fig.	
  S3.3).	
  Nevertheless,	
  CD2015	
  was	
  able	
  to	
  recover	
  and	
  eventually	
  
outcompete	
  the	
  non-­‐027	
  ribotype	
  strain	
  by	
  the	
  end	
  of	
  the	
  competitions.	
  At	
  the	
  day	
  
11	
  time	
  point	
  of	
  these	
  replicates,	
  the	
  ratios	
  continued	
  to	
  increase	
  resulting	
  in	
  CI’s	
  
close	
  to	
  or	
  >1	
  (Table	
  S3.1).	
  	
  	
  
	
  
	
  
	
  

	
  

118	
  

Competitive Index

104
103
102
101
100
10-1

CD2015
[CD3014(001)]

CD3017
[CD1014(014)]

CD4015
[CD2048(053)]

CD4010
[CD4004(002)]

027 Strain
[competitor strain (ribotype)]
	
  
Figure	
  3.6.	
  	
  Competitive	
  indices	
  of	
  ribotype	
  027	
  strains	
  relative	
  to	
  non-­‐027	
  
strains	
  in	
  the	
  presence	
  of	
  MBRA	
  fecal	
  communities.	
  Clindamycin-­‐treated	
  
bioreactors	
  were	
  inoculated	
  with	
  the	
  indicated	
  mixtures	
  of	
  strains	
  at	
  various	
  ratios	
  
and	
  the	
  abundance	
  of	
  each	
  strain	
  was	
  monitored	
  over	
  time	
  by	
  qPCR	
  to	
  measure	
  
ribotype	
  027:non-­‐027	
  ratios.	
  Plotted	
  here	
  is	
  the	
  competitive	
  index	
  of	
  the	
  ribotype	
  
027	
  strains	
  for	
  each	
  replicate	
  competition,	
  calculated	
  as	
  the	
  ratio	
  of	
  027:non-­‐027	
  at	
  
day	
  7	
  or	
  8	
  (see	
  Table	
  S1)	
  divided	
  by	
  the	
  ratio	
  at	
  day	
  0.	
  	
  Each	
  black	
  circle	
  represents	
  
an	
  individual	
  replicate	
  competition.	
  Where	
  the	
  non-­‐027	
  strain	
  was	
  below	
  the	
  limit	
  
of	
  detection,	
  the	
  ratio	
  was	
  determined	
  by	
  substituting	
  in	
  the	
  highest	
  CT	
  value	
  in	
  the	
  
linear	
  range	
  (detection	
  limit)	
  for	
  the	
  non-­‐027	
  value.	
  Black	
  bars	
  represent	
  the	
  mean	
  
of	
  replicates	
  for	
  each	
  competition.	
  	
  
	
  
	
  
One	
  possible	
  mechanism	
  for	
  ribotype	
  027	
  strains	
  to	
  out-­‐compete	
  the	
  non-­‐
027	
  strains	
  would	
  be	
  if	
  the	
  latter	
  were	
  inherently	
  unable	
  to	
  invade	
  the	
  complex	
  
MBRA	
  communities.	
  We	
  tested	
  this	
  hypothesis	
  by	
  inoculating	
  fecal	
  bioreactors	
  with	
  
individual	
  non-­‐027	
  strains	
  and	
  saw	
  that	
  these	
  strains	
  were	
  equally	
  able	
  to	
  invade	
  
the	
  microbiota	
  as	
  ribotype	
  027	
  strains	
  (data	
  not	
  shown).	
  In	
  addition,	
  the	
  minimal	
  

	
  

119	
  

inhibitory	
  concentrations	
  to	
  clindamycin	
  of	
  all	
  of	
  the	
  strains	
  used	
  in	
  this	
  study	
  
ranged	
  from	
  50	
  µg/ml	
  to	
  >100	
  mg/ml	
  (data	
  not	
  shown),	
  concentrations	
  several	
  fold	
  
higher	
  than	
  the	
  calculated	
  residual	
  clindamycin	
  in	
  the	
  reactors	
  at	
  the	
  time	
  of	
  C.	
  
difficile	
  inoculation	
  based	
  on	
  theoretical	
  washout	
  (<9	
  µg/ml).	
  Therefore,	
  the	
  
competition	
  outcome	
  was	
  not	
  reflective	
  of	
  differences	
  in	
  clindamycin	
  sensitivity.	
  	
  
Because	
  we	
  performed	
  our	
  competition	
  experiments	
  in	
  fecal	
  bioreactors	
  that	
  
had	
  been	
  treated	
  by	
  clindamycin	
  and	
  did	
  not	
  include	
  untreated	
  control	
  reactors,	
  we	
  
wanted	
  to	
  verify	
  that	
  the	
  communities	
  present	
  in	
  our	
  clindamycin-­‐treated	
  
competition	
  reactors	
  were	
  similar	
  to	
  those	
  previously	
  established	
  and	
  characterized	
  
in	
  our	
  in	
  vitro	
  invasion	
  model.	
  Therefore,	
  we	
  sequenced	
  the	
  V3-­‐V5	
  hypervariable	
  
region	
  of	
  16S	
  rRNA	
  gene	
  from	
  our	
  competition	
  bioreactor	
  samples	
  at	
  day	
  7,	
  just	
  
prior	
  to	
  C.	
  difficile	
  inoculation,	
  by	
  pyrosequencing	
  and	
  compared	
  these	
  sequences	
  to	
  
the	
  previous	
  data	
  we	
  had	
  collected	
  from	
  our	
  in	
  vitro	
  invasion	
  model	
  (Fig.	
  3.4	
  and	
  
3.5).	
  We	
  found	
  that	
  the	
  richness,	
  diversity,	
  and	
  evenness	
  of	
  the	
  competition	
  
communities	
  were	
  similar	
  to	
  those	
  of	
  the	
  other	
  clindamycin-­‐treated	
  communities	
  on	
  
day	
  6	
  (Fig.	
  S3.4).	
  When	
  we	
  compared	
  the	
  microbial	
  community	
  structures	
  of	
  the	
  
competition	
  bioreactor	
  samples	
  using	
  the	
  Bray-­‐Curtis	
  dissimilarity	
  measure	
  and	
  
plotted	
  them	
  with	
  the	
  previous	
  clindamycin-­‐treated	
  and	
  mock-­‐treated	
  samples	
  with	
  
NMDS,	
  we	
  found	
  that	
  they	
  grouped	
  together	
  with	
  the	
  clindamycin-­‐treated	
  
communities	
  (Fig.	
  S3.5).	
  These	
  community	
  comparisons	
  show	
  that	
  the	
  strains	
  were	
  
competed	
  in	
  the	
  presence	
  of	
  complex,	
  diverse	
  fecal	
  communities	
  and	
  not	
  community	
  
anomalies	
  made	
  up	
  of	
  unexpectedly	
  low	
  richness	
  or	
  diversity.	
  
	
  

	
  

	
  

120	
  

Ribotype	
  027	
  strains	
  display	
  a	
  competitive	
  advantage	
  in	
  vivo.	
  In	
  order	
  to	
  
address	
  if	
  ribotype	
  027	
  strains	
  are	
  capable	
  of	
  outcompeting	
  non-­‐027	
  ribotype	
  
strains	
  in	
  the	
  intestinal	
  tract,	
  we	
  competed	
  two	
  ribotype	
  027	
  and	
  non-­‐027	
  strains	
  in	
  
a	
  humanized	
  microbiota	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection.	
  Mice	
  were	
  treated	
  with	
  
an	
  antibiotic	
  cocktail	
  (36)	
  for	
  three	
  days,	
  followed	
  by	
  a	
  single	
  dose	
  of	
  clindamycin.	
  
24	
  hours	
  later	
  mice	
  were	
  gavaged	
  with	
  C.	
  difficile	
  spores.	
  Under	
  these	
  conditions,	
  
strains	
  CD3017	
  and	
  CD4015	
  (ribotype	
  027)	
  and	
  strains	
  CD1014	
  and	
  CD2048	
  
(ribotypes	
  014	
  and	
  053,	
  respectively)	
  were	
  able	
  to	
  transiently	
  colonize	
  the	
  
intestinal	
  tracts	
  of	
  mice	
  when	
  infected	
  individually	
  without	
  causing	
  severe	
  disease	
  
(Fig.	
  S3.6).	
  	
  
To	
  compare	
  the	
  relative	
  fitness	
  between	
  ribotype	
  027	
  and	
  non-­‐027	
  strains,	
  
antibiotic-­‐treated	
  animals	
  were	
  treated	
  with	
  a	
  mixture	
  of	
  104	
  spores	
  from	
  strains	
  
CD3017	
  (027)	
  and	
  CD1014	
  (014)	
  (ratio	
  of	
  1:14)	
  or	
  CD4015	
  (027)	
  and	
  CD2048	
  
(053)	
  (ratio	
  of	
  1:50).	
  The	
  abundance	
  of	
  each	
  strain	
  was	
  monitored	
  daily	
  by	
  selective	
  
plating	
  of	
  mouse	
  feces.	
  The	
  027	
  strain	
  competitive	
  indices	
  in	
  replicate	
  mice	
  for	
  both	
  
competition	
  groups	
  are	
  plotted	
  in	
  Fig.	
  3.7.	
  Competitive	
  indices	
  are	
  calculated	
  by	
  
dividing	
  the	
  027:non-­‐027	
  ratios	
  at	
  day	
  4	
  by	
  the	
  ratios	
  present	
  in	
  the	
  gavaged	
  spore	
  
mixtures.	
  	
  

	
  

121	
  

	
  

Competitive Index

1000

100

10

1
CD3017
[CD1014 (014)]

CD4015
[CD2048 (053)]

027 Strain
[competitor strain (ribotype)]

	
  
Figure	
  3.7.	
  	
  Competitive	
  indices	
  of	
  ribotype	
  027	
  strains	
  relative	
  to	
  non-­‐027	
  
strains	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection.	
  	
  After	
  antibiotic	
  treatment,	
  mice	
  
were	
  gavaged	
  with	
  mixtures	
  of	
  027	
  and	
  non-­‐027	
  ribotype	
  strain	
  spores.	
  	
  C.	
  difficile	
  
abundance	
  for	
  the	
  indicated	
  strains	
  was	
  determined	
  by	
  selective	
  plating	
  of	
  the	
  fecal	
  
samples.	
  Plotted	
  here	
  is	
  the	
  competitive	
  index	
  of	
  the	
  027	
  strains	
  for	
  each	
  replicate	
  
mouse	
  competition	
  as	
  the	
  ratio	
  of	
  027:non-­‐027	
  at	
  day	
  4	
  divided	
  by	
  the	
  ratio	
  at	
  day	
  0	
  
(ratio	
  in	
  spore	
  mixes).	
  Black	
  bars	
  represent	
  the	
  mean	
  of	
  replicates	
  for	
  each	
  
competition.	
  	
  The	
  dotted	
  line	
  on	
  the	
  Y-­‐axis	
  represents	
  the	
  upper	
  CI	
  limit	
  for	
  the	
  
CD3017/CD1014	
  competition	
  based	
  on	
  the	
  plating	
  limit	
  of	
  detection.	
  
	
  
In	
  both	
  competitions,	
  we	
  noted	
  that	
  the	
  ribotype	
  027	
  strains	
  had	
  a	
  
competitive	
  advantage	
  when	
  directly	
  competing	
  in	
  the	
  mouse	
  intestinal	
  tract	
  (Fig.	
  
3.7).	
  CD3017	
  displayed	
  a	
  dramatic	
  expansion	
  over	
  days	
  2-­‐4	
  when	
  competed	
  against	
  
CD1014;	
  at	
  day	
  4,	
  the	
  mean	
  CI	
  for	
  3017	
  was	
  86.7	
  (range=18.8	
  to	
  140.9).	
  CD4015	
  
also	
  displayed	
  a	
  competitive	
  index	
  that	
  showed	
  it	
  had	
  a	
  competitive	
  advantage	
  over	
  
CD2048,	
  and	
  although	
  the	
  CI	
  was	
  not	
  as	
  robust	
  as	
  observed	
  with	
  CD3017,	
  it	
  had	
  a	
  
competitive	
  advantage	
  with	
  a	
  CI	
  of	
  12.4	
  (range=	
  7.9	
  to	
  25.0).	
  These	
  data	
  

	
  

122	
  

demonstrate	
  that	
  ribotype	
  027	
  strains	
  have	
  a	
  competitive	
  advantage	
  over	
  non-­‐027	
  
strains	
  in	
  vivo.	
  	
  
	
  
	
  
Discussion	
  
Ribotype	
  027	
  strains	
  have	
  been	
  frequently	
  shown	
  to	
  be	
  overrepresented	
  in	
  hospital	
  
outbreaks	
  and	
  have	
  been	
  linked	
  to	
  increased	
  morbidity	
  and	
  mortality.	
  Although	
  this	
  
association	
  with	
  a	
  hypervirulent	
  state	
  is	
  controversial,	
  the	
  fact	
  that	
  ribotype	
  027	
  
strains	
  have	
  swept	
  across	
  the	
  globe	
  implies	
  they	
  have	
  acquired	
  an	
  increased	
  ability	
  
to	
  cause	
  disease.	
  We	
  hypothesized	
  that	
  differences	
  in	
  strain	
  physiology	
  could	
  give	
  
ribotype	
  027	
  strains	
  a	
  competitive	
  advantage	
  over	
  strains	
  of	
  other	
  ribotypes,	
  
thereby	
  leading	
  to	
  the	
  increased	
  prevalence	
  of	
  ribotype	
  027	
  strains.	
  We	
  used	
  the	
  
MBRA	
  C.	
  difficile	
  invasion	
  model	
  to	
  demonstrate	
  that	
  ribotype	
  027	
  strains	
  were	
  able	
  
to	
  outcompete	
  strains	
  of	
  other	
  ribotypes	
  in	
  the	
  presence	
  of	
  complex	
  fecal	
  bacterial	
  
communities.	
  We	
  then	
  demonstrated	
  similar	
  competitive	
  advantages	
  of	
  ribotype	
  
027	
  strains	
  to	
  strains	
  of	
  other	
  ribotypes	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection.	
  	
  
Our	
  work	
  demonstrates	
  that	
  ribotype	
  027	
  strains	
  can	
  directly	
  outcompete	
  
strains	
  of	
  other	
  ribotypes.	
  Because	
  we	
  used	
  four	
  independent	
  ribotype	
  027	
  strains	
  
competed	
  against	
  four	
  independent	
  strains	
  of	
  varying	
  ribotypes,	
  we	
  do	
  not	
  expect	
  
that	
  the	
  observed	
  increase	
  in	
  competitive	
  fitness	
  of	
  the	
  ribotype	
  027	
  strains	
  was	
  due	
  
to	
  strain	
  selection.	
  In	
  the	
  majority	
  of	
  competition	
  pairs	
  we	
  studied,	
  the	
  ribotype	
  027	
  
strains	
  became	
  the	
  dominant	
  C.	
  difficile	
  strain	
  in	
  the	
  community,	
  sometimes	
  leading	
  
to	
  the	
  complete	
  loss	
  of	
  the	
  non-­‐027	
  strain.	
  This	
  dominance	
  was	
  observed	
  in	
  all	
  

	
  

123	
  

bioreactor	
  communities	
  studied,	
  as	
  well	
  as	
  in	
  the	
  mouse	
  competition	
  between	
  the	
  
ribotype	
  027	
  strain	
  CD3017	
  and	
  the	
  ribotype	
  014	
  strain	
  CD1014.	
  However,	
  in	
  the	
  
second	
  mouse	
  competition	
  between	
  ribotype	
  027	
  strain	
  CD4015	
  and	
  ribotype	
  053	
  
strain	
  CD2048,	
  the	
  ratio	
  of	
  CD4015	
  to	
  CD2048	
  increased	
  over	
  time,	
  from	
  a	
  1:50	
  ratio	
  
at	
  the	
  beginning	
  of	
  the	
  experiment	
  to	
  a	
  mean	
  1:4	
  ratio	
  (range=1:10-­‐1:2)	
  at	
  the	
  end	
  
of	
  the	
  experiment,	
  but	
  CD4015	
  did	
  not	
  become	
  the	
  dominant	
  C.	
  difficile	
  strain	
  in	
  the	
  
community.	
  Although	
  we	
  interpret	
  these	
  results	
  to	
  indicate	
  that	
  CD4015	
  was	
  
outcompeting	
  CD2048	
  for	
  available	
  niche	
  space	
  and	
  would	
  likely	
  have	
  led	
  to	
  its	
  loss	
  
from	
  the	
  community	
  if	
  the	
  experiment	
  had	
  been	
  continued	
  further,	
  we	
  cannot	
  
exclude	
  the	
  hypothesis	
  that	
  both	
  strains	
  would	
  have	
  reached	
  a	
  steady	
  level	
  of	
  co-­‐
existence.	
  Interestingly,	
  when	
  comparing	
  CI’s	
  of	
  competition	
  pairs	
  in	
  the	
  MBRA	
  or	
  
mice,	
  similar	
  trends	
  were	
  observed,	
  with	
  the	
  CD3017	
  CI	
  higher	
  than	
  the	
  CD4015	
  CI	
  
in	
  each	
  model,	
  supporting	
  that	
  the	
  MBRA	
  model	
  can	
  recapitulate	
  C.	
  difficile	
  
dynamics	
  that	
  occur	
  in	
  vivo,	
  and	
  further	
  demonstrating	
  the	
  validity	
  of	
  the	
  MBRA	
  
model	
  as	
  a	
  precursor	
  for	
  in	
  vivo	
  experiments	
  (Fig.	
  S3.7).	
  	
  
However,	
  it	
  is	
  currently	
  unclear	
  what	
  physiological	
  differences	
  present	
  in	
  
ribotype	
  027	
  strains	
  allow	
  these	
  strains	
  to	
  outcompete	
  strains	
  of	
  other	
  ribotypes.	
  
Although	
  it	
  is	
  possible	
  that	
  ribotype	
  027	
  strains	
  are	
  capable	
  of	
  directly	
  antagonizing	
  
strains	
  of	
  other	
  ribotypes,	
  we	
  favor	
  the	
  model	
  that	
  ribotype	
  027	
  strains	
  are	
  better	
  
able	
  to	
  proliferate	
  within	
  the	
  intestinal	
  environment	
  and	
  thereby	
  indirectly	
  lead	
  to	
  
elimination	
  of	
  strains	
  of	
  other	
  ribotypes.	
  	
  
One	
  aspect	
  of	
  physiology	
  that	
  could	
  potentially	
  impact	
  competition	
  outcome	
  
is	
  inter-­‐strain	
  variability	
  in	
  rates	
  of	
  sporulation.	
  If	
  one	
  strain	
  had	
  a	
  higher	
  

	
  

124	
  

proportion	
  of	
  the	
  cells	
  in	
  its	
  population	
  enter	
  into	
  sporulation	
  during	
  the	
  course	
  of	
  
competition,	
  that	
  strain	
  would	
  have	
  less	
  cells	
  in	
  logarithmic	
  growth,	
  effectively	
  
reducing	
  its	
  competitive	
  fitness.	
  Although	
  there	
  have	
  been	
  some	
  reports	
  that	
  
ribotype	
  027	
  strains	
  sporulate	
  more	
  efficiently	
  than	
  other	
  strains	
  (e.g.,	
  (42,	
  43),	
  
larger	
  studies	
  comparing	
  multiple	
  isolates	
  of	
  different	
  ribotypes	
  have	
  found	
  that	
  
there	
  is	
  no	
  significant	
  correlation	
  between	
  sporulation	
  efficiency	
  and	
  ribotype	
  (44,	
  
45).	
  Because	
  sporulation	
  dynamics	
  of	
  individual	
  strains	
  co-­‐cultured	
  within	
  the	
  
MBRA	
  are	
  difficult	
  to	
  measure;	
  we	
  do	
  not	
  currently	
  have	
  data	
  to	
  determine	
  whether	
  
the	
  non-­‐027	
  strains	
  used	
  in	
  our	
  study	
  sporulate	
  to	
  higher	
  levels	
  than	
  the	
  027	
  
strains.	
  However,	
  when	
  we	
  assayed	
  sporulation	
  in	
  pure	
  culture	
  under	
  both	
  batch	
  
and	
  continuous-­‐culture	
  conditions,	
  we	
  did	
  not	
  observe	
  higher	
  rates	
  of	
  sporulation	
  of	
  
these	
  non-­‐027	
  strains	
  (data	
  not	
  shown).	
  The	
  dynamics	
  of	
  C.	
  difficile	
  sporulation	
  in	
  
the	
  context	
  of	
  growing	
  in	
  the	
  presence	
  of	
  a	
  complex	
  fecal	
  community	
  is	
  an	
  area	
  of	
  
current	
  and	
  future	
  investigation.	
  While	
  it	
  is	
  important	
  to	
  consider	
  the	
  impact	
  of	
  
sporulation	
  dynamics	
  on	
  competition	
  outcome,	
  we	
  do	
  not	
  believe	
  this	
  is	
  the	
  factor	
  
responsible	
  for	
  ribotype	
  027	
  strains	
  outcompeting	
  other	
  ribotypes	
  in	
  our	
  
experiments.	
  
A	
  second	
  aspect	
  of	
  physiology	
  that	
  could	
  play	
  a	
  role	
  in	
  competitive	
  fitness	
  is	
  
differences	
  in	
  germination.	
  Differential	
  germination	
  does	
  not	
  play	
  a	
  role	
  in	
  the	
  
competitive	
  advantage	
  of	
  ribotype	
  027	
  strains	
  in	
  the	
  MBRA,	
  since	
  competitions	
  
were	
  initiated	
  with	
  vegetative	
  cells.	
  In	
  contrast,	
  competition	
  in	
  the	
  mouse	
  model	
  
was	
  initiated	
  by	
  gavaging	
  a	
  mixture	
  of	
  spores.	
  Recent	
  work	
  published	
  by	
  Francis	
  et	
  
al.	
  demonstrates	
  that	
  murine	
  bile	
  acids	
  (muricholic	
  acids)	
  inhibit	
  C.	
  difficile	
  spore	
  

	
  

125	
  

germination	
  and	
  that	
  there	
  is	
  strain	
  variability	
  in	
  this	
  inhibition	
  (46).	
  In	
  addition,	
  
published	
  data	
  shows	
  there	
  is	
  significant	
  variability	
  in	
  rates	
  of	
  germination	
  and	
  the	
  
compounds	
  that	
  serve	
  as	
  germinants	
  among	
  large	
  sets	
  of	
  C.	
  difficile	
  strains	
  of	
  
varying	
  ribotyes,	
  at	
  least	
  in	
  vitro	
  (47).	
  However,	
  when	
  we	
  compared	
  the	
  day	
  one	
  
levels	
  of	
  C.	
  difficile	
  in	
  the	
  mice	
  gavaged	
  with	
  individual	
  strains	
  of	
  different	
  ribotypes	
  
to	
  the	
  levels	
  of	
  spores	
  present	
  in	
  the	
  initial	
  inoculum,	
  we	
  found	
  similar	
  increases	
  in	
  
the	
  abundance	
  of	
  C.	
  difficile	
  cells	
  across	
  the	
  different	
  strains	
  by	
  day	
  one	
  (Fig.	
  S3.6).	
  
Based	
  upon	
  these	
  results,	
  we	
  conclude	
  that	
  differences	
  in	
  germination	
  rates	
  are	
  
unlikely	
  to	
  play	
  a	
  role	
  in	
  the	
  ribotype	
  027	
  competitive	
  advantage.	
  	
  
Evidence	
  suggests	
  that	
  factors	
  of	
  colonization	
  resistance	
  are	
  important	
  in	
  
preventing	
  C.	
  difficile	
  infection,	
  and	
  therefore	
  must	
  be	
  overcome	
  in	
  order	
  for	
  C.	
  
diffcile	
  to	
  proliferate	
  in	
  the	
  colon	
  (4).	
  Some	
  of	
  these	
  factors	
  include	
  competition	
  for	
  
nutrients,	
  antagonism	
  by	
  production	
  of	
  inhibitory	
  compounds	
  (such	
  as	
  bacteriocins	
  
or	
  short	
  chain	
  fatty	
  acids),	
  and	
  microbiota-­‐dependent	
  modulation	
  of	
  host	
  immune	
  
functions	
  (reviewed	
  in	
  (5)).	
  Identifying	
  which	
  of	
  these	
  factors	
  is	
  driving	
  the	
  
competition	
  outcome	
  in	
  our	
  model	
  is	
  the	
  next	
  area	
  of	
  investigation.	
  Comparative	
  
genomic	
  studies	
  have	
  identified	
  potential	
  candidate	
  genes	
  that	
  could	
  provide	
  
ribotype	
  027	
  strains	
  an	
  increased	
  competitive	
  advantage	
  over	
  strains	
  of	
  other	
  
ribotypes.	
  One	
  such	
  genomic	
  difference	
  is	
  the	
  presence	
  of	
  the	
  thymidylate	
  synthase	
  
gene,	
  thyA,	
  in	
  ribotype	
  027	
  strains,	
  which	
  has	
  effectively	
  replaced	
  the	
  native,	
  
alternative	
  thymidylate	
  synthase	
  gene,	
  thyX,	
  in	
  the	
  form	
  of	
  a	
  four-­‐gene	
  insertion	
  
(34).	
  Escartin	
  et	
  al.	
  showed	
  that	
  ThyA	
  enzymes	
  have,	
  on	
  average,	
  10-­‐fold	
  faster	
  
catalytic	
  rates	
  than	
  ThyX	
  enzymes	
  in	
  vitro	
  and	
  are	
  able	
  to	
  confer	
  faster	
  genome	
  

	
  

126	
  

replication	
  rates	
  in	
  vivo	
  (48).	
  Based	
  upon	
  this	
  work,	
  we	
  hypothesize	
  that	
  the	
  
presence	
  of	
  thyA	
  in	
  ribotype	
  027	
  strains	
  may	
  confer	
  a	
  growth	
  advantage,	
  and	
  
therefore	
  play	
  a	
  role	
  in	
  competitive	
  fitness.	
  	
  We	
  have	
  also	
  begun	
  assessing	
  whether	
  
differences	
  in	
  carbohydrate	
  metabolism	
  between	
  ribotype	
  027	
  strains	
  and	
  strains	
  of	
  
other	
  ribotypes	
  could	
  provide	
  a	
  competitive	
  advantage.	
  Preliminary	
  studies	
  using	
  
phenotypic	
  arrays	
  (Biolog)	
  and	
  follow-­‐up	
  growth	
  studies	
  have	
  revealed	
  differences	
  
in	
  metabolism	
  of	
  some	
  carbohydrates	
  by	
  ribotype	
  027	
  strains	
  (C.D.R.	
  and	
  R.A.B.,	
  
unpublished	
  results).	
  	
  Moreover,	
  we	
  are	
  actively	
  investigating	
  whether	
  ribotype	
  027	
  
strains	
  require	
  less	
  of	
  a	
  disruption	
  of	
  the	
  microbiota	
  to	
  invade,	
  for	
  example	
  after	
  
shorter	
  antibiotic	
  regimens	
  or	
  after	
  lower	
  dosing	
  of	
  antibiotics,	
  which	
  would	
  
provide	
  further	
  evidence	
  as	
  to	
  why	
  these	
  strains	
  are	
  so	
  prevalent	
  in	
  many	
  clinical	
  
locations.	
  	
  
	
  
Fecal	
  MBRA	
  as	
  a	
  model	
  for	
  C.	
  difficile	
  invasion.	
  Our	
  goal	
  was	
  to	
  cultivate	
  complex	
  
fecal	
  microbial	
  communities	
  within	
  the	
  MBRA	
  that	
  could	
  resist	
  invasion	
  by	
  C.	
  
difficile,	
  recognizing	
  that	
  these	
  communities	
  would	
  not	
  be	
  1:1	
  translations	
  of	
  the	
  
starting	
  fecal	
  inoculum.	
  Relative	
  to	
  our	
  initial	
  fecal	
  inoculum,	
  we	
  observed	
  a	
  
significant	
  shift	
  in	
  the	
  ratio	
  of	
  Bacteroides:Firmicutes	
  phylum	
  members	
  (Fig.	
  3.4B).	
  
Bacteroides-­‐dominated	
  communities	
  have	
  been	
  observed	
  in	
  several	
  different	
  in	
  vitro	
  
bioreactor	
  models	
  (22,	
  49-­‐51),	
  which	
  often	
  use	
  media	
  of	
  similar	
  composition	
  for	
  
cultivation.	
  Several	
  parameters	
  can	
  affect	
  the	
  composition	
  of	
  the	
  microbial	
  
communities	
  that	
  are	
  established	
  in	
  fecal	
  bioreactors,	
  including	
  source	
  of	
  fecal	
  
material	
  and	
  how	
  it	
  is	
  processed,	
  the	
  media	
  composition	
  and	
  turnover	
  time	
  used	
  for	
  

	
  

127	
  

cultivation,	
  and	
  the	
  availability	
  of	
  surfaces	
  for	
  biofilm	
  formation	
  (reviewed	
  in	
  (52)).	
  
For	
  example,	
  we	
  chose	
  to	
  pool	
  fecal	
  samples	
  from	
  twelve	
  donors,	
  reasoning	
  that	
  this	
  
may	
  lead	
  to	
  an	
  in	
  vitro	
  community	
  that	
  was	
  more	
  representative	
  of	
  the	
  microbial	
  
diversity	
  present	
  amongst	
  different	
  individuals	
  than	
  could	
  be	
  achieved	
  from	
  a	
  single	
  
donor.	
  However,	
  by	
  choosing	
  this	
  pooling	
  strategy	
  we	
  may	
  have	
  selected	
  for	
  
communities	
  that	
  would	
  not	
  normally	
  co-­‐exist.	
  We	
  have	
  examined	
  the	
  differences	
  
between	
  MBRA	
  communities	
  formed	
  from	
  single	
  and	
  pooled	
  fecal	
  samples	
  and	
  
found	
  that	
  they	
  exhibit	
  similar	
  C.	
  difficile	
  invasion	
  dynamics	
  (manuscript	
  in	
  
preparation).	
  Although	
  modifying	
  different	
  aspects	
  of	
  the	
  operating	
  parameters	
  
could	
  lead	
  to	
  communities	
  with	
  higher	
  similarities	
  to	
  the	
  starting	
  fecal	
  inoculum,	
  
our	
  results	
  demonstrate	
  that	
  our	
  current	
  model	
  yields	
  complex	
  fecal	
  MBRA	
  
communities	
  that	
  resist	
  invasion	
  by	
  C.	
  difficile	
  when	
  unperturbed	
  and	
  are	
  
susceptible	
  when	
  disturbed	
  by	
  antibiotics.	
  	
  
Our	
  MBRA	
  model	
  does	
  not	
  promote	
  invasion	
  of	
  antibiotic-­‐treated	
  
communities	
  with	
  C.	
  difficile	
  spores	
  and	
  thus	
  we	
  cannot	
  monitor	
  spore	
  germination	
  
dynamics	
  in	
  the	
  presence	
  of	
  a	
  fecal	
  microbiota.	
  We	
  are	
  currently	
  attempting	
  to	
  
modify	
  the	
  model	
  to	
  enable	
  spore	
  germination	
  within	
  the	
  MBRAs.	
  Determining	
  
those	
  aspects	
  of	
  the	
  current	
  model	
  inhibitory	
  to	
  spore	
  germination	
  and	
  outgrowth	
  
may	
  also	
  provide	
  new	
  insights	
  into	
  the	
  dynamics	
  between	
  C.	
  difficile	
  and	
  the	
  
microbiota.	
  
In	
  spite	
  of	
  the	
  limitations	
  discussed	
  above,	
  the	
  in	
  vitro	
  model	
  that	
  we	
  
developed	
  allows	
  for	
  robust,	
  higher	
  throughput	
  studies	
  of	
  C.	
  difficile	
  invasion	
  on	
  
shorter	
  time	
  scales	
  than	
  can	
  be	
  accommodated	
  in	
  animal	
  models	
  and	
  other	
  more	
  

	
  

128	
  

complex	
  bioreactor	
  models.	
  Therefore,	
  this	
  model	
  can	
  serve	
  as	
  a	
  complement	
  to	
  
animal	
  studies	
  by	
  providing	
  a	
  platform	
  for	
  conducting	
  initial,	
  hypothesis-­‐generating	
  
experiments,	
  including	
  those	
  experiments	
  regarding	
  potential	
  therapeutic	
  
treatment	
  of	
  C.	
  difficile	
  infection.	
  	
  
	
  
Conclusions.	
  The	
  data	
  from	
  the	
  present	
  study	
  suggest	
  that	
  ribotype	
  027	
  strains	
  
have	
  an	
  ecological	
  advantage	
  over	
  other	
  C.	
  difficile	
  ribotypes	
  in	
  the	
  context	
  of	
  the	
  
intestinal	
  microbiota.	
  If	
  this	
  competitive	
  advantage	
  holds	
  true	
  in	
  the	
  human	
  colonic	
  
environment,	
  it	
  may	
  explain,	
  in	
  part,	
  the	
  epidemic	
  nature	
  of	
  ribotype	
  027	
  strains.	
  
Mixed	
  species	
  C.	
  difficile	
  infections	
  have	
  been	
  found	
  to	
  occur	
  in	
  7-­‐13%	
  of	
  patients	
  
infected	
  with	
  disease	
  ((53,	
  54),	
  and	
  references	
  therein).	
  However,	
  it	
  has	
  been	
  
difficult	
  to	
  explore	
  the	
  impact	
  of	
  mixed	
  infection	
  on	
  disease	
  progression	
  due	
  to	
  
limitations	
  in	
  the	
  ability	
  to	
  accurately	
  quantify	
  the	
  rate	
  at	
  which	
  mixed	
  infections	
  
occur,	
  or	
  the	
  dynamics	
  of	
  the	
  mixed	
  strains	
  over	
  time	
  within	
  individual	
  patients	
  (53,	
  
54).	
  While	
  direct	
  competition	
  between	
  strains	
  may	
  be	
  occurring	
  during	
  co-­‐infection	
  
in	
  patients,	
  we	
  suspect	
  that	
  ribotype	
  027	
  strains	
  are	
  able	
  to	
  outcompete	
  strains	
  of	
  
other	
  ribotypes	
  in	
  our	
  models	
  of	
  C.	
  difficile	
  infection	
  due	
  to	
  their	
  ability	
  to	
  better	
  
exploit	
  the	
  limited	
  resources	
  available	
  within	
  the	
  intestinal	
  communities.	
  Finally,	
  
this	
  work	
  further	
  demonstrates	
  that	
  aside	
  from	
  virulence	
  factors	
  such	
  as	
  toxin	
  
production	
  and	
  antibiotic	
  resistance,	
  the	
  physiology	
  of	
  C.	
  difficile	
  should	
  be	
  
considered	
  an	
  important	
  contributor	
  to	
  its	
  success	
  as	
  a	
  pathogen.	
  
	
  
	
  

	
  

129	
  

Acknowledgements	
  
The	
  authors	
  acknowledge	
  Sara	
  McNamara	
  (MDCH)	
  for	
  providing	
  C.	
  difficile	
  strains,	
  
Seth	
  Walk	
  (University	
  of	
  Michigan)	
  for	
  ribotyping	
  strains,	
  Robert	
  Stedtfeld	
  (MSU)	
  
for	
  designing	
  the	
  mini-­‐bioreactor	
  arrays,	
  Sara	
  Poe	
  and	
  Kathryn	
  Eaton	
  for	
  
collaboration	
  in	
  developing	
  the	
  humanized	
  microbiota	
  mice,	
  and	
  Vince	
  Young	
  
(University	
  of	
  Michigan)	
  and	
  Richard	
  Lenski	
  (MSU)	
  for	
  helpful	
  comments	
  regarding	
  
the	
  manuscript.	
  This	
  work	
  was	
  supported	
  by	
  award	
  5U19AI090872-­‐02	
  from	
  the	
  
National	
  Institutes	
  of	
  Allergy	
  and	
  Infectious	
  Diseases	
  to	
  RAB.	
  	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

130	
  

APPENDIX	
  

	
  

131	
  

Appendix	
  
	
  
Supplementary	
  Methods:	
  MBRA	
  Design	
  and	
  Operation.	
  	
  	
  
As	
  described	
  in	
  the	
  Methods,	
  MBRAs	
  were	
  fabricated	
  from	
  DSM	
  Somos	
  Watershed	
  
XC	
  11122	
  via	
  stereolithography	
  (FineLine	
  prototyping).	
  Six	
  reactors	
  with	
  an	
  
internal	
  volume	
  of	
  25	
  mL	
  were	
  designed	
  into	
  a	
  single	
  strip	
  with	
  200mm	
  x	
  47	
  mm	
  x	
  
36	
  mm	
  dimensions	
  (Fig.	
  3.1).	
  Reactors	
  were	
  drawn	
  with	
  25	
  x	
  25	
  x	
  40	
  mm	
  
dimensions,	
  including	
  10	
  mm	
  radial	
  blends	
  on	
  the	
  bottom	
  corners	
  (to	
  prevent	
  
buildup	
  of	
  cells	
  and	
  other	
  insoluble	
  materials),	
  and	
  5	
  mm	
  radial	
  blends	
  on	
  the	
  top.	
  
Reactors	
  were	
  spaced	
  32.5	
  mm	
  center	
  to	
  center	
  to	
  match	
  dimensions	
  on	
  the	
  stir	
  
plate.	
  Three	
  5.55	
  mm	
  diameter	
  holes	
  (influent,	
  effluent,	
  sampling)	
  were	
  placed	
  into	
  
the	
  top	
  of	
  each	
  reactor,	
  and	
  spaced	
  16	
  mm	
  apart	
  center	
  to	
  center.	
  Holes	
  were	
  
threaded	
  to	
  fit	
  conventional	
  leur	
  connectors.	
  Inner	
  walls	
  of	
  the	
  reactors	
  were	
  placed	
  
2	
  mm	
  from	
  the	
  bottom	
  of	
  the	
  strip,	
  5	
  mm	
  from	
  the	
  top	
  of	
  the	
  strip,	
  5	
  mm	
  from	
  the	
  
side	
  of	
  the	
  strip,	
  and	
  3.25	
  mm	
  into	
  the	
  strip.	
  A	
  1.25	
  x	
  16	
  x	
  31	
  mm	
  intrusion	
  was	
  
designed	
  into	
  the	
  bottom	
  corners	
  of	
  the	
  MBRA	
  for	
  fixing	
  into	
  a	
  custom	
  built	
  acrylic	
  
block	
  that	
  held	
  the	
  reactors	
  upright	
  and	
  properly	
  aligned	
  with	
  the	
  stir-­‐plate.	
  A	
  CAD	
  
file	
  (.stl)	
  that	
  can	
  be	
  used	
  directly	
  for	
  fabrication	
  via	
  stereolithography	
  is	
  available	
  
upon	
  request.	
  	
  	
  	
  	
  
Media	
  was	
  transferred	
  from	
  the	
  source	
  bottles	
  through	
  a	
  combination	
  of	
  1/8	
  
in	
  inner	
  diameter	
  (ID)	
  C-­‐flex	
  tubing	
  (Cole-­‐Parmer)	
  tubing	
  and	
  0.89	
  mm	
  ID	
  2-­‐stop	
  
Tygon	
  lab	
  tubing	
  supplied	
  to	
  the	
  reactors	
  via	
  a	
  24-­‐channel	
  peristaltic	
  pump	
  
(205S/CA,	
  Watson-­‐Marlow).	
  Waste	
  was	
  removed	
  from	
  the	
  reactors	
  through	
  a	
  

	
  

132	
  

combination	
  of	
  1/8	
  in	
  ID	
  C-­‐flex	
  tubing	
  tubing	
  and	
  1.14	
  mm	
  ID	
  2-­‐stop	
  Tygon	
  lab	
  
tubing	
  drawn	
  from	
  the	
  reactors	
  via	
  the	
  same	
  24-­‐channel	
  peristaltic	
  pump.	
  Reactors	
  
were	
  stirred	
  using	
  magnetic	
  stir	
  bars	
  driven	
  by	
  independent	
  magnets	
  on	
  a	
  60-­‐spot	
  
magnetic	
  stir	
  plate	
  (VarioMAG	
  HP	
  60,	
  Vario-­‐MAG	
  USA).	
  	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

133	
  

Table	
  S3.1.	
  	
  Competitive	
  indices	
  of	
  ribotype	
  027	
  strains	
  at	
  selected	
  time	
  points	
  
after	
  C.	
  difficile	
  inoculation	
  as	
  determined	
  by	
  quantitative	
  PCR.	
  	
  	
  
	
  
027	
  CIs	
  in	
  Fecal	
  Community	
  Background	
  Competitions	
  
	
  
	
  

CD2015	
  (027)	
  +	
  	
  
CD3014	
  (001)	
  

CD3017	
  (027)	
  +	
  
CD1014	
  (014)	
  

CD4015	
  (027)	
  +	
  	
  
CD2048	
  (053)	
  

CD4010	
  (027)	
  +	
  	
  
CD4004	
  (002)	
  

Days	
  post	
  C.	
  difficile	
  inoculation	
  
3	
  
7	
  
11	
  
0.06+	
  
0.52+	
  
0.22+	
  
3.30+	
  
3.19	
  
0.11	
  

0.75+	
  
1.19+	
  
0.49+	
  
22.39+*	
  
18.59	
  
3.36	
  

8.09+	
  
3.74+*	
  
0.88+	
  
ND	
  
ND	
  
36.17	
  

119.15	
  

36.93	
  

3.83	
  

11.88	
  

21.91	
  

33.05	
  

17.15	
  

22.84	
  

18.13	
  

232.86	
  

3993.21	
  

ND	
  

8.15	
  

86.92*	
  

23.43¢	
  

8.11+	
  

4.97+	
  

24.59+	
  

11.16+	
  

34.62+	
  

52.47+	
  

8.44+	
  

1.41+	
  

18.34+	
  

138.94*	
  

593.38*	
  

ND	
  

20.21	
  

21.26	
  

40.32	
  

19.43	
  

183.55	
  

107.20*	
  

15.24	
  

30.34	
  

9.14*	
  

12.97	
  

31.27	
  

50.09*	
  

27.22	
  

627.14*	
  

ND	
  

43.71*	
  

764.94*	
  

1386.61*	
  

	
  
ND=	
  not	
  determined	
  
*	
  Ratio	
  calculated	
  based	
  upon	
  limit	
  of	
  detection	
  for	
  non-­‐027	
  ribotype	
  	
  
+	
  CI’s	
  calculated	
  1	
  day	
  later	
  than	
  the	
  day	
  indicated	
  than	
  the	
  column	
  heading.	
  
¢CI’s	
  calculated	
  1	
  day	
  earlier	
  than	
  the	
  day	
  indicated	
  than	
  the	
  column	
  heading.	
  
Shaded	
  rows	
  indicate	
  replicates	
  where	
  C.	
  difficile	
  was	
  added	
  to	
  the	
  MBRAs	
  post-­‐	
  
cessation	
  of	
  clindamycin	
  dosing	
  24	
  hrs	
  later	
  than	
  the	
  other	
  replicates.	
  	
  
	
  
	
  
	
  
	
  
	
  

134	
  

Table	
  S3.2.	
  	
  Primers	
  used	
  for	
  qPCR.	
  
Target	
  Gene	
  

Primer	
  sequences	
  (Forward	
  &	
  Reverse)	
  

Citation	
  

C.	
  difficile	
  tcdA	
  

F:	
  AGC	
  TTT	
  CGC	
  TTT	
  AGG	
  CAG	
  TG	
  

This	
  study	
  

R:	
  ATG	
  GCT	
  GGG	
  TTA	
  AGG	
  TGT	
  TG	
  
Bacterial	
  16S	
  rRNA	
  	
   F:	
  ACT	
  CCT	
  ACG	
  GGA	
  GGC	
  AGC	
  AG	
  

(55)	
  

R:	
  ATT	
  ACC	
  GCG	
  GCT	
  GCT	
  GG	
  
C.	
  difficile	
  thyA	
  

F:	
  GAT	
  GGC	
  CAG	
  CCT	
  GCT	
  CAT	
  ACA	
  ATA	
  

This	
  study	
  

R:	
  TGT	
  TTC	
  ATC	
  AGC	
  CCA	
  GCT	
  ATC	
  CCA	
  
C.	
  difficile	
  thyX	
  

F:	
  CCA	
  GTT	
  GGG	
  ACA	
  GAC	
  GAA	
  AT	
  
R:	
  TGA	
  ACA	
  AGC	
  CCT	
  TGA	
  AAT	
  ACC	
  

	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

135	
  

This	
  study	
  

CT  (Bacterial  16S  rRNA)

23
22
21
20
19
18
17

0

2

4
6
8 10 12
Days  in  Culture

	
  
	
  
Figure	
  S3.1.	
  Bacterial	
  abundance	
  does	
  not	
  change	
  significantly	
  in	
  clindamycin-­‐
treated	
  reactors.	
  We	
  measured	
  the	
  relative	
  abundance	
  of	
  16S	
  rRNA	
  gene	
  copies	
  in	
  
bioreactor	
  samples	
  from	
  triplicate	
  clindamycin-­‐treated	
  (open	
  circles)	
  and	
  mock-­‐
treated	
  reactors	
  (closed-­‐squares)	
  using	
  quantitative	
  PCR	
  with	
  previously	
  described	
  
broad-­‐range	
  16S	
  rRNA	
  gene	
  qPCR	
  primers	
  (55).	
  We	
  have	
  reported	
  the	
  average	
  cycle	
  
threshold	
  (CT,	
  ±	
  standard	
  deviation)	
  where	
  the	
  qPCR	
  reactions	
  began	
  amplifying	
  
linearly.	
  We	
  did	
  not	
  determine	
  absolute	
  quantification	
  of	
  16S	
  rRNA	
  gene	
  copies	
  in	
  
the	
  samples	
  because	
  the	
  sample	
  populations	
  are	
  composed	
  of	
  mixtures	
  of	
  bacteria	
  
with	
  different	
  16S	
  rRNA	
  gene	
  copy	
  numbers.	
  Clindamycin	
  or	
  mock-­‐treatment	
  began	
  
on	
  day	
  2.5	
  in	
  culture	
  and	
  continued	
  twice	
  daily	
  through	
  day	
  6.	
  
	
  
	
  

	
  

136	
  

	
  

	
  
	
  
	
  
	
  

bp
NAP-2
NAP-1

500
bp
thyX	
  

1000
500

250
bp
thyA	
  

100

630
196
37
VPI 10463
negative

P-

unknown

N

N

3
P-

A

A

N

	
  

4
A
P
N -6
A
P
N -7
A
P8

	
  

500
bp
thyX	
  

bp
1000
500

250
bp
thyA	
  
100

	
  
Figure	
  S3.2.	
  	
  PCR	
  screen	
  of	
  DNA	
  samples	
  from	
  88	
  strains	
  of	
  C.	
  difficile	
  for	
  
detection	
  of	
  insert	
  containing	
  thyA	
  or	
  the	
  uninterrupted	
  thyX.	
  

	
  

137	
  

CD3017(027) + CD1014(014)

CD2015(027) + CD3014(001)
103

101
100
10-1
10-2

102

027:non-027 ratio

027:non-027 ratio

102

0

5

101
100
10-1
10-2

10

0

5

CD4015(027) + CD2048(053)

CD4010(027) + CD4004(002)
103

102

027:non-027 ratio

027:non-027 ratio

103

101
100
10-1
10-2

10

Days in Competition

Days in Competition

0

5

102
101
100
10-1

10

Days in Competition

0

5

10

Days in Competition

Figure	
  S3.3.	
  	
  Ratios	
  of	
  ribotype	
  027:non-­‐027	
  C.	
  difficile	
  strains	
  over	
  time	
  in	
  
MBRA	
  competitions.	
  	
  Each	
  plot	
  represents	
  a	
  different	
  competition	
  pair	
  of	
  one	
  
ribotype	
  027	
  and	
  one	
  non-­‐027	
  ribotype	
  strain;	
  ribotypes	
  indicated	
  in	
  parentheses	
  
above.	
  	
  Each	
  line	
  in	
  the	
  plots	
  represents	
  a	
  replicate	
  reactor,	
  combined	
  from	
  three	
  
independent	
  experiments.	
  	
  Where	
  the	
  non-­‐027	
  ribotype	
  strain	
  was	
  below	
  the	
  limit	
  
of	
  detection,	
  the	
  ratio	
  was	
  determined	
  by	
  substituting	
  in	
  the	
  highest	
  CT	
  value	
  in	
  the	
  
linear	
  range	
  (detection	
  limit)	
  for	
  the	
  non-­‐027	
  value	
  (open	
  circles).	
  	
  
	
  
	
  
	
  
	
  

	
  

	
  

138	
  

A.
#  of  OTUs  (3%  ID)

80
60
40
20
0

0

2

4
6
8
10
Days  in  Culture

12

0

2

4
6
8
10
Days  in  Culture

12

2

4
6
8
Days  in  Culture

12

B.
Inverse  Simpson  

16
12
8
4
0

C.
Simpson  Evenness  

0.3
0.2
0.1
0.0

0

10

	
  
	
  
Figure	
  S3.4.	
  	
  Comparison	
  of	
  the	
  community	
  structure	
  on	
  day	
  7	
  from	
  
clindamycin-­‐treated	
  reactors	
  used	
  for	
  C.	
  difficile	
  competition	
  experiments	
  to	
  
triplicate	
  mock-­‐treated	
  and	
  clindamycin-­‐treated	
  reactors	
  infected	
  with	
  
CD2015.	
  As	
  described	
  in	
  Fig.	
  3.4,	
  we	
  analyzed	
  the	
  16S	
  rRNA	
  gene	
  abundances	
  
(binned	
  into	
  OTUs	
  with	
  3%	
  sequence	
  identity)	
  from	
  three	
  mock-­‐treated	
  and	
  
clindamycin-­‐treated	
  communities	
  on	
  days	
  2,	
  4,	
  6,	
  8,	
  10	
  and	
  12	
  in	
  culture	
  and	
  
compared	
  these	
  to	
  samples	
  from	
  all	
  competitions	
  described	
  (except	
  one	
  replicate	
  
each	
  of	
  2015/3014	
  and	
  4010/4004	
  competitions;	
  these	
  are	
  not	
  plotted	
  due	
  to	
  
technical	
  failures	
  of	
  the	
  sample	
  analysis)	
  on	
  day	
  7	
  just	
  prior	
  to	
  the	
  addition	
  of	
  C.	
  
difficile.	
  We	
  plotted	
  the	
  number	
  of	
  OTUs	
  (A),	
  inverse	
  Simpson	
  microbial	
  diversity	
  
indicator	
  (B),	
  and	
  Simpson	
  evenness	
  indicator	
  (C)	
  in	
  the	
  clindamycin	
  (closed	
  
circles),	
  mock-­‐treated	
  (open	
  squares),	
  and	
  competition	
  (plus	
  symbols)	
  communities.	
  

	
  

139	
  



ï ï

!"#$%&'()%*



The	
  solid	
  lines	
  represent	
  the	
  mean	
  values	
  for	
  the	
  triplicate	
  clindamycin-­‐treated	
  and	
  
mock-­‐treated	
  samples	
  at	
  each	
  time	
  point,	
  which	
  are	
  reproduced	
  from	
  Fig.	
  3.4.	
  
	
  

+
-


*


,

,
+
*
 -

ï ï 
10'6$[LV

 

	
  
Figure	
  S3.5.	
  Similar	
  community	
  structure	
  changes	
  were	
  observed	
  in	
  response	
  
to	
  clindamycin-­‐treatment	
  in	
  competition	
  bioreactor	
  communities.	
  As	
  described	
  
in	
  Fig.	
  3.5,	
  we	
  plotted	
  the	
  Bray-­‐Curtis	
  dissimilarity	
  (3%	
  OTUs)	
  between	
  samples	
  
using	
  nonmetric	
  multidimensional	
  scaling	
  (NMDS).	
  Samples	
  were	
  plotted	
  from	
  
three	
  clindamycin-­‐treated	
  replicates	
  (1-­‐3;	
  closed	
  symbols)	
  and	
  three	
  mock-­‐treated	
  
replicates	
  (4-­‐6;	
  open	
  symbols)	
  every	
  two	
  days	
  from	
  day	
  2	
  (pre-­‐treatment)	
  through	
  
day	
  12.	
  The	
  numbers	
  in	
  boxes	
  indicate	
  the	
  points	
  for	
  the	
  indicated	
  reactors	
  on	
  day	
  2	
  
and	
  day	
  12	
  and	
  correspond	
  to	
  the	
  reactor	
  numbers	
  indicated	
  in	
  Fig.	
  3.2.	
  Intervening	
  
time	
  points	
  are	
  represented	
  by	
  symbols	
  and	
  are	
  connected	
  in	
  sequential	
  order	
  by	
  
lines.	
  Asterisks	
  represent	
  the	
  day	
  7	
  samples	
  from	
  the	
  competition	
  bioreactor	
  
communities	
  described	
  in	
  Fig.	
  S3.4.	
  The	
  ellipses	
  indicate	
  the	
  95%	
  confidence	
  
intervals	
  for	
  the	
  indicated	
  groups	
  (pre-­‐treatment;	
  mock-­‐treatment;	
  clindamycin-­‐
treatment.)	
  The	
  plot	
  stress	
  was	
  0.201.	
  
	
  
	
  
	
  
	
  

	
  

140	
  

108

CFU/g feces

107
106

CD3017(027)
CD1014(014)
CD4015(027)
CD2048(053)

105
104
103
102
101
100

1

2

3

4

Days post gavage

	
  
Figure	
  S3.6.	
  	
  	
  Levels	
  of	
  C.	
  difficile	
  strains	
  across	
  time	
  in	
  mouse	
  model	
  of	
  
infection	
  as	
  determined	
  by	
  plating	
  from	
  fecal	
  pellets.	
  	
  After	
  antibiotic	
  treatment,	
  
mice	
  were	
  gavaged	
  with	
  spores	
  from	
  each	
  individual	
  ribotype.	
  	
  Fecal	
  samples	
  were	
  
collected	
  daily	
  for	
  four	
  days.	
  	
  For	
  each	
  time	
  point,	
  fecal	
  samples	
  were	
  weighed,	
  
suspended	
  in	
  sterile	
  water,	
  heat	
  killed	
  for	
  30	
  min	
  at	
  65°C,	
  and	
  plated	
  on	
  BHIS	
  
containing	
  0.1%	
  taurocholic	
  acid.	
  	
  Plotted	
  here	
  are	
  the	
  mean	
  CFU/g	
  feces	
  from	
  
replicate	
  mice	
  for	
  each	
  strain	
  group	
  (with	
  standard	
  deviations	
  where	
  applicable;	
  
some	
  data	
  points	
  were	
  lost	
  due	
  to	
  technical	
  failure;	
  two	
  groups	
  lost	
  a	
  triplicate	
  
mouse	
  after	
  day	
  1).	
  	
  	
  
	
  

Competitive Index

10000
1000
100
10
1

CD3017
[CD1014(014)]

CD4015
[CD2048(053)]

	
  
	
  
Figure	
  S3.7.	
  	
  Competitive	
  indices	
  (CI)	
  of	
  two	
  competition	
  pairs	
  of	
  ribotype	
  027	
  
and	
  non-­‐027	
  C.	
  difficile	
  strains	
  in	
  the	
  MBRA	
  (circles)	
  and	
  Mouse	
  (triangles)	
  
models.	
  	
  When	
  comparing	
  the	
  relative	
  differences	
  between	
  the	
  CI’s	
  of	
  the	
  
competition	
  pairs	
  in	
  each	
  model,	
  the	
  mean	
  CI	
  of	
  the	
  CD3017/CD1014	
  competition	
  
pair	
  is	
  higher	
  than	
  the	
  CD4015/CD2048	
  competition	
  pair	
  in	
  both	
  models.	
  	
  

	
  

141	
  

REFERENCES	
  

	
  

142	
  

REFERENCES	
  
	
  
	
  
1.	
  

Yassin	
  SF,	
  Young-­‐Fadok	
  TM,	
  Zein	
  NN,	
  Pardi	
  DS.	
  2001.	
  Clostridium	
  difficile-­‐
associated	
  diarrhea	
  and	
  colitis.	
  Mayo	
  Clin.	
  Proc.	
  76:725–730.	
  

2.	
  

Bartlett	
  JG.	
  1992.	
  Antibiotic-­‐associated	
  diarrhea.	
  Clin.	
  Infect.	
  Dis.	
  15:573–
581.	
  

3.	
  

Wiström	
  J,	
  Norrby	
  SR,	
  Myhre	
  EB,	
  Eriksson	
  S,	
  Granström	
  G,	
  Lagergren	
  L,	
  
Englund	
  G,	
  Nord	
  CE,	
  Svenungsson	
  B.	
  2001.	
  Frequency	
  of	
  antibiotic-­‐
associated	
  diarrhoea	
  in	
  2462	
  antibiotic-­‐treated	
  hospitalized	
  patients:	
  a	
  
prospective	
  study.	
  J.	
  Antimicrob.	
  Chemother.	
  47:43–50.	
  

4.	
  

Wilson	
  KH,	
  Freter	
  R.	
  1986.	
  Interaction	
  of	
  Clostridium	
  difficile	
  and	
  
Escherichia	
  coli	
  with	
  microfloras	
  in	
  continuous-­‐flow	
  cultures	
  and	
  gnotobiotic	
  
mice.	
  Infection	
  and	
  Immunity	
  54:354–358.	
  

5.	
  

Britton	
  RA,	
  Young	
  VB.	
  2012.	
  Interaction	
  between	
  the	
  intestinal	
  microbiota	
  
and	
  host	
  in	
  Clostridium	
  difficile	
  colonization	
  resistance.	
  Trends	
  in	
  
Microbiology	
  20:313–319.	
  

6.	
  

Miller	
  BA,	
  Chen	
  LF,	
  Sexton	
  DJ,	
  Anderson	
  DJ.	
  2011.	
  Comparison	
  of	
  the	
  
Burdens	
  of	
  Hospital-­‐Onset,	
  Healthcare	
  Facility–Associated	
  Clostridium	
  
difficileInfection	
  and	
  of	
  Healthcare-­‐Associated	
  Infection	
  due	
  to	
  Methicillin-­‐
Resistant	
  Staphylococcus	
  aureusin	
  Community	
  Hospitals.	
  Infection	
  Control	
  
and	
  Hospital	
  Epidemiology	
  32:387–390.	
  

7.	
  

Loo	
  VG,	
  Poirier	
  L,	
  Miller	
  MA,	
  Oughton	
  M,	
  Libman	
  MD,	
  Michaud	
  S,	
  
Bourgault	
  A-­‐M,	
  Nguyen	
  T,	
  Frenette	
  C,	
  Kelly	
  M,	
  Vibien	
  A,	
  Brassard	
  P,	
  Fenn	
  
S,	
  Dewar	
  K,	
  Hudson	
  TJ,	
  Horn	
  R,	
  René	
  P,	
  Monczak	
  Y,	
  Dascal	
  A.	
  2005.	
  A	
  
predominantly	
  clonal	
  multi-­‐institutional	
  outbreak	
  of	
  Clostridium	
  difficile-­‐
associated	
  diarrhea	
  with	
  high	
  morbidity	
  and	
  mortality.	
  N.	
  Engl.	
  J.	
  Med.	
  
353:2442–2449.	
  

8.	
  

Miller	
  M,	
  Gravel	
  D,	
  Mulvey	
  M,	
  Taylor	
  G,	
  Boyd	
  D,	
  Simor	
  A,	
  Gardam	
  M,	
  
McGeer	
  A,	
  Hutchinson	
  J,	
  Moore	
  D,	
  Kelly	
  S.	
  2010.	
  Health	
  Care–Associated	
  
Clostridium	
  difficileInfection	
  in	
  Canada:	
  Patient	
  Age	
  and	
  Infecting	
  Strain	
  Type	
  
Are	
  Highly	
  Predictive	
  of	
  Severe	
  Outcome	
  and	
  Mortality.	
  Clin	
  Infect	
  Dis	
  
50:194–201.	
  

9.	
  

Pepin	
  J.	
  2005.	
  Mortality	
  attributable	
  to	
  nosocomial	
  Clostridium	
  difficile-­‐
associated	
  disease	
  during	
  an	
  epidemic	
  caused	
  by	
  a	
  hypervirulent	
  strain	
  in	
  
Quebec.	
  Canadian	
  Medical	
  Association	
  Journal	
  173:1037–1042.	
  

10.	
  

Hubert	
  B,	
  Loo	
  VG,	
  Bourgault	
  A-­‐M,	
  Poirier	
  L,	
  Dascal	
  A,	
  Fortin	
  E,	
  Dionne	
  M,	
  
Lorange	
  M.	
  2007.	
  A	
  portrait	
  of	
  the	
  geographic	
  dissemination	
  of	
  the	
  

	
  

143	
  

Clostridium	
  difficile	
  North	
  American	
  pulsed-­‐field	
  type	
  1	
  strain	
  and	
  the	
  
epidemiology	
  of	
  C.	
  difficile-­‐associated	
  disease	
  in	
  Québec.	
  Clin	
  Infect	
  Dis	
  
44:238–244.	
  
11.	
  

Bartlett	
  JG.	
  2006.	
  Narrative	
  review:	
  the	
  new	
  epidemic	
  of	
  Clostridium	
  difficile-­‐
associated	
  enteric	
  disease.	
  Ann.	
  Intern.	
  Med.	
  145:758–764.	
  

12.	
  

Kuijper	
  EJ,	
  van	
  Dissel	
  JT.	
  2008.	
  Spectrum	
  of	
  Clostridium	
  difficile	
  infections	
  
outside	
  health	
  care	
  facilities.	
  Canadian	
  Medical	
  Association	
  Journal	
  179:747–
748.	
  

13.	
  

McDonald	
  LC,	
  Killgore	
  GE,	
  Thompson	
  A,	
  Owens	
  RC,	
  Kazakova	
  SV,	
  Sambol	
  
SP,	
  Johnson	
  S,	
  Gerding	
  DN.	
  2005.	
  An	
  epidemic,	
  toxin	
  gene-­‐variant	
  strain	
  of	
  
Clostridium	
  difficile.	
  N.	
  Engl.	
  J.	
  Med.	
  353:2433–2441.	
  

14.	
  

Morgan	
  OW,	
  Rodrigues	
  B,	
  Elston	
  T,	
  Verlander	
  NQ,	
  Brown	
  DFJ,	
  Brazier	
  J,	
  
Reacher	
  M.	
  2008.	
  Clinical	
  Severity	
  of	
  Clostridium	
  difficile	
  PCR	
  Ribotype	
  027:	
  
A	
  Case-­‐Case	
  Study.	
  PLoS	
  One	
  3:e1812.	
  

15.	
  

Sirard	
  S,	
  Valiquette	
  L,	
  Fortier	
  LC.	
  2011.	
  Lack	
  of	
  Association	
  between	
  Clinical	
  
Outcome	
  of	
  Clostridium	
  difficile	
  Infections,	
  Strain	
  Type,	
  and	
  Virulence-­‐
Associated	
  Phenotypes.	
  Journal	
  of	
  Clinical	
  Microbiology	
  49:4040–4046.	
  

16.	
  

Cloud	
  J,	
  Noddin	
  L,	
  Pressman	
  A,	
  Hu	
  M,	
  Kelly	
  C.	
  2009.	
  Clostridium	
  difficile	
  
strain	
  NAP-­‐1	
  is	
  not	
  associated	
  with	
  severe	
  disease	
  in	
  a	
  nonepidemic	
  setting.	
  	
  
Clin.	
  Gastro.	
  Hep.	
  7:868–873.	
  

17.	
  

Walk	
  ST,	
  Micic	
  D,	
  Jain	
  R,	
  Lo	
  ES,	
  Trivedi	
  I,	
  Liu	
  EW,	
  Almassalha	
  LM,	
  Ewing	
  
SA,	
  Ring	
  C,	
  Galecki	
  AT.	
  2012.	
  Clostridium	
  difficile	
  ribotype	
  does	
  not	
  predict	
  
severe	
  infection.	
  Clin	
  Infect	
  Dis.	
  55:1661-­‐1668.	
  

18.	
  

Walker	
  AS,	
  Eyre	
  DW,	
  Wyllie	
  DH,	
  Dingle	
  KE,	
  Griffiths	
  D,	
  Shine	
  B,	
  Oakley	
  S,	
  
O'Connor	
  L,	
  Finney	
  J,	
  Vaughan	
  A,	
  Crook	
  DW,	
  Wilcox	
  MH,	
  Peto	
  TEA.	
  2013.	
  
Relationship	
  Between	
  Bacterial	
  Strain	
  Type,	
  Host	
  Biomarkers,	
  and	
  Mortality	
  
in	
  Clostridium	
  difficile	
  Infection.	
  Clin	
  Infect	
  Dis.	
  56:1589-­‐1600.	
  

19.	
  

Hensgens	
  MPM,	
  Goorhuis	
  A,	
  Dekkers	
  OM,	
  van	
  Benthem	
  BHB,	
  Kuijper	
  EJ.	
  
2013.	
  All-­‐Cause	
  and	
  Disease-­‐Specific	
  Mortality	
  in	
  Hospitalized	
  Patients	
  With	
  
Clostridium	
  difficile	
  Infection:	
  A	
  Multicenter	
  Cohort	
  Study.	
  Clin	
  Infect	
  Dis	
  
56:1108–1116.	
  

20.	
  

Waslawski	
  S,	
  Lo	
  ES,	
  Ewing	
  SA,	
  Young	
  VB,	
  Aronoff	
  DM,	
  Sharp	
  SE,	
  Novak-­‐
Weekley	
  SM,	
  Crist	
  AE,	
  Dunne	
  WM,	
  Hoppe-­‐Bauer	
  J,	
  Johnson	
  M,	
  Brecher	
  
SM,	
  Newton	
  DW,	
  Walk	
  ST.	
  2013.	
  Clostridium	
  difficile	
  ribotype	
  diversity	
  at	
  six	
  
health	
  care	
  institutions	
  in	
  the	
  United	
  States.	
  Journal	
  of	
  Clinical	
  Microbiology	
  
51:1938–1941.	
  

	
  

144	
  

21.	
  

He	
  M,	
  Miyajima	
  F,	
  Roberts	
  P,	
  Ellison	
  L,	
  Pickard	
  DJ,	
  Martin	
  MJ,	
  Connor	
  TR,	
  
Harris	
  SR,	
  Fairley	
  D,	
  Bamford	
  KB,	
  D'Arc	
  S,	
  Brazier	
  J,	
  Brown	
  D,	
  Coia	
  JE,	
  
Douce	
  G,	
  Gerding	
  D,	
  Kim	
  HJ,	
  Koh	
  TH,	
  Kato	
  H,	
  Senoh	
  M,	
  Louie	
  T,	
  Michell	
  S,	
  
Butt	
  E,	
  Peacock	
  SJ,	
  Brown	
  NM,	
  Riley	
  T,	
  Songer	
  G,	
  Wilcox	
  M,	
  Pirmohamed	
  
M,	
  Kuijper	
  E,	
  Hawkey	
  P,	
  Wren	
  BW,	
  Dougan	
  G,	
  Parkhill	
  J,	
  Lawley	
  TD.	
  2013.	
  
Emergence	
  and	
  global	
  spread	
  of	
  epidemic	
  healthcare-­‐associated	
  Clostridium	
  
difficile.	
  Nat	
  Genet	
  45:109–113.	
  

22.	
  

Freeman	
  J.	
  2003.	
  Effects	
  of	
  cefotaxime	
  and	
  desacetylcefotaxime	
  upon	
  
Clostridium	
  difficile	
  proliferation	
  and	
  toxin	
  production	
  in	
  a	
  triple-­‐stage	
  
chemostat	
  model	
  of	
  the	
  human	
  gut.	
  Journal	
  of	
  Antimicrobial	
  Chemotherapy	
  
52:96–102.	
  

23.	
  

Honda	
  H,	
  Hoyles	
  L,	
  Gibson	
  GR,	
  Farmer	
  S,	
  Keller	
  D,	
  McCartney	
  AL.	
  2011.	
  
Impact	
  of	
  GanedenBC30	
  (Bacillus	
  coagulans	
  GBI-­‐30,	
  6086)	
  on	
  population	
  
dynamics	
  of	
  the	
  human	
  gut	
  microbiota	
  in	
  a	
  continuous	
  culture	
  fermentation	
  
system.	
  Anaerobe.	
  17:36-­‐42.	
  

24.	
  

Baines	
  SD,	
  Crowther	
  GS,	
  Todhunter	
  SL,	
  Freeman	
  J,	
  Chilton	
  CH,	
  Fawley	
  
WN,	
  Wilcox	
  MH.	
  2013.	
  Mixed	
  infection	
  by	
  Clostridium	
  difficile	
  in	
  an	
  in	
  vitro	
  
model	
  of	
  the	
  human	
  gut.	
  Journal	
  of	
  Antimicrobial	
  Chemotherapy	
  68:1139–
1143.	
  

25.	
  

Sorg	
  JA,	
  Dineen	
  SS.	
  2005.	
  Laboratory	
  Maintenance	
  of	
  Clostridium	
  difficile.	
  
John	
  Wiley	
  &	
  Sons,	
  Inc.,	
  Hoboken,	
  NJ,	
  USA.	
  

26.	
  

Antonopoulos	
  DA,	
  Huse	
  SM,	
  Morrison	
  HG,	
  Schmidt	
  TM,	
  Sogin	
  ML,	
  Young	
  
VB.	
  2009.	
  Reproducible	
  Community	
  Dynamics	
  of	
  the	
  Gastrointestinal	
  
Microbiota	
  following	
  Antibiotic	
  Perturbation.	
  Infection	
  and	
  Immunity	
  
77:2367–2375.	
  

27.	
  

Consortium	
  THMP.	
  2012.	
  Structure,	
  function	
  and	
  diversity	
  of	
  the	
  healthy	
  
human	
  microbiome.	
  Nature	
  486:207–214.	
  

28.	
  

Schloss	
  PD.	
  2009.	
  A	
  High-­‐Throughput	
  DNA	
  Sequence	
  Aligner	
  for	
  Microbial	
  
Ecology	
  Studies.	
  PLoS	
  ONE	
  4:e8230.	
  

29.	
  

Schloss	
  PD,	
  Gevers	
  D,	
  Westcott	
  SL.	
  2011.	
  Reducing	
  the	
  Effects	
  of	
  PCR	
  
Amplification	
  and	
  Sequencing	
  Artifacts	
  on	
  16S	
  rRNA-­‐Based	
  Studies.	
  PLoS	
  ONE	
  
6:e27310.	
  

30.	
  

Edgar	
  RC,	
  Haas	
  BJ,	
  Clemente	
  JC,	
  Quince	
  C,	
  Knight	
  R.	
  2011.	
  UCHIME	
  
improves	
  sensitivity	
  and	
  speed	
  of	
  chimera	
  detection.	
  Bioinformatics	
  
27:2194–2200.	
  

31.	
  

Schloss	
  PD,	
  Westcott	
  SL.	
  2011.	
  Assessing	
  and	
  improving	
  methods	
  used	
  in	
  
operational	
  taxonomic	
  unit-­‐based	
  approaches	
  for	
  16S	
  rRNA	
  gene	
  sequence	
  

	
  

145	
  

analysis.	
  Applied	
  and	
  Environmental	
  Microbiology	
  77:3219–3226.	
  
32.	
  

White	
  JR,	
  Nagarajan	
  N,	
  Pop	
  M.	
  2009.	
  Statistical	
  Methods	
  for	
  Detecting	
  
Differentially	
  Abundant	
  Features	
  in	
  Clinical	
  Metagenomic	
  Samples	
  
5:e1000352.	
  

33.	
  

Oksanen	
  J,	
  Blanchet	
  FG,	
  Kindt	
  R,	
  Legendre	
  P,	
  Minchin	
  PR,	
  O'Hara	
  RB,	
  
Simpson	
  GL,	
  Solymos	
  P,	
  Stevens	
  MHH,	
  Wagner	
  H.	
  2013.	
  vegan:	
  Community	
  
Ecology	
  Package.	
  R	
  package	
  version	
  2.0-­‐9.	
  

34.	
  

Knetsch	
  CW,	
  Hensgens	
  MPM,	
  Harmanus	
  C,	
  van	
  der	
  Bijl	
  MW,	
  Savelkoul	
  
PHM,	
  Kuijper	
  EJ,	
  Corver	
  J,	
  van	
  Leeuwen	
  HC.	
  2011.	
  Genetic	
  markers	
  for	
  
Clostridium	
  difficile	
  lineages	
  linked	
  to	
  hypervirulence.	
  Microbiology	
  
157:3113–3123.	
  

35.	
  

Pfaffl	
  MW.	
  2001.	
  A	
  new	
  mathematical	
  model	
  for	
  relative	
  quantification	
  in	
  
real-­‐time	
  RT-­‐PCR.	
  Nucleic	
  Acids	
  Research	
  29:e45.	
  

36.	
  

Chen	
  X,	
  Katchar	
  K,	
  Goldsmith	
  JD,	
  Nanthakumar	
  N,	
  Cheknis	
  A,	
  Gerding	
  DN,	
  
Kelly	
  CP.	
  2008.	
  A	
  mouse	
  model	
  of	
  Clostridium	
  difficile-­‐associated	
  disease.	
  
Gastroenterology	
  135:1984–1992.	
  

37.	
  

Bartlett	
  JG.	
  1981.	
  Antimicrobial	
  agents	
  implicated	
  in	
  Clostridium	
  difficile	
  
toxin-­‐associated	
  diarrhea	
  of	
  colitis.	
  Johns	
  Hopkins	
  Med	
  J	
  149:6–9.	
  

38.	
  

Jernberg	
  C,	
  Löfmark	
  S,	
  Edlund	
  C,	
  Jansson	
  JK.	
  2010.	
  Long-­‐term	
  impacts	
  of	
  
antibiotic	
  exposure	
  on	
  the	
  human	
  intestinal	
  microbiota.	
  Microbiology	
  
156:3216–3223.	
  

39.	
  

CLARKE	
  KR.	
  1993.	
  Non-­‐parametric	
  multivariate	
  analyses	
  of	
  changes	
  in	
  
community	
  structure	
  18:117–143.	
  

40.	
  

Reeves	
  AE,	
  Theriot	
  CM,	
  Bergin	
  IL,	
  Huffnagle	
  GB,	
  Schloss	
  PD,	
  Young	
  VB.	
  
2011.	
  The	
  interplay	
  between	
  microbiome	
  dynamics	
  and	
  pathogen	
  dynamics	
  
in	
  a	
  murine	
  model	
  of	
  Clostridium	
  difficile	
  Infection.	
  gutmicrobes	
  2:145–158.	
  

41.	
  

Antharam	
  VC,	
  Li	
  EC,	
  Ishmael	
  A,	
  Sharma	
  A,	
  Mai	
  V,	
  Rand	
  KH,	
  Wang	
  GP.	
  2013.	
  
Intestinal	
  dysbiosis	
  and	
  depletion	
  of	
  butyrogenic	
  bacteria	
  in	
  Clostridium	
  
difficile	
  infection	
  and	
  nosocomial	
  diarrhea.	
  Journal	
  of	
  Clinical	
  Microbiology	
  
51:2884–2892.	
  

42.	
  

Merrigan	
  M,	
  Venugopal	
  A,	
  Mallozzi	
  M,	
  Roxas	
  B,	
  Viswanathan	
  VK,	
  Johnson	
  
S,	
  Gerding	
  DN,	
  Vedantam	
  G.	
  2010.	
  Human	
  Hypervirulent	
  Clostridium	
  difficile	
  
Strains	
  Exhibit	
  Increased	
  Sporulation	
  as	
  Well	
  as	
  Robust	
  Toxin	
  Production.	
  
Journal	
  of	
  Bacteriology	
  192:4904–4911.	
  

43.	
  

Akerlund	
  T,	
  Persson	
  I,	
  Unemo	
  M,	
  Noren	
  T,	
  Svenungsson	
  B,	
  Wullt	
  M,	
  

	
  

146	
  

Burman	
  LG.	
  2008.	
  Increased	
  Sporulation	
  Rate	
  of	
  Epidemic	
  Clostridium	
  
difficile	
  Type	
  027/NAP1.	
  Journal	
  of	
  Clinical	
  Microbiology	
  46:1530–1533.	
  
44.	
  

Burns	
  DA,	
  Heeg	
  D,	
  Cartman	
  ST,	
  Minton	
  NP.	
  2011.	
  Reconsidering	
  the	
  
sporulation	
  characteristics	
  of	
  hypervirulent	
  Clostridium	
  difficile	
  
BI/NAP1/027.	
  PLoS	
  ONE	
  6:e24894.	
  

45.	
  

Carlson	
  PE,	
  Walk	
  ST,	
  Bourgis	
  AET,	
  Liu	
  MW,	
  Kopliku	
  F,	
  Lo	
  E,	
  Young	
  VB,	
  
Aronoff	
  DM,	
  Hanna	
  PC.	
  2013.	
  The	
  relationship	
  between	
  phenotype,	
  ribotype,	
  
and	
  clinical	
  disease	
  in	
  human	
  Clostridium	
  difficile	
  isolates.	
  Anaerobe	
  24:109–
116.	
  

46.	
  

Francis	
  MB,	
  Allen	
  CA,	
  Sorg	
  JA.	
  2013.	
  Muricholic	
  Acids	
  Inhibit	
  Clostridium	
  
difficile	
  Spore	
  Germination	
  and	
  Growth.	
  PLoS	
  ONE	
  8:e73653.	
  

47.	
  

Heeg	
  D,	
  Burns	
  DA,	
  Cartman	
  ST,	
  Minton	
  NP.	
  2012.	
  Spores	
  of	
  Clostridium	
  
difficile	
  Clinical	
  Isolates	
  Display	
  a	
  Diverse	
  Germination	
  Response	
  to	
  Bile	
  Salts.	
  
PLoS	
  ONE	
  7:e32381.	
  

48.	
  

Escartin	
  F,	
  Skouloubris	
  S,	
  Liebl	
  U,	
  Myllykallio	
  H.	
  2008.	
  Flavin-­‐dependent	
  
thymidylate	
  synthase	
  X	
  limits	
  chromosomal	
  DNA	
  replication.	
  Proceedings	
  of	
  
the	
  National	
  Academy	
  of	
  Sciences	
  105:9948–9952.	
  

49.	
  

Van	
  den	
  Abbeele	
  P,	
  Grootaert	
  C,	
  Marzorati	
  M,	
  Possemiers	
  S,	
  Verstraete	
  
W,	
  Gerard	
  P,	
  Rabot	
  S,	
  Bruneau	
  A,	
  Aidy	
  El	
  S,	
  Derrien	
  M,	
  Zoetendal	
  E,	
  
Kleerebezem	
  M,	
  Smidt	
  H,	
  Van	
  de	
  Wiele	
  T.	
  2010.	
  Microbial	
  Community	
  
Development	
  in	
  a	
  Dynamic	
  Gut	
  Model	
  Is	
  Reproducible,	
  Colon	
  Region	
  Specific,	
  
and	
  Selective	
  for	
  Bacteroidetes	
  and	
  Clostridium	
  Cluster	
  IX.	
  Applied	
  and	
  
Environmental	
  Microbiology	
  76:5237–5246.	
  

50.	
  

Rajilic-­‐Stojanovic	
  M,	
  Maathuis	
  A,	
  Heilig	
  HGHJ,	
  Venema	
  K,	
  de	
  Vos	
  WM,	
  
Smidt	
  H.	
  2010.	
  Evaluating	
  the	
  microbial	
  diversity	
  of	
  an	
  in	
  vitro	
  model	
  of	
  the	
  
human	
  large	
  intestine	
  by	
  phylogenetic	
  microarray	
  analysis.	
  Microbiology	
  
156:3270–3281.	
  

51.	
  

Macfarlane	
  G,	
  Macfarlane	
  S,	
  Gibson	
  G.	
  1998.	
  Validation	
  of	
  a	
  Three-­‐Stage	
  
Compound	
  Continuous	
  Culture	
  System	
  for	
  Investigating	
  the	
  Effect	
  of	
  
Retention	
  Time	
  on	
  the	
  Ecology	
  and	
  Metabolism	
  of	
  Bacteria	
  in	
  the	
  Human	
  
Colon.	
  Microbial	
  Ecology	
  35:180–187.	
  

52.	
  

Payne	
  AN,	
  Zihler	
  A,	
  Chassard	
  C,	
  Lacroix	
  C.	
  2012.	
  Advances	
  and	
  perspectives	
  
in	
  in	
  vitro	
  human	
  gut	
  fermentation	
  modeling.	
  Trends	
  in	
  Biotechnology	
  30:17–
25.	
  

53.	
  

Eyre	
  DW,	
  Cule	
  ML,	
  Griffiths	
  D,	
  Crook	
  DW,	
  Peto	
  TEA,	
  Walker	
  AS,	
  Wilson	
  DJ.	
  
2013.	
  Detection	
  of	
  Mixed	
  Infection	
  from	
  Bacterial	
  Whole	
  Genome	
  Sequence	
  
Data	
  Allows	
  Assessment	
  of	
  Its	
  Role	
  in	
  Clostridium	
  difficile	
  Transmission.	
  PLoS	
  

	
  

147	
  

Comp	
  Biol	
  9:e1003059.	
  
54.	
  

Behroozian	
  AA,	
  Chludzinski	
  JP,	
  Lo	
  ES,	
  Ewing	
  SA,	
  Waslawski	
  S,	
  Newton	
  
DW,	
  Young	
  VB,	
  Aronoff	
  DM,	
  Walk	
  ST.	
  2013.	
  Detection	
  of	
  Mixed	
  Populations	
  
of	
  Clostridium	
  difficilefrom	
  Symptomatic	
  Patients	
  Using	
  Capillary-­‐Based	
  
Polymerase	
  Chain	
  Reaction	
  Ribotyping.	
  Infection	
  Control	
  and	
  Hospital	
  
Epidemiology	
  34:961–966.	
  

55.	
  

Fierer	
  N,	
  Jackson	
  JA,	
  Vilgalys	
  R,	
  Jackson	
  RB.	
  2005.	
  Assessment	
  of	
  soil	
  
microbial	
  community	
  structure	
  by	
  use	
  of	
  taxon-­‐specific	
  quantitative	
  PCR	
  
assays.	
  Appl	
  Microbiol	
  71:4117–4120.	
  

	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

148	
  

CHAPTER	
  4	
  
	
  
Differential	
  Metabolism	
  of	
  Trehalose	
  by	
  Epidemic	
  Ribotypes	
  of	
  Clostridum	
  
difficile	
  	
  
	
  
	
  
	
  
Introduction	
  
Clostridium	
  difficile	
  is	
  one	
  of	
  the	
  primary	
  causative	
  agents	
  of	
  antibiotic-­‐associated	
  
diarrhea	
  (1).	
  	
  	
  While	
  Bartlett	
  et	
  al.	
  first	
  described	
  this	
  association	
  almost	
  four	
  
decades	
  ago	
  in	
  1978,	
  it	
  is	
  still	
  poorly	
  understood	
  why	
  disease	
  generally	
  only	
  occurs	
  
following	
  antibiotic	
  treatment	
  (2).	
  	
  Several	
  factors	
  from	
  the	
  perspectives	
  of	
  the	
  host	
  
and	
  microbiota	
  likely	
  contribute	
  to	
  the	
  complex	
  dynamics	
  of	
  both	
  resistance	
  and	
  
susceptibility	
  to	
  C.	
  difficile	
  infection	
  (CDI).	
  	
  However,	
  mechanisms	
  of	
  colonization	
  
resistance	
  and	
  competitive	
  inhibition	
  are	
  believed	
  to	
  play	
  a	
  role	
  (reviewed	
  in	
  (3)).	
  	
  	
  
Competition	
  for	
  nutrients	
  is	
  one	
  of	
  the	
  mechanisms	
  governing	
  colonization	
  
resistance,	
  and	
  although	
  evidence	
  to	
  support	
  this	
  in	
  the	
  case	
  of	
  establishment	
  of	
  CDI	
  
exists,	
  we	
  do	
  not	
  have	
  a	
  complete	
  understanding	
  of	
  the	
  role	
  of	
  nutrient	
  availability	
  
(4-­‐6).	
  	
  Therefore,	
  there	
  is	
  still	
  much	
  to	
  be	
  learned	
  in	
  terms	
  of	
  how	
  C.	
  difficile	
  
metabolism	
  and	
  competition	
  for	
  nutrients	
  contribute	
  to	
  colonization	
  and	
  
establishment	
  of	
  disease.	
  	
  Moreover,	
  recent	
  increases	
  in	
  the	
  incidence	
  and	
  severity	
  
of	
  CDI	
  correlate	
  with	
  the	
  emergence	
  of	
  epidemic-­‐associated	
  lineages	
  of	
  C.	
  difficile,	
  
primarily	
  ribotype	
  027	
  	
  (RT	
  027)	
  (7-­‐10).	
  	
  	
  In	
  chapter	
  3	
  we	
  show	
  that	
  RT	
  027	
  strains	
  
are	
  able	
  to	
  out-­‐compete	
  strains	
  of	
  other	
  ribotypes	
  in	
  the	
  presence	
  of	
  complex	
  fecal	
  

	
  

149	
  

communities	
  in	
  an	
  in	
  vitro	
  model	
  of	
  C.	
  difficile	
  invasion.	
  	
  	
  Furthermore,	
  we	
  speculate	
  
that	
  there	
  are	
  aspects	
  of	
  RT	
  027	
  strain	
  physiology	
  related	
  to	
  nutrient	
  utilization	
  that	
  
play	
  a	
  role	
  in	
  this	
  increased	
  competitive	
  fitness.	
  	
  	
  
	
  

In	
  this	
  chapter,	
  I	
  will	
  present	
  work	
  that	
  was	
  performed	
  to	
  investigate	
  carbon	
  

source	
  utilization	
  of	
  C.	
  difficile	
  with	
  particular	
  emphasis	
  on	
  identifying	
  carbon	
  
sources	
  that	
  RT	
  027	
  strains	
  differentially	
  metabolize.	
  	
  The	
  primary	
  nutrients	
  and	
  
carbon	
  sources	
  utilized	
  by	
  C.	
  difficile	
  during	
  outgrowth	
  in	
  the	
  gastrointestinal	
  tract	
  
are	
  not	
  well	
  explored.	
  	
  Identifying	
  compounds	
  that	
  support	
  C.	
  difficile	
  growth	
  in	
  the	
  
context	
  of	
  the	
  intestinal	
  environment	
  would	
  increase	
  our	
  understanding	
  in	
  this	
  area	
  
and	
  provide	
  insight	
  into	
  knowing	
  which	
  nutritional	
  factors	
  are	
  important	
  for	
  C.	
  
difficile	
  gut	
  colonization.	
  	
  	
  Additionally,	
  elucidating	
  if	
  RT	
  027	
  strains	
  have	
  the	
  ability	
  
to	
  utilize	
  a	
  wider	
  range	
  of	
  nutrients,	
  or	
  specific	
  nutrients	
  more	
  efficiently,	
  might	
  
help	
  to	
  explain	
  the	
  competitive	
  advantage	
  of	
  RT	
  027	
  strains.	
  	
  	
  Moreover,	
  it	
  would	
  
provide	
  further	
  evidence	
  to	
  support	
  the	
  hypothesis	
  that	
  these	
  strains	
  have	
  become	
  
more	
  prevalent	
  due	
  to	
  differences	
  in	
  physiology	
  rather	
  than,	
  or	
  in	
  addition	
  to,	
  
differences	
  related	
  to	
  traditional	
  virulence	
  factors.	
  	
  	
  	
  
	
  

Most	
  of	
  the	
  experimental	
  work	
  in	
  this	
  chapter	
  focuses	
  on	
  C.	
  difficile	
  

metabolism	
  of	
  a	
  glucose	
  disaccharide	
  sugar,	
  trehalose.	
  	
  Trehalose	
  is	
  a	
  naturally	
  
occurring	
  sugar	
  consisting	
  of	
  two	
  glucose	
  subunits	
  linked	
  by	
  an	
  α,α-­‐1,1-­‐glycosidic	
  
bond	
  (11).	
  	
  Trehalose	
  synthesis	
  occurs	
  in	
  organisms	
  in	
  all	
  three	
  domains	
  of	
  life	
  (12).	
  	
  
Initially,	
  trehalose	
  was	
  thought	
  to	
  primarily	
  serve	
  as	
  an	
  energy	
  storage	
  molecule,	
  
however,	
  many	
  more	
  biological	
  functions	
  of	
  this	
  molecule	
  have	
  since	
  been	
  
discovered	
  (reviewed	
  in	
  (11,	
  13)).	
  	
  	
  These	
  functions	
  include	
  incorporation	
  in	
  cellular	
  

	
  

150	
  

structural	
  components,	
  transport,	
  intracellular	
  signaling,	
  and	
  regulation	
  (11,	
  13).	
  	
  
Furthermore,	
  because	
  of	
  its	
  particular	
  chemical	
  characteristics,	
  intracellular	
  
accumulation	
  of	
  trehalose	
  provides	
  stabilization	
  of	
  biological	
  molecules,	
  providing	
  
protection	
  against	
  osmotic,	
  heat,	
  cold,	
  and	
  oxidative	
  stresses	
  (14-­‐17).	
  	
  	
  	
  
	
  

Availability	
  of	
  trehalose	
  in	
  the	
  gastrointestinal	
  tract	
  has	
  not	
  been	
  well	
  

defined	
  and	
  it	
  is	
  not	
  included	
  in	
  studies	
  quantifying	
  other	
  intestinally	
  relevant	
  
sugars	
  such	
  as	
  glucose,	
  sucrose,	
  maltose,	
  and	
  lactose.	
  	
  Nonetheless,	
  there	
  are	
  several	
  
pieces	
  of	
  evidence	
  supporting	
  its	
  presence	
  and	
  potential	
  role	
  in	
  colonization.	
  	
  	
  First,	
  
brush	
  border	
  enterocytes	
  of	
  the	
  small	
  intestine	
  express	
  trehalase,	
  the	
  enzyme	
  
required	
  to	
  breakdown	
  trehalose	
  into	
  its	
  glucose	
  subunits	
  which	
  can	
  then	
  be	
  
absorbed	
  and	
  metabolized	
  (18).	
  	
  Historically,	
  the	
  primary	
  dietary	
  sources	
  of	
  
trehalose	
  were	
  mushrooms,	
  insects,	
  lobsters	
  and	
  crabs,	
  honey,	
  and	
  baker’s	
  yeast	
  
(13).	
  	
  More	
  recently,	
  consumption	
  of	
  trehalose	
  is	
  increasing	
  due	
  to	
  increased	
  use	
  in	
  
commercial	
  food	
  production	
  since	
  being	
  granted	
  GRAS	
  (Generally	
  Recognized	
  As	
  
Safe)	
  status	
  by	
  the	
  US	
  FDA	
  in	
  2000.	
  	
  Furthermore,	
  other	
  countries/regions	
  have	
  
approved	
  its	
  use	
  as	
  a	
  food	
  additive	
  including	
  Korea	
  and	
  Taiwan	
  in	
  1998,	
  the	
  UK	
  in	
  
1991,	
  Canada	
  in	
  2005,	
  and	
  Europe	
  in	
  2001.	
  	
  	
  An	
  extensive	
  report	
  published	
  by	
  
Hayashibara	
  International	
  Inc.	
  addressing	
  the	
  safety	
  of	
  trehalose	
  based	
  on	
  several	
  
toxicity	
  studies	
  prompted	
  this	
  approval	
  in	
  several	
  countries	
  (13).	
  	
  Trehalose	
  has	
  
unique	
  chemical	
  properties	
  that	
  make	
  it	
  useful	
  in	
  the	
  food	
  industry,	
  including	
  use	
  as	
  
a	
  stabilizing	
  agent	
  and	
  texturizer	
  in	
  various	
  foods,	
  cryoprotectant	
  in	
  freeze	
  dried	
  
foods,	
  and	
  a	
  sweetening	
  agent.	
  	
  Indeed,	
  estimates	
  regarding	
  consumption	
  of	
  
trehalose	
  by	
  the	
  average	
  adult	
  in	
  the	
  US	
  are	
  as	
  high	
  as	
  16	
  g/day	
  (13).	
  	
  	
  In	
  addition,	
  a	
  

	
  

151	
  

rare	
  condition	
  of	
  trehalose	
  intolerance	
  in	
  some	
  individuals	
  has	
  been	
  reported,	
  which	
  
is	
  caused	
  by	
  insufficient	
  trehalase	
  activity	
  in	
  the	
  small	
  intestine,	
  and	
  results	
  in	
  
intestinal	
  symptoms	
  similar	
  to	
  those	
  of	
  lactose	
  intolerance	
  upon	
  ingestion	
  of	
  
trehalose-­‐containing	
  foods	
  (19,	
  20).	
  	
  
	
  

A	
  recent	
  proteomic	
  analysis	
  of	
  C.	
  difficile	
  proteins	
  expressed	
  in	
  a	
  pig	
  ileal-­‐

ligated	
  loop	
  model	
  reported	
  a	
  2-­‐fold	
  increase	
  in	
  TreA	
  expression	
  in	
  vivo	
  compared	
  
to	
  in	
  vitro,	
  supporting	
  the	
  presence	
  of	
  trehalose	
  in	
  the	
  intestinal	
  environment	
  (21).	
  	
  
Moreover,	
  the	
  role	
  of	
  trehalose	
  in	
  colonization	
  of	
  other	
  intestinal	
  pathogens	
  has	
  
been	
  shown	
  (22).	
  	
  Martindale	
  et	
  al.	
  used	
  an	
  infant	
  rat	
  gut	
  model	
  to	
  investigate	
  genes	
  
important	
  for	
  intestinal	
  colonization	
  of	
  Escherichia	
  coli	
  (23).	
  	
  They	
  identified	
  a	
  treB	
  
mutant	
  of	
  Escherichia	
  coli	
  that	
  was	
  defective	
  in	
  colonization;	
  TreB	
  is	
  a	
  trehalose-­‐
specific	
  component	
  of	
  the	
  phosphotransferase	
  system	
  required	
  for	
  uptake	
  of	
  
trehalose.	
  	
  	
  
	
  

Finally,	
  as	
  discussed	
  above,	
  many	
  bacterial	
  organisms	
  are	
  able	
  to	
  synthesize	
  

and	
  maintain	
  intracellular	
  stores	
  of	
  trehalose.	
  	
  Indeed,	
  if	
  any	
  members	
  of	
  the	
  
intestinal	
  microbiota	
  had	
  intracellular	
  stores	
  of	
  trehalose,	
  this	
  would	
  present	
  a	
  
source	
  of	
  trehalose	
  upon	
  microbial	
  lysis	
  after	
  antibiotic	
  treatment.	
  	
  	
  Taken	
  together,	
  
all	
  of	
  this	
  evidence	
  supports	
  that	
  trehalose	
  is	
  a	
  relevant	
  nutrient	
  in	
  the	
  
gastrointestinal	
  environment	
  and	
  has	
  the	
  potential	
  to	
  play	
  a	
  role	
  in	
  C.	
  difficile	
  
colonization.	
  	
  
	
  

Very	
  little	
  work	
  has	
  been	
  done	
  regarding	
  metabolism	
  of	
  trehalose	
  by	
  C.	
  

difficile.	
  	
  Genome	
  annotation	
  data	
  suggest	
  that	
  C.	
  difficile	
  solely	
  uses	
  trehalose	
  as	
  a	
  
carbon	
  source,	
  and	
  that	
  the	
  genes	
  required	
  to	
  synthesize	
  trehalose	
  are	
  absent.	
  	
  It	
  

	
  

152	
  

only	
  harbors	
  treA,	
  the	
  gene	
  that	
  codes	
  for	
  the	
  trehalase	
  enzyme,	
  and	
  treR,	
  the	
  treA	
  
repressor.	
  	
  It	
  is	
  not	
  surprising	
  that	
  it	
  does	
  not	
  appear	
  to	
  be	
  able	
  to	
  also	
  synthesize	
  
trehalose	
  for	
  use	
  in	
  energy	
  storage	
  or	
  stress	
  resistance,	
  as	
  this	
  functional	
  capacity	
  
occurs	
  primarily	
  in	
  Gram-­‐negative	
  bacteria	
  (12).	
  	
  In	
  terms	
  of	
  regulation	
  of	
  trehalose	
  
utilization	
  genes,	
  however,	
  Antunes	
  et	
  al.	
  used	
  in	
  silico	
  analysis	
  to	
  identify	
  predicted	
  
CcpA-­‐dependent	
  carbon	
  catabolite-­‐regulated	
  genes	
  in	
  C.	
  difficile,	
  including	
  treR	
  (24).	
  
	
  

Here	
  we	
  show	
  that	
  RT	
  027	
  strains,	
  and	
  strains	
  of	
  another	
  highly	
  prevalent	
  

epidemic-­‐associated	
  ribotype	
  (RT	
  078),	
  grow	
  to	
  higher	
  cell	
  densities	
  on	
  trehalose.	
  	
  
In	
  the	
  case	
  of	
  RT	
  027	
  this	
  is	
  potentially	
  due	
  to	
  an	
  identified	
  amino	
  acid	
  substitution	
  
in	
  TreR	
  within	
  these	
  strains.	
  	
  This	
  is	
  supported	
  by	
  the	
  observation	
  that	
  RT	
  027	
  
strains	
  have	
  increased	
  treA	
  expression.	
  	
  In	
  addition,	
  a	
  treA	
  knockout	
  mutant	
  in	
  
CD630	
  was	
  generated	
  and	
  preliminary	
  mouse	
  studies	
  using	
  the	
  wild-­‐type	
  and	
  
mutant	
  strains	
  suggest	
  that	
  trehalose	
  plays	
  a	
  role	
  in	
  colonization	
  and	
  competitive	
  
fitness.	
  	
  Should	
  this	
  result	
  translate	
  into	
  the	
  human	
  gastrointestinal	
  tract,	
  it	
  is	
  
possible	
  that	
  the	
  ability	
  of	
  RT	
  027	
  strains	
  to	
  grow	
  better	
  on	
  trehalose	
  would	
  give	
  
them	
  a	
  competitive	
  advantage.	
  	
  Furthermore,	
  this	
  physiological	
  trait	
  could	
  partially	
  
explain	
  the	
  prevalence	
  of	
  RT	
  027	
  strains.	
  	
  	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

153	
  

Materials	
  and	
  Methods	
  
C.	
  difficile	
  strains	
  used	
  in	
  this	
  study	
  and	
  growth	
  conditions	
  
All	
  strains	
  (except	
  CD630	
  and	
  CD196)	
  were	
  provided	
  to	
  the	
  Britton	
  laboratory	
  by	
  
the	
  Michigan	
  Department	
  of	
  Community	
  Health,	
  who	
  determined	
  the	
  toxinotypes	
  
and	
  did	
  the	
  PFGE	
  typing	
  (Table	
  4.1).	
  	
  	
  	
  Strains	
  were	
  collected	
  from	
  infected	
  patients	
  
at	
  either	
  Southeast	
  Michigan	
  or	
  Mid-­‐Michigan	
  Hospitals	
  between	
  11/2007	
  and	
  
5/2008.	
  	
  Strains	
  CD630,	
  CD630Δerm	
  and	
  CD196	
  were	
  kindly	
  provided	
  by	
  the	
  
laboratory	
  of	
  Lincoln	
  Sonnenshein	
  (Tufts	
  University).	
  	
  Ribotyping	
  of	
  the	
  strains	
  was	
  
determined	
  by	
  Seth	
  Walk	
  (previously	
  of	
  the	
  Young	
  laboratory,	
  University	
  of	
  
Michigan).	
  	
  All	
  studies	
  requiring	
  growth	
  of	
  C.	
  difficile	
  were	
  carried	
  out	
  in	
  a	
  Coy	
  
anaerobic	
  chamber	
  (5%	
  hydrogen,	
  90%	
  nitrogen,	
  5%	
  CO2	
  atmosphere),	
  incubated	
  at	
  
37°C.	
  	
  	
  
	
  
Table	
  4.1.	
  Characterization	
  of	
  strains	
  used	
  in	
  this	
  study.	
  	
  
	
  
Strain	
   toxinotype	
   PFGE	
  type	
  (NAP	
  status)	
   Ribotype	
  
CD2015	
  
III	
  
MI_NAP1	
  
027	
  
CD196	
  
III	
  
MI-­‐NAP1	
  
027	
  
CD4015	
  
III	
  
MI-­‐NAP1	
  
027	
  
CD4001	
  
III	
  
MI-­‐UN13	
  
027	
  
CD4010	
  
III	
  
MI-­‐UN13	
  
027	
  
CD630	
  
ND	
  
ND	
  
012	
  
CD3014	
  
0	
  
MI-­‐NAP2	
  
001	
  
CD4011	
  
0	
  
MI-­‐NAP2	
  
001	
  
CD2048	
  
0	
  
MI-­‐NAP3	
  
053	
  
CD1007	
  
0	
  
MI-­‐NAP3	
  
053	
  
CD1014	
  
0	
  
MI-­‐NAP4	
  
014	
  
CD2012	
  
0	
  
MI-­‐NAP6	
  
002	
  
CD1015	
  
V	
  
MI-­‐NAP7	
  
078	
  
CD1018	
  
V	
  
MI-­‐NAP7	
  
078	
  
CD2001	
  
V	
  
MI-­‐NAP8	
  
078	
  
	
  
	
  

154	
  

	
  
Identifying	
  nutritional	
  compounds	
  that	
  increase	
  growth	
  of	
  Clostridium	
  
difficile.	
  	
  Several	
  different	
  plates	
  are	
  available	
  to	
  screen	
  compounds	
  that	
  can	
  serve	
  
as	
  carbon,	
  nitrogen,	
  phosphorus	
  and	
  sulfur	
  sources	
  as	
  well	
  as	
  nutritional	
  
supplements	
  such	
  as	
  peptides,	
  fatty	
  acids,	
  and	
  vitamins.	
  	
  A	
  list	
  of	
  the	
  different	
  plates	
  
(PM1-­‐PM10)	
  and	
  their	
  compounds	
  can	
  be	
  found	
  at:	
  
http://www.biolog.com/pdf/pm_lit/PM1-­‐PM10.pdf.	
  	
  	
  Selected	
  medium	
  was	
  
inoculated	
  (1:100)	
  with	
  overnight	
  culture	
  (CD630	
  or	
  CD2015)	
  and	
  then	
  100	
  µl	
  
added	
  to	
  the	
  wells	
  of	
  the	
  entire	
  Phenotype	
  MicroArray	
  plate.	
  	
  	
  Growth	
  was	
  
monitored	
  using	
  a	
  Tecan	
  Sunrise	
  spectrophotometer,	
  reading	
  every	
  20	
  minutes	
  
(OD600)	
  for	
  24	
  hrs	
  under	
  anaerobic	
  conditions	
  at	
  37°C.	
  	
  	
  Two	
  types	
  of	
  media	
  were	
  
used	
  for	
  these	
  experiments-­‐	
  a	
  defined	
  medium	
  developed	
  for	
  C.	
  difficile	
  ((25)),	
  and	
  
modified	
  sporulation	
  medium.	
  	
  The	
  defined	
  medium	
  was	
  made	
  following	
  the	
  
published	
  recipe	
  (Table	
  4.2)	
  with	
  a	
  few	
  modifications;	
  glucose	
  was	
  omitted	
  for	
  all	
  
Phenotype	
  MicroArray	
  plate	
  experiments,	
  and	
  for	
  plates	
  PM3-­‐8,	
  only	
  essential	
  
amino	
  acids	
  were	
  added	
  (see	
  table	
  legend).	
  	
  	
  The	
  other	
  medium	
  used	
  is	
  a	
  variation	
  of	
  
the	
  C.	
  difficile	
  sporulation	
  medium	
  published	
  by	
  Wilson	
  et	
  al.	
  (26),	
  named	
  MSM	
  for	
  
“modified	
  smorulation	
  medium”.	
  	
  MSM	
  contained:	
  22.5	
  g/L	
  trypticase	
  peptone	
  
(VWR),	
  1	
  g/L	
  (NH4)2SO4,	
  and	
  1.5	
  g/L	
  tris	
  base	
  (Invitrogen);	
  adjusted	
  to	
  pH	
  7.5	
  and	
  
autoclaved	
  at	
  121°	
  C	
  for	
  30	
  min.	
  	
  	
  The	
  sporulation	
  medium	
  was	
  chosen	
  because	
  it	
  is	
  
simple,	
  composed	
  only	
  of	
  peptones	
  and	
  salts,	
  and	
  was	
  diluted	
  four-­‐fold	
  in	
  order	
  to	
  
provide	
  a	
  low	
  level	
  of	
  growth	
  and	
  allow	
  for	
  detectability	
  of	
  increased	
  yields.	
  	
  Growth	
  
yields	
  were	
  determined	
  as	
  the	
  maximum	
  OD600	
  of	
  the	
  growth	
  curve	
  for	
  individual	
  

	
  

155	
  

wells.	
  	
  Compounds	
  that	
  increased	
  the	
  maximum	
  OD600	
  by	
  at	
  least	
  1.5-­‐fold	
  were	
  
concluded	
  to	
  confer	
  a	
  growth	
  advantage.	
  	
  	
  
	
  
Table	
  4.2.	
  	
  Defined	
  medium	
  ingredients	
  and	
  concentrations.	
  	
  *these	
  amino	
  acids	
  
were	
  the	
  only	
  ones	
  added	
  for	
  medium	
  used	
  in	
  Biolog	
  Phenotype	
  MicroArray	
  plates	
  
PM3-­‐8.	
  
	
  
	
  Amino	
  acids:	
  
histidine	
  
tryptophan*	
  
glycine	
  
tyrosine	
  
arginine	
  
phenylalanine	
  
methionine	
  
threonine	
  
alanine	
  
lysine	
  
serine	
  
valine*	
  
isoleucine*	
  
aspartic	
  acid	
  
leucine*	
  
cysteine*	
  
proline*	
  
glutamic	
  acid	
  

	
  

	
  
	
  
	
  

	
  
mg/L	
  
100	
  

Vitamins:	
   	
  
thiamin	
  
calcium-­‐D-­‐
pantothenate	
  
nicotinamide	
  
rioflavin	
  
pyridoxine	
  
p-­‐Aminobenzoic	
  acid	
  
folic	
  acid	
  
biotin	
  
B12	
  

100	
  
100	
  
100	
  
200	
  
200	
  
200	
  
200	
  
200	
  
300	
  
300	
  
300	
  
300	
  
300	
  
400	
  
500	
  
600	
  
900	
  

	
  minerals:	
  
KH2PO4	
  
Na2HPO4	
  
NaCl	
  
CaCl2.2H2O	
  
MgCl2.6H2O	
  
MnCl2.4H2O	
  
(NH4)2SO4	
  
FeSO4.7H2O	
  
CoCl2.6H2O	
  
NaHCO3	
  
	
  

	
  
	
  
	
  
	
  

	
  
	
  
	
  
	
  
	
  

	
  

156	
  

	
   mg/L	
  
1	
  
1	
  
1	
  
1	
  
1	
  
0.05	
  
0.0125	
  
0.0125	
  
0.005	
  
	
  
	
  
300	
  
1500	
  
900	
  
26	
  
20	
  
10	
  
40	
  
4	
  
1	
  
5000	
  
	
  

Growth	
  experiments	
  further	
  investigating	
  C.	
  difficile	
  utilization	
  of	
  compounds	
  
identified	
  in	
  Phenotype	
  MicroArray	
  plates.	
  	
  Compounds	
  initially	
  identified	
  to	
  
increase	
  growth	
  in	
  Phenotype	
  MicroArray	
  plates	
  (PM1	
  and	
  PM2)	
  were	
  used	
  to	
  
further	
  assess	
  C.	
  difficile	
  growth	
  using	
  additional	
  strains.	
  	
  Although	
  Biolog	
  does	
  not	
  
provide	
  the	
  specific	
  concentrations	
  of	
  the	
  components	
  in	
  their	
  phenotype	
  
microarray	
  plates,	
  they	
  suggest	
  follow-­‐up	
  studies	
  be	
  done	
  in	
  a	
  range	
  of	
  5-­‐50mM	
  in	
  
their	
  carbon	
  source	
  plates	
  (PM1-­‐2)	
  (per	
  email	
  correspondence	
  with	
  Biolog	
  technical	
  
support).	
  	
  The	
  defined	
  medium	
  was	
  used	
  (Table	
  4.2;	
  all	
  amino	
  acids)	
  and	
  
supplemented	
  with	
  N-­‐acetyl-­‐glucosamine	
  (25mM),	
  N-­‐acetyl-­‐neuraminic	
  acid	
  
(10mM),	
  alanine	
  and	
  hydroxyl-­‐proline	
  (12	
  mM	
  each),	
  mannose	
  (25mM),	
  or	
  
trehalose	
  (25mM).	
  	
  Cultures	
  were	
  grown	
  in	
  96-­‐well	
  plates	
  (Corning	
  Costar	
  
#CLS3595)	
  plates	
  by	
  adding	
  200	
  μL	
  medium	
  (+/-­‐	
  compound)	
  to	
  selected	
  wells	
  and	
  
inoculating	
  with	
  10	
  µL	
  MSM	
  overnight	
  culture	
  of	
  selected	
  strains.	
  	
  Growth	
  was	
  
monitored	
  using	
  a	
  Tecan	
  Sunrise	
  spectrophotometer,	
  reading	
  every	
  20	
  minutes	
  
(OD600)	
  for	
  24	
  hrs	
  under	
  anaerobic	
  conditions	
  (Coy	
  anaerobic	
  chamber	
  with	
  3%	
  
hydrogen,	
  97%	
  nitrogen	
  atmosphere)	
  at	
  37°C.	
  	
  	
  	
  	
  
	
  
Trehalose	
  growth	
  experiments.	
  	
  Selected	
  strains	
  were	
  grown	
  in	
  DM	
  (Table	
  4.2;	
  all	
  
amino	
  acids)	
  supplemented	
  with	
  the	
  specified	
  trehalose	
  concentrations.	
  	
  	
  Medium	
  
(200	
  µL)	
  was	
  added	
  to	
  selected	
  wells	
  of	
  96-­‐well	
  plate	
  and	
  wells	
  were	
  inoculated	
  
with	
  10	
  µL	
  of	
  overnight	
  culture	
  of	
  indicated	
  strains.	
  	
  Growth	
  was	
  monitored	
  using	
  a	
  
Tecan	
  Sunrise	
  spectrophotometer,	
  reading	
  every	
  20	
  minutes	
  (OD600)	
  for	
  24	
  hrs	
  

	
  

157	
  

under	
  anaerobic	
  conditions	
  (Coy	
  anaerobic	
  chamber	
  with	
  3%	
  hydrogen,	
  97%	
  
nitrogen	
  atmosphere)	
  at	
  37°C.	
  	
  	
  
	
  
RT-­‐qPCR	
  analysis	
  of	
  treA	
  expression.	
  
Cultures.	
  	
  Selected	
  C.	
  difficile	
  strains	
  were	
  grown	
  overnight	
  in	
  5	
  ml	
  DM	
  (table	
  4.1),	
  
inoculated	
  from	
  fresh	
  BHIS	
  agar	
  plates	
  (made	
  as	
  previously	
  described	
  except	
  
without	
  the	
  addition	
  of	
  cysteine,	
  (27)).	
  	
  	
  	
  Expression	
  cultures	
  were	
  started	
  by	
  
subculturing	
  100	
  µL	
  of	
  the	
  overnight	
  culture	
  into	
  10	
  mL	
  fresh	
  DM	
  supplemented	
  
with	
  25mM	
  trehalose	
  (by	
  addition	
  of	
  1M	
  stock	
  solution	
  of	
  trehalose	
  dissolved	
  in	
  
water	
  and	
  filter-­‐sterilized).	
  	
  	
  Cultures	
  were	
  incubated	
  (anaerobic,	
  37°C)	
  and	
  OD600	
  
(optical	
  density)	
  monitored	
  until	
  an	
  OD	
  of	
  ~0.1-­‐0.2	
  was	
  reached.	
  	
  The	
  remaining	
  
culture	
  (~8	
  mL)	
  was	
  mixed	
  with	
  an	
  equal	
  volume	
  of	
  ice	
  cold	
  ethanol:acetone	
  (1:1)	
  
mix,	
  placed	
  on	
  ice	
  for	
  10	
  min,	
  and	
  stored	
  at	
  -­‐80°C.	
  	
  	
  	
  	
  
	
  
RNA	
  extractions.	
  	
  Samples	
  (culture,	
  ethanol,	
  acetone	
  mix)	
  were	
  thawed	
  on	
  ice,	
  and	
  
then	
  pelleted	
  by	
  centrifugation	
  for	
  10	
  min	
  at	
  4°C,	
  and	
  the	
  supernatant	
  removed.	
  	
  Cell	
  
pellets	
  were	
  washed	
  with	
  500	
  µL	
  TE	
  buffer	
  (10mM	
  Tris,	
  1mM	
  EDTA,	
  pH	
  7.6).	
  	
  From	
  
this	
  point	
  RNA	
  was	
  extracted	
  using	
  Qiagen	
  RNeasy	
  Kit	
  (Qiagen	
  #74104),	
  following	
  a	
  
modified	
  version	
  of	
  the	
  recommended	
  protocol,	
  starting	
  with	
  resuspension	
  in	
  1	
  mL	
  
RLT	
  buffer	
  (to	
  which	
  1:100	
  dilution	
  of	
  β-­‐mercaptoethanol	
  is	
  added).	
  	
  	
  Samples	
  were	
  
then	
  transferred	
  to	
  2	
  ml	
  screw-­‐top	
  tubes	
  containing	
  ~200	
  µl	
  0.1	
  mm	
  silica	
  beads	
  
(Biospec	
  Products).	
  	
  The	
  samples	
  were	
  homogenized	
  by	
  bead-­‐beating	
  (BioSpec	
  
Products)	
  on	
  the	
  homogenize	
  setting	
  two	
  times	
  for	
  1	
  min	
  (placed	
  on	
  ice	
  for	
  1	
  min	
  
	
  

158	
  

between),	
  centrifuged	
  for	
  15	
  min	
  at	
  21000	
  X	
  g,	
  and	
  the	
  supernatant	
  was	
  transferred	
  
to	
  a	
  new	
  tube.	
  	
  Buffer	
  RLT	
  (+	
  1:100	
  β-­‐mercaptoethanol)	
  was	
  added	
  to	
  a	
  total	
  volume	
  
of	
  900	
  µL,	
  then	
  vortexed.	
  	
  500	
  µL	
  100%	
  ethanol	
  was	
  added,	
  the	
  sample	
  vortexed,	
  
and	
  then	
  transferred	
  to	
  RNeasy	
  spin	
  column.	
  	
  The	
  extraction	
  protocol	
  from	
  here	
  
followed	
  the	
  handbook	
  protocol.	
  	
  After	
  the	
  final	
  wash	
  with	
  RPE	
  buffer,	
  RNA	
  was	
  
eluted	
  off	
  the	
  columns	
  with	
  20µL	
  RNase-­‐free	
  water.	
  	
  1	
  µL	
  RNase	
  Inhibitor	
  (Roche	
  
#03335399001)	
  was	
  added	
  to	
  the	
  extracted	
  RNA;	
  samples	
  were	
  stored	
  at	
  -­‐80°C.	
  	
  	
  
Samples	
  were	
  treated	
  with	
  DNase	
  to	
  remove	
  any	
  contaminating	
  genomic	
  DNA	
  using	
  
Ambion	
  Turbo	
  DNA-­‐free	
  Kit	
  (Ambion	
  AM1907).	
  	
  	
  
	
  
Reverse	
  Transcription.	
  	
  cDNA	
  of	
  RNA	
  samples	
  was	
  synthesized	
  using	
  Invitrogen	
  
Superscript	
  III	
  reverse	
  transcriptase	
  (Invitrogen	
  #18080-­‐093)	
  following	
  the	
  
recommended	
  protocol.	
  	
  Other	
  reaction	
  components	
  used	
  were	
  random	
  primers	
  
(Promega	
  #C1181),	
  and	
  10mM	
  dNTP’s	
  (Promega	
  #U1515).	
  	
  Approximately	
  1	
  ng	
  of	
  
each	
  RNA	
  sample	
  was	
  used	
  for	
  cDNA	
  synthesis.	
  	
  	
  No	
  reverse	
  transcriptase	
  controls	
  
were	
  also	
  set-­‐up	
  by	
  replacing	
  the	
  1	
  µL	
  reverse	
  transcriptase	
  with	
  1	
  µL	
  water	
  in	
  
those	
  reactions;	
  all	
  other	
  components	
  were	
  the	
  same.	
  	
  	
  
	
  
Real-­‐Time	
  PCR	
  reactions.	
  	
  Real-­‐time	
  PCR	
  reactions	
  were	
  performed	
  in	
  triplicate	
  
and	
  contained	
  the	
  following	
  components:	
  1	
  µl	
  cDNA	
  (undiluted	
  (treA),	
  1:100	
  
dilution	
  in	
  sterile	
  water	
  (16S	
  rRNA),	
  or	
  1:100	
  dilution	
  of	
  –RT	
  control	
  sample),	
  10	
  µL	
  
Power	
  SYBR	
  Green	
  PCR	
  Master	
  Mix	
  (ABI,	
  Carlsbad,	
  CA),	
  0.25	
  µL	
  each	
  primer	
  (20	
  
µM)	
  (Supplementary	
  Table	
  4.3),	
  and	
  8.5	
  µL	
  Milli-­‐Q	
  water.	
  	
  Triplicate	
  water	
  controls	
  
	
  

159	
  

were	
  also	
  run;	
  1	
  µL	
  water	
  replaced	
  the	
  cDNA	
  template.	
  	
  Real-­‐time	
  PCR	
  was	
  
performed	
  using	
  an	
  Eppendorf	
  Mastercycler	
  PCR	
  machine	
  under	
  the	
  following	
  
conditions:	
  95°C	
  10	
  min,	
  40	
  cycles	
  of	
  95°C	
  for	
  15	
  sec	
  followed	
  by	
  60°C	
  for	
  1	
  min.	
  A	
  
20	
  min	
  melting	
  curve	
  was	
  also	
  performed	
  from	
  60°C	
  to	
  95°C.	
  	
  Standard	
  curves	
  of	
  
cDNA	
  were	
  run	
  to	
  determine	
  primer	
  efficiencies.	
  	
  The	
  template	
  used	
  for	
  standard	
  
curves	
  was	
  the	
  sample	
  with	
  the	
  lowest	
  CT	
  values,	
  diluted	
  4	
  logs	
  into	
  sterile	
  water.	
  	
  
Primer	
  efficiencies	
  (E)	
  were	
  calculated	
  by	
  the	
  method	
  published	
  by	
  Pfaffl	
  et	
  al.	
  (28);	
  
treA	
  E=	
  1.95;	
  16s	
  rRNA	
  E=	
  1.93.	
  	
  Expression	
  of	
  treA	
  at	
  the	
  stopping	
  point	
  of	
  each	
  
culture	
  was	
  determined	
  using	
  the	
  average	
  of	
  the	
  triplicate	
  CT	
  values	
  from	
  each	
  
sample	
  and	
  the	
  following	
  equation:	
  [EtreA(1.95)^(25-­‐sample	
  CT)]/[E16s(1.93)^(10-­‐
sample	
  CT)]	
  (28).	
  	
  A	
  baseline	
  expression	
  CT	
  signal	
  for	
  all	
  samples	
  was	
  used	
  which	
  
was	
  approximately	
  the	
  average	
  CT	
  value	
  for	
  all	
  strains	
  in	
  medium	
  supplemented	
  
with	
  glucose	
  instead	
  of	
  trehalose.	
  	
  This	
  was	
  done	
  to	
  normalize	
  the	
  variation	
  seen	
  in	
  
the	
  baseline	
  expression	
  of	
  treA	
  in	
  control	
  cultures	
  (those	
  without	
  trehalose)	
  where	
  
aberrant	
  expression	
  levels	
  in	
  some	
  replicate	
  cultures	
  were	
  observed.	
  	
  	
  Aberrantly	
  
high	
  baseline	
  expression	
  effectively	
  would	
  otherwise	
  decrease	
  the	
  fold-­‐induction	
  of	
  
treA	
  in	
  those	
  samples	
  and	
  not	
  be	
  representative	
  of	
  the	
  true	
  levels	
  of	
  treA	
  transcripts	
  
in	
  the	
  presence	
  of	
  trehalose.	
  	
  	
  
	
  
	
  
	
  
	
  
	
  

	
  

160	
  

Table	
  4.3.	
  	
  Primers	
  used	
  in	
  this	
  study.	
  
Target	
  Gene	
  (reference)	
  

Sequence	
  

C.	
  difficile	
  treA	
  (this	
  work)	
  

Fwd:	
  tacgctgatggtcctcgtat	
  

	
  

Rev:	
  cgcctcctttataatctgttttc	
  

C.	
  difficile	
  16s	
  rRNA	
  	
  
{Rinttila:2004da}	
  
	
  

Fwd:	
  ttgagcgatttacttcggtaaaga	
  

treA	
  IBS1.2_Intron	
  retarget	
  (this	
  work)	
  

atatcaagcttttgcaacccacgtcgatcgtgaataga

Rev:	
  ccatcctgtactggctcacct	
  

agattattgtgcgcccagatagggtg	
  
treA	
  EBS1_Intron	
  retarget	
  (this	
  work)	
  

cagattgtacaaatgtggtgataacagataagtcatta
ttattaacttacctttctttgt	
  

treA	
  EBS2_Intron	
  retarget	
  (this	
  work)	
  

cgcaagtttctaatttcggttttctatcgatagaggaaa
gtgtct	
  

treA	
  knock-­‐out	
  screen	
  (this	
  work)	
  

qPCR-­‐	
  wild-­‐type	
  specific	
  

qPCR-­‐	
  treA	
  mutant	
  specific	
  

Fwd:	
  gcaacaatgatggtataggtgatataaatgg	
  
	
  
Rev:	
  ggaacagaaccatcaggtttagca	
  
	
  
Fwd:	
  ggttgactcctatgtacgtttct	
  
	
  
Rev:	
  caaagtcactcattgtcccatattt	
  
	
  
Fwd:	
  tcctcctttctattaggcattcttg	
  
	
  
Rev:	
  ggagaacctatgggaacgaaac	
  

	
  
	
  
	
  
	
  
	
  

	
  

161	
  

Alignments	
  of	
  TreA	
  amino	
  acid	
  sequences	
  from	
  C.	
  difficile	
  clinical	
  isolates	
  and	
  
other	
  Gram-­‐positive	
  bacteria.	
  	
  Whole	
  genome	
  sequencing	
  (Illumina	
  MiSeq)	
  was	
  
conducted	
  on	
  several	
  of	
  the	
  C.	
  difficile	
  clinical	
  isolates	
  obtained	
  from	
  the	
  MDCH	
  used	
  
in	
  this	
  study	
  by	
  the	
  MSU	
  Research	
  and	
  Technology	
  Support	
  Facility.	
  	
  	
  Genome	
  
sequence	
  files	
  were	
  assembled	
  de	
  novo	
  using	
  the	
  next	
  generation	
  sequencing	
  
assembly	
  algorithm,	
  Velvet	
  (www.ebi.ac.uk/~zerbino/velvet/).	
  	
  BLAST	
  (Basic	
  Local	
  
Alignment	
  Search	
  Tool)	
  databases	
  were	
  made	
  for	
  each	
  genome	
  and	
  the	
  treR	
  
sequences	
  pulled	
  out	
  of	
  the	
  assembled	
  contigs	
  by	
  searching	
  for	
  sequences	
  matching	
  
the	
  CD630	
  (Integrated	
  Microbial	
  Genomes	
  (IMG);	
  https://img.jgi.doe.gov)	
  treR	
  
sequence.	
  	
  	
  The	
  treR	
  sequences	
  of	
  other	
  Gram-­‐positive	
  organisms	
  were	
  obtained	
  
from	
  the	
  public	
  database	
  of	
  IMG.	
  	
  All	
  alignments	
  were	
  done	
  using	
  ClustalW2	
  
(https://www.ebi.ac.uk/Tools/msa/clustalw2/).	
  	
  	
  
	
  
Construction	
  of	
  a	
  treA	
  knockout	
  mutant	
  in	
  the	
  C.	
  difficile	
  parent	
  strain,	
  
CD630Δerm.	
  	
  We	
  used	
  a	
  gene	
  knockout	
  system	
  similar	
  to	
  the	
  TagetTron	
  system	
  
available	
  through	
  Sigma-­‐Alderich	
  to	
  generate	
  a	
  treA	
  gene	
  knockout	
  mutant.	
  	
  	
  This	
  
system	
  uses	
  a	
  group-­‐II	
  intron,	
  retargeted	
  to	
  the	
  gene	
  of	
  interest,	
  to	
  generate	
  
functional	
  knockouts.	
  	
  We	
  used	
  the	
  protocol	
  previously	
  described	
  (29,	
  30).	
  	
  Primers	
  
were	
  designed	
  to	
  target	
  the	
  treA	
  gene	
  (CD3091)	
  of	
  CD630	
  using	
  the	
  treA	
  gene	
  
sequence	
  obtained	
  from	
  IMG	
  (https://img.jgi.doe.gov).	
  	
  	
  A	
  primer	
  design	
  script	
  was	
  
developed,	
  which	
  is	
  a	
  variation	
  of	
  the	
  one	
  available	
  from	
  Sigma	
  for	
  TargetTron	
  
primer	
  design	
  (http://www.sigma-­‐genosys.com/targetron/),	
  and	
  was	
  modified	
  to	
  
be	
  optimized	
  to	
  the	
  C.	
  difficile	
  genome.	
  	
  The	
  “intronator”	
  script	
  is	
  available	
  upon	
  

	
  

162	
  

request.	
  Primers	
  used	
  are	
  listed	
  in	
  Table	
  4.3.	
  	
  They	
  are	
  designed	
  to	
  target	
  the	
  intron	
  
to	
  insert	
  at	
  bp	
  177	
  of	
  treA.	
  	
  Two	
  plasmid	
  templates	
  (pBL64,	
  pBL65)	
  were	
  used	
  to	
  
PCR	
  amplify	
  the	
  retargeted	
  intron	
  fragment	
  and	
  ligate	
  it	
  into	
  pBL100	
  (30).	
  	
  Plasmids	
  
and	
  erythromycin-­‐sensitive	
  CD630	
  derivative	
  strain	
  (CD630Δerm)	
  were	
  generously	
  
provided	
  to	
  us	
  by	
  Dr.	
  Lincoln	
  Sonenshein’s	
  lab	
  (Tufts	
  University).	
  	
  The	
  treA-­‐targeted	
  
pBL100	
  plasmid	
  was	
  then	
  transformed	
  into	
  Escherichia	
  coli	
  SD46	
  (provided	
  by	
  Dr.	
  
Craig	
  Ellermeier,	
  University	
  of	
  Iowa),	
  and	
  then	
  mated	
  into	
  CD630Δerm.	
  	
  The	
  
resulting	
  treA	
  insertion	
  mutant	
  was	
  verified	
  by	
  PCR	
  using	
  primers	
  designed	
  to	
  flank	
  
the	
  treA	
  insertion	
  site,	
  resulting	
  in	
  a	
  350bp	
  product	
  for	
  the	
  wild-­‐type	
  gene	
  and	
  a	
  
2.4kbp	
  product	
  for	
  the	
  gene	
  knockout	
  (Table	
  4.3).	
  	
  	
  	
  	
  
	
  
Colonization	
  and	
  competition	
  of	
  CD630Δerm	
  wild-­‐type	
  and	
  treA	
  mutant	
  
strains	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection.	
  	
  The	
  previously	
  published	
  
cefoperazone	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection	
  was	
  used	
  for	
  these	
  experiments	
  
(31).	
  They	
  were	
  conducted	
  in	
  the	
  laboratory	
  of	
  Dr.	
  Vince	
  Young	
  (University	
  of	
  
Michigan,	
  Department	
  of	
  Microbiology	
  and	
  Immunology),	
  by	
  Dr.	
  Mark	
  
Koenigsknecht,	
  a	
  postdoctoral	
  research	
  associate.	
  	
  	
  Four	
  groups	
  of	
  C57BL/6	
  mice	
  (6-­‐
8wks	
  in	
  age)	
  were	
  used	
  in	
  these	
  experiments;	
  One	
  group	
  (3	
  mice)	
  served	
  as	
  no-­‐
antibiotic	
  control,	
  two	
  groups	
  (5	
  mice	
  each)	
  were	
  infected	
  with	
  wild	
  type	
  CD630erm	
  
or	
  treA	
  knockout	
  mutant	
  strain,	
  and	
  the	
  fourth	
  group	
  (5	
  mice)	
  was	
  infected	
  with	
  a	
  
mixture	
  of	
  the	
  wild-­‐type	
  and	
  mutant	
  strains.	
  	
  Mice	
  were	
  treated	
  with	
  cefoperazone	
  
(0.5mg/ml)	
  in	
  their	
  drinking	
  water	
  for	
  5	
  days,	
  followed	
  by	
  two	
  days	
  of	
  fresh	
  water	
  
before	
  oral	
  gavage	
  of	
  1	
  x	
  104	
  C.	
  difficile	
  spores/mouse.	
  	
  	
  Spores	
  were	
  cultivated	
  by	
  

	
  

163	
  

spread	
  plating	
  overnight	
  BHIS	
  cultures	
  of	
  C.	
  difficile	
  on	
  BHIS	
  medium	
  and	
  incubating	
  
anaerobically	
  at	
  37°C	
  for	
  3	
  days.	
  	
  Cells	
  were	
  scraped	
  from	
  the	
  plates	
  and	
  
resuspended	
  in	
  sterile	
  water,	
  heat-­‐treated	
  at	
  60°C	
  for	
  30	
  min	
  to	
  kill	
  vegetative	
  cells,	
  
and	
  the	
  number	
  of	
  viable	
  spores	
  were	
  enumerated	
  by	
  plating	
  appropriate	
  serial	
  
dilutions	
  on	
  BHIS	
  supplemented	
  with	
  0.1%	
  taurocholic	
  acid.	
  Spore	
  preparations	
  
were	
  diluted	
  in	
  sterile	
  water	
  to	
  yield	
  ~105	
  spores/ml,	
  then	
  mixed,	
  when	
  
appropriate,	
  prior	
  to	
  gavaging	
  a	
  total	
  of	
  ~104	
  spores/mouse.	
  	
  	
  Mice	
  were	
  weighed	
  
and	
  observed	
  daily	
  for	
  disease	
  symptoms	
  and	
  morbidity.	
  Fecal	
  samples	
  were	
  
collected	
  daily	
  and	
  transferred	
  into	
  an	
  anaerobic	
  chamber	
  within	
  two	
  hours	
  of	
  
collection.	
  	
  Fecal	
  samples	
  were	
  weighed,	
  suspended	
  in	
  anaerobic	
  PBS,	
  serially	
  
diluted,	
  and	
  plated	
  on	
  TCCFA	
  agar	
  plates	
  +/-­‐	
  10μg/ml	
  erythromycin.	
  	
  After	
  24-­‐48	
  
hrs	
  of	
  anaerobic	
  incubation	
  at	
  37°C,	
  colonies	
  were	
  counted	
  and	
  the	
  CFU/g	
  feces	
  was	
  
determined.	
  
	
  
Quantitative	
  PCR	
  Analysis	
  of	
  Competitions	
  and	
  Calculations	
  of	
  Competitive	
  
Index.	
  	
  For	
  mice	
  gavaged	
  with	
  a	
  mixture	
  of	
  wild	
  type	
  and	
  treA	
  mutant	
  strains,	
  the	
  
strain	
  ratios	
  were	
  determined	
  by	
  qPCR	
  instead	
  of	
  plating.	
  	
  Strains	
  ratios	
  determined	
  
by	
  qPCR	
  will	
  be	
  more	
  reflective	
  of	
  the	
  actual	
  ratios	
  since	
  primers	
  are	
  designed	
  to	
  
specifically	
  target	
  each	
  strain,	
  while	
  determination	
  by	
  plating	
  relies	
  on	
  subtraction	
  
of	
  the	
  colonies	
  on	
  selective	
  plates	
  from	
  the	
  total	
  on	
  non-­‐selective	
  plates,	
  and	
  will	
  be	
  
affected	
  by	
  inherent	
  plating	
  error,	
  making	
  small	
  differences	
  in	
  ratios	
  more	
  difficult	
  
to	
  resolve.	
  	
  DNA	
  was	
  extracted	
  from	
  fecal	
  samples	
  using	
  bead	
  beating	
  followed	
  by	
  
modified	
  cleanup	
  with	
  a	
  Qiagen	
  DNEasy	
  Tissue	
  Kit	
  as	
  described	
  (32).	
  	
  DNA	
  

	
  

164	
  

concentrations	
  were	
  determined	
  by	
  spectrophotometry	
  at	
  260	
  and	
  280	
  nm	
  
(Nanodrop),	
  and	
  ranged	
  from	
  18-­‐48	
  ng/μL.	
  	
  The	
  mixed	
  spores	
  preparation	
  (1	
  x	
  105	
  
spores/ml)	
  used	
  to	
  gavage	
  mice	
  was	
  treated	
  with	
  0.1%	
  taurocholate	
  at	
  room	
  
temperature	
  for	
  20	
  minutes	
  to	
  induce	
  germination	
  and	
  then	
  bead-­‐beated	
  as	
  above.	
  	
  
The	
  supernatant	
  was	
  transferred	
  to	
  a	
  fresh	
  tube	
  and	
  used	
  as	
  DNA	
  template	
  in	
  the	
  
qPCR	
  analysis.	
  	
  Primers	
  were	
  designed	
  to	
  specifically	
  target	
  the	
  wild-­‐type	
  strain	
  
(flank	
  the	
  intron	
  insertion	
  site)	
  or	
  the	
  treA	
  mutant	
  (intron-­‐specific)	
  (table	
  4.3).	
  	
  
Real-­‐time	
  PCR	
  reactions	
  were	
  set-­‐up	
  by	
  combining	
  the	
  following	
  components:	
  10	
  µL	
  
Power	
  Sybr	
  Green	
  PCR	
  Master	
  Mix	
  (ABI,	
  Carlsbad,	
  CA),	
  0.2	
  µL	
  each	
  primer	
  (20	
  µM),	
  
8.6	
  µL	
  Mili-­‐Q	
  water,	
  1	
  µL	
  DNA	
  sample.	
  	
  All	
  PCR	
  reactions	
  were	
  performed	
  in	
  
technical	
  triplicate	
  and	
  the	
  CT	
  values	
  are	
  an	
  average	
  of	
  the	
  triplicate	
  data	
  points.	
  The	
  
amplification	
  efficiency	
  (E)	
  of	
  each	
  primer	
  set	
  was	
  determined	
  by	
  plotting	
  the	
  CT	
  
values	
  of	
  a	
  standard	
  curve	
  generated	
  by	
  serial	
  4-­‐log	
  dilutions	
  of	
  the	
  mixed-­‐spore	
  
supernatant	
  diluted	
  into	
  a	
  DNA	
  sample	
  extracted	
  from	
  a	
  mouse	
  fecal	
  pellet	
  collected	
  
before	
  C.	
  difficile	
  gavage	
  (community	
  background	
  DNA).	
  Primer	
  efficiencies	
  were	
  
calculated	
  using	
  the	
  method	
  described	
  by	
  Pfaffl	
  et	
  al.;	
  E=	
  10(-­‐1/slope)	
  (28);	
  
Efficiencywild-­‐type	
  =	
  1.96,	
  Efficiencymutant	
  =	
  2.03.	
  	
  Competitive	
  Indices	
  (CI)	
  were	
  
calculated	
  by	
  dividing	
  the	
  end	
  point	
  wild-­‐type:mutant	
  ratio	
  by	
  the	
  ratio	
  at	
  T0	
  (ratio=	
  
Emut^CT_mut/Ewt^CT_wt).	
  	
  
	
  
	
  
	
  
	
  

	
  

165	
  

Results	
  and	
  Discussion	
  
Identification	
  of	
  carbon	
  sources	
  that	
  C.	
  difficile	
  is	
  able	
  to	
  use	
  for	
  growth.	
  	
  
Phenotype	
  MicroArray	
  plates	
  can	
  be	
  used	
  to	
  screen	
  various	
  compounds	
  for	
  their	
  
ability	
  to	
  increase	
  the	
  growth	
  rate	
  or	
  growth	
  yield	
  of	
  selected	
  strains.	
  	
  Phenotype	
  
MicroArray	
  plates	
  are	
  96-­‐well	
  plates	
  in	
  which	
  each	
  well	
  contains	
  a	
  different	
  
compound;	
  these	
  can	
  be	
  used	
  to	
  screen	
  not	
  only	
  for	
  phenotypic	
  effects	
  on	
  strain	
  
growth	
  but	
  also	
  other	
  aspects	
  of	
  physiology	
  such	
  as	
  sporulation,	
  toxin	
  production,	
  
or	
  biofilm	
  formation.	
  	
  The	
  focus	
  of	
  this	
  work	
  was	
  on	
  identifying	
  nutrients	
  which	
  C.	
  
difficile	
  is	
  able	
  to	
  use	
  to	
  increase	
  growth,	
  although	
  these	
  plates	
  could	
  also	
  be	
  used	
  to	
  
identify	
  compounds	
  that	
  inhibit	
  growth.	
  	
  Strains	
  CD630	
  and	
  CD2015	
  were	
  used	
  for	
  
initial	
  Phenotype	
  MicroArray	
  growth	
  studies	
  (Table	
  4.1).	
  	
  This	
  was	
  done	
  because	
  we	
  
wanted	
  to	
  identify	
  compounds	
  beneficial	
  for	
  C.	
  difficile	
  in	
  general	
  as	
  well	
  as	
  ones	
  
that	
  might	
  be	
  specific	
  to	
  the	
  epidemic	
  RT	
  027	
  strains.	
  	
  Two	
  different	
  media	
  were	
  
used	
  for	
  initial	
  experiments	
  in	
  order	
  to	
  detect	
  compounds	
  that	
  may	
  provide	
  an	
  
advantage	
  in	
  one	
  medium	
  but	
  may	
  not	
  be	
  detected	
  in	
  the	
  other.	
  	
  Table	
  4.4	
  
summarizes	
  the	
  results	
  from	
  just	
  the	
  two	
  carbon	
  source	
  plates	
  (PM1	
  and	
  PM2).	
  	
  Out	
  
of	
  190	
  compounds	
  tested,	
  a	
  total	
  of	
  16	
  were	
  found	
  to	
  confer	
  a	
  growth	
  advantage	
  to	
  
both	
  strains,	
  5	
  compounds	
  to	
  only	
  CD630,	
  and	
  3	
  compounds	
  to	
  only	
  CD2015,	
  (Table	
  
4.4).	
  	
  However,	
  these	
  results	
  are	
  representative	
  of	
  only	
  1	
  replicate	
  of	
  each	
  plate	
  for	
  
each	
  strain	
  and	
  each	
  media	
  type.	
  	
  More	
  replicates	
  or	
  media	
  types	
  might	
  have	
  shown	
  
more	
  conserved	
  overlap	
  in	
  the	
  utilization	
  of	
  these	
  compounds	
  by	
  these	
  two	
  strains.	
  	
  
Moreover,	
  some	
  compounds	
  may	
  have	
  increased	
  growth	
  for	
  either,	
  or	
  both,	
  strains	
  
but	
  may	
  have	
  not	
  met	
  the	
  1.5-­‐fold	
  cut-­‐off	
  value.	
  	
  This	
  was	
  an	
  initial	
  screen	
  for	
  

	
  

166	
  

identification	
  of	
  compounds	
  to	
  follow-­‐up	
  in	
  subsequent	
  and	
  better-­‐replicated	
  
growth	
  studies.	
  	
  	
  Therefore,	
  these	
  Phenotype	
  MicroArray	
  plate	
  experiments	
  are	
  not	
  
an	
  exhaustive	
  study	
  into	
  the	
  metabolic	
  capabilities	
  of	
  C.	
  difficile	
  and	
  are	
  likely	
  to	
  
overlook	
  some	
  compounds	
  that	
  may	
  be	
  of	
  interest.	
  	
  	
  	
  
	
  
Table	
  4.4.	
  Compounds	
  that	
  conferred	
  at	
  least	
  a	
  1.5-­‐fold	
  growth	
  yield	
  
advantage	
  Biolog	
  PM1	
  and	
  PM2	
  plates.a	
  	
  	
  
	
  
	
  
	
  
	
  
CD630	
  
CD2015	
  
	
  
Well	
  
Compound	
  
MSM	
   DM-­‐G	
   MSM	
   DM-­‐G	
  
PM1:	
   A3	
   N-­‐acetyl-­‐D-­‐glucosamine*	
  
+	
  
+	
  
ND	
  
+	
  
	
  
A8	
  
L-­‐proline	
  
+	
  
-­‐	
  
ND	
  
-­‐	
  
	
  
A10	
  
D-­‐trehalose*	
  
+	
  
+	
  
ND	
  
+	
  
	
  
A11	
  
D-­‐mannose*	
  
+	
  
+	
  
ND	
  
+	
  
	
  
B2	
  
D-­‐sorbitol	
  
-­‐	
  
-­‐	
  
ND	
  
+	
  
	
  
B11	
  
D-­‐mannitol	
  
+	
  
+	
  
ND	
  
+	
  
	
  
C7	
  
D-­‐fructose	
  
+	
  	
  
-­‐	
  
ND	
  
+	
  
	
  
C9	
  
α-­‐D-­‐glucose	
  
+	
  
+	
  
ND	
  
+	
  
	
  
D7	
  
α-­‐Keto-­‐Butyric	
  acid	
  
+	
  
+	
  
ND	
  
-­‐	
  
	
  
G3	
  
L-­‐serine	
  
+	
  
+	
  
ND	
  
-­‐	
  
	
  
G4	
  
L_threonine	
  
+	
  
-­‐	
  
ND	
  
-­‐	
  
	
  
H1	
  
glycyl-­‐L-­‐proline	
  
+	
  
-­‐	
  
ND	
  
-­‐	
  
PM2:	
   B2	
   N-­‐acetyl-­‐neuraminic	
  acid*	
  
-­‐	
  
+	
  
+	
  
+	
  
	
  
B6	
  
D-­‐arabitol	
  
-­‐	
  
-­‐	
  
-­‐	
  
+	
  
	
  
B8	
  
arbutin	
  
+	
  
+	
  
+	
  
+	
  
	
  
C4	
  
D-­‐melezilose	
  
+	
  
+	
  
+	
  
+	
  
	
  
D2	
  
salicin	
  
+	
  
+	
  
+	
  
+	
  
	
  
D6	
  
D-­‐tagatose	
  
+	
  
+	
  
-­‐	
  
+	
  
	
  
E5	
  
D-­‐glucosamine	
  
+	
  
+	
  
+	
  
+	
  
	
  
E8	
  
b-­‐hydroxy-­‐butyric	
  acid	
  
-­‐	
  
+	
  
-­‐	
  
+	
  
	
  
E10	
  
α-­‐keto	
  valeric	
  acid	
  
-­‐	
  
+	
  
+	
  
+	
  
	
  
G8	
  
hydroxy-­‐L-­‐proline*	
  
-­‐	
  
-­‐	
  
+	
  
+	
  
	
  
G10	
  
L-­‐leucine	
  
-­‐	
  
+	
  
+	
  
+	
  
	
  
G12	
  
L-­‐methionine	
  
+	
  
-­‐	
  
-­‐	
  
+	
  
	
  
aEach	
  +/-­‐	
  symbol	
  represents	
  results	
  of	
  an	
  individual	
  experiment	
  where	
  growth	
  was	
  

(+)	
  or	
  was	
  not	
  (-­‐)	
  increased	
  by	
  at	
  least	
  1.5-­‐fold	
  that	
  of	
  the	
  control.	
  	
  DM=	
  defined	
  
medium;	
  MSM=	
  modified	
  sporulation	
  medium.	
  
*These	
  compounds	
  were	
  selected	
  for	
  individual	
  growth	
  analysis	
  experiments.	
  	
  	
  
	
  
	
  

167	
  

Plates	
  PM3-­‐8	
  were	
  also	
  tested	
  for	
  increased	
  growth.	
  	
  The	
  results	
  are	
  
presented	
  in	
  Supplementary	
  Table	
  4.1;	
  however,	
  we	
  did	
  not	
  do	
  any	
  follow-­‐up	
  
growth	
  studies	
  on	
  the	
  compounds	
  identified	
  in	
  these	
  plates.	
  	
  
	
  
Compounds	
  identified	
  in	
  Phenotype	
  MicroArray	
  plates	
  increase	
  growth	
  yield	
  
of	
  several	
  C.	
  difficile	
  strains.	
  	
  Several	
  compounds	
  that	
  conferred	
  growth	
  
advantages	
  were	
  selected	
  for	
  follow-­‐up	
  experiments	
  (starred	
  in	
  Table	
  4.4).	
  These	
  
compounds	
  were	
  selected	
  based	
  on	
  their	
  potential	
  for	
  availability	
  in	
  the	
  intestinal	
  
environment.	
  	
  For	
  example,	
  N-­‐acetyl-­‐glucosamine	
  is	
  a	
  building	
  block	
  of	
  bacterial	
  
peptidoglycan.	
  	
  This	
  compound	
  may	
  have	
  increased	
  availability	
  during	
  periods	
  of	
  
bacterial	
  lysis	
  due	
  to	
  antibiotic	
  activity.	
  	
  N-­‐acetyl-­‐neuraminic	
  acid	
  is	
  a	
  building	
  block	
  
of	
  mucin,	
  a	
  substance	
  secreted	
  by	
  the	
  intestinal	
  epithelium.	
  	
  Reduction	
  in	
  bacterial	
  
levels	
  within	
  the	
  gastrointestinal	
  tract	
  following	
  antibiotic	
  treatment	
  may	
  increase	
  
access	
  to	
  this	
  compound.	
  	
  Additionally,	
  we	
  know	
  that	
  C.	
  difficile	
  is	
  capable	
  of	
  
generating	
  ATP	
  through	
  Stickland	
  reactions,	
  the	
  coupled	
  reduction	
  and	
  oxidation	
  of	
  
amino	
  acids	
  (4).	
  	
  Alanine	
  and	
  hydroxy-­‐proline	
  have	
  been	
  demonstrated	
  to	
  serve	
  as	
  a	
  
Stickland	
  pair	
  for	
  this	
  reaction.	
  	
  Mannose	
  is	
  a	
  six-­‐carbon	
  sugar	
  that	
  is	
  naturally	
  
occurring	
  and	
  often	
  a	
  component	
  of	
  cell-­‐surface	
  glycoproteins	
  as	
  well	
  as	
  present	
  in	
  
various	
  dietary	
  sources.	
  	
  	
  Finally,	
  as	
  presented	
  in	
  the	
  introduction,	
  trehalose	
  is	
  a	
  
glucose	
  disaccharide	
  that	
  has	
  several	
  lines	
  of	
  evidence	
  for	
  potential	
  gut-­‐availability.	
  	
  	
  
Several	
  C.	
  difficile	
  strains	
  (3	
  RT	
  027	
  and	
  4	
  RT	
  non-­‐027)	
  were	
  tested	
  for	
  the	
  
ability	
  to	
  grow	
  to	
  higher	
  cell	
  densities	
  on	
  these	
  selected	
  nutrient	
  sources;	
  the	
  results	
  
are	
  presented	
  in	
  figure	
  4.1.	
  	
  	
  The	
  defined	
  medium	
  was	
  used	
  for	
  these	
  experiments	
  

	
  

168	
  

and	
  compound	
  concentrations	
  were	
  chosen	
  based	
  on	
  the	
  concentration	
  range	
  given	
  
by	
  Biolog	
  (5-­‐50mM).	
  	
  N-­‐acetylglucosamine	
  (NAG)	
  and	
  N-­‐acetylneuraminic	
  acid	
  
(NANA)	
  increased	
  growth	
  yield	
  for	
  all	
  four	
  of	
  the	
  RT	
  non-­‐027	
  strains.	
  	
  Of	
  the	
  three	
  
RT	
  027	
  strains,	
  only	
  CD3017	
  increased	
  in	
  the	
  presence	
  of	
  NAG.	
  Interestingly,	
  
CD2015	
  did	
  not	
  have	
  increased	
  growth	
  in	
  these	
  experiments	
  in	
  the	
  presence	
  of	
  NAG	
  
or	
  NANA,	
  while	
  it	
  did	
  have	
  an	
  increase	
  in	
  the	
  Phenotype	
  MicroArray	
  plates.	
  	
  This	
  
may	
  be	
  an	
  indication	
  that	
  CD2015	
  is	
  more	
  sensitive	
  to	
  growth	
  conditions	
  for	
  
utilization	
  of	
  this	
  compound,	
  or	
  that	
  the	
  concentrations	
  used	
  in	
  the	
  follow-­‐up	
  
growth	
  cultures	
  were	
  lower	
  than	
  in	
  the	
  Phenotype	
  MicroArray	
  plates.	
  	
  The	
  amino	
  
acids	
  alanine	
  and	
  hydroxy-­‐proline	
  significantly	
  increased	
  growth	
  yield	
  only	
  for	
  
CD2015.	
  	
  Mannose	
  increased	
  growth	
  yield	
  for	
  all	
  strains	
  except	
  CD3014.	
  	
  All	
  strains	
  
tested	
  had	
  significant	
  increases	
  in	
  growth	
  yield	
  in	
  the	
  presence	
  of	
  trehalose.	
  	
  
Interestingly,	
  however,	
  when	
  comparing	
  the	
  fold-­‐increase	
  of	
  growth	
  in	
  trehalose-­‐
supplemented	
  medium	
  to	
  that	
  in	
  unsupplemented	
  medium,	
  the	
  RT027	
  strains	
  had	
  
an	
  average	
  of	
  3-­‐fold	
  increase	
  compared	
  to	
  1.2-­‐fold	
  increase	
  of	
  the	
  non-­‐RT	
  027	
  
strains	
  (except	
  for	
  CD630,	
  which	
  had	
  a	
  5-­‐fold	
  increase).	
  Taken	
  together,	
  these	
  data	
  
illustrate	
  the	
  proof	
  of	
  concept	
  that	
  Phenotype	
  MicroArray	
  plates	
  can	
  successfully	
  be	
  
used	
  to	
  identify	
  compounds	
  beneficial	
  to	
  C.	
  difficile	
  growth.	
  
	
  

	
  

169	
  

Maximum OD600

0.6

CD196 (027)
CD2015 (027)
CD3017 (027)
CD630 (012)
CD3014 (001)
CD4011 (001)
CD1007 (053)

0.5
0.4
0.3
0.2
0.1

H

N
yd Ala AN
ro ni A
xy ne
-P +
ro
lin
e
M
an
no
se
Tr
eh
al
os
e

G
A
N

D

M

-G

0.0

	
  
Figure	
  4.1.	
  Results	
  of	
  growth	
  yield	
  experiments	
  using	
  compounds	
  selected	
  
from	
  Phenotype	
  MicroArray	
  plates.	
  Results	
  are	
  reported	
  as	
  the	
  average	
  of	
  three	
  
independent	
  growth	
  curves	
  with	
  associated	
  standard	
  deviations.	
  (DM-­‐G=	
  defined	
  
medium	
  without	
  glucose,	
  NAG=	
  N-­‐acetyl-­‐glucosamine,	
  NANA=	
  N-­‐acetyl-­‐neuraminic	
  
acid)	
  
	
  
Characterization	
  of	
  growth	
  phenotypes	
  of	
  C.	
  difficile	
  strains	
  grown	
  on	
  
trehalose.	
  	
  With	
  the	
  observation	
  that	
  RT	
  027	
  strains	
  had	
  higher	
  fold-­‐increases	
  than	
  
RT	
  non-­‐027	
  strains	
  in	
  the	
  presence	
  of	
  trehalose	
  in	
  the	
  follow-­‐up	
  studies	
  presented	
  
above,	
  we	
  further	
  investigated	
  the	
  growth	
  dynamics	
  of	
  these	
  strains	
  in	
  the	
  presence	
  
of	
  this	
  sugar.	
  	
  	
  	
  Four	
  strains	
  were	
  selected	
  (two	
  RT	
  027	
  and	
  two	
  RT	
  non-­‐027)	
  and	
  
grown	
  on	
  a	
  range	
  of	
  trehalose	
  concentrations	
  to	
  determine	
  if	
  the	
  differences	
  in	
  
growth	
  yield	
  were	
  concentration-­‐dependent.	
  	
  	
  The	
  results	
  are	
  presented	
  in	
  Figure	
  
4.2.	
  	
  It	
  was	
  observed	
  that	
  the	
  RT	
  027	
  strain	
  cell	
  densities	
  began	
  to	
  increase	
  at	
  lower	
  
levels	
  of	
  trehalose	
  than	
  the	
  RT	
  non-­‐027	
  strains,	
  showing	
  significant	
  differences	
  as	
  
low	
  as	
  750uM	
  (p-­‐value<0.05,	
  student’s	
  t-­‐test),	
  while	
  the	
  RT	
  non-­‐027	
  strains	
  did	
  not	
  

	
  

170	
  

have	
  significant	
  differences	
  in	
  cell	
  densities	
  even	
  at	
  50mM	
  trehalose,	
  although	
  they	
  
were	
  trending	
  higher.	
  The	
  variation	
  in	
  culture	
  final	
  OD’s	
  is	
  high	
  at	
  these	
  higher	
  
concentrations,	
  however,	
  and	
  so	
  repetition	
  of	
  growth	
  experiments	
  would	
  likely	
  
result	
  in	
  significant	
  differences.	
  	
  Nonetheless,	
  it	
  is	
  clear	
  that	
  RT	
  027	
  strains	
  increase	
  
growth	
  yield	
  in	
  the	
  presence	
  of	
  trehalose	
  at	
  much	
  lower	
  concentrations.	
  	
  	
  
	
  

Maximum OD600

0.3

0
750uM
1.5uM
3mM
6.25mM
12.5mM
25mM

0.2

0.1

50mM
0.0

CD2015
(027)

CD3017
(027)

CD3014
(001)

CD4011
(001)

	
  
Figure	
  4.2.	
  Maximum	
  growth	
  yield	
  of	
  C.	
  difficile	
  strains	
  grown	
  in	
  the	
  presence	
  
of	
  a	
  range	
  of	
  trehalose	
  concentrations	
  in	
  defined	
  medium.	
  	
  Ribotype	
  027	
  strains	
  
increase	
  growth	
  yield	
  at	
  lower	
  trehalose	
  concentrations.	
  	
  Results	
  are	
  reported	
  as	
  the	
  
average	
  of	
  three	
  independent	
  growth	
  curves	
  with	
  associated	
  standard	
  deviations.	
  
	
  
	
  

Using	
  a	
  concentration	
  where	
  there	
  were	
  significant	
  differences	
  between	
  the	
  

growth	
  yields	
  of	
  RT	
  027	
  and	
  non-­‐027	
  strains,	
  the	
  growth	
  of	
  several	
  more	
  C.	
  difficile	
  
strains	
  of	
  other	
  ribotypes	
  was	
  assessed.	
  	
  The	
  results	
  are	
  presented	
  in	
  figure	
  4.3,	
  
showing	
  growth	
  yields	
  of	
  C.	
  difficile	
  strains	
  grown	
  in	
  DM	
  with	
  and	
  without	
  
supplementation	
  with	
  10mM	
  trehalose.	
  	
  	
  All	
  five	
  RT	
  027	
  strains	
  used	
  had	
  
significantly	
  higher	
  maximum	
  OD’s	
  in	
  trehalose-­‐supplemented	
  DM	
  (p-­‐value	
  <0.05,	
  

	
  

171	
  

student’s	
  t-­‐test),	
  while	
  all	
  six	
  RT	
  non-­‐027	
  strains	
  did	
  not.	
  	
  We	
  also	
  tested	
  three	
  RT	
  
078	
  strains	
  (another	
  epidemic-­‐associated	
  RT	
  of	
  high	
  prevalence),	
  and	
  observed	
  that	
  
they	
  also	
  had	
  significantly	
  higher	
  growth	
  yields	
  in	
  the	
  presence	
  of	
  10mM	
  trehalose.	
  	
  	
  
	
  

Maximum OD600

0.6

0.4

DM
10mM
trehalose

0.2

C
D
2
C 015
D (
C 19 0
D 6 2
C 401 (0 7)
D 5 2
C 400 (0 7)
D 1 27
4
C 010 (02 )
D
C 63 (0 7)
D 0 27
C 301 (0 )
D 4 1
C 204 (0 2)
D 8 0
C 100 (0 1)
D 7 5
C 101 (0 3)
D 4 5
C 201 (0 3)
D 2 1
C 101 (0 4)
D 5 0
C 101 (0 2)
D 8 78
20 ( )
01 07
(0 8)
78
)

0.0

Strain (ribotype)

	
  
	
  
Figure	
  	
  4.3.	
  Growth	
  yield	
  of	
  C.	
  difficile	
  strains	
  belonging	
  to	
  several	
  ribotype	
  
groups.	
  	
  Strains	
  were	
  grown	
  in	
  defined	
  medium	
  (DM)	
  +/-­‐	
  10mM	
  trehalose.	
  	
  	
  RT027	
  
and	
  RT078	
  strains	
  grow	
  to	
  higher	
  cell	
  densities	
  on	
  10mM	
  trehalose.	
  	
  Results	
  are	
  
reported	
  as	
  the	
  average	
  of	
  independent	
  replicates	
  with	
  associated	
  standard	
  
deviations.	
  
	
  
Construction	
  of	
  a	
  treA	
  mutant	
  and	
  analysis	
  of	
  the	
  growth	
  phenotype.	
  	
  The	
  
ability	
  to	
  utilize	
  trehalose	
  as	
  a	
  carbon	
  source	
  is	
  dependent	
  on	
  a	
  cell’s	
  ability	
  to	
  break	
  
the	
  disaccharide	
  bond,	
  generating	
  two	
  molecules	
  of	
  glucose	
  that	
  can	
  then	
  be	
  
shuttled	
  into	
  the	
  glycolytic	
  pathway	
  of	
  central	
  metabolism.	
  	
  In	
  C.	
  difficile,	
  a	
  
trehalose-­‐6-­‐phosphate	
  hydrolase	
  enzyme,	
  TreA,	
  fulfills	
  this	
  function.	
  	
  Using	
  a	
  group-­‐

	
  

172	
  

II	
  intron	
  based	
  gene	
  knockout	
  system,	
  a	
  treA	
  knockout	
  mutant	
  was	
  constructed	
  in	
  
the	
  CD630Δerm	
  background.	
  	
  	
  The	
  treA	
  mutant	
  and	
  CD630	
  parent	
  were	
  then	
  grown	
  
in	
  DM	
  supplemented	
  with	
  glucose,	
  trehalose,	
  and	
  the	
  other	
  glucose	
  disaccharide	
  
sugars,	
  cellobiose	
  and	
  maltose	
  (figure	
  4.4).	
  	
  The	
  growth	
  experiments	
  confirmed	
  that	
  
the	
  mutant	
  is	
  unable	
  to	
  metabolize	
  trehalose;	
  there	
  was	
  no	
  increase	
  in	
  growth	
  yield	
  
in	
  DM	
  supplemented	
  with	
  25mM	
  trehalose	
  compared	
  to	
  unsupplemented	
  DM,	
  while	
  
the	
  WT	
  strain	
  had	
  a	
  4-­‐fold	
  increase.	
  	
  It	
  is	
  worth	
  noting	
  that	
  while	
  CD630	
  is	
  not	
  a	
  RT	
  
027	
  strain,	
  it	
  does	
  grow	
  better	
  than	
  other	
  non-­‐027	
  RT	
  strains	
  on	
  trehalose,	
  
presumably	
  due	
  to	
  its	
  lab	
  adaption	
  to	
  rich	
  media	
  (BHIS),	
  which	
  contains	
  trehalose	
  
because	
  of	
  the	
  addition	
  of	
  yeast	
  extract.	
  	
  This	
  supports	
  that	
  treA	
  encodes	
  a	
  protein	
  
required	
  for	
  trehalose	
  metabolism.	
  	
  Moreover,	
  knockout	
  of	
  treA	
  does	
  not	
  affect	
  
metabolism	
  of	
  glucose,	
  cellobiose,	
  or	
  maltose,	
  since	
  the	
  growth	
  yields	
  of	
  the	
  WT	
  and	
  
mutant	
  strains	
  were	
  essentially	
  the	
  same	
  in	
  media	
  supplemented	
  with	
  these	
  
different	
  sugars.	
  	
  This	
  observation	
  is	
  evidence	
  that	
  the	
  glusosidase	
  activity	
  of	
  TreA	
  is	
  
specific	
  to	
  trehalose	
  and	
  not	
  other	
  glucose	
  disaccharides.	
  	
  Moreover,	
  the	
  phenotype	
  
of	
  this	
  mutant	
  is	
  specific	
  to	
  its	
  ability	
  to	
  utilize	
  trehalose	
  and	
  therefore	
  can	
  be	
  used	
  
for	
  in	
  vivo	
  or	
  in	
  vitro	
  experiments	
  to	
  study	
  the	
  importance	
  of	
  trehalose	
  metabolism	
  
to	
  C.	
  difficile	
  colonization	
  and	
  virulence.	
  	
  	
  	
  
	
  

	
  

173	
  

Maximum OD600

0.5
0.4
0.3

WT
treA KO

0.2
0.1
0.0

cellobiose glucose maltose trehalose

DM

	
  
Figure	
  4.4.	
  	
  	
  Growth	
  yields	
  of	
  CD630	
  wild-­‐type	
  (WT)	
  and	
  treA	
  knock-­‐out	
  
mutant	
  in	
  defined	
  medium	
  (DM)	
  supplemented	
  with	
  glucose	
  and	
  glucose	
  
disaccharide	
  sugars	
  (25mM).	
  
	
  
Alignments	
  of	
  treR,	
  the	
  repressor	
  of	
  trehalase	
  (treA),	
  reveal	
  a	
  conserved	
  
leucine	
  residue	
  that	
  is	
  substituted	
  with	
  isoleucine	
  in	
  RT	
  027	
  strains.	
  	
  	
  
The	
  treA	
  (locus	
  tag	
  CD3091)	
  gene	
  is	
  located	
  on	
  the	
  chromosome	
  immediately	
  
downstream	
  of	
  a	
  gene	
  that	
  codes	
  for	
  its	
  cognate	
  transcriptional	
  repressor,	
  TreR	
  
(locus	
  tag	
  CD3090)	
  (Figure	
  4.5).	
  	
  	
  
	
  

treR	
  (CD3090)

treA	
  (CD3091)
	
  

Figure	
  4.5.	
  	
  Trehalose	
  utilization	
  genes	
  located	
  on	
  the	
  C.	
  difficile	
  chromosome.	
  	
  
The	
  trehalase	
  gene,	
  treA,	
  is	
  located	
  directly	
  down	
  stream	
  of	
  treR,	
  the	
  trehalase	
  
transcriptional	
  repressor.	
  	
  
	
  
	
  

	
  

174	
  

	
  

We	
  hypothesized	
  that	
  the	
  ability	
  of	
  RT	
  027	
  and	
  078	
  strains	
  to	
  grow	
  better	
  on	
  

trehalose	
  was	
  due	
  to	
  a	
  genetic	
  difference	
  in	
  this	
  trehalose-­‐utilization	
  locus.	
  	
  Whole	
  
genome	
  sequencing	
  of	
  several	
  of	
  the	
  clinical	
  isolates	
  used	
  in	
  the	
  trehalose	
  growth	
  
studies	
  above	
  was	
  conducted.	
  	
  By	
  aligning	
  the	
  gene	
  sequences	
  pulled	
  out	
  of	
  those	
  
genome	
  sequence	
  files,	
  we	
  identified	
  a	
  conserved	
  amino	
  acid	
  substitution	
  in	
  TreR	
  of	
  
the	
  RT	
  027	
  strains	
  that	
  is	
  not	
  present	
  in	
  the	
  other	
  RT	
  strains	
  (figure	
  4.6).	
  	
  This	
  
substitution	
  corresponds	
  to	
  a	
  cytidine	
  to	
  adenine	
  mutation	
  in	
  the	
  gene	
  sequence	
  at	
  
bp	
  514,	
  resulting	
  in	
  a	
  leucine	
  to	
  isoleucine	
  substitution	
  at	
  AA172.	
  	
  	
  	
  
	
  
	
  

	
  
	
  
Figure	
  4.6.	
  Alignments	
  of	
  TreR	
  amino	
  acid	
  sequences	
  for	
  several	
  C.	
  difficile	
  
strains	
  of	
  various	
  ribotypes.	
  	
  RT027	
  strains	
  have	
  a	
  conserved	
  L172I	
  amino	
  acid	
  
substitution	
  (circled	
  in	
  red)	
  not	
  present	
  in	
  the	
  non-­‐RT027	
  strains	
  or	
  the	
  RT078	
  
strain.	
  	
  
	
  
	
  
	
  

In	
  order	
  to	
  investigate	
  the	
  conservation	
  of	
  this	
  residue	
  in	
  TreR	
  proteins	
  of	
  

other	
  organisms,	
  we	
  aligned	
  the	
  TreR	
  amino	
  acid	
  sequences	
  from	
  10	
  other	
  Gram-­‐

	
  

175	
  

positive	
  bacteria	
  (figure	
  4.7).	
  	
  	
  Indeed,	
  this	
  leucine	
  residue	
  was	
  conserved	
  across	
  all	
  
of	
  the	
  organisms	
  included	
  in	
  our	
  alignments,	
  suggesting	
  it	
  is	
  has	
  structural	
  or	
  
functional	
  importance.	
  	
  	
  	
  
	
  
	
  

	
  

Figure	
  4.7.	
  Alignments	
  of	
  TreR	
  amino	
  acid	
  sequences	
  from	
  several	
  Gram-­‐
positive	
  bacterial	
  organisms	
  showing	
  conservation	
  of	
  the	
  leucine	
  residue.	
  	
  	
  
	
  
	
  
	
  
	
  

The	
  only	
  available	
  crystal	
  structure	
  of	
  a	
  TreR	
  protein	
  homologous	
  to	
  that	
  in	
  

C.	
  difficile	
  was	
  published	
  by	
  Rezacove	
  et	
  al.	
  who	
  studied	
  the	
  C-­‐terminal	
  (effector-­‐
binding)	
  domain	
  of	
  TreR	
  in	
  Bacillus	
  subtilis	
  (33).	
  	
  This	
  conserved	
  leucine	
  (AA169)	
  is	
  
located	
  in	
  Helix	
  H3	
  of	
  the	
  structure	
  (Figure	
  4.8).	
  	
  	
  
	
  

	
  

176	
  

Rezacove	
  et	
  al.,	
  Proteins,	
  2007	
  	
  

leucine
Trehalose-­‐6-­‐P	
  binding	
  pocket

	
  

Figure	
  4.8.	
  	
  	
  Ribbon	
  diagram	
  of	
  the	
  C-­‐terminal	
  (effector-­‐binding	
  domain)	
  of	
  
TreR	
  from	
  Bacillus	
  subtilis	
  showing	
  the	
  locations	
  of	
  the	
  predicted	
  trehalose-­‐6-­‐
phosephate	
  binding	
  pocket	
  and	
  conserved	
  leucine	
  residue	
  (Leu-­‐169).	
  	
  
Structure	
  was	
  determined	
  and	
  published	
  by	
  Rezacova	
  et	
  al.	
  (33).	
  
	
  
	
  
	
  

TreR	
  functions	
  as	
  a	
  dimer	
  of	
  dimers,	
  and	
  this	
  helix	
  (H3)	
  was	
  shown	
  to	
  be	
  

important	
  for	
  interaction	
  between	
  subunits	
  of	
  the	
  two	
  dimers.	
  	
  When	
  trehalose	
  is	
  
absent,	
  the	
  dimers	
  form	
  a	
  tetramer	
  and	
  bind	
  to	
  two	
  operator	
  sequences	
  on	
  the	
  
chromosome,	
  upstream	
  of	
  the	
  trehalose	
  operon.	
  	
  These	
  14	
  bp	
  inverted	
  repeats	
  are	
  
located	
  32	
  bp	
  apart	
  within	
  the	
  promoter	
  sequence	
  of	
  the	
  trehalsoe	
  operon	
  in	
  B.	
  
subtilis	
  (34).	
  	
  When	
  trehalose	
  is	
  present	
  in	
  the	
  medium,	
  it	
  gets	
  phosphorylated	
  to	
  
trehalose-­‐6-­‐phoasphate	
  during	
  uptake	
  into	
  the	
  cell	
  by	
  way	
  of	
  a	
  trehalose-­‐specific	
  
phosphotransferase	
  transporter	
  system.	
  	
  Trehalose-­‐6-­‐phosphate	
  then	
  binds	
  to	
  TreR,	
  
causing	
  a	
  conformational	
  change	
  that	
  dissociates	
  TreR	
  from	
  the	
  DNA,	
  allowing	
  
expression	
  of	
  treA.	
  	
  Additionally,	
  the	
  location	
  of	
  the	
  leucine	
  in	
  the	
  structure	
  lies	
  very	
  
	
  

177	
  

close	
  (just	
  a	
  few	
  amino	
  acids	
  away)	
  from	
  the	
  predicted	
  effector-­‐binding	
  pocket,	
  
more	
  specifically,	
  where	
  the	
  phosphoryl	
  group	
  of	
  trehalose-­‐6-­‐phosephate	
  (the	
  
effector)	
  is	
  predicted	
  to	
  interact	
  with	
  key	
  amino	
  acid	
  residues	
  (33).	
  	
  Both	
  leucine	
  
and	
  isoleucine	
  are	
  hydrophobic	
  amino	
  acids	
  containing	
  branched,	
  non-­‐polar	
  side	
  
chains.	
  	
  While	
  the	
  chemistry	
  of	
  these	
  two	
  amino	
  acids	
  is	
  similar,	
  the	
  slight	
  
differences	
  in	
  their	
  structures	
  may	
  allow	
  for	
  a	
  change	
  in	
  structure	
  or	
  function	
  of	
  the	
  
protein.	
  	
  Since	
  this	
  amino	
  acid	
  is	
  highly	
  conserved,	
  it	
  is	
  likely	
  that	
  a	
  substitution	
  to	
  
an	
  amino	
  acid	
  with	
  significantly	
  different	
  chemistry	
  would	
  result	
  in	
  a	
  structural	
  or	
  
functional	
  change	
  that	
  could	
  be	
  deleterious,	
  potentially	
  even	
  resulting	
  in	
  complete	
  
loss	
  of	
  function.	
  	
  	
  Based	
  on	
  the	
  location	
  of	
  this	
  substitution,	
  we	
  predict	
  that	
  the	
  RT	
  
027	
  TreR	
  could	
  have	
  a	
  change	
  in	
  function,	
  which	
  allows	
  increased	
  induction	
  of	
  treA	
  
expression.	
  	
  This	
  could	
  occur	
  by	
  destabilization	
  of	
  the	
  TreR	
  tetramer,	
  since	
  AA172	
  is	
  
located	
  in	
  helix	
  H3,	
  which	
  is	
  important	
  for	
  dimer-­‐dimer	
  interactions.	
  	
  This	
  would	
  
allow	
  for	
  the	
  dimers	
  of	
  TreR	
  to	
  more	
  readily	
  dissociate	
  upon	
  binding	
  trehalose-­‐6-­‐
phosphate.	
  	
  Alternatively,	
  TreR	
  could	
  have	
  a	
  higher	
  affinity	
  for	
  trehalose-­‐6-­‐
phosphate,	
  because	
  of	
  AA172’s	
  close	
  proximity	
  to	
  the	
  effector-­‐binding	
  pocket,	
  which	
  
would	
  also	
  allow	
  for	
  TreR	
  to	
  more	
  readily	
  dissociate	
  from	
  the	
  operator	
  region.	
  	
  
Therefore,	
  we	
  hypothesize	
  that	
  RT	
  027	
  strains	
  are	
  able	
  to	
  grow	
  to	
  higher	
  cell	
  
densities	
  on	
  lower	
  levels	
  of	
  trehalose	
  due	
  to	
  increased	
  expression	
  of	
  treA.	
  	
  	
  
	
  

Since	
  the	
  RT	
  078	
  strains	
  do	
  not	
  have	
  this	
  amino	
  acid	
  substitution,	
  the	
  basis	
  

for	
  their	
  increased	
  growth	
  on	
  trehalose	
  must	
  be	
  the	
  result	
  of	
  some	
  other	
  genetic	
  
difference.	
  	
  	
  Through	
  personal	
  correspondence	
  with	
  Wilco	
  Knetsch	
  (lab	
  of	
  Dr.	
  Ed	
  
Kuijper,	
  Leiden	
  University	
  Medical	
  Center),	
  they	
  have	
  shared	
  that	
  they	
  recently	
  

	
  

178	
  

identified	
  a	
  genetic	
  insertion	
  in	
  RT	
  078	
  strains,	
  which	
  contains	
  genes	
  annotated	
  as	
  
trehalose-­‐utilization	
  genes	
  (unpublished	
  data).	
  	
  These	
  genes	
  may	
  be	
  the	
  genetic	
  
basis	
  for	
  the	
  trehalose	
  growth	
  phenotype	
  of	
  RT	
  078	
  strains;	
  however,	
  it	
  is	
  possible	
  
that	
  other	
  genetic	
  characteristics	
  are	
  playing	
  a	
  role.	
  	
  	
  	
  Regardless,	
  the	
  observation	
  of	
  
two	
  independent	
  lineages	
  of	
  epidemic-­‐associated	
  C.	
  difficile	
  acquiring	
  the	
  ability	
  to	
  
grow	
  better	
  on	
  trehalose	
  supports	
  that	
  this	
  metabolic	
  ability	
  may	
  play	
  a	
  role	
  in	
  C.	
  
difficile	
  colonization	
  and	
  disease.	
  	
  	
  	
  
	
  
RT	
  027	
  strains	
  have	
  increased	
  treA	
  expression.	
  	
  In	
  order	
  to	
  test	
  the	
  hypothesis	
  
that	
  RT	
  027	
  strains	
  have	
  increased	
  treA	
  expression,	
  we	
  grew	
  four	
  RT	
  027	
  strains,	
  
and	
  four	
  strains	
  belonging	
  to	
  different	
  RT	
  groups,	
  in	
  DM	
  supplemented	
  with	
  25mM	
  
trehalose	
  and	
  compared	
  the	
  levels	
  of	
  treA	
  mRNA	
  present	
  during	
  mid-­‐log	
  phase	
  of	
  
growth	
  using	
  RT	
  q-­‐PCR.	
  	
  The	
  CT	
  values	
  of	
  each	
  replicate	
  were	
  all	
  normalized	
  to	
  an	
  
average	
  baseline	
  CT	
  value	
  of	
  replicates	
  grown	
  in	
  medium	
  without	
  trehalose.	
  	
  The	
  
results	
  are	
  presented	
  in	
  figure	
  4.9.	
  	
  The	
  mean	
  values	
  of	
  the	
  four	
  RT	
  027	
  strains	
  are	
  
2599,	
  3339,	
  3451,	
  and	
  3752.	
  	
  The	
  mean	
  values	
  of	
  the	
  four	
  RT	
  non-­‐027	
  strains	
  are	
  
392,	
  780,	
  1395,	
  and	
  603.	
  	
  The	
  difference	
  treA	
  expression	
  in	
  RT	
  027	
  strains	
  
compared	
  to	
  RT	
  the	
  non-­‐027	
  strains	
  in	
  these	
  conditions	
  is	
  significant;	
  the	
  p-­‐value	
  
(student’s	
  t-­‐test)	
  calculated	
  from	
  groups	
  of	
  mean	
  values	
  from	
  the	
  RT	
  027	
  and	
  non-­‐
027	
  groups	
  is	
  0.00026.	
  	
  There	
  is	
  a	
  large	
  amount	
  of	
  variability	
  between	
  replicates	
  
within	
  some	
  strains,	
  and	
  this	
  may	
  be	
  due	
  to	
  the	
  concentration	
  of	
  trehalose	
  used	
  in	
  
these	
  experiments.	
  	
  This	
  concentration	
  is	
  potentially	
  close	
  to	
  a	
  threshold	
  
concentration	
  where	
  treA	
  expression	
  is	
  induced,	
  since	
  there	
  is	
  a	
  slight	
  increase	
  in	
  

	
  

179	
  

growth	
  yield	
  at	
  this	
  concentration	
  in	
  the	
  RT	
  non-­‐027	
  strains	
  shown	
  in	
  Figure	
  4.2.	
  	
  A	
  
lower	
  concentration	
  of	
  trehalose	
  (≤10mM)	
  might	
  yield	
  more	
  consistent	
  results,	
  
however,	
  the	
  lower	
  cell	
  densities	
  of	
  the	
  cultures	
  would	
  make	
  obtaining	
  adequate	
  
amounts	
  of	
  RNA	
  more	
  challenging.	
  	
  Regardless,	
  there	
  is	
  still	
  a	
  strong	
  correlation	
  
between	
  strains	
  that	
  have	
  increased	
  growth	
  on	
  trehalose	
  and	
  increased	
  expression	
  
of	
  treA.	
  	
  	
  One	
  exception	
  is	
  the	
  RT	
  078	
  strain,	
  CD1015,	
  which	
  has	
  significantly	
  
increased	
  growth	
  on	
  10mM	
  trehalose	
  yet	
  I	
  did	
  not	
  observe	
  increased	
  treA	
  
expression	
  in	
  this	
  strain.	
  	
  	
  It	
  would	
  be	
  interesting	
  to	
  test	
  other	
  RT	
  078	
  strains	
  to	
  
determine	
  if	
  they	
  also	
  do	
  not	
  have	
  increased	
  treA	
  levels.	
  	
  If	
  not,	
  this	
  might	
  provide	
  
insight	
  into	
  the	
  genetic	
  basis	
  for	
  their	
  growth	
  phenotype.	
  	
  For	
  example,	
  it	
  might	
  
indicate	
  that	
  the	
  trehalase	
  enzyme	
  they	
  carry	
  has	
  increased	
  catalytic	
  efficiency	
  or	
  
affinity	
  for	
  trehalose-­‐6-­‐phosphate.	
  	
  Alternatively,	
  it	
  might	
  provide	
  evidence	
  for	
  the	
  
presence	
  of	
  an	
  additional	
  trehalase	
  gene,	
  which	
  is	
  suggested	
  in	
  the	
  unique	
  insertion	
  
in	
  these	
  strains	
  that	
  is	
  discussed	
  above.	
  	
  	
  	
  	
  	
  
	
  
	
  

	
  

180	
  

Fold increase in
treA expression

10000

1000

C
D
20
C 15
D
40 (02
7)
1
C 5 (0
D
40 27
10 )
(0
C
D
19 27)
6
C
(
D
30 027
)
14
C
(0
D
0
10
15 1)
C
(
D
10 078
)
14
(0
C
D
63 14)
0
(0
12
)

100

Strain (ribotype)

	
  
Figure	
  4.9.	
  Expression	
  of	
  treA	
  in	
  several	
  C.	
  difficile	
  strains	
  grown	
  in	
  DM+25mM	
  
trehalose	
  as	
  determined	
  by	
  RT-­‐qPCR.	
  	
  Plotted	
  points	
  represent	
  the	
  fold-­‐increase	
  
of	
  each	
  replicate,	
  normalized	
  to	
  an	
  average	
  baseline	
  expression	
  in	
  medium	
  without	
  
trehalose.	
  	
  Replicates	
  of	
  each	
  strain	
  are	
  combined	
  data	
  from	
  2	
  or	
  3	
  independent	
  
experiments;	
  black	
  bars	
  represent	
  the	
  means.	
  	
  The	
  RT	
  027	
  strains	
  (four	
  on	
  left	
  side	
  
of	
  graph)	
  expressed	
  treA	
  to	
  higher	
  levels	
  under	
  the	
  conditions	
  tested.	
  	
  	
  The	
  p-­‐value	
  
(student’s	
  t-­‐test)	
  between	
  groups	
  of	
  RT	
  027	
  and	
  RT	
  non-­‐027	
  strain	
  averages	
  is	
  
0.00026.	
  	
  	
  
	
  
	
  
A	
  treA	
  knockout	
  mutant	
  of	
  C.	
  difficile	
  displays	
  a	
  decrease	
  in	
  colonization	
  levels	
  
in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection	
  and	
  a	
  decrease	
  in	
  competitive	
  fitness	
  
compared	
  to	
  the	
  wild	
  type	
  strain.	
  	
  In	
  order	
  to	
  investigate	
  if	
  the	
  ability	
  to	
  utilize	
  
trehalose	
  plays	
  a	
  role	
  in	
  C.	
  difficile	
  colonization	
  in	
  vivo,	
  we	
  infected	
  groups	
  of	
  mice	
  
with	
  either	
  the	
  wild-­‐type	
  CD630Δerm,	
  a	
  treA	
  knockout	
  mutant	
  of	
  CD630Δerm,	
  or	
  a	
  
mixture	
  of	
  the	
  two	
  strains.	
  	
  Mice	
  were	
  treated	
  with	
  an	
  antibiotic,	
  cefoperazone,	
  for	
  
several	
  days	
  in	
  order	
  to	
  induce	
  susceptibility	
  to	
  C.	
  difficile	
  infection,	
  then	
  gavaged	
  

	
  

181	
  

with	
  104	
  spores	
  of	
  the	
  desired	
  strain/s.	
  	
  Fecal	
  samples	
  were	
  collected	
  daily	
  and	
  C.	
  
difficile	
  was	
  enumerated	
  by	
  either	
  plating	
  (wild-­‐type,	
  mutant,	
  and	
  competition	
  
groups)	
  or	
  qPCR	
  (competition	
  group).	
  	
  	
  	
  The	
  CFU/g	
  feces	
  as	
  determined	
  by	
  plating	
  
on	
  C.	
  difficile	
  selective	
  media	
  of	
  the	
  mice	
  infected	
  with	
  either	
  the	
  wild-­‐type	
  or	
  
mutant	
  strains	
  is	
  plotted	
  is	
  figure	
  4.10.	
  	
  	
  The	
  fecal	
  levels	
  ranged	
  from	
  1.2	
  x	
  107	
  to	
  1.8	
  
x	
  108	
  CFU/g	
  feces	
  across	
  all	
  five	
  days	
  for	
  the	
  mice	
  infected	
  with	
  wild-­‐type	
  strain,	
  and	
  
the	
  range	
  for	
  the	
  mutant-­‐infected	
  mice	
  is	
  from	
  1.0	
  x	
  107	
  to	
  8.6	
  x	
  107	
  CFU/g	
  feces.	
  	
  	
  
Interestingly,	
  there	
  is	
  a	
  significant	
  difference	
  between	
  the	
  CFU/g	
  feces	
  in	
  the	
  wild-­‐
type	
  and	
  mutant	
  groups	
  on	
  days	
  1	
  and	
  2	
  (p=0.02	
  and	
  0.03,	
  respectively)	
  and	
  close	
  to	
  
significant	
  difference	
  on	
  day	
  3	
  (p=0.08),	
  with	
  the	
  mutant	
  colonizing	
  the	
  mice	
  to	
  
lower	
  levels	
  on	
  those	
  days	
  (~2-­‐fold	
  difference	
  between	
  means	
  on	
  all	
  three	
  days).	
  	
  
	
  
	
  
	
  
	
  

	
  

182	
  

CFU/g feces

109

wild-type

108

mutant
107

106

1

2

3

4

5

Day post-gavage

Figure	
  4.10.	
  	
  CFU/g	
  feces	
  of	
  wild-­‐type	
  and	
  treA	
  mutant	
  C.	
  difficile	
  infected	
  mice.	
  	
  
Two	
  groups	
  of	
  five	
  mice	
  were	
  gavaged	
  with	
  spores	
  of	
  either	
  wild-­‐type	
  CD630Δerm	
  
or	
  a	
  treA	
  knockout	
  mutant.	
  	
  Fecal	
  pellets	
  were	
  collected	
  daily	
  and	
  C.	
  difficile	
  
quanitifed	
  by	
  plating	
  on	
  C.	
  difficile	
  selective	
  media.	
  	
  Results	
  are	
  plotted	
  as	
  the	
  mean	
  
and	
  standard	
  deviation	
  of	
  the	
  data	
  points	
  from	
  each	
  group	
  on	
  the	
  indicated	
  days.	
  	
  	
  
*p-­‐value	
  <0.05	
  
	
  
	
  
	
  

In	
  the	
  mice	
  infected	
  with	
  a	
  mixture	
  of	
  wild-­‐type	
  and	
  mutant	
  spores,	
  a	
  

competitive	
  index	
  was	
  calculated	
  to	
  assess	
  the	
  advantage	
  of	
  having	
  a	
  functional	
  treA	
  
gene.	
  	
  Competitive	
  indices	
  (CI’s)	
  were	
  determined	
  by	
  dividing	
  the	
  wild-­‐type:mutant	
  
ratio	
  within	
  individual	
  mice	
  at	
  days	
  1,	
  2,	
  3,	
  and	
  5,	
  and	
  dividing	
  by	
  the	
  ratio	
  in	
  the	
  
gavaged	
  spore	
  mixture	
  (ratio	
  =	
  0.29).	
  	
  Quantitative	
  PCR	
  using	
  primers	
  specific	
  to	
  
either	
  strain	
  was	
  used	
  to	
  determine	
  the	
  strain	
  ratios.	
  	
  The	
  CI’s	
  are	
  plotted	
  in	
  figure	
  
4.11.	
  	
  Each	
  data	
  point	
  represents	
  the	
  CI	
  of	
  the	
  wild-­‐type	
  strain	
  in	
  an	
  individual	
  
mouse	
  on	
  the	
  indicated	
  day;	
  lines	
  connect	
  the	
  CI’s	
  within	
  each	
  mouse	
  across	
  time.	
  	
  A	
  
CI	
  of	
  one	
  would	
  indicate	
  no	
  change	
  in	
  ratio,	
  while	
  a	
  CI	
  greater	
  than	
  1	
  indicates	
  an	
  
increase	
  in	
  ratio	
  and	
  therefore	
  a	
  competitive	
  advantage	
  of	
  the	
  wild-­‐type	
  strain.	
  	
  
When	
  tracking	
  the	
  CI’s	
  across	
  time	
  in	
  the	
  individual	
  mice,	
  four	
  of	
  the	
  five	
  mice	
  had	
  

	
  

183	
  

CI’s	
  of	
  >1	
  on	
  or	
  before	
  day	
  3.	
  	
  Mouse	
  3	
  had	
  a	
  consistent	
  increase	
  in	
  CI	
  from	
  days	
  2	
  to	
  
5,	
  indicating	
  a	
  clear	
  competitive	
  advantage	
  for	
  the	
  wild-­‐type	
  strain.	
  	
  Mouse	
  1	
  had	
  
CI’s	
  >1	
  from	
  days	
  2	
  to	
  5	
  as	
  well,	
  however,	
  they	
  were	
  decreasing	
  as	
  the	
  experiment	
  
progressed.	
  	
  Mice	
  2	
  and	
  4	
  had	
  CI’s	
  >1	
  on	
  days	
  2	
  and	
  3,	
  respectively,	
  however,	
  they	
  
dropped	
  below	
  1	
  by	
  day	
  5.	
  	
  Mouse	
  5	
  CI’s	
  were	
  below	
  1	
  throughout	
  the	
  experiment	
  
with	
  the	
  lowest	
  CI	
  on	
  day	
  5	
  (0.6).	
  	
  In	
  general,	
  there	
  was	
  a	
  trend	
  of	
  decreasing	
  CI’s	
  
toward	
  the	
  end	
  of	
  the	
  experiment.	
  
	
  

One	
  observation	
  from	
  the	
  plating	
  data	
  was	
  that	
  the	
  total	
  number	
  of	
  C.	
  difficile	
  

in	
  all	
  mice	
  had	
  a	
  significant	
  decrease	
  in	
  CFU/g	
  feces	
  after	
  day	
  5	
  (figure	
  4.12).	
  	
  	
  There	
  
was	
  larger	
  drop	
  (up	
  to	
  over	
  2	
  logs)	
  in	
  levels	
  of	
  C.	
  difficile	
  in	
  the	
  wild-­‐type	
  and	
  
competition	
  mouse	
  groups,	
  while	
  the	
  drop	
  in	
  levels	
  in	
  the	
  mutant-­‐infected	
  mouse	
  
group	
  was	
  lower,	
  yet	
  still	
  significant	
  (up	
  to	
  13-­‐fold	
  decrease).	
  	
  This	
  observation	
  
indicates	
  that	
  the	
  intestinal	
  microbiota	
  of	
  the	
  mice	
  are	
  likely	
  recovering	
  from	
  the	
  
antibiotic-­‐induced	
  perturbation	
  and	
  resistance	
  to	
  C.	
  difficile	
  is	
  being	
  restored,	
  
therefore	
  resulting	
  in	
  C.	
  difficile	
  washout.	
  	
  Changes	
  in	
  the	
  intestinal	
  environment,	
  
including	
  nutrient	
  availability,	
  are	
  likely	
  in	
  this	
  situation	
  and	
  would	
  affect	
  the	
  
outcome	
  of	
  these	
  experiments,	
  particularly	
  the	
  competition	
  dynamics	
  of	
  the	
  wild-­‐
type	
  and	
  mutant	
  strains.	
  	
  Therefore,	
  it	
  is	
  reasonable	
  to	
  draw	
  conclusions	
  from	
  our	
  
data	
  based	
  on	
  the	
  earlier	
  time	
  points	
  in	
  the	
  experiment.	
  	
  	
  
	
  

In	
  summary,	
  there	
  were	
  significant	
  differences	
  in	
  the	
  wild-­‐type	
  and	
  mutant	
  

levels	
  of	
  C.	
  difficile	
  in	
  the	
  individually-­‐infected	
  mice	
  on	
  days	
  1	
  and	
  2.	
  	
  In	
  addition,	
  CI’s	
  
in	
  the	
  competition	
  mice	
  within	
  the	
  first	
  3	
  days	
  of	
  infection	
  indicated	
  a	
  competitive	
  

	
  

184	
  

advantage	
  of	
  the	
  wild-­‐type	
  strain.	
  	
  	
  Combined,	
  these	
  data	
  support	
  the	
  hypothesis	
  
that	
  trehalose	
  plays	
  a	
  role	
  in	
  colonization	
  and	
  competitive	
  fitness	
  in	
  vivo.	
  	
  	
  	
  
	
  

In	
  consideration	
  of	
  the	
  variability	
  in	
  wild-­‐type	
  CI’s	
  among	
  mice	
  in	
  the	
  

competition	
  group,	
  and	
  the	
  washout	
  of	
  total	
  C.	
  difficile	
  levels	
  in	
  all	
  mouse	
  groups	
  
beyond	
  day	
  5	
  of	
  the	
  experiment,	
  the	
  data	
  are	
  not	
  as	
  robust	
  as	
  desired	
  and	
  warrant	
  a	
  
repeat	
  of	
  the	
  experiment.	
  	
  Nonetheless,	
  a	
  recent	
  study	
  published	
  by	
  Ng	
  et	
  al.	
  
reported	
  similar	
  differences	
  in	
  colonization	
  levels	
  of	
  wild-­‐type	
  C.	
  difficile	
  and	
  a	
  
mutant	
  deficient	
  in	
  sialic	
  acid	
  consumption	
  in	
  a	
  mouse	
  model,	
  with	
  the	
  mutant	
  
colonizing	
  at	
  levels	
  2.3-­‐fold	
  lower	
  than	
  the	
  wild-­‐type	
  (35).	
  	
  	
  Unlike	
  sialic	
  acid,	
  which	
  
is	
  a	
  component	
  of	
  mucin,	
  trehalose	
  availability	
  in	
  the	
  human	
  gastrointestinal	
  tract	
  is	
  
at	
  least	
  partially	
  dependent	
  on	
  diet	
  composition.	
  	
  Therefore,	
  administration	
  of	
  
trehalose	
  to	
  mice	
  during	
  colonization	
  and	
  competition	
  studies	
  is	
  reasonable	
  and	
  
potentially	
  more	
  closely	
  replicate	
  the	
  nutritional	
  availability	
  of	
  trehalose	
  in	
  humans.	
  	
  	
  
Future	
  mouse	
  experiments	
  will	
  include	
  groups	
  of	
  mice	
  fed	
  trehalose	
  based	
  on	
  the	
  
estimated	
  average	
  daily	
  consumption	
  of	
  trehalose	
  reported	
  in	
  the	
  literature.	
  	
  We	
  
expect	
  the	
  data	
  from	
  these	
  mice	
  to	
  reflect	
  an	
  even	
  stronger	
  support	
  for	
  the	
  
importance	
  of	
  the	
  ability	
  of	
  C.	
  difficile	
  to	
  metabolize	
  trehalose	
  in	
  vivo.	
  	
  	
  
	
  

	
  

185	
  

Competitive Index

10

mouse 1
mouse 2
mouse 3
mouse 4
mouse 5

1

0.1

1

2

3

5

Day post-gavage

	
  
Figure	
  4.11.	
  	
  Competitive	
  indices	
  of	
  wild-­‐type	
  CD630Δerm	
  when	
  competed	
  
against	
  the	
  treA	
  knockout	
  mutant	
  in	
  a	
  conventional	
  mouse	
  model	
  of	
  C.	
  diffiicile	
  
infection.	
  	
  Mice	
  were	
  gavaged	
  with	
  a	
  mixture	
  of	
  wild-­‐type	
  and	
  mutant	
  spores,	
  and	
  
fecal	
  pellets	
  were	
  collected	
  daily.	
  	
  Competitive	
  indices	
  were	
  determined	
  by	
  qPCR	
  
using	
  primers	
  specific	
  to	
  either	
  strain	
  and	
  calculated	
  by	
  dividing	
  the	
  wild-­‐
type:mutant	
  ratio	
  at	
  the	
  indicated	
  day	
  by	
  the	
  ratio	
  of	
  the	
  spore	
  mix.	
  	
  	
  
	
  
109

CFU/g feces

108

wild-type
mutant
competiiton

107
106
105

9

7

6

5

4

3

2

1

104

Day post-gavage

	
  
Figure	
  4.12.	
  	
  Total	
  CFU/g	
  feces	
  of	
  C.	
  difficile	
  in	
  each	
  group	
  of	
  mice.	
  Three	
  groups	
  
of	
  five	
  mice	
  were	
  gavaged	
  with	
  spores	
  of	
  either	
  wild-­‐type	
  CD630Δerm,	
  a	
  treA	
  
knockout	
  mutant,	
  or	
  a	
  mixture	
  of	
  both	
  strains.	
  	
  Fecal	
  pellets	
  were	
  collected	
  daily	
  and	
  
C.	
  difficile	
  quanitifed	
  by	
  plating	
  on	
  C.	
  difficile	
  selective	
  media.	
  	
  Results	
  are	
  plotted	
  as	
  
the	
  mean	
  and	
  standard	
  deviation	
  of	
  the	
  data	
  points	
  from	
  each	
  group	
  on	
  the	
  
indicated	
  days.	
  

	
  

186	
  

	
  
Summary	
  and	
  Future	
  Directions	
  
The	
  primary	
  focus	
  of	
  the	
  work	
  in	
  this	
  chapter	
  was	
  to	
  investigate	
  carbon	
  sources	
  of	
  C.	
  
difficile,	
  with	
  an	
  emphasis	
  on	
  ones	
  that	
  RT	
  027	
  strains	
  are	
  able	
  to	
  differentially	
  
utilize,	
  compared	
  to	
  other	
  ribotypes.	
  	
  	
  Using	
  Biolog	
  Phenotype	
  Microarray	
  plates,	
  
several	
  compounds	
  that	
  increased	
  growth	
  yield	
  of	
  C.	
  difficile	
  were	
  identified.	
  	
  	
  
Ribotype	
  027	
  strains	
  were	
  shown	
  to	
  grow	
  to	
  higher	
  cell	
  densities	
  on	
  one	
  of	
  these	
  
compounds,	
  trehalose,	
  as	
  well	
  as	
  strains	
  of	
  another	
  epidemic	
  ribotype,	
  RT	
  078.	
  	
  	
  
Genetic	
  analysis	
  identified	
  an	
  amino	
  acid	
  substitution	
  in	
  treR,	
  the	
  repressor	
  of	
  treA	
  
(trehalase),	
  which	
  is	
  conserved	
  among	
  ribotype	
  027	
  clinical	
  isolates.	
  	
  Moreover,	
  
increases	
  in	
  treA	
  expression	
  in	
  ribotype	
  027	
  strains,	
  presumably	
  due	
  to	
  this	
  
mutation	
  in	
  treR,	
  have	
  been	
  observed	
  by	
  RT-­‐qPCR.	
  	
  
Using	
  a	
  treA	
  knockout	
  mutant	
  of	
  C.	
  difficile,	
  we	
  tested	
  the	
  importance	
  of	
  
ability	
  to	
  metabolize	
  trehalose	
  in	
  a	
  mouse	
  model	
  of	
  C.	
  difficile	
  infection.	
  	
  The	
  results	
  
of	
  this	
  preliminary	
  mouse	
  experiment	
  suggest	
  that	
  trehalose	
  utilization	
  is	
  important	
  
for	
  reaching	
  maximum	
  levels	
  of	
  colonization	
  in	
  vivo	
  and	
  may	
  also	
  increase	
  
competitive	
  fitness.	
  	
  	
  
	
  

The	
  treA	
  mutant	
  generated	
  in	
  this	
  study	
  represents	
  the	
  first	
  experimental	
  

proof	
  of	
  the	
  function	
  of	
  treA,	
  the	
  gene	
  annotated	
  to	
  encode	
  a	
  trehalose-­‐6-­‐phosphate	
  
hydrolase	
  in	
  C.	
  difficile.	
  	
  	
  Not	
  only	
  was	
  it	
  shown	
  that	
  the	
  mutant	
  is	
  unable	
  to	
  grow	
  on	
  
trehalose,	
  but	
  that	
  its	
  ability	
  to	
  metabolize	
  other	
  glucose	
  disaccharide	
  sugars	
  was	
  
unaffected.	
  	
  	
  	
  
	
  

Currently,	
  the	
  phenotypic	
  effects	
  of	
  the	
  leucine	
  to	
  isoleucine	
  amino	
  acid	
  

substitution	
  in	
  TreR	
  of	
  RT	
  027	
  strains	
  is	
  being	
  investigated	
  using	
  a	
  different	
  model	
  
	
  

187	
  

organism,	
  Lactococcus	
  lactis.	
  	
  Genetic	
  tools	
  in	
  C.	
  difficile	
  are	
  relatively	
  limited,	
  
however,	
  the	
  Britton	
  lab	
  has	
  recently	
  developed	
  and	
  optimized	
  a	
  recombineering	
  
system	
  for	
  making	
  targeted	
  mutations	
  in	
  lactic	
  acid	
  bacteria.	
  	
  	
  The	
  TreR	
  of	
  L.	
  lactis	
  
also	
  contains	
  the	
  highly	
  conserved	
  leucine	
  residue,	
  as	
  shown	
  in	
  figure	
  4.7.	
  	
  	
  Using	
  
recombineering,	
  I	
  have	
  made	
  a	
  treR	
  mutant	
  of	
  L.	
  lactis	
  NZ9000	
  that	
  encodes	
  the	
  
same	
  leucine	
  to	
  isoleucine	
  amino	
  acid	
  substitution	
  as	
  that	
  in	
  the	
  TreR	
  of	
  C.	
  difficile	
  
RT	
  027	
  strains.	
  	
  Currently,	
  growth	
  and	
  expression	
  experiments	
  are	
  being	
  conducted	
  
to	
  see	
  if	
  the	
  amino	
  acid	
  substitution	
  displays	
  a	
  phenotype.	
  	
  	
  If	
  this	
  is	
  demonstrated,	
  it	
  
will	
  provide	
  further	
  support	
  that	
  this	
  genetic	
  change	
  in	
  the	
  RT	
  027	
  strains	
  
contributes	
  to	
  the	
  trehalose	
  growth	
  phenotype.	
  	
  	
  Moreover,	
  additional	
  experiments	
  
are	
  planned	
  to	
  investigate	
  if	
  there	
  are	
  differences	
  in	
  the	
  ability	
  of	
  the	
  RT	
  027	
  TreR	
  
and	
  the	
  RT	
  non-­‐027	
  TreR	
  to	
  bind	
  the	
  treR	
  or	
  treA	
  promoter	
  regions	
  using	
  enzyme	
  
mobility	
  shift	
  assays.	
  	
  We	
  hypothesize	
  that	
  the	
  TreR	
  of	
  RT	
  027	
  strains	
  will	
  dissociate	
  
from	
  DNA	
  when	
  lower	
  levels	
  of	
  trehalose-­‐6-­‐phosphate	
  are	
  added	
  compared	
  to	
  the	
  
TreR	
  of	
  RT	
  non-­‐027	
  strains.	
  	
  	
  This	
  would	
  also	
  provide	
  support	
  for	
  the	
  functional	
  
effect	
  of	
  the	
  identified	
  amino	
  acid	
  substitution.	
  
	
  

In	
  conclusion,	
  the	
  work	
  in	
  this	
  chapter	
  suggests	
  that	
  trehalose	
  is	
  one	
  of	
  the	
  

nutrients	
  utilized	
  by	
  C.	
  difficile	
  in	
  vivo	
  and	
  that	
  epidemic-­‐associated	
  ribotype	
  strains	
  
are	
  able	
  to	
  utilize	
  trehalose	
  more	
  efficiently.	
  	
  	
  This	
  work	
  not	
  only	
  contributes	
  to	
  
understanding	
  the	
  nutritional	
  factors	
  involved	
  in	
  C.	
  diffcile	
  gut	
  colonization,	
  but	
  also	
  
to	
  understanding	
  the	
  physiology	
  of	
  epidemic	
  strains.	
  	
  	
  Although	
  a	
  potential	
  genetic	
  
basis	
  for	
  the	
  trehalose	
  advantage	
  of	
  RT	
  078	
  strains	
  was	
  not	
  identified,	
  the	
  
observation	
  that	
  this	
  second	
  epidemic-­‐associated	
  and	
  prevalent	
  ribotype	
  group	
  

	
  

188	
  

grows	
  better	
  on	
  trehalose	
  is	
  intriguing.	
  	
  It	
  further	
  supports	
  that	
  the	
  ability	
  to	
  utilize	
  
trehalose	
  is	
  important	
  for	
  C.	
  difficile	
  colonization	
  and	
  might	
  indirectly	
  contribute	
  to	
  
virulence.	
  	
  	
  
	
  

There	
  are	
  several	
  potential	
  applications	
  of	
  this	
  insight.	
  	
  	
  For	
  example,	
  

including	
  a	
  trehalose-­‐consuming	
  organism	
  in	
  a	
  prophylactic	
  probiotic	
  treatment	
  
during	
  antibiotic	
  treatment	
  may	
  help	
  to	
  reduce	
  development	
  of	
  C.	
  difficile	
  infection.	
  	
  
Additionally,	
  altering	
  the	
  diet	
  of	
  patients	
  at	
  risk	
  for	
  C.	
  difficile	
  infection,	
  by	
  reducing	
  
consumption	
  of	
  trehalose,	
  may	
  also	
  help	
  to	
  reduce	
  disease	
  development.	
  	
  	
  	
  
Ultimately,	
  the	
  more	
  we	
  understand	
  about	
  all	
  aspects	
  of	
  this	
  complicated	
  disease,	
  
the	
  better	
  we	
  will	
  be	
  able	
  to	
  treat	
  and	
  prevent	
  it.	
  	
  
	
  
Acknowledgments	
  	
  
There	
  are	
  several	
  people	
  that	
  contributed	
  to	
  the	
  work	
  in	
  this	
  chapter	
  that	
  I	
  would	
  
like	
  to	
  thank.	
  	
  Dr.	
  Linc	
  Sonnenshein,	
  Dr.	
  Joe	
  Sorg,	
  Dr.	
  Craig	
  Ellermeier,	
  and	
  Dr.	
  
Laruent	
  Bouillaut	
  kindly	
  provided	
  C.	
  difficile	
  strains,	
  protocols,	
  and	
  valuable	
  
experimental	
  advice.	
  	
  	
  Sara	
  McNamara	
  (Michigan	
  Department	
  of	
  Community	
  Health)	
  
provided	
  us	
  with	
  clinical	
  isolate	
  strains	
  of	
  C.	
  difficile.	
  	
  	
  Dr.	
  Seth	
  Walk	
  for	
  ribotyping	
  
many	
  of	
  our	
  strains.	
  	
  Finally,	
  Dr.	
  Mark	
  Koenigsknecht	
  (and	
  Dr.	
  Vince	
  Young)	
  for	
  
conducting	
  the	
  conventional	
  mouse	
  experiments.	
  	
  	
  
	
  
	
  
	
  
	
  

	
  

189	
  

APPENDIX	
  

	
  

190	
  

Appendix	
  
Table	
  S4.1.	
  Compounds	
  that	
  conferred	
  at	
  least	
  a	
  1.5-­‐fold	
  growth	
  yield	
  advantage	
  
(relative	
  to	
  unsupplemented	
  medium	
  control)	
  for	
  either	
  one	
  or	
  both	
  strains	
  in	
  the	
  
Biolog	
  PM3-­‐8.	
  	
  Defined	
  medium	
  was	
  used	
  for	
  all	
  plates.*	
  
	
  
	
  
Well	
  
Compound	
  
CD630	
   CD2015	
  
PM3:	
   A6	
  
biuret	
  
-­‐-­‐	
  
+-­‐	
  
	
  
A7	
  
L-­‐alanine*	
  
-­‐-­‐	
  
+-­‐	
  
	
  
A8	
  
L-­‐Arginine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
A11	
  
L-­‐cysteine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
B5	
  
L-­‐leucine	
  
++	
  
++	
  
	
  
B10	
  
L-­‐serine	
  
+-­‐	
  
+-­‐	
  
	
  
B11	
  
L-­‐threonine	
  
+-­‐	
  
++	
  
	
  
B12	
  
L-­‐tryptophan	
  
-­‐-­‐	
  
++	
  
	
  
C12	
  
L-­‐ornithine	
  
+-­‐	
  
++	
  
	
  
D9	
  
ethanolamine	
  
+-­‐	
  
+-­‐	
  
	
  
D12	
  
agmatine	
  
++	
  
+-­‐	
  
	
  
E8	
  
D-­‐glucosamine	
  
++	
  
+-­‐	
  
	
  
E9	
  
D-­‐galactosamine	
  
+-­‐	
  
-­‐-­‐	
  
	
  
E11	
  
n-­‐acetyl-­‐glucosamine	
  
+-­‐	
  
+-­‐	
  
	
  
F6	
  
Guanine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
F8	
  
Thymine	
  
+-­‐	
  
+-­‐	
  
	
  
F12	
  
Inosine	
  
+-­‐	
  
+-­‐	
  
	
  
G2	
  
xanthosine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
G8	
  
γ-­‐amino-­‐n-­‐butyric	
  acid	
  
-­‐-­‐	
  
+-­‐	
  
	
  
G10	
   D,L-­‐α-­‐Amino-­‐Caprylic	
  Acid	
  
+-­‐	
  
+-­‐	
  
	
  
G12	
  
α-­‐Amino-­‐N-­‐Valeric	
  Acid	
  
+-­‐	
  
+-­‐	
  
	
  
H4	
  
alanine-­‐glycine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
H6	
  
alanine-­‐leucine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
H7	
  
alanine-­‐threonine	
  
-­‐-­‐	
  
++	
  
	
  
H9	
  
Glycine-­‐Glutamate	
  
-­‐-­‐	
  
+-­‐	
  
	
  
H11	
  
Glycine-­‐methionine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
H12	
  
Methionine-­‐Alanine	
  
-­‐-­‐	
  
++	
  
PM4:	
   F4	
  
tetrathionate	
  
-­‐-­‐	
  
+-­‐	
  
	
  
F10	
  
L-­‐cysteic	
  acid	
  
-­‐-­‐	
  
+-­‐	
  
	
  
G7	
  
L-­‐methionine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
G8	
  
D-­‐methionine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
H4	
  
D,	
  L-­‐	
  lipoamide	
  
-­‐-­‐	
  
+-­‐	
  
PM5:	
   A5	
  
L-­‐asparagine	
  
+-­‐	
  
-­‐-­‐	
  
	
  
B1	
  
L-­‐glutamine	
  
+-­‐	
  
-­‐-­‐	
  
	
  
B8	
  
L-­‐phenylalanine	
  
-­‐-­‐	
  
+-­‐	
  
	
  
B11	
  
guanosine	
  
+-­‐	
  
-­‐-­‐	
  
	
  
C12	
  
deoxy-­‐inosine	
  
+-­‐	
  
+-­‐	
  
	
  
F12	
  
thymidine	
  
+-­‐	
  
-­‐-­‐	
  
	
  

191	
  

Table	
  S4.1	
  (cont’d)	
  
	
  
	
  
	
  
	
  
	
  
	
  
PM6:	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
PM7:	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
PM8:	
  

G1	
  
oxaloacetic	
  acid	
  
G3	
  
cyano-­‐cobalamine	
  
G4	
   ρ-­‐amino-­‐benzoic	
  acid	
  
G5	
  
folic	
  acid	
  
H7	
  
D,	
  L-­‐	
  carnitine	
  
H8	
  
choline	
  
B1	
  
alanine-­‐serine	
  
B2	
  
alanine-­‐threonine	
  
B6	
  
arginine-­‐arginine	
  
B8	
  
arginine-­‐glycine	
  
B12	
  
arginine-­‐lysine	
  
D10	
  
glutamate-­‐serine	
  
D12	
  
glutamate-­‐tyrosine	
  
E4	
  
glycine-­‐cystein	
  
G12	
  
lsoleucine-­‐serine	
  
H1	
   isoleucine-­‐tryptophan	
  
H4	
  
leucine-­‐alanine	
  
H6	
  
leucine-­‐asparagine	
  
H9	
  
leucine-­‐isoleucine	
  
H10	
  
leucine-­‐leucine	
  
H11	
  
leucine-­‐methionine	
  
H12	
   leucine-­‐phenylalanine	
  
B3	
  
lysine-­‐threonine	
  
B10	
   methinonine-­‐glutamine	
  
B12	
   methionine-­‐histidine	
  
C2	
  
methionine-­‐leucine	
  
C3	
  
methionine-­‐lysine	
  
D5	
  
proline-­‐asparagine	
  
E1	
  
serine-­‐alanine	
  
E8	
  
serine-­‐serine	
  
E12	
  
threonine-­‐arginine	
  
F1	
  
threonine-­‐glutamine	
  
F2	
  
threonine-­‐glycine	
  
F3	
  
threonine-­‐leucine	
  
F4	
   threonine-­‐methionine	
  
F7	
   trpyptophan-­‐arginine	
  
G10	
  
tyrosine-­‐leucine	
  
H5	
  
valine-­‐asparagine	
  
H9	
  
valine-­‐leucine	
  
H12	
   γ-­‐glutamate-­‐glycine	
  
E5	
  
serine-­‐methionine	
  

+-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
-­‐-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
+-­‐	
  
+-­‐	
  
-­‐-­‐	
  
+-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
+	
  
-­‐	
  
-­‐	
  
-­‐	
  
+	
  
+	
  
+	
  
+	
  
+	
  
+	
  
+	
  
-­‐	
  
-­‐	
  
+	
  
+	
  
+	
  
+	
  
+	
  
+-­‐	
  

+-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
+-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
-­‐-­‐	
  
+-­‐	
  
-­‐-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
+-­‐	
  
-­‐	
  
+	
  
+	
  
+	
  
-­‐	
  
-­‐	
  
-­‐	
  
-­‐	
  
-­‐	
  
-­‐	
  
-­‐	
  
+	
  
+	
  
-­‐	
  
-­‐	
  
-­‐	
  
-­‐	
  
-­‐	
  
-­‐-­‐	
  

	
  

*Each	
  +/-­‐	
  symbol	
  represents	
  results	
  of	
  an	
  individual	
  experiment	
  where	
  growth	
  was	
  
(+)	
  or	
  was	
  not	
  (-­‐)	
  increased	
  by	
  at	
  least	
  1.5-­‐fold	
  that	
  of	
  the	
  control.	
  	
  	
  
	
  

192	
  

REFERENCES	
  

	
  

193	
  

REFERENCES	
  
	
  
1.	
  

Yassin	
  SF,	
  Young-­‐Fadok	
  TM,	
  Zein	
  NN,	
  Pardi	
  DS.	
  2001.	
  Clostridium	
  difficile-­‐
associated	
  diarrhea	
  and	
  colitis.	
  Mayo	
  Clin.	
  Proc.	
  76:725–730.	
  

2.	
  

Bartlett	
  JG,	
  Chang	
  TW,	
  Gurwith	
  M,	
  Gorbach	
  SL,	
  Onderdonk	
  AB.	
  1978.	
  
Antibiotic-­‐associated	
  pseudomembranous	
  colitis	
  due	
  to	
  toxin-­‐producing	
  
clostridia.	
  N.	
  Engl.	
  J.	
  Med.	
  298:531–534.	
  

3.	
  

Britton	
  RA,	
  Young	
  VB.	
  2014.	
  Role	
  of	
  the	
  intestinal	
  microbiota	
  in	
  resistance	
  to	
  
colonization	
  by	
  Clostridium	
  difficile.	
  Gastroenterology	
  146:1547–1553.	
  

4.	
  

Jackson	
  S,	
  Calos	
  M,	
  Myers	
  A,	
  Self	
  WT.	
  2006.	
  Analysis	
  of	
  Proline	
  Reduction	
  in	
  
the	
  Nosocomial	
  Pathogen	
  Clostridium	
  difficile.	
  Journal	
  of	
  Bacteriology	
  
188:8487–8495.	
  

5.	
  

Nakamura	
  S,	
  Nakashio	
  S,	
  Yamakawa	
  K,	
  Tanabe	
  N,	
  Nishida	
  S.	
  2013.	
  
Carbohydrate	
  Fermentation	
  by	
  Clostridium	
  difficile.	
  Microbiol.	
  Immunol.	
  
26:107–111.	
  

6.	
  

Wilson	
  KH,	
  Perini	
  F.	
  1988.	
  Role	
  of	
  competition	
  for	
  nutrients	
  in	
  suppression	
  
of	
  Clostridium	
  difficile	
  by	
  the	
  colonic	
  microflora.	
  Infection	
  and	
  Immunity	
  
56:2610–2614.	
  

7.	
  

Wiström	
  J,	
  Norrby	
  SR,	
  Myhre	
  EB,	
  Eriksson	
  S,	
  Granström	
  G,	
  Lagergren	
  L,	
  
Englund	
  G,	
  Nord	
  CE,	
  Svenungsson	
  B.	
  2001.	
  Frequency	
  of	
  antibiotic-­‐
associated	
  diarrhoea	
  in	
  2462	
  antibiotic-­‐treated	
  hospitalized	
  patients:	
  a	
  
prospective	
  study.	
  J.	
  Antimicrob.	
  Chemother.	
  47:43–50.	
  

8.	
  

Loo	
  VG,	
  Poirier	
  L,	
  Miller	
  MA,	
  Oughton	
  M,	
  Libman	
  MD,	
  Michaud	
  S,	
  
Bourgault	
  A-­‐M,	
  Nguyen	
  T,	
  Frenette	
  C,	
  Kelly	
  M,	
  Vibien	
  A,	
  Brassard	
  P,	
  Fenn	
  
S,	
  Dewar	
  K,	
  Hudson	
  TJ,	
  Horn	
  R,	
  René	
  P,	
  Monczak	
  Y,	
  Dascal	
  A.	
  2005.	
  A	
  
predominantly	
  clonal	
  multi-­‐institutional	
  outbreak	
  of	
  Clostridium	
  difficile-­‐
associated	
  diarrhea	
  with	
  high	
  morbidity	
  and	
  mortality.	
  N.	
  Engl.	
  J.	
  Med.	
  
353:2442–2449.	
  

9.	
  

Ricciardi	
  R,	
  Rothenberger	
  DA,	
  Madoff	
  RD,	
  Baxter	
  NN.	
  2007.	
  Increasing	
  
Prevalence	
  and	
  Severity	
  of	
  Clostridium	
  difficile	
  Colitis	
  in	
  Hospitalized	
  Patients	
  
in	
  the	
  United	
  States.	
  Arch	
  Surg	
  142:624–631.	
  

10.	
  

He	
  M,	
  Miyajima	
  F,	
  Roberts	
  P,	
  Ellison	
  L,	
  Pickard	
  DJ,	
  Martin	
  MJ,	
  Connor	
  TR,	
  
Harris	
  SR,	
  Fairley	
  D,	
  Bamford	
  KB,	
  D'Arc	
  S,	
  Brazier	
  J,	
  Brown	
  D,	
  Coia	
  JE,	
  
Douce	
  G,	
  Gerding	
  D,	
  Kim	
  HJ,	
  Koh	
  TH,	
  Kato	
  H,	
  Senoh	
  M,	
  Louie	
  T,	
  Michell	
  S,	
  
Butt	
  E,	
  Peacock	
  SJ,	
  Brown	
  NM,	
  Riley	
  T,	
  Songer	
  G,	
  Wilcox	
  M,	
  Pirmohamed	
  
M,	
  Kuijper	
  E,	
  Hawkey	
  P,	
  Wren	
  BW,	
  Dougan	
  G,	
  Parkhill	
  J,	
  Lawley	
  TD.	
  2013.	
  

	
  

194	
  

Emergence	
  and	
  global	
  spread	
  of	
  epidemic	
  healthcare-­‐associated	
  Clostridium	
  
difficile.	
  Nat	
  Genet	
  45:109–113.	
  
11.	
  

Elbein	
  AD.	
  2003.	
  New	
  insights	
  on	
  trehalose:	
  a	
  multifunctional	
  molecule.	
  
Glycobiology	
  13:17R–27.	
  

12.	
  

Arg	
  x000FC	
  elles	
  JC.	
  2000.	
  Physiological	
  roles	
  of	
  trehalose	
  in	
  bacteria	
  and	
  
yeasts:	
  a	
  comparative	
  analysis.	
  Archives	
  of	
  Microbiology	
  174:217–224.	
  

13.	
  

Richards	
  AB,	
  Krakowka	
  S,	
  Dexter	
  LB,	
  Schmid	
  H,	
  Wolterbeek	
  APM,	
  
Waalkens-­‐Berendsen	
  DH,	
  Shigoyuki	
  A,	
  Kurimoto	
  M.	
  2002.	
  Trehalose:	
  a	
  
review	
  of	
  properties,	
  history	
  of	
  use	
  and	
  human	
  tolerance,	
  and	
  results	
  of	
  
multiple	
  safety	
  studies.	
  Food	
  Chem.	
  Toxicol.	
  40:871–898.	
  

14.	
  

Benaroudj	
  N.	
  2001.	
  Trehalose	
  Accumulation	
  during	
  Cellular	
  Stress	
  Protects	
  
Cells	
  and	
  Cellular	
  Proteins	
  from	
  Damage	
  by	
  Oxygen	
  Radicals.	
  Journal	
  of	
  
Biological	
  Chemistry	
  276:24261–24267.	
  

15.	
  

Kandror	
  O,	
  DeLeon	
  A,	
  Goldberg	
  AL.	
  2002.	
  Trehalose	
  synthesis	
  is	
  induced	
  
upon	
  exposure	
  of	
  Escherichia	
  coli	
  to	
  cold	
  and	
  is	
  essential	
  for	
  viability	
  at	
  low	
  
temperatures.	
  Proc.	
  Natl.	
  Acad.	
  Sci.	
  U.S.A.	
  99:9727–9732.	
  

16.	
  

Strøm	
  AR,	
  Kaasen	
  I.	
  1993.	
  Trehalose	
  metabolism	
  in	
  Escherichia	
  coli:	
  stress	
  
protection	
  and	
  stress	
  regulation	
  of	
  gene	
  expression.	
  Molecular	
  Microbiology	
  
8:205–210.	
  

17.	
  

Iturriaga	
  G,	
  Suárez	
  R,	
  Nova-­‐Franco	
  B.	
  2009.	
  Trehalose	
  Metabolism:	
  From	
  
Osmoprotection	
  to	
  Signaling.	
  IJMS	
  10:3793–3810.	
  

18.	
  

Lebenthal	
  A,	
  Lebenthal	
  E.	
  1999.	
  The	
  ontogeny	
  of	
  the	
  small	
  intestinal	
  
epithelium.	
  JPEN	
  J	
  Parenter	
  Enteral	
  Nutr	
  23:S3–6.	
  

19.	
  

Arola	
  H,	
  Koivula	
  T,	
  Karvonen	
  AL,	
  Jokela	
  H,	
  Ahola	
  T,	
  Isokoski	
  M.	
  1999.	
  Low	
  
trehalase	
  activity	
  is	
  associated	
  with	
  abdominal	
  symptoms	
  caused	
  by	
  edible	
  
mushrooms.	
  Scand	
  J	
  Gastroenterol	
  34:898–903.	
  

20.	
  

Murray	
  IA,	
  Coupland	
  K,	
  Smith	
  JA,	
  Ansell	
  ID,	
  Long	
  RG.	
  2000.	
  Intestinal	
  
trehalase	
  activity	
  in	
  a	
  UK	
  population:	
  establishing	
  a	
  normal	
  range	
  and	
  the	
  
effect	
  of	
  disease.	
  Br.	
  J.	
  Nutr.	
  83:241–245.	
  

21.	
  

Janvilisri	
  T,	
  Scaria	
  J,	
  Teng	
  C-­‐H,	
  McDonough	
  SP,	
  Gleed	
  RD,	
  Fubini	
  SL,	
  Zhang	
  
S,	
  Akey	
  B,	
  Chang	
  Y-­‐F.	
  2012.	
  Temporal	
  differential	
  proteomes	
  of	
  Clostridium	
  
difficile	
  in	
  the	
  pig	
  ileal-­‐ligated	
  loop	
  model.	
  PLoS	
  ONE	
  7:e45608.	
  

22.	
  

Staib	
  L,	
  Fuchs	
  TM.	
  2014.	
  From	
  food	
  to	
  cell:	
  nutrient	
  exploitation	
  strategies	
  of	
  
enteropathogens.	
  Microbiology.	
  160:1020-­‐1039.	
  

	
  

195	
  

23.	
  

Martindale	
  J,	
  Stroud	
  D,	
  Moxon	
  ER,	
  Tang	
  CM.	
  2000.	
  Genetic	
  analysis	
  of	
  
Escherichia	
  coli	
  K1	
  gastrointestinal	
  colonization.	
  Molecular	
  Microbiology	
  
37:1293–1305.	
  

24.	
  

Antunes	
  A,	
  Camiade	
  E,	
  Monot	
  M,	
  Courtois	
  E,	
  Barbut	
  F,	
  Sernova	
  NV,	
  
Rodionov	
  DA,	
  Martin-­‐Verstraete	
  I,	
  Dupuy	
  B.	
  2012.	
  Global	
  transcriptional	
  
control	
  by	
  glucose	
  and	
  carbon	
  regulator	
  CcpA	
  in	
  Clostridium	
  difficile.	
  Nucleic	
  
Acids	
  Research	
  40:10701–10718.	
  

25.	
  

KARASAWA	
  T,	
  Ikoma	
  S,	
  Yamakawa	
  K,	
  Nakamura	
  S.	
  1995.	
  A	
  defined	
  growth	
  
medium	
  for	
  Clostridium	
  difficile.	
  Microbiology	
  (Reading,	
  Engl.)	
  141:371–375.	
  

26.	
  

Wilson	
  KH,	
  Kennedy	
  MJ,	
  Fekety	
  FR.	
  1982.	
  Use	
  of	
  sodium	
  taurocholate	
  to	
  
enhance	
  spore	
  recovery	
  on	
  a	
  medium	
  selective	
  for	
  Clostridium	
  difficile.	
  
Journal	
  of	
  Clinical	
  Microbiology	
  15:443–446.	
  

27.	
  

Sorg	
  JA,	
  Dineen	
  SS.	
  2005.	
  Laboratory	
  Maintenance	
  of	
  Clostridium	
  difficile.	
  
John	
  Wiley	
  &	
  Sons,	
  Inc.,	
  Hoboken,	
  NJ,	
  USA.	
  

28.	
  

Pfaffl	
  MW.	
  2001.	
  A	
  new	
  mathematical	
  model	
  for	
  relative	
  quantification	
  in	
  
real-­‐time	
  RT-­‐PCR.	
  Nucleic	
  Acids	
  Research	
  29:e45.	
  

29.	
  

Bouillaut	
  L,	
  McBride	
  SM,	
  Sorg	
  JA.	
  2011.	
  Genetic	
  manipulation	
  of	
  Clostridium	
  
difficile.	
  Curr	
  Protoc	
  Microbiol	
  Chapter	
  9:Unit	
  9A.2.	
  

30.	
  

Bouillaut	
  L,	
  Self	
  WT,	
  Sonenshein	
  AL.	
  2013.	
  Proline-­‐Dependent	
  Regulation	
  of	
  
Clostridium	
  difficile	
  Stickland	
  Metabolism.	
  Journal	
  of	
  Bacteriology	
  195:844–
854.	
  

31.	
  

Theriot	
  CM,	
  Koumpouras	
  CC,	
  Carlson	
  PE,	
  Bergin	
  II,	
  Aronoff	
  DM,	
  Young	
  
VB.	
  2011.	
  Cefoperazone-­‐treated	
  mice	
  as	
  an	
  experimental	
  platform	
  to	
  assess	
  
differential	
  virulence	
  of	
  Clostridium	
  difficile	
  strains.	
  gutmicrobes	
  2:326–334.	
  

32.	
  

Antonopoulos	
  DA,	
  Huse	
  SM,	
  Morrison	
  HG,	
  Schmidt	
  TM,	
  Sogin	
  ML,	
  Young	
  
VB.	
  2009.	
  Reproducible	
  Community	
  Dynamics	
  of	
  the	
  Gastrointestinal	
  
Microbiota	
  following	
  Antibiotic	
  Perturbation.	
  Infection	
  and	
  Immunity	
  
77:2367–2375.	
  

33.	
  

Rezácová	
  P,	
  Krejciríková	
  V,	
  Borek	
  D,	
  Moy	
  SF,	
  Joachimiak	
  A,	
  Otwinowski	
  
Z.	
  2007.	
  The	
  crystal	
  structure	
  of	
  the	
  effector-­‐binding	
  domain	
  of	
  the	
  trehalose	
  
repressor	
  TreR	
  from	
  Bacillus	
  subtilis	
  168	
  reveals	
  a	
  unique	
  quarternary	
  
assembly.	
  Proteins	
  69:679–682.	
  

34.	
  

Bürklen	
  L,	
  Schöck	
  F,	
  Dahl	
  MK.	
  1998.	
  Molecular	
  analysis	
  of	
  the	
  interaction	
  
between	
  the	
  Bacillus	
  subtilis	
  trehalose	
  repressor	
  TreR	
  and	
  the	
  tre	
  operator.	
  
Mol.	
  Gen.	
  Genet.	
  260:48–55.	
  

	
  

196	
  

35.	
  

Ng	
  KM,	
  Ferreyra	
  JA,	
  Higginbottom	
  SK,	
  Lynch	
  JB,	
  Kashyap	
  PC,	
  Gopinath	
  S,	
  
Naidu	
  N,	
  Choudhury	
  B,	
  Weimer	
  BC,	
  Monack	
  DM,	
  Sonnenburg	
  JL.	
  2013.	
  
Microbiota-­‐liberated	
  host	
  sugars	
  facilitate	
  post-­‐antibiotic	
  expansion	
  of	
  enteric	
  
pathogens.	
  Nature	
  502:96–99.	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

197	
  

CHAPTER	
  5	
  
	
  
Discussion	
  and	
  Conclusions	
  
	
  
Discussion	
  
The	
  intestinal	
  microbiota	
  plays	
  a	
  fundamental	
  role	
  in	
  preventing	
  disease	
  caused	
  by	
  
C.	
  difficile,	
  illustrated	
  by	
  the	
  fact	
  that	
  disease	
  generally	
  only	
  occurs	
  after	
  antibiotic	
  
treatment.	
  	
  However,	
  these	
  interactions	
  are	
  inherently	
  complex	
  due	
  to	
  the	
  variation	
  
in	
  microbiota	
  across	
  individuals,	
  diversity	
  among	
  infecting	
  strains,	
  and	
  underlying	
  
complexities	
  of	
  the	
  intestinal	
  environment.	
  	
  Therefore,	
  much	
  work	
  is	
  still	
  needed	
  to	
  
understand	
  all	
  of	
  the	
  intricacies	
  related	
  to	
  the	
  development	
  of	
  CDI.	
  	
  	
  In	
  addition,	
  
there	
  is	
  a	
  lack	
  of	
  understanding	
  about	
  why	
  recently	
  emerged	
  epidemic	
  strains	
  of	
  C.	
  
difficile	
  have	
  spread	
  globally	
  and	
  become	
  prevalent	
  in	
  such	
  a	
  short	
  time	
  frame.	
  	
  The	
  
work	
  presented	
  in	
  this	
  thesis	
  aims	
  to	
  address	
  some	
  of	
  these	
  questions.	
  	
  In	
  chapter	
  
two,	
  we	
  developed	
  a	
  novel	
  gut	
  microbiota	
  mini-­‐bioreactor	
  system	
  in	
  which	
  we	
  can	
  
establish	
  continuously	
  cultured	
  complex	
  fecal	
  communities,	
  and	
  adapted	
  this	
  
system	
  as	
  an	
  in	
  vitro	
  C.	
  difficile	
  infection	
  model	
  in	
  chapter	
  3.	
  	
  We	
  then	
  investigated	
  
the	
  competitive	
  dynamics	
  of	
  epidemic	
  and	
  non-­‐epidemic	
  C.	
  difficile	
  strains	
  in	
  the	
  
presence	
  of	
  complex	
  communities	
  in	
  both	
  the	
  bioreactors	
  and	
  a	
  mouse	
  infection	
  
model,	
  showing	
  that	
  these	
  newly	
  emerged	
  epidemic	
  strains	
  have	
  a	
  competitive	
  
advantage.	
  	
  	
  A	
  potential	
  physiological	
  basis	
  for	
  this	
  competitive	
  advantage	
  is	
  
identified	
  in	
  chapter	
  4,	
  where	
  I	
  demonstrate	
  that	
  epidemic	
  strains	
  grow	
  better	
  on	
  

	
  

198	
  

trehalose,	
  and	
  that	
  this	
  ability	
  may	
  be	
  conferred	
  by	
  a	
  single	
  amino	
  acid	
  substitution	
  
in	
  a	
  regulatory	
  gene	
  involved	
  in	
  metabolism	
  of	
  this	
  sugar.	
  	
  	
  	
  
This	
  work	
  contributes	
  to	
  the	
  field	
  of	
  C.	
  difficile	
  in	
  three	
  ways.	
  	
  First,	
  we	
  have	
  
developed	
  a	
  novel	
  C.	
  difficile	
  infection	
  model	
  that	
  could	
  be	
  used	
  to	
  study	
  many	
  
different	
  aspects	
  of	
  C.	
  difficile	
  physiology	
  and	
  interactions	
  with	
  intestinal	
  microbiota.	
  	
  
These	
  include	
  but	
  certainly	
  aren’t	
  limited	
  to	
  vegetative	
  growth,	
  sporulation	
  and	
  
germination	
  dynamics,	
  gene	
  expression	
  including	
  toxin	
  production,	
  and	
  the	
  effects	
  
of	
  different	
  community	
  assemblages	
  on	
  these	
  processes.	
  	
  Second,	
  the	
  work	
  in	
  
chapter	
  4	
  broadens	
  our	
  understanding	
  of	
  C.	
  difficile	
  metabolism	
  and	
  the	
  nutrients	
  
that	
  potentially	
  play	
  a	
  role	
  in	
  intestinal	
  colonization.	
  	
  	
  Lastly,	
  identification	
  of	
  the	
  
trehalose	
  growth	
  advantage	
  and	
  treR	
  mutation	
  in	
  epidemic	
  strains	
  provides	
  insight	
  
into	
  physiological	
  attributes	
  that	
  potentially	
  explain	
  not	
  only	
  why	
  these	
  strains	
  are	
  
so	
  prevalent	
  and	
  cause	
  more	
  disease,	
  but	
  also	
  insight	
  into	
  the	
  evolution	
  of	
  this	
  
unique	
  pathogen.	
  	
  	
  
Since	
  competitive	
  interactions	
  between	
  the	
  microbiota	
  and	
  C.	
  difficile	
  are	
  so	
  
integral	
  to	
  disease	
  development,	
  evolutionary	
  adaptations	
  affecting	
  nutrient	
  
competition	
  are	
  likely	
  contributing	
  to	
  the	
  evolution	
  of	
  this	
  organism.	
  	
  In	
  the	
  case	
  of	
  
RT	
  027	
  strains,	
  the	
  ability	
  to	
  grow	
  better	
  on	
  trehalose	
  may	
  have	
  contributed	
  to	
  its	
  
recent	
  emergence;	
  however,	
  it	
  is	
  likely	
  that	
  there	
  are	
  other	
  contributing	
  factors.	
  	
  
Genetic	
  mutations	
  affecting	
  nutrient	
  utilization	
  could	
  provide	
  a	
  fitness	
  advantage,	
  
resulting	
  in	
  several	
  outcomes.	
  	
  Firstly,	
  a	
  mutation	
  might	
  allow	
  for	
  more	
  efficient	
  
growth	
  on	
  a	
  particular	
  nutrient	
  by	
  affecting	
  either	
  nutrient	
  uptake	
  or	
  metabolism,	
  
such	
  as	
  I	
  have	
  demonstrated	
  here	
  for	
  trehalose	
  utilization.	
  	
  In	
  this	
  example,	
  the	
  new	
  

	
  

199	
  

strain	
  is	
  more	
  fit	
  than	
  the	
  other	
  C.	
  difficile	
  strains,	
  yet	
  still	
  less	
  fit	
  than	
  the	
  competing	
  
member/s	
  of	
  the	
  microbiota.	
  	
  	
  The	
  competitive	
  dynamics	
  observed	
  in	
  chapter	
  3	
  
exemplify	
  this	
  scenario	
  since	
  the	
  colonization	
  levels	
  and	
  invasion	
  of	
  individual	
  
strains	
  in	
  the	
  in	
  vitro	
  and	
  in	
  vivo	
  models	
  were	
  similar.	
  	
  	
  Alternatively,	
  there	
  may	
  be	
  
mutations	
  that	
  allow	
  the	
  new	
  strain	
  to	
  more	
  readily	
  invade	
  the	
  intestinal	
  
environment,	
  not	
  only	
  outcompeting	
  the	
  other	
  C.	
  difficile	
  strains	
  but	
  also	
  members	
  
of	
  the	
  microbiota	
  that	
  previously	
  competed	
  for	
  a	
  similar	
  niche.	
  	
  	
  	
  In	
  this	
  situation,	
  the	
  
evolved	
  strain	
  is	
  more	
  fit	
  than	
  the	
  other	
  C.	
  difficile	
  strains	
  and	
  also	
  the	
  competing	
  
microbiota	
  members,	
  therefore	
  affecting	
  the	
  background	
  community	
  structure.	
  	
  In	
  
the	
  case	
  of	
  epidemic	
  C.	
  difficile,	
  some	
  evidence	
  of	
  this	
  may	
  be	
  demonstrated	
  by	
  the	
  
observation	
  that	
  mice	
  infected	
  with	
  epidemic	
  (RT	
  027)	
  strains	
  had	
  persistent	
  
dysbiosis	
  of	
  the	
  intestinal	
  microbiota,	
  different	
  from	
  the	
  non-­‐epidemic	
  strain	
  
infected	
  mice	
  (1).	
  	
  While	
  the	
  authors	
  speculate	
  that	
  dysbiosis	
  is	
  due	
  to	
  the	
  epidemic	
  
strain’s	
  impact	
  on	
  the	
  inflammatory	
  response	
  of	
  the	
  host	
  which	
  in	
  turn	
  affects	
  the	
  
microbiota,	
  it	
  is	
  possible	
  that	
  these	
  strains	
  directly	
  induce	
  changes	
  to	
  the	
  microbiota	
  
through	
  competition.	
  	
  Even	
  though	
  it	
  does	
  not	
  appear	
  that	
  this	
  is	
  occurring	
  in	
  our	
  
humanized	
  mouse	
  infection	
  model,	
  it	
  would	
  be	
  interesting	
  to	
  look	
  more	
  closely	
  at	
  
the	
  communities	
  of	
  epidemic	
  and	
  non-­‐epidemic	
  C.	
  difficile	
  infected	
  mice	
  and	
  see	
  if	
  
they	
  are	
  also	
  different.	
  	
  In	
  a	
  third	
  scenario,	
  the	
  new	
  strain	
  could	
  acquire	
  mutations	
  
that	
  allow	
  it	
  to	
  fill	
  a	
  new	
  intestinal	
  niche,	
  for	
  example,	
  by	
  being	
  able	
  to	
  metabolize	
  a	
  
carbon	
  source	
  not	
  utilized	
  by	
  other	
  C.	
  difficile	
  strains.	
  	
  In	
  this	
  scenario,	
  it	
  would	
  be	
  
possible	
  for	
  both	
  strains	
  to	
  co-­‐colonize	
  since	
  they	
  would	
  no	
  longer	
  be	
  competing	
  for	
  
the	
  same	
  niche.	
  	
  	
  In	
  addition,	
  this	
  could	
  affect	
  the	
  background	
  community	
  as	
  

	
  

200	
  

described	
  in	
  the	
  second	
  example	
  above,	
  as	
  the	
  new	
  strain	
  could	
  now	
  be	
  out-­‐
competing	
  different	
  microbiota	
  members	
  for	
  this	
  other	
  nutrient.	
  	
  The	
  Biolog	
  
experiments	
  reported	
  in	
  chapter	
  4	
  identified	
  several	
  compounds	
  whose	
  utilization	
  
may	
  be	
  unique	
  to	
  RT	
  027	
  strains,	
  however,	
  much	
  more	
  work	
  would	
  need	
  to	
  be	
  done	
  
to	
  support	
  these	
  findings	
  by	
  screening	
  additional	
  strains	
  and	
  growth	
  conditions.	
  	
  	
  	
  
	
  

Several	
  comparative	
  genomic	
  analysis	
  studies	
  have	
  been	
  conducted	
  in	
  an	
  

effort	
  to	
  understand	
  C.	
  difficile	
  strain	
  evolution	
  and	
  also	
  identify	
  genetic	
  factors	
  that	
  
potentially	
  contribute	
  to	
  the	
  epidemic	
  phenotype	
  of	
  newly	
  emerged	
  strains	
  (2-­‐6).	
  	
  	
  
Stabler	
  et	
  al.	
  used	
  whole	
  genome	
  sequence	
  analysis	
  of	
  three	
  C.	
  difficile	
  strains;	
  a	
  
historic	
  non-­‐epidemic	
  RT	
  012	
  isolate	
  (CD630),	
  and	
  historic	
  non-­‐epidemic	
  RT	
  027	
  
isolate	
  (CD196),	
  and	
  a	
  recent	
  epidemic	
  RT	
  027	
  isolate	
  (R20291)	
  (2).	
  	
  They	
  identified	
  
234	
  genes	
  unique	
  to	
  the	
  RT	
  027	
  strains,	
  many	
  of	
  which	
  are	
  associated	
  with	
  motility,	
  
antibiotic	
  resistance,	
  and	
  toxicity,	
  and	
  an	
  additional	
  5	
  genetic	
  regions	
  unique	
  to	
  just	
  
the	
  recent	
  RT	
  027	
  isolate.	
  	
  	
  While	
  the	
  impacts	
  of	
  some	
  of	
  these	
  traits	
  on	
  virulence	
  or	
  
colonization	
  have	
  been	
  investigated,	
  such	
  as	
  motility,	
  for	
  the	
  most	
  part	
  most	
  have	
  
not.	
  	
  Moreover,	
  inclusion	
  of	
  recent	
  clinical	
  isolates	
  of	
  other	
  RT’s	
  would	
  have	
  been	
  
beneficial,	
  as	
  some	
  of	
  these	
  genes	
  may	
  not	
  be	
  unique	
  to	
  just	
  RT	
  027	
  strains.	
  	
  
	
  

Knetcsh	
  et	
  al.	
  reported	
  several	
  genes	
  linked	
  to	
  epidemic	
  lineages	
  of	
  C.	
  difficile,	
  

which	
  may	
  have	
  physiological	
  implications	
  (7).	
  	
  	
  These	
  include	
  an	
  antibiotic	
  
synthesis	
  gene	
  set	
  in	
  RT	
  078	
  strains,	
  and	
  an	
  insertion	
  in	
  RT	
  027	
  strains	
  containing	
  
genes	
  involved	
  in	
  generation	
  of	
  thymidine.	
  	
  	
  We	
  also	
  observed	
  this	
  genetic	
  insertion	
  
in	
  RT	
  027	
  strains,	
  as	
  reported	
  in	
  chapter	
  3.	
  	
  Their	
  analysis,	
  however,	
  showed	
  that	
  
these	
  genetic	
  elements	
  are	
  not	
  unique	
  to	
  the	
  RT	
  078	
  and	
  027	
  strains,	
  but	
  are	
  found	
  

	
  

201	
  

in	
  other	
  closely	
  related	
  ribotype	
  strains.	
  	
  Thus,	
  these	
  might	
  not	
  play	
  a	
  significant	
  
role	
  in	
  their	
  epidemic	
  phenotype,	
  although	
  it	
  would	
  be	
  interesting	
  to	
  test	
  this	
  
hypothesis.	
  	
  	
  	
  	
  	
  	
  
	
  	
  

Fluoroquinolone	
  resistance	
  is	
  implicated	
  to	
  be	
  an	
  important	
  factor	
  in	
  the	
  

emergence	
  of	
  RT	
  027	
  strains,	
  as	
  it	
  was	
  shown	
  to	
  have	
  independently	
  evolved	
  in	
  two	
  
separate	
  clades	
  of	
  post-­‐epidemic	
  RT	
  027	
  isolates	
  (6).	
  	
  In	
  the	
  same	
  study,	
  comparison	
  
of	
  pre-­‐	
  and	
  post-­‐epidemic	
  RT	
  027	
  isolates	
  identified	
  several	
  SNPs	
  unique	
  to	
  the	
  
post-­‐epidemic	
  strains,	
  none	
  of	
  which	
  suggested	
  any	
  obvious	
  phenotypic	
  function	
  
(6).	
  	
  	
  This	
  study	
  only	
  focused	
  on	
  genetic	
  comparisons	
  of	
  strains	
  within	
  the	
  RT	
  027	
  
clade,	
  and	
  so	
  would	
  not	
  be	
  able	
  to	
  identify	
  genetic	
  elements	
  novel	
  to	
  the	
  RT	
  027	
  
lineage	
  as	
  a	
  whole,	
  which	
  may	
  also	
  be	
  important.	
  	
  Again,	
  comparison	
  of	
  several	
  
recent	
  clinical	
  isolates	
  of	
  different	
  RT	
  groups	
  with	
  a	
  focus	
  on	
  identifying	
  genes	
  or	
  
SNPs	
  unique	
  to	
  just	
  RT	
  027	
  or	
  other	
  epidemic	
  RT’s	
  is	
  warranted.	
  	
  
	
  

Many	
  challenges	
  exist	
  when	
  interpreting	
  the	
  results	
  of	
  these	
  genomic	
  

comparison	
  studies.	
  	
  Many	
  identified	
  genetic	
  elements,	
  which	
  are	
  conserved	
  among	
  
a	
  specific	
  strain	
  set,	
  are	
  hypothetical	
  proteins	
  or	
  otherwise	
  functionally	
  
uncharacterized.	
  	
  Moreover,	
  our	
  understanding	
  of	
  the	
  functions	
  of	
  specific	
  genes,	
  
mutations	
  within	
  these	
  genes,	
  or	
  regulatory	
  networks	
  as	
  whole	
  is	
  largely	
  not	
  
sufficient	
  to	
  predict	
  the	
  phenotypic	
  effects	
  of	
  these	
  mutations,	
  especially	
  for	
  C.	
  
difficile.	
  	
  Ultimately,	
  much	
  work	
  will	
  need	
  to	
  be	
  done	
  to	
  directly	
  test	
  if	
  any	
  of	
  these	
  
genes	
  or	
  mutations	
  play	
  a	
  role	
  in	
  C.	
  difficile	
  colonization	
  or	
  virulence.	
  	
  	
  	
  	
  	
  
	
  
	
  

	
  

	
  

202	
  

Conclusions	
  
We	
  have	
  only	
  begun	
  to	
  scratch	
  the	
  surface	
  when	
  it	
  comes	
  to	
  understanding	
  the	
  
complexities	
  of	
  C.	
  difficile	
  disease.	
  	
  	
  However,	
  recent	
  increases	
  in	
  the	
  incidence	
  and	
  
severity	
  of	
  CDI	
  have	
  fueled	
  the	
  need	
  to	
  address	
  this	
  issue	
  in	
  order	
  to	
  develop	
  better	
  
ways	
  of	
  preventing	
  and	
  treating	
  this	
  infection.	
  	
  As	
  a	
  result,	
  research	
  in	
  the	
  areas	
  of	
  C.	
  
difficile	
  physiology,	
  infection	
  model	
  development,	
  and	
  generation	
  of	
  C.	
  difficile	
  
genetic	
  tools	
  are	
  advancing	
  at	
  an	
  ever-­‐increasing	
  pace.	
  	
  Indeed,	
  the	
  work	
  presented	
  
in	
  this	
  thesis	
  is	
  no	
  exception.	
  	
  Combined	
  with	
  the	
  contributions	
  from	
  research	
  
investigating	
  the	
  human	
  microbiome	
  structure	
  and	
  function,	
  new	
  insights	
  into	
  the	
  
dynamics	
  of	
  CDI	
  are	
  being	
  made	
  all	
  the	
  time.	
  	
  Undoubtedly,	
  we	
  are	
  moving	
  closer	
  to	
  
discovering	
  novel,	
  more	
  effective	
  ways	
  of	
  preventing	
  and	
  treating	
  infection	
  caused	
  
by	
  C.	
  difficile.	
  	
  	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  

	
  

203	
  

REFERENCES	
  

	
  

204	
  

REFERENCES	
  
	
  
1.	
  

Lawley	
  TD,	
  Clare	
  S,	
  Walker	
  AW,	
  Stares	
  MD,	
  Connor	
  TR,	
  Raisen	
  C,	
  Goulding	
  
D,	
  Rad	
  R,	
  Schreiber	
  F,	
  Brandt	
  C,	
  Deakin	
  LJ,	
  Pickard	
  DJ,	
  Duncan	
  SH,	
  Flint	
  HJ,	
  
Clark	
  TG,	
  Parkhill	
  J,	
  Dougan	
  G.	
  2012.	
  Targeted	
  Restoration	
  of	
  the	
  Intestinal	
  
Microbiota	
  with	
  a	
  Simple,	
  Defined	
  Bacteriotherapy	
  Resolves	
  Relapsing	
  
Clostridium	
  difficile	
  Disease	
  in	
  Mice.	
  PLoS	
  Pathog	
  8:e1002995.	
  

2.	
  

Stabler	
  RA,	
  He	
  M,	
  Dawson	
  L,	
  Martin	
  M,	
  Valiente	
  E,	
  Corton	
  C,	
  Lawley	
  TD,	
  
Sebaihia	
  M,	
  Quail	
  MA,	
  Rose	
  G,	
  Gerding	
  DN,	
  Gibert	
  M,	
  Popoff	
  MR,	
  Parkhill	
  J,	
  
Dougan	
  G,	
  Wren	
  BW.	
  2009.	
  Comparative	
  genome	
  and	
  phenotypic	
  analysis	
  of	
  
Clostridium	
  difficile	
  027	
  strains	
  provides	
  insight	
  into	
  the	
  evolution	
  of	
  a	
  
hypervirulent	
  bacterium.	
  Genome	
  Biol	
  10:R102.	
  

3.	
  

He	
  M,	
  Sebaihia	
  M,	
  Lawley	
  TD,	
  Stabler	
  RA,	
  Dawson	
  LF,	
  Martin	
  MJ,	
  Holt	
  KE,	
  
Seth-­‐Smith	
  HMB,	
  Quail	
  MA,	
  Rance	
  R,	
  Brooks	
  K,	
  Churcher	
  C,	
  Harris	
  D,	
  
Bentley	
  SD,	
  Burrows	
  C,	
  Clark	
  L,	
  Corton	
  C,	
  Murray	
  V,	
  Rose	
  G,	
  Thurston	
  S,	
  van	
  
Tonder	
  A,	
  Walker	
  D,	
  Wren	
  BW,	
  Dougan	
  G,	
  Parkhill	
  J.	
  2010.	
  Evolutionary	
  
dynamics	
  of	
  Clostridium	
  difficile	
  over	
  short	
  and	
  long	
  time	
  scales.	
  Proceedings	
  of	
  
the	
  National	
  Academy	
  of	
  Sciences	
  107:7527–7532.	
  

4.	
  

Stabler	
  RA,	
  Dawson	
  LF,	
  Valiente	
  E,	
  Cairns	
  MD,	
  Martin	
  MJ,	
  Donahue	
  EH,	
  
Riley	
  TV,	
  Songer	
  JG,	
  Kuijper	
  EJ,	
  Dingle	
  KE,	
  Wren	
  BW.	
  2012.	
  Macro	
  and	
  Micro	
  
Diversity	
  of	
  Clostridium	
  difficile	
  Isolates	
  from	
  Diverse	
  Sources	
  and	
  
Geographical	
  Locations.	
  PLoS	
  ONE	
  7:e31559.	
  

5.	
  

Knetsch	
  CW,	
  Terveer	
  EM,	
  Lauber	
  C,	
  Gorbalenya	
  AE,	
  Harmanus	
  C,	
  Kuijper	
  
EJ,	
  Corver	
  J,	
  van	
  Leeuwen	
  HC.	
  2012.	
  Comparative	
  analysis	
  of	
  an	
  expanded	
  
Clostridium	
  difficile	
  reference	
  strain	
  collection	
  reveals	
  genetic	
  diversity	
  and	
  
evolution	
  through	
  six	
  lineages.	
  Infection,	
  Genetics	
  and	
  Evolution	
  12:1577–
1585.	
  

6.	
  

He	
  M,	
  Miyajima	
  F,	
  Roberts	
  P,	
  Ellison	
  L,	
  Pickard	
  DJ,	
  Martin	
  MJ,	
  Connor	
  TR,	
  
Harris	
  SR,	
  Fairley	
  D,	
  Bamford	
  KB,	
  D'Arc	
  S,	
  Brazier	
  J,	
  Brown	
  D,	
  Coia	
  JE,	
  
Douce	
  G,	
  Gerding	
  D,	
  Kim	
  HJ,	
  Koh	
  TH,	
  Kato	
  H,	
  Senoh	
  M,	
  Louie	
  T,	
  Michell	
  S,	
  
Butt	
  E,	
  Peacock	
  SJ,	
  Brown	
  NM,	
  Riley	
  T,	
  Songer	
  G,	
  Wilcox	
  M,	
  Pirmohamed	
  M,	
  
Kuijper	
  E,	
  Hawkey	
  P,	
  Wren	
  BW,	
  Dougan	
  G,	
  Parkhill	
  J,	
  Lawley	
  TD.	
  2013.	
  
Emergence	
  and	
  global	
  spread	
  of	
  epidemic	
  healthcare-­‐associated	
  Clostridium	
  
difficile.	
  Nat	
  Genet	
  45:109–113.	
  

7.	
  

Knetsch	
  CW,	
  Hensgens	
  MPM,	
  Harmanus	
  C,	
  van	
  der	
  Bijl	
  MW,	
  Savelkoul	
  
PHM,	
  Kuijper	
  EJ,	
  Corver	
  J,	
  van	
  Leeuwen	
  HC.	
  2011.	
  Genetic	
  markers	
  for	
  
Clostridium	
  difficile	
  lineages	
  linked	
  to	
  hypervirulence.	
  Microbiology	
  157:3113–
3123.	
  	
  

	
  

205