INTERACTIONS  BETWEEN  BIOMASS  FEEDSTOCK  CHARACTERISTICS  AND  BIOENERGY   PRODUCTION:  FROM  THE  LANDSCAPE  TO  THE  MOLECULAR  SCALE     By     Rebecca  Garlock  Ong                 A  DISSERTATION     Submitted  to   Michigan  State  University   in  partial  fulfillment  of  the  requirements   for  the  degree  of     DOCTOR  OF  PHILOSOPHY   Chemical  Engineering       2011     ABSTRACT   INTERACTIONS  BETWEEN  BIOMASS  FEEDSTOCK  CHARACTERISTICS  AND  BIOENERGY   PRODUCTION:  FROM  THE  LANDSCAPE  TO  THE  MOLECULAR  SCALE     By     Rebecca  Garlock  Ong       The  choices  that  are  made  with  respect  to  the  efficient  development  and  operation  of   any   bioenergy   conversion   process   are   inherently   linked   to   the   physical   and   chemical   properties   of   the   feedstock.   These   interactions   can   be   examined   at   a   variety   of   scales   ranging   from   the   landscape   scale   where   the   availability   of   feedstock   can   affect   the   appropriate   energy   generation   method   for   a   given   region,   to   the   molecular   scale   where   small   variations   in   cell   wall   components   can   have   a   large   impact   on   process   yields.   Because   biomass   is   an   inherently   heterogeneous   material,   it   is   necessary   to   develop   a   broad   understanding   of   how   different   characteristics   impact   the   conversion   process:   how   differences   in   biomass   classification   can   help   us   make   generalizations   about   new   feedstocks,   whether   different   varieties   of   the   same   species  have  inherent  differences  that  alter  their  relative  efficiencies  of  conversion,  and  what   variations  are  possible  depending  on  which  portion  of  the  plant  is  used  for  a  feedstock.   At   the   landscape   scale,   the   distribution   of   usable   crop   residues   would   be   one   factor   influencing   the   decision   of   where   to   locate   a   lignocellulosic   biorefinery.   In   Mainland   China,   594   million   metric   tons   of   crop   residues   are   produced   each   year,   however   only   125   million   tons   are   available   for   energy   generation,   either   in   rural   homes   or   in   a   larger   facility.   Based   on   residue   availability,  Henan  in  particular  would  be  the  most  likely  site  for  a  biorefinery,  with  potential  for   other  locations  in  central,  eastern,  and  northeastern  China.         Plant  materials  can  interact  with  pretreatment  and  enzymatic  hydrolysis  at  a  variety  of   scales.   Plant   classification   largely   determines   cell   wall   chemistry,   and   there   are   distinct   differences   between   the   way   that   dicots   and   grasses   interact   with   pretreatment   and   enzymatic   hydrolysis.  Mixed-­‐species  feedstocks  that  have  a  higher  mass  contribution  by  grasses  are  more   digestible   and   also   generate   higher   sugar   yields   compared   to   those   dominated   by   dicots.   In   contrast,   the   differences   between   different   varieties   of   the   same   species,   in   this   case   switchgrass   (when   grown   under   the   same   environmental   conditions),   showed   much   smaller   differences  in  digestibility,  optimal  pretreatment  conditions,  and  enzyme  combinations.  At  the   next  scale  down,  the  different  portions  of  the  plant  also  have  distinctly  different  structures  and   compositions   that   affect   their   amenability   toward   pretreatment.   This   could   be   one   consideration   toward   determining   how   to   best   harvest   lignocellulosics   on   the   field.   For   corn   stover,  the  best  scenario  involved  harvesting  the  fractions  in  order  of  decreasing  lignin  content:   husk   >   leaves   >   stem   >   cob,   an   order   that   is   feasible   using   currently   available   harvesting   equipment  and  methods.     Klason   lignin   content   was   related   to   decreased   glucan   digestibility   in   both   the   mixed-­‐ species  materials  and  in  corn  stover  fractions.  However,  no  effect  due  to  lignin  content  or  lignin   monomer   composition   was   observed   for   genetically   modified   poplar.   Of   the   samples   tested,   the  C4H::F5H  poplar  that  was  modified  to  have  highly  linear  and  extractable  lignin  with  mostly   TM syringyl  residues  showed  the  greatest  improvement  following  AFEX  pretreatment.  However   the  increase  was  not  substantial  and  for  these  types  of  materials  it  may  be  more  effective  to   employ   an   liquid   ammonia   pretreatment   that   is   able   to   extract   the   lignin   and   simultaneously   modify  the  cellulose  crystal  structure.                     DEDICATION   I  dedicate  this  dissertation  to  the  people  I  love  most  and  who  have  generously  helped  to  keep   me   sane   during   my   doctoral   work:   my   husband   Benjamin   Ong,   who   loved,   supported,   encouraged,  and  perhaps  most  importantly,  made  sure  I  ate  while  writing  this  dissertation;  to   my   parents   Dayle   and   Nancy   Garlock,   sister   Robin,   and   brother   Kevin,   who   have   given   me   unconditional  support  and  love  my  entire  life  and  expressed  their  pride  in  me  for  accomplishing   this   grand   and   sometimes   painful   task;   to   my   grad   school   roommate   of   four   years,   Sarah   Coefield,  who  took  me  out  to  pet  owls,  helped  me  laugh  when  I  was  discouraged,  and  is  one  of   my  best  friends;  and  most  of  all  to  Jesus   Christ,   my   savior,  who  taught  me  during  my  time  in   graduate   school   more   about   humility,   patience,   and   what   to   do   when   everything   goes   wrong   (i.e.  pray  like  crazy)  than  I  ever  realized  that  I  needed  to  learn.       iv       ACKNOWLEDGMENTS   This   dissertation   would   never   have   been   completed   without   the   generous   contributions   and  support  of  a  great  many  people.     First,   I   would   like   to   thank   Dr.   Bruce   E.   Dale   for   his   guidance   as   my   Ph.D.   advisor,   the   revisions   of   many   publications,   and   the   opportunity   to   work   in   the   Biomass   Conversion   Research   Lab   where   I   could   put   into   practice   my   background   in   Chemical   Engineering   and   Plant   Biology.  I  would  also  like  to  thank  my  dissertation  committee:  Drs.  Kyung-­‐Hwan  Han,  Kenneth   Keegstra,   and   David   Hodge   for   their   time,   comments,   and   edits   during   all   the   stages   of   my   doctoral  degree  program.  They  have  made  this  work  substantially  better  for  their  contributions.     I   would   also   like   to   thank   all   of   the   members   of   the   Biomass   Conversion   Research   Laboratory  for  their  assistance,  ideas,  insights,  comments,  and  critiques  of  my  research  during   the  past  five  years,  and  my  colleagues  from  the  Great  Lakes  Bioenergy  Research  Center  for  their   insights  and  collaboration.  I  would  particularly  like  to  thank  Venkatesh  Balan  for  his  leadership   and   guidance   of   the   laboratory,   and   his   help   at   ordering   much   needed   pieces   of   equipment;   Bryan  Bals,  Shishir  Chundawat,  Ming  Woei  Lau,  Dahai  Gao,  Leonardo  da  Costa  Sousa,  Mingjie  Jin,   and  Pragnya  Eranki  for  being  sounding  boards  for  experimental  design  and  analysis;  Isaac  Wong,   Rohini   Bala   Chandran,   Christa   Gunawan,   Pete   Donald,   Derek   Marshall,   Nirmal   Uppugundula,   and   James   Humpula   for   assistance   in   conducting   various   aspects   of   the   experiments   and   helping   to   repair   equipment;   Cliff   Foster,   Mike   Allen,   and   David   Main,   for   training   me   on   methods  and  allowing  me  to  conduct  experiments  in  their  laboratories;  Stacey  Vanderwulp  and   the  W.K.  Kellogg  Biological  Station,   Natalia  de  Leon  at  UW-­‐Madison,   Ceres  Inc.,  Bill  Widdicombe   and   the   MSU   Agronomy   Farm,   Paul   Bloese   and   the   MSU   Tree   Research   Center,   Kyung-­‐Hwan   v       Han  and  members  of  his  laboratory  at  MSU,  Rebecca  Van  Acker  and  others  at  VIB  Department   of   Plant   Systems   at   Ghent   University,   and   Shawn   Mansfield   at   UBC   for   providing   the   many   different   feedstocks   used   for   these   experiments;   and   Genencor,   A   Danisco   Division   for   generously  providing  a  number  of  the  enzymes  used  for  these  experiments.   I  would  also  like  to  acknowledge  Dow  Chemical  and  the  Hong  Kong  University  of  Science   and  Technology  for  funding  my  project  on  crop  residues  in  China  and  giving  me  the  once-­‐in-­‐a-­‐ lifetime   opportunity   to   live   overseas   in   a   cross-­‐cultural   environment.   I   am   especially   grateful   to   Christopher   Chao   and   Moses   Ng   for   hosting   me   during   my   stay   at   HKUST,   as   well   as   Chris   Chan   from   Dow,   David   Zweig   and   Meggy   Wan   from   HKUST,   and   Glenn   Shive   from   the   Hong   Kong-­‐ America  Center,  for  their  assistance  and  involvement  during  my  time  in  Hong  Kong.     A   number   of   funding   sources   made   these   projects   possible:   Dow   Chemical;   the   DOE   Great   Lakes   Bioenergy   Research   Center   (through   the   DOE   Office   of   Science   BER   DE-­‐FC02-­‐ 07ER64494);   the   Michigan   State   University   Research   Foundation;   and   the   Consortium   for   Applied  Fundamentals  in  Innovation  (through  the  DOE  Office  of  the  Biomass  Program  DE-­‐FG36-­‐ 07GO17102).   vi       TABLE  OF  CONTENTS     LIST  OF  TABLES  ......................................................................................................................  xi   LIST  OF  FIGURES  ....................................................................................................................   xv   KEY  TO  SYMBOLS  AND  ABBREVIATIONS  ...............................................................................  xxii   CHAPTER  1  :    INTRODUCTION   .................................................................................................  1   CHAPTER   2   :       LANDSCAPE   SCALE:   ANALYSIS   OF   KEY   CROP   RESIDUE   AVAILABILITY   FOR   BIOENERGY  IN  MAINLAND  CHINA  .........................................................................................  13   2.1.  Introduction  ...............................................................................................................  13   2.2.  Materials  and  methods  ..............................................................................................  15   2.2.1.  Land  area  and  population  data   ...................................................................................  15   2.2.2.  Crop  production  data  .................................................................................................  16   2.2.3.  Crop  residue  yield  calculations  and  mapping  .............................................................  17   2.2.4.  Straw  used  for  pulp  and  paper  production  ................................................................  19   2.2.5.  Straw  used  for  animal  feed   .........................................................................................  21   2.2.6.  Straw  returned  to  the  field  .........................................................................................  23   2.2.7.  Provincial  distribution  of  usable  crop  residues  ..........................................................  24   2.3.  Results  and  discussion  ................................................................................................  25   2.3.1.  Total  crop  residue  production  ....................................................................................  25   2.3.2.  Distribution  of  crops  within  Mainland  China  ..............................................................  29   2.3.3.  Influence  of  residue-­‐to-­‐grain  ratio  on  spatial  distribution  .........................................  33   2.3.4.  Usable  amount  of  crop  residues  .................................................................................  36   2.4.  Conclusion   .................................................................................................................  41   CHAPTER   3   :       CLASSIFICATION   SCALE:   INFLUENCE   OF   VARIABLE   SPECIES   COMPOSITION   ON   TM THE   SACCHARIFICATION   OF   AFEX   PRETREATED   BIOMASS   FROM   UNMANAGED   FIELDS   IN   COMPARISON  TO  CORN  STOVER   ...........................................................................................  42   3.1.  Introduction  ...............................................................................................................  42   3.2.  Materials  and  methods  ..............................................................................................  45   3.2.1.  Sample  harvest  and  preparation  ................................................................................  45   3.2.2.  Composition  analysis  ..................................................................................................  46   3.2.3.   Effect   of   pretreatment   conditions   on   hydrolysis   yields   from   early   successional   samples  .................................................................................................................................  47   vii       TM 3.2.4.  AFEX   pretreatment  for  comparison  of  early  successional  old  field  replicates  with   late  successional  old  field  corn  stover  samples  ....................................................................  49   3.2.5.  Enzymatic  hydrolysis  ..................................................................................................  49   3.2.6.  Oligomeric  sugar  analysis  ...........................................................................................  51   3.2.7.  HPLC  analysis  ..............................................................................................................  51   3.2.8.  In  vitro  rumen  digestibility  and  neutral  detergent  fiber  determination  ....................  52   3.2.9.  Statistical  analysis  .......................................................................................................  52   3.3.  Results  .......................................................................................................................  53   3.3.1.  Feedstock  characteristics  and  plot  yields  ...................................................................  53   3.3.2.   Relationship   of   AFEX TM   pretreatment   conditions   to   hydrolysis   yields   from   early   successional  samples  ............................................................................................................  56   3.3.3.  In  vitro  digestibility  .....................................................................................................  58   3.3.4.   Comparison   of   early   successional   old   field   replicates   to   late   successional   old   field   (LSF)  and  corn  stover  (CS)  samples  .......................................................................................  60   3.4.  Discussion  ..................................................................................................................  63   TM CHAPTER  4  :      SPECIES  SCALE:  OPTIMIZATION  OF  AFEX   PRETREATMENT  CONDITIONS  AND   ENZYME   MIXTURES   TO   MAXIMIZE   SUGAR   RELEASE   FROM   UPLAND   AND   LOWLAND   SWITCHGRASS  ......................................................................................................................  70   4.1.  Introduction  ...............................................................................................................  70   4.2.  Materials  and  methods  ..............................................................................................  72   4.2.1.  Feedstock   ....................................................................................................................  72   4.2.2.  Pre-­‐wash  .....................................................................................................................  72   4.2.3.  Composition  analysis  ..................................................................................................  73   4.2.4.  Design  of  experiments  ................................................................................................  75   4.2.4.1.  Response  surface  optimization  of  pretreatment  conditions  ...................................  75   4.2.4.2.  Mixture  optimization  of  hydrolysis  enzymes  ...........................................................  76   TM 4.2.5.  AFEX  pretreatment  ................................................................................................  78   4.2.6.  Enzymatic  hydrolysis  ..................................................................................................  78   4.2.7.  Soluble  total  and  oligomeric  sugar  analysis   ................................................................  80   4.2.8.  HPLC  analysis  ..............................................................................................................  80   4.3.  Results  and  discussion  ................................................................................................  81   4.3.1.  Switchgrass  characteristics  .........................................................................................  81   4.3.2.  Pretreatment  parameter  optimization  .......................................................................  82   4.3.3.  Commercial  enzyme  mixture  optimization  ................................................................  91   4.4.  Conclusions  ..............................................................................................................  100   viii       CHAPTER   5   :       COMPONENT   SCALE:   OPTIMIZING   HARVEST   OF   CORN   STOVER   FRACTIONS   TM BASED   ON   OVERALL   SUGAR   YIELDS   FOLLOWING   AFEX   PRETREATMENT   AND   ENZYMATIC   HYDROLYSIS  .......................................................................................................................  101   5.1.  Introduction  .............................................................................................................  101   5.2.  Materials  and  methods  ............................................................................................  103   5.2.1.  Harvest  and  milling  ...................................................................................................  103   5.2.2.  Composition  analysis  ................................................................................................  104   TM 5.2.3.  AFEX  treatment  ....................................................................................................  105   5.2.4.  Enzymatic  hydrolysis  ................................................................................................  105   5.2.5.  Sugar  analysis  ...........................................................................................................  107   5.2.6.  Statistical  analysis  .....................................................................................................  107   5.2.7.  Empirical  modeling  of  harvest  scenarios  ..................................................................  108   5.3.  Results  .....................................................................................................................  108   5.3.1.  Composition  analysis  ................................................................................................  108   TM 5.3.2.  AFEX  pretreatment  and  hydrolysis  ......................................................................  109   5.3.3.  Statistical  analysis  .....................................................................................................  115   5.3.4.  Optimization  of  harvest  scenarios  ............................................................................  118   5.4.  Discussion  ................................................................................................................  120   5.4.1.  Composition  analysis  ................................................................................................  120   TM 5.4.2.  AFEX  pretreatment  ..............................................................................................  120   5.4.3.  Statistical  analysis  .....................................................................................................  123   5.4.4.  Empirical  modeling  of  harvest  scenarios  ..................................................................  125   5.5.  Conclusions  ..............................................................................................................  130   TM CHAPTER   6   :       MOLECULAR   SCALE:   AFEX   PRETREATMENT   OF   POPLAR   MODIFIED   FOR   LIGNIN  CONTENT  AND  COMPOSITION  ................................................................................  132   6.1.  Introduction  .............................................................................................................  132   6.2.  Materials  and  methods  ............................................................................................  135   6.2.1.  Feedstock   ..................................................................................................................  135   6.2.2.  Composition  analysis  ................................................................................................  136   6.2.3.  Acidic  and  alkaline  digestibility  assay  .......................................................................  136   TM 6.2.4.  AFEX  pretreatment  ..............................................................................................  137   6.2.5.  Enzyme  optimization  -­‐  enzymatic  hydrolysis  and  sugar  analysis   ..............................  138   6.2.6.  Poplar  comparison  -­‐  enzymatic  hydrolysis  and  sugar  analysis  .................................  139   6.2.7.  Statistical  analysis  .....................................................................................................  139   6.3.  Results  and  discussion  ..............................................................................................  141   ix       6.3.1.  Enzyme  optimization  ................................................................................................  141   6.3.2.  Influence  of  4CL  downregulation  on  cell  wall  composition  ......................................  142   6.3.3.  Influence  of  CCR  downregulation  on  cell  wall  composition  .....................................  148   6.3.4.   Influence   of   composition   and   pretreatment   on   enzymatic   digestibility   –   F5H   upregulation  .......................................................................................................................  150   6.3.5.  Influence  of  composition  and    pretreatment  on  enzymatic  digestibility  –  4CL  and  CCR   downregulation  ..................................................................................................................  154   6.4.  Conclusions  ..............................................................................................................  162   CHAPTER  7  :    CONCLUSIONS  AND  RECOMMENDATIONS  .....................................................  165   7.1.  Conclusions  ..............................................................................................................  165   7.2.  Recommendations  for  future  research  .....................................................................  167   APPENDIX  A  :  SUPPLEMENTARY  INFORMATION  FOR  CHAPTER  2  ........................................  171   APPENDIX  B  :  SUPPLEMENTARY  INFORMATION  FOR  CHAPTER  3   .........................................  177   APPENDIX  C  :  SUPPLEMENTARY  INFORMATION  FOR  CHAPTER  4  .........................................  184   APPENDIX  D  :  SUPPLEMENTARY  INFORMATION  FOR  CHAPTER  6  ........................................  191   REFERENCES  .......................................................................................................................  202   x       LIST  OF  TABLES       Table   1.1:   Key   differences   in   the   cell   wall   chemistry   for   different   plant   classifications.   Information   on   hemicelluloses   is   from   [15].   For   the   lignin   subunits,   (+)   represents   relative   abundance  within  the  lignin  polymer  and  (-­‐)  represents  absence.  .............................................  6   Table  1.2:  Publication  status  of  thesis  chapters  and  other  relevant  work.  ............................  12   Table   2.1:   Residue-­‐to-­‐grain   ratios   of   common   crops   from   various   sources.   All   data   are   specific   to   Mainland   China   except   those   from   Kim   and   Dale   [76],   which   are   based   on   U.S.   values  and  are  provided  for  reference.  .....................................................................................  18   Table   2.2:   Estimates   on   total   crop   residue   yields   and   usable   amounts   of   crop   residues   in   Mainland  China  from  various  sources.  The  usable  residue  values  for  [89,  101]  only  take  into   account  the  amount  of  collectable  residues  and  not  competing  uses.  .....................................  26   Table  2.3:  Estimated  use  of  crop  residues  in  Mainland  China  from  various  sources.  ............  37   Table  3.1:  Species  composition,  biomass  yield,  and  distribution  for  the  GLBRC  old-­‐field  and   LTER  replicates.  .........................................................................................................................  54   Table   3.2:   Composition   analysis   data   as   %   of   total   dry   matter   (DM).    The  standard  error  is   reported   in   parenthesis   and   represents   three   replicates.   LSF   =   late   successional   old   field   sample.  ......................................................................................................................................  55   TM Table   3.3:   In   vitro   rumen   digestibility   of   untreated   and   AFEX -­‐treated   early   successional   old  field  samples.  ......................................................................................................................  59   Table  4.1:  Composition  analysis  data  for  untreated   Alamo  and  Shawnee  switchgrass  (%  of   total   dry   biomass).  Washed  switchgrass  samples  had  been  sequentially  washed  three  times   with   80-­‐90°C   water   in   order   to   remove   the   majority   of   the   soluble   sugars.   Values   with   different   superscripts   in   each   row   were   statistically   different   based   on   Tukey’s   pair-­‐wise   comparisons  with  α  =  0.05.  .......................................................................................................  74   Table   4.2:   Response   surface   optimization   of   pretreatment   parameters   in   terms   of   total   monomeric  and  oligomeric  glucose  and  xylose  release  following  enzymatic  hydrolysis.    A  =   ammonia  loading  (g  NH3:g  DM),  B  =  water  loading  (g  H2O:g  DM),  C  =  temperature  (°C),    D  =   residence  time  (min).  ................................................................................................................  84   xi       TM Table  4.3:  Comparison  of  literature  on  optimal  AFEX   pretreatment  conditions  and  sugar   yields   from   switchgrass.   Predicted   yields   are   sugar   yields   as   predicted   by   the   response   surface   regression   model.   Values   in   parenthesis   represent   monomeric   +   oligomeric   sugars.   Pred.  =  predicted  .......................................................................................................................  87   Table  4.4:  Mixture  regression  of  enzymes  and  total  protein  loading  in  terms  of  sugar  release   following  enzymatic  hydrolysis  of  Alamo  switchgrass.   S  =  Spezyme®  CP;  B  =  Novozyme®  188   (β-­‐glucosidase);  X  =  Multifect®  Xylanase;  P  =  Multifect®  Pectinase;  Amt  =  Enzyme  loading  in   terms  of  total  protein.  ...............................................................................................................  92   Table  4.5:  Mixture  regression  of  enzymes  and  total  protein  loading  in  terms  of  sugar  release   following  enzymatic  hydrolysis  of  Shawnee  switchgrass.  S  =  Spezyme®  CP;  B  =  Novozyme®   188  (β-­‐glucosidase);  X  =  Multifect®  Xylanase;  P  =  Multifect®  Pectinase;  Amt  =  Enzyme  loading   in  terms  of  total  protein.  ...........................................................................................................  93   Table   4.6:   Base   case   and   optimized   commercial   enzyme   mixtures   for   monomeric   and   monomeric   +   oligomeric   (total)   sugar   yields   (both   glucose   +   xylose)   from   Alamo   and   Shawnee  switchgrass.     Enzyme   proportions   are   expressed   as   a   percentage   of   total   protein   in   the   enzyme   mixture.   Sugar   yields   are   presented   as   %   of   sugars   theoretically   available   in   untreated,  dry  biomass.  Sugar  yields  in  parentheses  represent  total  monomeric  +  oligomeric   sugars,  and  those  not  in  parentheses  represent  monomeric  sugar  yields.  Enzyme  activities  for   each  enzyme  mixture  are  estimated  based  on  activities  per  mL  reported  by  Dien  et  al.  [194]   for  each  of  the  commercial  enzyme  preparations.  ...................................................................  95   Table  5.1:    Enzymatic  hydrolysis  xylanase  loading  in  terms  of  xylan  content  of  each  fraction.  .................................................................................................................................................  106   Table  5.2:  Corn  stover  composition  for  early  and  late  harvest  stover  fractions.  Values  with   different   superscripts   in   a   column   were   statistically   different   using   Tukey’s   pairwise   comparison  with  α  =  0.05.    The  ‘other’  column  determined  by  difference  from  100%.  .........  109   Table  5.3:  Analysis  of  variance  for  factors  influencing  sugar  yields.  .....................................  114   Table   5.4:   Optimized   harvest   scenarios   based   on   desired   sugar   and   ammonia   fiber   expansion  ammonia  loading.  .................................................................................................  118   Table   5.5:   Estimated   yields   for   70%   collection   of   selectively   harvested   corn   stover   (SHCS)   TM following  AFEX ,  enzymatic  hydrolysis  and  fermentation.  ................................................  119   xii       Table   5.6:   Estimated   yields   for   30%   collection   of   selectively   harvested   corn   stover   (SHCS)   TM following  AFEX ,  enzymatic  hydrolysis  and  fermentation.  ................................................  119   Table   6.1:   Pearson   coefficients   for   4CL   poplar   sample   cell   wall   components.   Data   were   analyzed  as  the  combined  wildtype  and  transgenic  samples.  ..............................................  146   Table   6.2:   Pearson   coefficients   for   CCR   poplar   sample   cell   wall   components.   Data   were   analyzed  as  the  combined  wildtype  and  transgenic  samples.  ..............................................  149   Table   6.3:   Pearson   coefficients   for   24   h   and   168   h   enzymatic   hydrolysis   sugar   yields   from   untreated  and  pretreated  4CL  poplar  samples.  Data  were  analyzed  as  the  conglomerate  of  all   control  and  transgenic  samples.  AFEX TM -­‐pretreatment  conditions  were  1:1  g  NH3:g  DM;  1:1  g   H2O:g  DM;  180°C  and  20  min.  ................................................................................................  156   Table   6.4:   Pearson   coefficients   for   24   h   and   168   h   enzymatic   hydrolysis   sugar   yields   from   untreated   and   pretreated   CCR   poplar   samples.   Data  were  analyzed  as  the  conglomerate  of   TM all   control   and   transgenic   samples.   AFEX -­‐pretreatment   conditions   were   1:1   g   NH3:g   DM;   1:1  g  H2O:g  DM  for  all  pretreated  samples.  ............................................................................  159   Table   A.1:   Paper   and   paperboard   production   (million   metric   tons)   in   2006.   Values   for   provinces  with  greater  than  1  million  metric  tons  of  production  (and  Jiangxi)  are  from  [93].  All   other  values  are  estimated  from  [92].  ....................................................................................  171   Table  A.2:  Total  and  non-­‐pastured  ruminant  animals  (million  head),  grazing  and  pastureland   area  (million  ha),  and  carrying  capacity  per  province  in  2006.  Data  on  ruminant  production   and   pastureland   area   are   from   [70],   and   carrying   capacity   data   are   from   [95],   except   for   Heilongjiang,  Jilin,  Sichuan,  Guizhou  and  Gansu,  which  were  estimated.   ...............................  172   Table  A.3:  Estimated  amount  of  vegetable  sown  area  cropped  only  with  vegetables,  either   single-­‐cropped   or   triple-­‐cropped   (not   multi-­‐cropped   with   other   types   of   crops)   for   different   regions  in  Mainland  China.  Regions  defined  based  on  Qiu  et  al.  [96].  ...................................  173   Table   A.4:   National   and   provincial   sown,   cultivated,   and   fallow   land   area   of   cereals,   legumes,  tubers,  oilseeds  and  cotton  crops  in  Mainland  China  in  2006.  ..............................  174   Table  A.5:  Crop   residue   production   and   use   by   province.   The  amount  that  is  usable  as  fuel   includes   the   amount   used   for   rural   energy   and   assumes   this   amount   is   either   replaced   by   combustion   of   crop   straw   for   electricity   or   that   this   energy   is   replaced   by   some   other   means.   Values  for  the  amount  of  crop  residues  burned  on-­‐field  are  from  Wang  and  Zhang  [100].  ...  175   xiii       Table  B.1:  Species  composition  of  GLBRC  early  successional  old  field  replicates.  ...............  177   Table  B.2  :  Species  composition  of  LTER  late  successional  old  field  replicates.  ...................  178   Table  B.3  :  Pretreatment  conditions  and  total  sugar  yields  for  each  design  point  for  the  early   successional  samples.  Three  extra  design  points  were  included  in  an  attempt  to  improve  the   model   fit.   The   center   design   point   (red)   was   replicated   three   times.   The   pretreatment   condition  chosen  for  further  experiments  is  highlighted  in  teal.    The  top  five  sugar  yields  for   each  sample  and  the  highest  total  sugar  yield  for  each  sample  are  highlighted  in  orange  and   yellow,  respectively.  One  sample  of  E87  (gray)  was  excluded  from  the  analysis  as  a  statistical   outlier.  .....................................................................................................................................  181   Table  B.4:  Response  surface  model  coefficients  for  pretreatment  optimization  of  the  early   successional   samples   in   terms   of   total   monomeric   and   oligomeric   glucose   and   xylose   release  following  enzymatic  hydrolysis.   A  =  ammonia  loading  (g:g  DM),  B  =  water  loading  (g:g   2 2 2 DM),  C  =  temperature  (°C),  D  =  residence  time  (min).    Pred.  R   =  predictive  R  value;  Adj.  R   =   2 adjusted  R  value.  ...................................................................................................................  182   Table   C.1:   Experimental   levels   and   additional   design   points   for   Box-­‐Behnken   response   TM surface  optimization  of  AFEX  pretreatment  conditions.  ..................................................  186   Table   C.2:   ANOVA   of   pretreatment   optimization   regression   model   for   total   sugar   yields   (monomers  +  oligomers  of  glucose  and  xylose)  from  Alamo  and  Shawnee  switchgrass.  ....  187   Table  C.3:  ANOVA  of  Alamo  switchgrass  enzyme  mixture  regression.  Total  glucose  and  total   xylose  refer  total  monomeric  +  oligomeric  sugar  yields.  ........................................................  188   Table   C.4:   ANOVA   of   Shawnee   switchgrass   enzyme   mixture   regression.   Total   glucose   and   total  xylose  refer  to  the  total  monomeric  +  oligomeric  sugar  yields.  .....................................  189   Table   D.1:   Composition   data   for   the   NM-­‐6   poplar,   and   F5H,   CCR,   and   4CL   control   and   transgenic  lines.  ......................................................................................................................  191   Table   D.2:   Variance   within   the   fully   nested   ANOVA   for   each   line,   related   to   downregulation   strength  (4CL  only),  parent  line,  or  sample  pool  for  the  4CL  and  CCR  samples.  An  ANOVA  was   unable   to   be   performed   for   the   4CL   lignin   monomers   as   information   on   the   individual   data   replicates  was  not  provided.  ...................................................................................................  196   xiv       LIST  OF  FIGURES       Figure  1.1:  Worldwide  primary  energy  consumption:  1980  –  2008.  For  interpretation  of  the   references  to  color  in  this  and  all  other  figures,  the  reader  is  referred  to  the  electronic  version   of  this  dissertation.  ......................................................................................................................  1   Figure  1.2:  Generic  process  flow  diagram  of  the  cellulosic  ethanol  conversion  process.  ........  3   Figure   1.3:   Common   bioenergy   plants   and   model   laboratory   species   arranged   in   groups   according  to  botanical  classification.  .........................................................................................  5   Figure  1.4:    Lignin  synthesis  pathway  for  the  monolignols  p-­‐coumaryl,  sinapyl  and  coniferyl   alcohol.  From  Vanholme  et  al.  [21].  PAL  =  phenylalanine  ammonia-­‐lyase;  C4H  =  cinnamate  4-­‐ hydroxylase;   4CL   =   4-­‐coumarate:CoA   ligase;   C3H   =   p-­‐coumarate   3-­‐hydroxylase;   HCT   =   p-­‐ hydroxycinnamoyl-­‐CoA:quinate/   shikimate   p-­‐hydroxycinnamoyltransferase;   CCoAOMT   =   caffeoyl-­‐CoA   O-­‐methyltransferase;   CCR   =   cinnamoyl-­‐CoA   reductase;   F5H   =   ferulate   5-­‐ hydroxylase;   COMT   =   caffeic   acid   O-­‐methyltransferase;   CAD   =   cinnamyl   alcohol   dehydrogenase.  ...........................................................................................................................  8   Figure  1.5:  Scales  of  interaction  between  plant  feedstocks  and  energy  production  processes.  .....................................................................................................................................................  9   Figure  2.1:  Map  of  Mainland  China  with  provinces,  autonomous  regions,  and  municipalities.  ...................................................................................................................................................  15   Figure   2.2:   Estimated   crop   residue   yields   in   China   in   2006:   (A,   B)   Total   yields   for   each   6 prefecture  and  province  (10  Mg);  (C,  D)  Residue  yield  per  total  land  area  (Mg/ha);  (E,  F)  Per   capita   crop   residue   yield   (kg/person).   Residues   included   cereals,   legumes,   tubers,   cotton,   and   oilseed   crops,   and   the   amounts   were   estimated   using   the   average   harvest   indices   for   the   respective  crops  reported  by  Xie  and  colleagues  [74,  75].  ........................................................  28   6 Figure  2.3:  Spatial  distribution  of  total  residue  yields  (10   Mg)  for  different  types  of  crops  in   2006:  (A)  wheat,  (B)  rice,  (C)  corn,  (D)  cotton,  (E)  oilseeds,  and  (F)  legumes  and  tubers.   ......  30   Figure  2.4:  Amount  and  proportion  of  the  different  types  of  crop  residues  for  each  province   in  Mainland  China.  ....................................................................................................................  31   Figure  2.5:   Cultivated  land  as  the  proportion  of  total  land  area  in  a  prefecture  or  province,   and   estimated   2006   crop   residue   yields   per   prefecture   or   province   cultivated   land   area   xv       (Mg/ha)   as   affected   by   the   chosen   harvest   indices.   (A,B)   Proportion   of   total   land   area   as   cultivated  land;  (C,  D)  Harvest  indices  from  Xie  et  al.  [74,  75];  and  (E,  F)  Harvest  indices  from   the  Ministry  of  Agriculture  report  [78].  .....................................................................................  35   Figure  2.6:  Estimated  ideal  use  of  crop  residues  in  each  province  in  Mainland  China  in  2006.   Values  for  the  amount  of  residues  used  as  animal  feed  and  returned  to  the  field  are  based  on   ideal   scenarios   without   overgrazing   or   adversely   effects   on   soil   health.   Red   circles   (at   the   same  scale  as  the  pie  graphs)  are  used  to  show  the  amount  of  additional  residues  needed  in   provinces  that  are  do  not  produce  the  amount  required  for  the  ideal  scenario.  Data  on  field   burning   of   crop   residue   are   from   [100].   A   table   listing   all   the   values   for   residue   use   and   on   field  burning  is  provided  in  the  supplemental  information  (Table  A.5).  ...................................  38   Figure   3.1:   Relationship   between   enzymatic   hydrolysis   glucose   and   xylose   yields   for   the   GLBRC   old   field   replicates.   (A)  Forb-­‐dominated  samples  (E7  and  E21).    (B)  Grass-­‐dominated   samples   (E60,   E83,   and   E87).     Yields   are   calculated   as   the   total   monomeric   and   oligomeric   sugar  solubilized  based  on  the  total  sugar  theoretically  available  in  the  untreated  dry  biomass.   For   the   E60,   E83,   and   E87   regressions,   (p   =   0.000).     Each   data   point   represents   one   of   30   different  pretreatment  conditions  (except  for  E87,  which  only  has  data  for  29  conditions  due   to  one  significant  outlier).  .........................................................................................................  57   Figure   3.2:   Comparative   monomeric   and   oligomeric   sugar   yields.     (A)   Glucose   yields.   (B)   Xylose   yields.   (C)   Total   sugar   (glucose   +   xylose)   yields.   Oligomeric   sugars   are   reported   in   monomeric   equivalents.   The   maximum   theoretical   sugar   yield   is   the   maximum   amount   of   glucose,  xylose  or  total  sugars  that  could  be  released  from  the  untreated  dry  biomass.    Total   (monomeric  and  oligomeric)  sugar  yields  with  different  letters  are  statistically  different  based   on   Tukey’s   test   (p   <   0.05).   E7,   E21,   E60,   E83,   E87   =   early   successional   old   field   replicates   from   the  GLBRC  intensive  site;  LSF  =  LTER  late-­‐successional  old  field  combined  sample;  CS  =  corn   stover.   Each   sample   was   subjected   to   the   same   pretreatment   and   enzymatic   hydrolysis   conditions.  .................................................................................................................................  61   Figure  3.3:  Correlation  of  glucose  (A)  and  xylose  (B)  percent  conversion  (g  sugar  released/g   theoretically  available  in  untreated  dry  biomass)  to  Klason  lignin  content  on  a  cell  wall  basis   for   the   early   successional   old   field   replicates,   late-­‐successional   old   field   sample,   and   corn   stover.  The  solid  line  represents  the  correlation  when  the  old  field  replicate  E60  (represented   by   an   open   circle)   is   included   in   the   analysis   and   the   dashed   line   indicates   the   correlation   when  sample  E60  is  removed  from  the  analysis.  ......................................................................  62   Figure  4.1:  Relationship  between  enzymatic  hydrolysis  glucose  and  xylose  yields  for  Alamo   and   Shawnee   switchgrass.   (A)   Monomeric   sugar   yields.     (B)   Total   monomeric   +   oligomeric   sugar   yields.     Yields   are   calculated   as   the   total   sugar   solubilized   based   on   the   total   sugar   xvi       theoretically   available   in   the   untreated   dry   biomass.   Each   data   point   represents   one   of   30   (Shawnee)  or  32  (Alamo)  different  pretreatment  experiments.  ...............................................  82   Figure   4.2:   Histograms   and   scatter   plots   showing   deviation   of   predicted   sugar   yields   from   actual   sugar   yields   for   the   pretreatment   regression   models.   (A)   Shawnee   histogram.   (B)   Alamo  histogram.  (C)  Shawnee  scatter  plot.  (D)  Alamo  scatter  plot.   ........................................  85   TM Figure   4.3:   Contour   plots   showing   the   interactive   effect   of   pairs   of   AFEX   pretreatment   parameters   on   monomeric   +   oligomeric   glucose   yields   from   (A)   Alamo   and   (B)   Shawnee   switchgrass.  The  two  pretreatment  parameters  not  shown  in  each  sub-­‐figure  were  held  at  the   optimal   level.   Hydrolysis   was   conducted   at   50°C,   200   rpm,   and   1%   glucan   loading   using   30   FPU  Spezyme®  CP  and  15  CBU  Novozyme®  188  per  g  glucan,  with  72  h  sampling.  ................  88   TM Figure   4.4:   Contour   plots   showing   the   interactive   effect   of   pairs   of   AFEX   pretreatment   parameters   on   monomeric   +   oligomeric   xylose   yields   from   (A)   Alamo   and   (B)   Shawnee   switchgrass.  The  two  pretreatment  parameters  not  shown  in  each  sub-­‐figure  were  held  at  the   optimal   level.   Hydrolysis   was   conducted   at   50°C,   200   rpm,   and   1%   glucan   loading   using   30   FPU  Spezyme®  CP  and  15  CBU  Novozyme®  188  per  g  glucan,  with  72  h  sampling.  ................  89   Figure   4.5:   Ternary   plots   based   on   the   enzyme   mixture   regression   model   showing   the   interactive  effect  of  Spezyme®  CP,  Multifect®  Xylanase  and  Multifect®  Pectinase  on  sugar   TM yields   from   AFEX   pretreated   Alamo   (A-­‐C)   and   Shawnee   (D-­‐F)   switchgrass.   (A,   D)   monomeric   glucose;   (B,   E)   monomeric   xylose;   (C,   F)   monomeric   +   oligomeric   xylose.   Percentages  of  each  enzyme  are  with  respect  to  total  protein  loading,  which  was  held  at  30   mg  per  g  glucan  for  these  figures.  β-­‐glucosidase  was  held  at  0%  in  all  plots.  Enzyme  loadings   were  constrained  to  Spezyme  CP  >  20%,  Multifect  Xylanase  <  80%  and  Multifect  Pectinase  <   80%  of  total  enzyme  protein  added  to  hydrolysis.  ....................................................................  96   Figure  4.6:  Monomeric  glucose  (A)  and  xylose  (B)  release  from  optimally  pretreated  Alamo   and  Shawnee  switchgrass  hydrolyzed  with  base  enzyme  loading  (15  FPU  Spezyme®  CP  and   30   CBU   Novozyme®   188   per   g   glucan)   compared   to   with   our   optimized   enzyme   mixture   for   each   variety   containing   Spezyme®   CP,   Multifect®   Xylanase   and   Multifect®   Pectinase.   Enzyme   loading   was   27   mg   protein   per   g   glucan   and   samples   were   taken   at   1   h,   24   h   and     168  h.  .........................................................................................................................................  98   xvii       TM Figure  5.1:  Effect  of  ammonia  fiber  expansion  (AFEX )  pretreatment  ammonia  loading  and   xylanase   addition   on   enzymatic   hydrolysis   monomeric   sugar   yields.     Glucose   yields   are   reported   in   part   A   and   xylose   yields   are   in   part   B.     All   AFEX TM   runs   were   kept   at   constant   moisture   content   (60%   dry-­‐weight   basis),   temperature   (90°C)   and   residence   time   (5   min).     Yields  are  in  terms  of  sugar  available  in  untreated  dry  biomass.  ............................................  111   TM Figure   5.2:   Effect   of   ammonia   fiber   expansion   (AFEX )   pretreatment   temperature   on   TM enzymatic   hydrolysis   monomeric   sugar   yields.     All   AFEX   runs   were   kept   at   a   constant   -­‐1 moisture  content  (60%  dry-­‐weight  basis),  ammonia  loading  (1.0  g  NH3  g  dry  biomass)  and   residence  time  (5  min).    Yields  are  in  terms  of  sugar  available  in  untreated  dry  biomass.    Glu  =   glucose,  Xyl  =  xylose,  MTSY  =  maximum  theoretical  sugar  yield.  ............................................  112   TM Figure   5.3:   Effect   of   ammonia   fiber   expansion   (AFEX )   moisture   content   and   residence   time   on   enzymatic   hydrolysis   monomeric   sugar   yields.    Base  AFEX TM  conditions:    moisture   -­‐1 content  (60%  dry-­‐weight  basis),  ammonia  loading  (1.0  g  NH3  g  dry  biomass),  temperature   (90°C)   and   residence   time   (5   min).     Yields   are   in   terms   of   sugar   available   in   untreated   dry   biomass.     MC   =   moisture   content,   RT   =   residence   time,   Glu   =   glucose,   Xyl   =   xylose,   MTSY   =   maximum  theoretical  sugar  yield.  ...........................................................................................  113   TM Figure  5.4:  Interaction  effect  plot  of  AFEX  parameters,  stover  fraction  and  harvest  period   on   monomeric   (A)   glucose   and   (B)   xylose   yields   (g   sugar   per   kg   untreated   dry   biomass)   following   72   h   enzymatic   hydrolysis.     MC   =   moisture   content,   RT   =   residence   time,   DWB   =   dry-­‐weight  basis,  N  =  no  xylanase  added,  Y  =  xylanase  added  (10%  of  total  cellulase  protein).  .................................................................................................................................................  115   Figure   5.5:   Estimated   dry   matter   distribution   for   70%   and   30%   (dry-­‐weight   basis)   harvest   of   late  harvest  corn  stover.    Percentages  of  the  individual  fractions  harvested  are  based  on  the   total  amount  of  each  fraction  available.  .................................................................................  117   TM Figure   6.1:   Ternary   diagrams   for   enzyme   optimization   experiments   on   AFEX -­‐treated   poplar   (NM-­‐6)   using   either   (A)   15   or   (B)   30   mg   total   protein   per   g   glucan.   Hydrolysis   performed   was   in   microplates   at   0.2%   glucan   loading   with   quadruplicates   of   each   enzyme   combination.   Glucose   yields   are   reported   as   the   percentage   of   the   theoretically   available   glucan  (hemicellulose  glucan  and  cellulose)  in  untreated,  dry  biomass  that  was  released.  ...  140   xviii       Figure  6.2:  Structural  carbohydrate  and  acetyl  bromide  lignin  content  of  the  control  and  4CL   downregulated   transgenics,   arranged   by   strength   of   downregulation   (weak,   medium,   or   strong)  and  parent  line.  Each   data   point   represents   one   pool.     The   (+)   symbols   represent   the   average   structural   carbohydrate   content   across   pools   for   each   transgenic   line,   and   the   (x)   symbol   represents   the   same   for   acetyl-­‐bromide   lignin.   Values   with   different   letters   for   the   average   line   structural   carbohydrates   or   lignin   are   statistically   different   based   on   Tukey’s   pairwise  comparisons  (95%  CI)  (p  <  0.05).  ..............................................................................  143   Figure   6.3:   Boxplot   of   lignin,   total   glucose   (from   hemicellulose   glucan   and   crystalline   cellulose),  and  xylose  within  the  cell  wall  for  the  different  strengths  of  4CL  downregulation   compared   to   the   control.   Boxplots   with   different   letters   within   each   component   are   statistically  different  based  on  Tukey’s  pairwise  comparisons  (95%  CI)  (P  <  0.05).  ................  144   Figure   6.4:   95%   confidence   intervals   around   the   mean   S:G   ratio   for   each   4CL   control   and   transgenic  line.  UCL  and  LCL  lines  represent  the  upper  and  lower  confidence  limit  for  all  the   control   samples.   Stars   represent   lines   that   were   statistically   different   from   all   the   control   samples.  Ctrl  =  control.  ............................................................................................................  147   Figure   6.5:   95%   confidence   intervals   around   the   mean   S:G   ratio   for   each   CCR   control   and   transgenic  line.  UCL  and  LCL  lines  represent  the  upper  and  lower  confidence  limit  for  all  the   control   samples.   Stars   represent   lines   that   were   statistically   different   from   all   the   control   samples.   ...................................................................................................................................  150   Figure  6.6:  Pretreatment  and  enzymatic  digestibility  assay  for  control  and  C4H::F5H  poplar   samples.   Green   boxplots   represent   no   pretreatment,   red   represents   acidic,   and   blue   represents  alkaline.  Boxplots  with  different  letters  have  statistically  different  glucose  release   based  on  Tukey’s  pairwise  comparisons  (95%  CI),  (p  <  0.05).  .................................................  151   Figure  6.7:  (A)  Glucose  and  (B)  xylose  yields  from  control  and  C4H::F5H  poplar  samples  for   TM   different  AFEX pretreatment  conditions.  Sugar  yields  (168  h)  with  different  letters  within   each  subplot  are  statistically  different  based  on  Tukey’s  pairwise  comparisons  (95%  CI),  (p  <   0.05).   All   samples   were   pretreated   using   1:1   g   NH3:g   DM   and   1:1   g   H2O:g   DM.   Enzymatic   hydrolysis  was  conducted  at  200  rpm  and  1.25%  total  sugar  loading  with  24  mg  total  protein   per   g   cell   wall   sugars   (80%   Accellerase®   1500,   10%   Accellerase®   XY,   10%   Multifect®   Pectinase).  Ctrl  =  Control;  F5H  =  C4H::F5H.  .............................................................................  153   Figure  6.8:  Correlations  between  sugar  yields  and  cell  wall  composition  (g/g  DM)  for  4CL  (A-­‐ H)   and   CCR   (I-­‐L)   poplar   samples   across   all   pretreatment   conditions.   The   top   row   shows   correlation   of   sample   xylose   content   (A-­‐B)   and   glucose   content   (C-­‐D)   on   glucose   release   from   xix       4CL  poplar.  The  effect  of  lignin  content  on  glucose  yields  (E-­‐F,  I-­‐J)  and  xylose  yields  (G-­‐H,  K-­‐L)   is   shown   for   4CL   and   CCR   poplar.   Untreated   poplar   samples   are   shown   in   black   and   pretreated  samples  are  shown  in  red.  Sugar  yields  are  expressed  in  terms  of  the  total  sugar   theoretically  present  in  the  untreated  dry  biomass.    Lines  represent  the  linear  regression  for   the   respective   parameters.   The   dashed   lines   in   (K)   and   (L)   represent   all   of   the   pretreated   samples.   The   two   solid   lines   represent   the   separate   regression   for   transgenic   samples   and   control  samples.  ......................................................................................................................  158   Figure  6.9:  Influence  of  intial  rate  of  hydrolysis  at  24  h  on  168  hr  glucose  and  xylose  yields   from  the  control  and  transgenic  (A)  4CL  and  (B)  CCR  control  poplar.  Xylose  yields  are  shown   in  red  and  glucose  yields  in  black.  Solid  ines  represent  regressions  on  the  given  sugar  yields   for   separate   untreated   and   pretreated   groupings.   Dashed   lines   in   each   figure   represent   the   linear  regression  for  untreated  and  pretreated  samples  as  one  group.  .................................  160   Figure   A.1:   Residue   density   based   on   prefecture   cultivated   area   (Mg/ha)   for   different   crops   in  2006:  (A)  wheat,  (B)  rice,  (C)  corn,  (D)  cotton,  (E)  oilseeds,  and  (F)  legumes  and  tubers.  176   Figure   B.1:   Range   of   total   sugar   yields   and   location   of   the   chosen   pretreatment   condition   within  the  range  for  each  early  successional  old  field  sample.   The  red  circle  represents  the   location  of  the  pretreatment  condition  that  was  chosen  as  a  basis  for  further  experimentation   within  the  range  of  the  sugar  yield  raw  data.  .........................................................................  183   TM Figure   C.1:   Contour   plots   showing   the   interactive   effect   of   pairs   of   AFEX   pretreatment   parameters  on  monomeric  glucose  yields  from  (A)  Alamo  and  (B)  Shawnee  switchgrass.  The   two   pretreatment   parameters   not   shown   in   each   sub-­‐figure   were   held   at   the   optimal   level.   Enzymatic  hydrolysis  was  conducted  at  50°C,  200  rpm,  and  1%  glucan  loading  using  30  FPU   Spezyme®  CP  and  15  CBU  Novozyme®  188  per  g  glucan,  with  72  h  sampling.  .....................  184   TM Figure   C.2:   Contour   plots   showing   the   interactive   effect   of   pairs   of   AFEX   pretreatment   parameters  on  monomeric  xylose  yields  from  (A)  Alamo  and  (B)  Shawnee  switchgrass.   The   two   pretreatment   parameters   not   shown   in   each   sub-­‐figure   were   held   at   the   optimal   level.   Enzymatic  hydrolysis  was  conducted  at  50°C,  200  rpm,  and  1%  glucan  loading  using  30  FPU   Spezyme®  CP  and  15  CBU  Novozyme®  188  per  g  glucan,  with  72  h  sampling.  .....................  185   Figure   D.1:   Hemicellulose   glucose   vs.   mannose   content   for   4CL   poplar   samples.   The   equation  represents  the  linear  regression  of  the  data.    Glucose  is  derived  from  hemicellulose   and  does  not  include  cellulose-­‐derived  or  soluble  glucose.  ....................................................  197   xx       Figure   D.2:   Untreated   wildtype   and   4CL   downregulated   poplar   (A)   glucose   and   (B)   xylose   yields.  Sugar  yields  with  different  letters  within  each  subplot  are  statistically  different  based   on  Tukey’s  pairwise  comparisons  (95%  CI),  (p  <  0.05)  and  are  not  comparable  between  24  h   and   168   h   data.   Enzymatic   hydrolysis   was   conducted   at   200   rpm   and   1.25%   total   sugar   loading   with   24   mg   total   protein   per   g   cell   wall   sugars   (80%   Accellerase®   1500,   10%   Accellerase®  XY,  10%  Multifect®  Pectinase).  .........................................................................  198   Figure   D.3:  Pretreated   wildtype   and   4CL   downregulated   poplar   (A)   glucose   and   (B)   xylose   yields.  Sugar  yields  with  different  letters  within  each  subplot  are  statistically  different  based   on  Tukey’s  pairwise  comparisons  (95%  CI),  (p  <  0.05)  and  are  not  comparable  between  24  h   and  168  h  data.  All  samples  were  pretreated  using  1:1  g  NH3:g  DM;  1:1  g  H2O:g  DM;  180°C  for   20  min.  Enzymatic  hydrolysis  was  conducted  at  200  rpm  and  1.25%  total  sugar  loading  with   24   mg   total   protein   per   g   cell   wall   sugars   (80%   Accellerase®   1500,   10%   Accellerase®   XY,   10%   Multifect®  Pectinase).  ............................................................................................................  199   Figure  D.4:  Control  and  CCR  downregulated  poplar  (A)  glucose  and  (B)  xylose  yields.  Sugar   yields   with   different   letters   within   each   subplot   are   statistically   different   based   on   Tukey’s   pairwise  comparisons  (95%  CI),  (p  <  0.05)  and  are  not  comparable  between  24  h  and  168  h   data.  Samples  were  pretreated  using  1:1  g  NH3:DM  and  1:1  g  H2O:g  DM.  Enzymatic  hydrolysis   was  conducted  at  200  rpm  and  1.25%  total  sugar  loading  with  24  mg  total  protein  per  g  cell   wall  sugars  (80%  Accellerase®  1500,  10%  Accellerase®  XY,  10%  Multifect®  Pectinase).  .....  200       xxi       KEY  TO  SYMBOLS  AND  ABBREVIATIONS     ABSL  =  acetyl  bromide  soluble  lignin   4CL  =  4-­‐coumarate:CoA  ligase   TM AFEX  =  ammonia  fiber  expansion  pretreatment   ANOVA  =  analysis  of  variance   CAD  =  cinnamyl  alcohol  dehydrogenase   COMT  =  caffeic  acid  o-­‐methyltransferase   CCR  =  cinnamoyl-­‐CoA  reductase   CS  =  corn  stover   DM  =  dry  matter   dwb  =  dry  weight  basis   F5H  =  ferulate-­‐5-­‐hydroxylase   GLBRC  =  Great  Lakes  Bioenergy  Research  Center   HPLC  =  high-­‐performance  liquid  chromatography   KBS  =  W.K.  Kellogg  Biological  Station   LSF  =  late  successional  old  field  combined  sample   MANOVA  =  multivariate  analysis  of  variance   MESP  =  minimum  ethanol  selling  price   MTSY  =  maximum  theoretical  sugar  yield   NDF  =  neutral  detergent  fiber   LTER  =  Long-­‐Term  Ecological  Research   R1-­‐R5  =  GLBRC  intensive  site  old  field  replicates   SF1-­‐SF3=  LTER  late-­‐successional  old  field  treatments   S:G  =  syringyl:guaicyl  lignin  ratio   SHCS  =  selectively  harvested  corn  stover   SOC  =  soil  organic  carbon xxii         CHAPTER  1 :     INTRODUCTION   The   issues   related   to   the   development,   acquisition,   and   use   of   fossil   fuels   are   becoming   increasingly   global   as   third   world   countries   understandably   attempt   to   become   more   developed   and   worldwide   primary   energy   consumption   continues   to   climb   (Figure  1.1).       Figure  1.1:  Worldwide  primary  energy  consumption:  1980  –  2008.   For   interpretation   of   the   references   to   color   in   this   and   all   other   figures,   the   reader   is   referred   to   the   electronic   version   of   this   dissertation.     These   issues,   including   climate   change,   energy   security,   and   pollution,   are   compounded   by   the   increasing   global   population.   Bioenergy,   or   the   production   of   energy   from   plant   and   algal   feedstocks,   is   one   form   of   alternative   energy   and   perhaps   the   one   most   surrounded   by   confusion  and  controversy.  Barriers  to  large-­‐scale  bioenergy  production  include  the  supposed   competition   between   food   and   fuel   production   [1-­‐3]   and   uncertainty   surrounding   the   long-­‐ 1       term   effects   of   growing   bioenergy   crops   on   global   emissions   [4-­‐6].   But   there   are   a   number   of   benefits   surrounding   bioenergy   production   that   should   not   be   overlooked.   First,   bioenergy   produced  from  lignocellulosic  plant  materials,  such  as  waste  paper;  agricultural  and  forestry   residues;   and   dedicated   herbaceous   and   woody   energy   crops;   does   not   necessarily   have   to   compete   with   or   displace   food   production.   Second,   the   generation   of   usable   energy   from   lignocellulosic   materials   is   a   near-­‐term   solution.   If   progress   on   current   projects   continues,   there   could   be   significant   quantities   produced   within   the   next   five   years   [7].   Finally,   certain   methods,  infrastructure,  and  supply  chains  for  biofuel  production  from  lignocellulosics,  could   readily   transition   into   markets   such   as   animal   feed,   bio-­‐based   chemicals,   polymers,   and   pharmaceuticals,  for  which  renewable  sources  other  than  biomass  do  not  exist  [8-­‐11].     The   term   lignocellulose   refers   to   the   three   main   classes   of   compounds   found   within   plant   cell   walls:   cellulose,   hemicellulose,   and   lignin.   Ethanol   produced   from   cellulosics   is   derived  from  the  long-­‐chain  carbohydrates,  cellulose  and  hemicellulose.  Except  for  the  source   material  and  greater  complexity  of  the  process,  the  ethanol  that  is  produced  from  a  cellulosic   material   is   identical   to   ethanol   produced   from   starchy   materials,   such   as   corn   or   wheat   grain,   or   from   sucrose-­‐rich   materials,   such   as   sugar   cane   or   sugar   beets.   The   biochemical   process   for   conversion   of   cellulosics   to   ethanol   requires   three   key   steps:   pretreatment   (a   chemical   and/or   mechanical   process   necessary   to   disrupt   the   cell   wall   structure),   followed   by   enzymatic   saccharification   to   release   cell   wall   sugars   and   subsequent   or   simultaneous   fermentation  to  convert  these  sugars  into  ethanol  (Figure  1.2).     2         Figure  1.2:  Generic  process  flow  diagram  of  the  cellulosic  ethanol  conversion  process.     TM 1 Ammonia   fiber   expansion   (AFEX )  is   one   pretreatment   that   is   used   to   disrupt   the   TM plant   cell   wall   prior   to   enzymatic   conversion.   AFEX   is   an   alkali   pretreatment   that   uses   anhydrous   or   conc-­‐entrated   ammonia   as   a   reactant.   Ammonia   and   water   react   with   the   biomass,   cleaving   internal   bonds   via   ammonolysis   and   hydrolysis   reac-­‐tions   and   solubilizing   components   within   the   cell   wall   that   are   later   deposited   on   the   biomass   surface.   This   TM effectively   opens   the   cell   wall   structure,   allowing   greater   enzyme   access.   AFEX   reactions   are   typically   operated   at   temperatures   from   60   –   180°C,   for   less   than   an   hour,   and   at   pressures  ranging  from  100  –  600  psi.  At  the  end  of  the  residence  time,  the  reactor  is  vented,                                                                                                               1 TM  AFEX  is  a  registered  trademark  of  trademark  of  MBI  International,  Lansing,  Michigan   3       releasing   excess   ammonia   gas   and   water   vapor   to   be   recycled,   and   the   system   is   cooled.   TM AFEX   is   a   unique   pretreatment   process   in   that   it   uses   a   much   smaller   amount   of   water   compared  to  many  of  the  other  forms  of  pretreatment.  Because  of  this,  there  are  no  separate   liquid  streams  produced  and  unless  the  biomass  is  washed  following  pretreatment,  all  of  the   biomass  components  are  present  in  the  subsequent  enzymatic  hydrolysis  step.     As  mentioned  earlier,  potential  cellulosic  feedstocks  for  bioethanol  production  include   waste  paper,  agricultural  and  forestry  residues,  and  dedicated  woody  and  herbaceous  energy   crops.  It  is  important  for  the  biorefinery  choose  the  best  feedstocks  for  their  process  and  to   fully   make   use   of   any   feedstock   that   is   selected.   This   is   because   the   feedstock   and   associated   handling   costs   are   predicted   to   contribute   the   greatest   amount   to   the   cost   of   producing   cellulosic   ethanol,   more   than   the   amount   contributed   by   pretreatment,   enzymatic   hydrolysis,   and   fermentation   combined   [12],   and   the   relative   importance   of   the   feedstock   will   only   increase   as   the   liquid   biofuel   industry   matures.   So   research   on   improving   plant   materials   and   their  interactions  with  processing  can  decrease  the  overall  costs.  Because  different  feedstocks   do  not  perform  equally  well  with  a  given  pretreatment  or  energy  conversion  method  [13,  14],   it   is   necessary   to   know   which   feedstocks   are   suitable   for   a   conversion   process   before   a   biorefinery   location   is   decided   upon.   Otherwise   a   company   may   make   the   decision   based   solely  on  other  important  factors  such  as  transportation,  available  labor  force,  governmental   incentives,  and  available  feedstock  supply.  Later  they  may  discover  that  the  feedstock  that  is   available  does  not  generate  profitable  yields.     Unfortunately,   it   is   currently   too   time-­‐consuming   and   labor-­‐intensive   to   test   and   optimize  each  potential  feedstock  that  could  be  used  by  the  biorefinery.  It  is  more  realistic  to   4       research   the   detailed   interactions   between   the   conversion   process   and   a   handful   of   feedstocks   and   then   make   generalizations   for   similar   feedstocks.   The   most   common   way   to   classify   bioenergy   feedstocks   is   based   on   the   relative   compositions   of   lignin,   hemicellulose,   cellulose,   and   various   minor   components.   However,   while   the   relative   amounts   of   these   components   impact   the   overall   yields   and   feedstock   digestibility,   the   variability   between   feedstocks   cannot   be   entirely   explained   by   the   relative   amounts   of   each   component.   The   exact   interactions   that   are   possible   between   processing   conditions   and   cell   wall   characteristics   are   not   completely   understood   and   are   difficult   to   predict   because   of   the   complexity  and  variability  of  the  plant  cell  wall.         Figure   1.3:   Common   bioenergy   plants   and   model   laboratory   species   arranged   in   groups   according  to  botanical  classification.     5       Table   1.1:   Key   differences   in   the   cell   wall   chemistry   for   different   plant   classifications.   Information   on   hemicelluloses   is   from   [15].   For   the   lignin   subunits,   (+)   represents   relative   abundance  within  the  lignin  polymer  and  (-­‐)  represents  absence.   Class   Major   Hemicellulose   Commelinids     (Grasses  and  Relatives)   Glucurono-­‐ arabinoxylan   Non-­‐Commelinid   Monocots  and  Dicots     Glucuronoxylan   (Forbs  and  Hardwoods)   Gymnosperms     (Softwoods)   Galacto-­‐ glucomannan   Proportion  of   Are  Xylans   Lignin  Subunits   Hemicellulose   Esterified  with   S   G   H   in  2°  Cell  Wall   Ferulic  Acid?   40-­‐50%   Yes  (mostly)   +++   ++   +   20-­‐30%   No   +++   ++   +/-­‐   10-­‐30%   No   -­‐   +++   ++     Another   way   to   group   bioenergy   feedstocks   is   in   terms   of   their   botanical   classification   that  automatically  incorporates  certain  distinct,  and  more  detailed,  differences  in  the  cell  wall   structure   and   chemistry   (Figure   1.3,   Table   1.1).   Typical   bioenergy   feedstocks   will   fall   within   one   of   four   major   groups:   commelinid   monocots;   non-­‐commelinid   monocots   and   herbaceous   dicots;   hardwoods;   softwoods;   and   a   potential   fifth   group,   “woody”   commelinids   such   as   bamboo   and   palms.   Due   to   these   distinct   differences   in   cell   wall   chemistry   that   result   in   differences  in  the  organization  and  ultrastructure  of  the  cell  wall  between  each  of  the  classes,   one   would   expect   species   within   the   same   group   to   behave   more   alike   in   terms   of   interactions   with   pretreatment   chemistry   and   enzymatic   conversion   compared   to   species   from  a  different  class.     The   most   well   known   differences   in   cell   wall   chemistry   between   the   classes   are   hemicellulose   and   lignin   composition.   Hemicelluloses   are   a   diverse   class   of   amorphous   carbohydrates   that   cross-­‐link   cellulose   microfibrils   and   lignin   chains   within   the   cell   wall.   6       Different   classes   of   plants   have   distinctly   different   types   of   hemicelluloses   within   their   cell   walls.   In   commelinids   (grasses   and   related   species),   the   primary   hemicelluloses   are   heteroxylans   such   as   glucurono-­‐arabinoxylan   [16-­‐18],   which   is   cross-­‐linked   to   lignin   via   ferulate   and   diferulate   bridges   [19,   20].   The   ester   linkages   between   the   arabinose   and   ferulate  molecules  are  known  to  be  readily  cleaved  under  alkaline  conditions  [19]  and  this  is   one  proposed  mode  of  action  for  the  improved  digestibility  of  alkaline  pretreated  grasses  [21].   The   hemicelluloses   in   non-­‐commelinid   monocot   and   dicot   cell   walls,   including   all   hardwood   tree   species,   are   primarily   xylans   (4-­‐O-­‐methyl-­‐glucuronoxylans),   xyloglucans,   and   some   glucomannans   [16,   18].   In   softwoods   the   hemicelluloses   are   mainly   galactomannans   or   galactoglucomannans  [18].     A   second   difference   between   the   different   plant   classes,   are   the   ratios   of   specific   lignin  monomers  that  are  present  in  the  lignin  polymer  chains.  The  three  lignin  subunits:  p-­‐   hydroxyphenyl   (H),   guaiacyl   (G),   and   syringyl   (S),   are   synthesized   via   a   complex   enzymatic   pathway  [22]  (Figure   1.4)  and  then  transferred  into  the  cell  wall  where  they  are  oxidatively   coupled   to   form   the   complex   lignin   polymer   [23,   24].   Commelinids   have   similar   levels   of   G-­‐   and   S-­‐units   with   significant   amounts   of   H-­‐units.   Non-­‐commelinid   monocots   and   dicots   have   principally   G-­‐   and   S-­‐units   with   trace   amounts   of   H-­‐units   [23]   .   Softwoods   have   primarily   G-­‐ units   and   low   levels   of   H-­‐units   [25].   While   differences   in   the   lignin   monomer   composition   within  the  plant  cell  wall  have  been  shown  to  have  little  effect  on  enzymatic  digestibility  [26],   there   can   be   an   effect   on   pretreatment   or   other   chemical   processes   due   to   changes   in   cleavable   linkages.   The   increase   in   resistance   to   Kraft   pulping   by   COMT   down-­‐regulated   poplar   was   attributed   to   an   increase   in   proportion   of   G   units   that   resulted   in   a   decrease   in   7         Figure   1.4:     Lignin   synthesis   pathway   for   the   monolignols   p-­‐coumaryl,   sinapyl   and   coniferyl   alcohol.   From  Vanholme  et  al.  [21].  PAL  =  phenylalanine  ammonia-­‐lyase;  C4H  =  cinnamate  4-­‐ hydroxylase;   4CL   =   4-­‐coumarate:CoA   ligase;   C3H   =   p-­‐coumarate   3-­‐hydroxylase;   HCT   =   p-­‐ hydroxycinnamoyl-­‐CoA:quinate/   shikimate   p-­‐hydroxycinnamoyltransferase;   CCoAOMT   =   caffeoyl-­‐CoA   O-­‐methyltransferase;   CCR   =   cinnamoyl-­‐CoA   reductase;   F5H   =   ferulate   5-­‐ hydroxylase;  COMT  =  caffeic  acid  O-­‐methyltransferase;  CAD  =  cinnamyl  alcohol  dehydrogenase.   8       β-­‐  O-­‐4  linkages,  which  are  easily  cleavable  by  chemical  means  [27],  and  increase  in  resistant  5-­‐ 5  biphenyl  structures  [24].     Landscape Biomass availability in the landscape and the impact of biomass class on the choice of energy generation method Classification Trends in plant classes with respect to interactions with pretreatment and saccharification Species Species or variety interactions with pretreatment and saccharification Component Plant fraction (organ) interaction with pretreatment and enzymatic hydrolysis and subsequent effect on logistics Molecule Variations in composition and orientation of individual cell wall components and interactions with process parameters Figure  1.5:  Scales  of  interaction  between  plant  feedstocks  and  energy  production  processes.     The   interaction   of   the   feedstock   with   the   conversion   process   can   be   examined   at   a   number   of   different   scales   (Figure   1.5).   At   the   largest   scale,   the   landscape   scale,   the   availability  of  different  feedstocks  can  determine  the  choice  of  energy  generation  method  for   a   region   and   the   best   location   for   a   new   facility   based   on   the   economics,   logistics   and   environmental   impacts.   At   the   classification   scale,   feedstocks   from   the   different   botanical   classes  can  be  compared  and  generalizations  and  trends  related  to  their  ease  of  conversion   can   be   determined.   This   could   allow   one   to   predict   which   related   species   would   perform   well   or  poorly  with  a  given  process,  assuming  similar  environmental  factors,  location  and  maturity.   9         At   the   smaller   scales,   feedstock   interactions   with   pretreatment   parameters   and   enzyme   mixture  and  loading  can  be  examined  at  different  levels,  comparing  different  species  within   the   same   class   or   varieties   within   the   same   species   (species   scale),   different   organs   or   tissues   of   the   same   plant   (component   scale),   and   plant   materials   with   specific,   divergent   ultrastructural  or  molecular  properties  (molecular  scale).       For  this  research,  one  area  of  potential  interest  was  examined  at  each  scale  of  interaction:   • Landscape   scale:   For   this   chapter,   the   quantity   and   spatial   distribution   of   key   crop   residues  in  Mainland  China  were  determined,  as  well  as  the  amounts  of  these  residues   that   are   usable   for   energy   generation   and   their   distribution   throughout   the   country.   Improved   understanding   of   feedstock   distribution   and   availability   allows   for   a   fuller   understanding  of  optimal  locations  for  placement  of  future  biorefineries.   • Classification  scale:  For  this  chapter,  mixed-­‐species  feedstocks  comprised  of  different   ratios   of   plants   from   different   species   and   different   plant   classifications   were   compared   with   respect   to   their   interaction   with   pretreatment   and   hydrolysis   processing   conditions.   By   understanding   how   the   different   classifications   of   plants   interact  with  processing,  one  can  either  focus  on  species  that  are  amenable  to  a  given   method,  or  seek  to  alter  either  the  methods  or  feedstocks  in  order  to  better  process   those  materials  that  are  less  amenable.   • Species   scale:   For   this   chapter,   optimal   pretreatment   conditions   and   enzyme   combinations  were  determined  and  compared  for  two  varieties  of  switchgrass  grown   under  the  same  environmental  conditions.  Within  a  given  species,  there  can  be  a  great   10       deal   of   variability   in   composition   and   processing   characteristics.   In   order   to   better   understand   actual   differences   due   to   genotype,   it   is   necessary   to   minimize   the   environmental  differences  during  plant  growth  and  development.   • Component   scale:  For  this  chapter,  different  fractions  of  corn  stover  were  compared   with   respect   to   their   response   to   pretreatment   and   hydrolysis,   and   the   effect   of   different   selective   harvesting   scenarios   on   theoretical   ethanol   yields   was   examined.   Different   parts   of   the   same   plant   can   have   very   different   characteristics.   By   better   understanding  how  effectively  they  are  processed  and  the  subsequent  effect  on  yields,   we  can  devise  better  methods  for  harvesting  these  materials  for  biofuel  production.   • Molecular   scale:  For  this  chapter,  different  poplar  samples  that  had  been  genetically   modified   for   altered   lignin   content   or   composition   were   tested   and   compared   with   respect  to  their  interactions  with  pretreatment.  Certain  cell  wall  components  may  be   more   beneficial   or   more   restrictive   to   biomass   processing   than   others.   By   using   transgenic  feedstocks  which  have  been  modified  it  is  possible  to  determine  whether  1)   using  materials  that  have  been  modified  in  this  way  has  value  for  the  biorefinery  that   could   lead   to   further   development   and   planting   of   these   materials,   or   2)   whether   there   are   certain   traits   that   should   be   focused   on   more   carefully   for   future   transgenic   work  on  other  potential  feedstocks.       The  publication  status  of  each  chapter  and  other  relevant  work  by  the  author  are  listed  in   Table  1.2.   11       Table  1.2:  Publication  status  of  thesis  chapters  and  other  relevant  work.   Publication   Ref.   Status   Chapter   Title   2   Analysis  of  key  crop  residue  availability  for  bioenergy   in  Mainland  China     Not   Submitted   -­‐   Submitted   -­‐   Influence  of  variable  species  composition  on  the   3   TM saccharification  of  AFEX  pretreated  biomass  from   unmanaged  fields  in  comparison  to  corn  stover   TM 4   Optimization  of  AFEX  pretreatment  conditions  and   Published   enzyme  mixtures  to  maximize  sugar  release  from   upland  and  lowland  switchgrass   [28]   Optimizing  harvest  of  corn  stover  fractions  based  on   TM 5   overall  sugar  yields  following  AFEX and  enzymatic  hydrolysis     Published   [29]   6   AFEX  pretreatment  of  poplar  modified  for  lignin   content  and  composition     Not   Submitted   -­‐   Comparative  material  balances  around  pretreatment   technologies  for  the  conversion  of  switchgrass  to   soluble  sugars   -­‐   Published   [30]   Published   [31]    pretreatment   TM TM -­‐     AFEX  pretreatment  and  enzymatic  conversion  of   black  locust  (Robinia  pseudoacacia  L.)  to  soluble   sugars     12         CHAPTER  2 :      LANDSCAPE  SCALE:  ANALYSIS  OF  KEY  CROP  RESIDUE  AVAILABILITY  FOR  BIOENERGY  IN   MAINLAND  CHINA   2.1.  Introduction   Based  on  international  statistics,  the  top  three  world  energy  consumers  are  the  United   States,  the  European  Union,  and  Mainland  China,  each  contributing  to  roughly  one-­‐fifth  of  the   world’s   primary   energy   consumption   [32].   However,   while   the   demand   for   petroleum   has   been  steadily  decreasing  in  most  developed  countries,  in  the  developing  world  demand  has   been  increasing.  Mainland  China  is  currently  ranked  second  behind  the  United  States  in  world   petroleum  demand  [33]  however  limited  petroleum  reserves  are  forcing  increased  reliance  on   imports   [34].   Coal   is   the   most   abundant   source   of   energy   in   China,   and   coal   consumption   has   also  increased  rapidly  over  the  last  decade  [32],  however  acute  environmental  pollution  and   human   health   problems   accompany   its   use.   In   response   to   these   concerns,   the   Chinese   government  has  implemented  a  number  of  policies  to  develop  the  nation’s  renewable  energy   resources.  In  2007,  the  ‘Medium  and  Long-­‐term  Development  Program  for  Renewable  Energy’   set   two   goals   within   each   class   of   renewable   energy:   biomass,   wind,   solar,   hydro   and   geothermal,  with  the  intention  to  reach  the  first  goal  by  2010  and  the  second  by  2020  [34].     In  accordance  with  these  goals,  China  has  begun  developing  liquid  renewable  fuels.  By   2005,  production  capacity  of  grain-­‐based  bioethanol  reached  1  million  tons  per  year,  with  the   intent  of  reaching  2.0  million  tons  per  year  by  2010,  increasing  to  10.0  million  tons  per  year   by   2020   [34].   However,   because   of   issues   with   food   security,   it   is   necessary   to   meet   these   goals   using   feedstocks   that   do   not   compete   with   the   production   of   food,   either   through   demand   for   the   same   feedstocks   or   competition   for   arable   land.   While   there   are   options   13       being   pursued   in   China   using   starch   or   sugar-­‐based   ethanol   production,   of   these   options   only   cassava  doesn’t  compete  with  food  crops  because  it  is  not  a  staple  food  and  can  be  grown  on   marginal   land   [35].   This   leaves   bioethanol   produced   from   lignocellulosic   materials   as   the   best   option   for   meeting   China’s   goals   for   renewable   liquid   fuel.   Two   of   the   best   options   for   lignocellulosic   ethanol   feedstocks   are   crop   straw   residues   left   following   grain   harvest   and   dedicated   energy   crops   or   mixed-­‐species   feedstocks   that   can   be   grown   on   abandoned   or   degraded  lands.  Because  feedstock  supply  to  the  biorefinery  is  one  of  the  largest  issues  facing   commercialization  of  cellulosic  ethanol  in  China  [36],  the  location  of  the  biorefinery  will  be  a   critical  decision.  It  is  necessary  to  locate  a  biorefinery  where  there  is  a  stable  and  accessible   supply  of  feedstock,  especially  for  straws  that  are  not  dense  and  can  have  high  storage  and   transportation   costs   [36,   37].   As   an   alternative   to   liquid   fuel   production,   straw   can   also   be   burned   to   provide   electricity,   displacing   coal   and   some   of   its   negative   impacts   [38].   However,   just  as  for  the  biorefinery,  feedstock  supply  is  also  an  important  issue  for  a  power  plant.   Most  of  the  research  on  crop  residue  production  in  China  has  been  at  the  national  or   provincial  scale,  as  that  data  is  more  readily  available  and  easier  to  analyze.  Our  goal  was  to   determine   the   production   of   crop   residues   at   the   next   administrative   level   down,   the   prefecture   level,   in   an   attempt   to   improve   the   spatial   resolution   and   better   inform   on   potential  locations  for  biofuel  production  in  Mainland  China.  We  also  evaluated  the  effect  of   residue-­‐to-­‐grain   ratios   on   the   spatial   distribution   of   crop   residues.   In   order   to   estimate   the   distribution   of   usable   crop   residues   in   Mainland   China,   we   estimated   the   amount   of   crop   residues  in  each  province  used  for  pulp  and  paper  production  and  animal  feed,  returned  to   the  field  as  fertilizer,  and  burned  on-­‐field.     14       Figure  2.1:  Map  of  Mainland  China  with  provinces,  autonomous  regions,  and  municipalities.       2.2.  Materials  and  methods   2.2.1.  Land  area  and  population  data   A   labeled   map   of   China   is   provided   for   ease   of   reference   (Figure   2.1).   Data   on   population   and   total   and   specific   crop   sown   area   for   each   city   prefecture,   autonomous   prefecture,   was   determined   from   the   individual   province,   autonomous   region,   and   municipality  yearbooks  [39-­‐69].  Data  on  population,  total  sown  land  area,  and  total  cultivated   land  area  for  the  provinces,  autonomous  regions,  and  municipalities  were  obtained  from  the   China   Statistical   Yearbook   [70].   For   simplicity,   from   this   point   forward   city   prefectures   and   15       autonomous  prefectures  will  both  be  referred  to  as  prefectures,  and  autonomous  regions  and   municipalities  will  also  be  referred  to  as  provinces.  Prefecture  total  land  area  and  cultivated   land   area   data   (where   available)   were   obtained   from   the   China   Statistical   Yearbook   for   Regional   Economy,   and   the   provincial   values   were   calculated   by   adding   the   prefecture   values   [71].  Prefecture  values  for  the  2006  Fujian  total  land  area;  and  the  Liaoning,  Shanghai,  Shanxi,   Tianjin,  and  Yunnan  city  prefecture  cultivated  land  areas  were  obtained  from  the  China  Data   Online   database   [72].   Data   for   Fujian   cultivated   area   and   Shenyang   prefecture   in   Liaoning,   were  unavailable  for  2006,  so  2005  data  were  used  for  these  values  [72].     2.2.2.  Crop  production  data   Due  to  limitations  on  available  prefecture-­‐level  data  for  cereal  crop  production  in  all  of   the  provinces,  the  most  recent  year  that  could  be  modeled  was  2006.  Crop  yield  data  were   collected  for  each  province  and  prefecture  from  a  variety  of  sources  for  the  following  crops:   all   cereals   including   rice,   wheat,   corn,   millet,   sorghum,   and   miscellaneous   cereals;   legumes   and   soybeans;   tubers   and   potatoes;   cotton;   oilseeds   including   canola,   peanut,   sesame,   sunflower,   flaxseed,   and   miscellaneous   oilseeds   [39-­‐69,   71].   According   to   the   Chinese   classification   system,   legumes,   tubers,   and   cereals   are   all   considered   grain   crops   and   any   reference   in   this   paper   will   include   these   categories.   Because   in   many   cases   there   was   not   enough   prefecture   level   data   available   to   estimate   the   yield   of   the   sugar   crops   (sugarcane   and  sugar  beets)  and  tobacco,  it  was  decided  to  not  include  them  in  this  analysis.    In  some  cases  data  on  the  miscellaneous  cereals  or  miscellaneous  oilseeds  were  not   provided  for  a  given  prefecture.  This  value  was  then  determined  by  taking  the  difference  of   16       the  total  value  minus  the  provided  subcategories.  When  the  value  of  the  total  cereals  was  not   given,   the   value   of   the   miscellaneous   cereals   was   determined   by   subtracting   all   known   cereals,   legumes   and   tubers   from   the   total   grain   value.   In   cases   where   it   was   necessary   to   estimate   the   yields   of   important   crops   based   on   limitations   in   available   data,   the   provincial   crop  yield  per  hectare  values  [73],  sown  area,  and  yield  values  for  each  unknown  crop  type   were  used  to  derive  the  unknown  values  based  on  the  total  grain  yield  and  sown  area  or  total   oilseed   yield   and   sown   area   for   each   prefecture   within   the   province.   Yields   were   simultaneously   calculated   for   all   unknowns   within   the   category   (grains   or   oilseeds)   by   solving   to   minimize   the   differences   between   the   sown   areas   and   total   yields   for   all   crop   residues   being  estimated  and  the  actual  values  for  the  both  the  province  and  the  prefectures.  In  some   cases   where   the   yields   were   not   provided   and   they   could   not   be   estimated,   those   crops   with   yields  of  less  than  100,000  tons  for  a  given  province  were  neglected  from  the  analysis.  Of  the   crops   analyzed   that   were   not   provided   and   had   yields   >   100,000   tons   for   the   entire   province,   only  prefecture  values  for  corn  and  wheat  in  Zhejiang  and  rice  in  Inner  Mongolia  could  not  be   estimated.  These  were  reported  at  the  prefecture  scale  as  part  of  the  miscellaneous  cereals.       2.2.3.  Crop  residue  yield  calculations  and  mapping   Crop   residue   yields   were   calculated   using   residue-­‐to-­‐grain   ratios   reported   in   literature   [74,   75].   For   miscellaneous   cereal   and   miscellaneous   oilseed   yields   that   did   not   have   a   residue-­‐to-­‐grain   ratio,   we   used   what   seemed   to   be   common   values   among   those   reported   (Table   2.1),   namely,   1.5   for   cereals   and   2.0   for   oilseeds.   Because   there   is   a   large   range   in   values   for   reported   residue-­‐to-­‐grain   ratios   (Table   2.1)   we   also   compared   our   results   to   the   17       Table   2.1:   Residue-­‐to-­‐grain   ratios   of   common   crops   from   various   sources.   All   data   are   specific   to   Mainland   China   except   those   from   Kim   and   Dale   [76],   which   are   based   on   U.S.   values  and  are  provided  for  reference.   Residue-­‐  to-­‐Grain  Ratios  of  Various  Crops  Provided  in  Literature   Source   [77]   [78]   [79]   Year   1990   1995   1999   2007   Rice   Wheat   Corn   Millet   Sorghum   Misc.  Cereals   Legumes   Soybeans   1.32   1.72   1.37   1.61   1.59       1.30   0.623   1.366   2.0       1.0     1.5   0.97   0.68   1.03   0.73   1.37   1.25   1.51     1.44     1.60             Tubers     Potatoes   Cotton   Oilseeds   Canola   Peanut   Sesame   Sunflower   Flaxseed     1.62     2.99   1.35   5.88       Sugarcane   b   Sugar  beet   Fiber  crops   Tobacco   Cited  by:     1.81   1.06   a   0.5   0.61     3.0   2.0             [81,  82]   [36]   [83]   [76]   -­‐   -­‐   -­‐   1   1   2       1.5   1.5     1.32   1.72   1.27   1.61   1.59     1.30     1.4   1.3   1.0     1.3   1.3       2006-­‐ Used   Avg.   2010   1.00   1.0   1.0   1.17   1.1   1.4   1.04   2.0   1.6     1.6   1.5     2.0   1.6     1.6   1.5     1.6   1.5   1.50   1.6   1.5   c   0.5   0.5   1         2.91     2.87   1.14   2.01       0.5   3.0   2.0   3.0   1.5   3.0   3.0   2.0   0.5   2.5   2.0   2.3   1.4   2.6   2.0   2.0     3   2   2   2   2       0.40     1.35   2.94   5.88                             0.1   0.2   0.1   0.8         0.71   0.1   1.7   1.0   0.2   1.9   1.1   0.1       0.45             0.55     3.00   5.51       3.00   1.01   1.52     0.64     0.6         0.1   0.25   a [80]   [74,  75]             1.7           [84-­‐88]     [89]               Gao  et  al.  reported  a  large  number  of  harvest  indices  for  various  crops  and  then  chose  the   most  cited  value  to  perform  their  analysis  of  crop  residue  production.  We  also  report  the   average  of  the  values  they  cited.   b Sugarcane  residue  only  consists  of  the  leaf  and  doesn’t  take  into  account  any  bagasse   remaining  following  processing.   c Average  value  of  those  reported  for  potatoes  and  sweet  potatoes.   18         most   commonly   cited   dataset   from   a   joint   Chinese   Ministry   of   Agriculture   and   U.S.   Department   of   Energy   project   [78].   The   total   residue   yields   for   each   prefecture   or   province   equaled:         n R = ∑ ri ⋅ Ci   i=1 (2.1)   where  R  =  the  total  residue  yield  (tons),  ri  =  the  residue  ratio  for  the  i-­‐th  crop,  Ci  =  the  crop   yield  for  the  i-­‐th  crop  (tons),  and  n  =  the  total  number  of  crops.  Province  crop  residue  yields   were   calculated   in   the   same   way   based   on   provincial   crop   production   data   rather   than   summing  the  prefecture-­‐level  data  for  each  province.  Residue  densities  were  calculated  with   respect  to  cultivated  land  area  as  opposed  to  crop  sown  area,  as  sown  area  is  counted  twice   when  double  crops  are  cultivated,  once  for  each  crop.  By  calculating  based  on  sown  area,  a   deceptively  low  crop  residue  density  (kg/ha)  would  be  generated  for  a  given  location  if  there   is  a  large  amount  of  double-­‐  or  triple-­‐cropping.  Crop  density  in  terms  of  total  prefecture  or   province  area  is  also  presented  for  comparison.   Maps   were   generated   using   DIVA-­‐GIS   7.4   mapping   software   and   shapefiles   of   the   prefecture-­‐level  and  province-­‐level  administrative  regions  in  China  [90].       2.2.4.  Straw  used  for  pulp  and  paper  production   The  amount  of  straw  used  for  pulp  and  paper  production  in  each  province  in  2006  was   calculated  using  data  from  the  China  Paper  Association  Paper  Industry  Reports  [91-­‐93].  Total   paper   and   paperboard   production   were   reported   in   the   industry   reports   for   provinces   with   greater   than   1.0   million   tons   of   production   (Hebei,   Shandong,   Jiangsu,   Zhejiang,   Fujian,   19       Guangdong,   Henan,   Hubei,   Hunan,   Anhui,   Guangxi,   Jiangxi,   and   Sichuan).     The   production   values  for  the  other  provinces  were  estimated  from  a  figure  in  the  2005  report  and  assumed   to   be   roughly   equal   to   the   production   in   2006   (Table   A.1).   Assumptions   are   as   follows:   1)   provinces  were  assigned  to  regions  (east,  central,  and  west)  and  the  sum  of  provincial  values   in  a  given  region  were  required  to  equal  the  total  for  the  region,  as  provided  in  literature;  2)   there   was   no   production   of   paperboard   in   Hainan,   Qinghai,   and   Tibet;   3)   the   amount   produced  in  Shanxi,  Inner  Mongolia,  Yunnan,  Xinjiang,  Chongqing,  and  Tianjin  were  equal;  4)   the  amount  produced  in  Shaanxi  and  Ningxia  were  equal;  6)  the  amount  produced  in  Liaoning   and  Heilongjiang  were  equal  and  slightly  greater  than  that  produced  in  Shaanxi  and  Ningxia;  6)   values   for   the   other   provinces   were   chosen   in   order   to   satisfy   the   initial   assumption   and   based  on  their  relationship  to  production  in  other  provinces  .  An  additional  assumption  was   that  the  production  of  pulp  and  paper  was  performed  in  the  province  from  which  the  straw   was  harvested.  For  pulp  production  this  is  a  reasonable  assumption  as  straw  is  not  dense  and   therefore   costly   to   transport   long   distances.   However,   this   may   be   an   oversimplification   for   paper  production  as  pulp  could  feasibly  be  transported  across  province  lines.   Total   pulp   consumption   in   2006   was   59.92   million   tons,   and   1.085   tons   of   paper   were   produced  for  every  ton  of  pulp  consumed.  Of  this,  12.9  million  tons  of  pulp  were  produced   from   non-­‐wood   sources,   equal   to   21.5%   of   the   total   pulp   consumed.   The   amount   of   straw-­‐ based   pulp   (rice   and   wheat)   was   estimated   as   60%   of   the   non-­‐wood   pulp   based   on   the   decreasing  trend  in  this  proportion  from  1995  (81%)  to  2000  (69%)  [94].  It  is  estimated  that   2.25  to  2.5  tons  of  crop  straw  are  needed  to  produce  one  ton  of  pulp  [84,  94].  We  assume  a   conservative  estimate  of  2.5  tons  of  straw  per  ton  of  pulp.  It  was  assumed  that  these  ratios   20       were  constant  for  every  province.  These  values  were  then  used  to  calculate  the  tons  of  crop   straw   consumed   for   paper   and   paperboard   production   in   each   province   using   the   following   equation:   Si = Pi ⋅ p ⋅ n ⋅ s ⋅ c   (2.2)   where  Si  is  the  total  straw  consumed  for  pulp  and  paper  in  the  ith  province,  Pi  is  the  total  pulp   and   paperboard   production   for   the   ith   province,   p   is   the   tons   of   pulp   consumed   per   ton   of   paper  produced  (0.922),  n  is  the  proportion  of  pulp  from  non-­‐wood  sources  (0.215),  s  is  the   proportion  of  non-­‐wood  pulp  from  rice  and  wheat  straw  (0.6),  and  c  is  the  tons  of  straw  input   per  ton  of  pulp  output  (2.5).     2.2.5.  Straw  used  for  animal  feed   A  previous  estimate  on  the  amount  of  straw  needed  for  animal  feed  production  took   into  account  only  the  number  of  cattle  in  each  province  and  also  assumed  that  provinces  with   a   large   area   of   pastureland   would   not   require   the   use   of   crop   straw   as   an   animal   feed   in   support   of   cattle   production   [78].   There   are   two   issues   with   these   assumptions.   First,   as   all   ruminants   can   consume   straw,   sheep   and   goats   should   also   be   taken   into   account   in   an   estimate  of  the  amount  of  crop  straw  needed  for  animal  feed.  Second,  many  of  the  provinces   with   significant   amounts   of   range   and   pastureland   also   have   issues   with   overgrazing,   which   has   led   to   loss   of   vegetation,   erosion,   and   dust   storms.   So   it   should   also   not   be   automatically   assumed   that   rangeland   in   these   provinces   can   or   should   support   the   amount   of   ruminant   animals   that   are   currently   being   produced.   For   our   estimates   we   calculated   the   number   of   ruminant   animals   (cattle,   buffalo,   sheep,   and   goats)   that   could   be   sustainably   produced   on   21       rangeland   based   on   the   carrying   capacity.   Any   animals   unable   to   be   supported   by   pasture   were  assumed  to  be  fed  a  daily  recommended  amount  of  crop  straw.   Data  for  the  number  of  ruminant  animals  and  the  area  of  grazing  and  pastureland  for   each  province  in  2006  are  from  the  China  Statistical  Yearbook  [70]  (Table   A.2).  Only  provinces   with   over   1   million   ha   of   pastureland   were   taken   into   consideration   for   animal   grazing.   Dryland  carrying  capacity  (sheep  per  ha)  for  Xinjiang,  Inner  Mongolia,  Ningxia,  Gansu,  Qinghai,   and  Shaanxi  were  taken  from  Shen  [95].  The  carrying  capacities  for  Jilin  and  Heilongjiang  were   assumed   to   be   equal   to   Inner   Mongolia.   The   average   of   the   carrying   capacities   of   Xinjiang,   Ningxia,  Gansu,  and  Qinghai  was  used  for  Tibet,  Sichuan,  and  Guizhou.  We  estimated  the  total   number   of   ruminants   in   terms   of   “animal   units”,   where   a   cow   or   buffalo   equals   1.0   animal   unit   and   a   sheep   or   goat   equals   0.2   animal   units.   The   carrying   capacity   for   each   province   was   converted  to  animal  units  by  dividing  by  5  (Table  A.2).    The  number  of  ruminants  that  could   be   supported   by   the   pastureland   was   calculated   by   multiplying   the   carrying   capacity   by   the   area   of   pastureland.   This   value   was   subtracted   from   the   total   number   of   ruminants   to   give   the  number  of  ruminants  fed  on  crop  straw  in  each  province  (Table  A.2).   There   is   a   maximum   amount   of   straw   that   should   be   fed   to   ruminant   animals   to   maintain  their  health.  Li  et  al.  report  a  daily  recommended  amount  of  3.49  kg  of  ammoniated   straw   per   cow   per   day   [78],   which   equals   1.27   tons   of   straw   consumed   per   cow   (or   animal   unit)  per  year.  Multiplying  this  value  by  the  number  of  ruminant  animals  that  are  not  pasture-­‐ fed  gives  an  estimate  of  the  amount  of  crop  straw  that  would  be  needed  in  each  province  to   support  them,  assuming  that  the  number  of  ruminants  in  each  province  throughout  the  year   is   roughly   equal   to   the   number   of   ruminants   at   year   end   as   reported   in   literature,   that   22       logistics  will  allow  continuous  feeding  of  crop  straw  throughout  the  year,  and  that  crop  straws   would  be  preferentially  fed  to  animals  over  feed  and  grain,  up  to  the  maximum  amount.     2.2.6.  Straw  returned  to  the  field   For   the   estimate   of   crop   residues   that   should   to   be   returned   to   the   field   in   each   province,   we   used   values   reported   by   Cui   et   al.   for   different   regions   in   China:   2.25   t/ha   for   northeast  China  (Heilongjiang,  Jilin,  and  Liaoning)  and  the  Qinghai-­‐Tibet  region  (Qinghai  and   Tibet);   and   3.0   t/ha   for   all   other   regions   [80].   However   estimating   the   amount   that   should   be   left  behind  becomes  complicated  due  to  the  large  amount  of  double-­‐  and  triple-­‐cropping  that   occurs,  particularly  in  central  and  southern  China.  By  some  estimates,  multi-­‐cropping  systems   account   for   over   1/3   of   China’s   total   cropland   [96].     Because   sown   area   counts   each   crop   separately  though  sown  on  the  same  land  area,  if  the  value  of  sown  land  area  was  used  to   estimate   the   amount   of   straw   that   should   be   returned   per   hectare,   this   would   result   in   double-­‐  or  triple-­‐counting  multi-­‐cropped  land,  resulting  in  excessive  return  of  straw  to  fields   that  are  multi-­‐cropped.  So  for  our  study  we  attempted  to  account  for  multi-­‐cropped  land  for   the  crops  of  interest.  First,  the  provincial  cultivated  land  area  for  the  crops  of  interest  (cereals,   legumes,   tubers,   oilseeds   and   cotton)   was   calculated   by   subtracting   the   sown   area   of   fiber   crops,   sugar   crops,   and   tobacco,   which   were   all   assumed   to   be   all   single-­‐cropping   systems,   from  the  total  province  cultivated  land  area.  Because  vegetables  are  commonly  multi-­‐cropped   with   the   crops   of   interest   [96],   the   vegetable   sown   area   could   not   directly   subtracted.   As   a   result  the  proportion  of  vegetable  sown  land  area  that  was  solely  planted  with  vegetables  for   each  province  (Table  A.3)  was  estimated  based  on  the  regions  defined  by  Qiu  et  al.  [96].     23       Using   our   assumed   proportions,   we   estimated   that   31.3%   of   the   national   vegetable   sown   area   was   single-­‐   or   triple-­‐cropped,   which   is   similar   to   31.1%   that   was   estimated   by   another   study   [96].   Assuming   that   the   remaining   vegetable   sown   area   overlapped   with   our   crops   of   interest,   the   estimate   of   single-­‐   and   triple-­‐cropped   vegetable   sown   area   was   subtracted   from   the   total   province   cultivated   land   area,   in   addition   to   the   other   crop   types   mentioned  previously.  This  resulted  in  an  estimate  of  the  cultivated  land  area  for  our  crops  of   interest  for  each  province.  For  provinces  where  the  reported  sown  land  area  of  the  crops  of   interest   was   less   than   the   calculated   cultivated   land   area,   the   excess   cultivated   land   was   assumed  to  be  either  in  fallow  rotation  or  a  cropping  system  that  was  not  accounted  for.  In   this   case,   the   sown   area   of   the   crops   of   interest   (as   reported   in   literature)   was   used   to   calculate  the  amount  of  crop  straw  that  should  be  left  on  the  field.  For  provinces  where  the   sown  area  was  greater  than  the  calculated  cultivated  land  area  for  the  crops  of  interest,  the   difference  was  assumed  to  be  due  to  multi-­‐cropping  and  the  cultivated  land  area  was  used  to   determine  the  amount  of  crop  straw  that  should  be  left  on  the  field.  The  estimated  cultivated   land  area  for  the  crops  of  interest  are  listed  in  Table  A.4.  Sown  area  of  individual  farm  crops   for  each  province  and  total  cultivated  land  area  for  each  province  in  2006  were  taken  from   the  China  Statistical  Yearbook  [70].         2.2.7.  Provincial  distribution  of  usable  crop  residues   The   amount   of   residues   able   to   be   used   for   energy   in   each   province   was   calculated   by   subtracting   the   amount   of   residues   required   for   pulp   and   paper   production,   animal   feed,   and   return  to  the  field,  from  the  total  amount  of  residues  produced.  It  is  difficult  to  estimate  the   24       amount   of   straw   used   for   rural   energy   in   Mainland   China.   While   the   government   reports   statistics   for   the   amount   of   rural   energy   provided   by   straw   combustion   for   each   province,   these  values  are  generally  believed  to  overestimate  the  use  of  crop  residues  as  a  fuel  source   [97].    Many  of  the  other  estimates  on  the  proportion  of  crop  residues  used  for  rural  energy   are   based   on   values   from   2000   [97-­‐99].   We   chose   not   to   use   these   values   to   estimate   the   amount   of   biomass   used   for   rural   energy   in   2006,   as   the   proportion   of   biomass   used   for   rural   fuel   in   each   province   has   likely   decreased   with   the   increased   availability   of   coal   and   commercial   electricity.   Using   these   values   would   overestimate   the   amount   of   biomass   needed   for   rural   energy.   As   a   result   our   value   for   usable   residues   includes   the   amount   currently   being   used   for   rural   energy.   Alternatively,   one   could   use   estimates   of   on-­‐field   combustion  of  crop  residues  as  a  measure  of  the  amount  of  usable  residues  in  each  province.   We  compare  values  from  Wang  and  Zhang  [100]  to  a  scenario  where  sufficient  crop  residues   are  left  on  the  field  to  maintain  soil  health  and  where  land  is  not  overgrazed  (Table  A.5).       2.3.  Results  and  discussion   2.3.1.  Total  crop  residue  production   Various   studies   have   estimated   crop   residue   production   within   Mainland   China   (Table   2.2).  These  amounts  range  from  the  most  conservative  estimate,  433.0  million  metric  tons  in   2006,  to  the  most  liberal  estimate,  939.3  million  metric  tons  in  1999.  Our  study  value  for  the   total  crop  residue  production  in  China  is  at  the  lower  end  of  the  range,  with  an  estimate  of   593.5   million   dry   metric   tons   produced   in   2006.   The   reason   for   the   differences   in   crop   estimates  is  at  least  partly  due  to  the  types  of  crop  residues  that  were  included  in  the  analysis 25       Table   2.2:   Estimates   on   total   crop   residue   yields   and   usable   amounts   of   crop   residues   in   Mainland   China   from   various   sources.   The   usable   residue   values   for   [89,   101]   only   take   into   account   the   amount  of  collectable  residues  and  not  competing  uses.   Total  Crop   Residues     Usable   Residues     (10  Mg)   (10  Mg)   433.0   533.0   557.5   593.5   604.0   620.3   627.3   636.2   754.7   774.0   788.6   841.8   939.3   175.9   452.8   -­‐   124.7   254.1   -­‐   -­‐   -­‐   114.7   -­‐   -­‐   686.0   551.4   6 Year  of   Estimate   Reference   2006   2007   1995   2006   1995   2002   1994-­‐2004   2006   2007   2008   -­‐   2005   1999   6 [80]   [89]   [86]   This  work   [78]   [87]   [84]   [86]   [36]   [88]   [102]   [101]   [85]       Estimates  with  the  lowest  numbers  either  examined  a  more  limited  range  of  feedstocks  [78,   80],  or  they  tended  to  use  more  conservative  estimates  for  the  residue-­‐to-­‐grain  ratios  [80,  89].   Our  value  is  low  for  both  of  these  reasons.     Our  estimate  of  total  usable  residues  is  also  at  the  lower  end  of  those  reported  with   only   124.7   million   dry   metric   tons   usable   for   fuel.   One   reason   for   this   is   that   a   number   of   the   other   estimates   only   take   into   consideration   the   amount   of   residues   that   are   able   to   be   harvested  based  on  a  collection  coefficient  and  do  not  consider  competing  uses  [89,  101].  Of   the  other  five  studies  that  report  a  usable  amount  of  residues,  only  the  study  by  Liao  et  al.  [85]   26       is   significantly   higher   than   the   rest.   They   assumed   that   the   proportion   of   usable   residues   was   the   same   as   was   determined   in   the   MOA/DOE   joint   report   [78],   however   their   estimate   of   crop  residue  production  is  so  large  as  to  make  their  estimate  of  usable  residues  seem  unlikely.   When   we   recalculated   the   total   yield   of   crop   residues   based   on   their   data,   we   find   that   there   is  an  error  in  their  calculations.  When  calculated  directly  from  their  reported  data,  the  total   amount   of   crop   residues   is   584.1   million   metric   tons,   over   350   million   tons   less   than   their   reported  value  of  939.3  million  metric  tons  [85].  Their  other  calculations  are  equally  suspect   based   on   the   inability   to   match   their   reported   values   (such   as   the   amount   of   unused   residue)   with  our  calculations  based  on  their  numbers.   The   distribution   of   total   yields   across   Mainland   China   is   shown   in   Figure   2.2.   On   a   mass  basis,  most  of  the  crop  residues  are  localized  in  the  Huang-­‐Huai-­‐Hai  plain  and  Yangtze   River  region  (Henan,  Shandong,  Anhui,  Jiangsu),  and  in  Heilongjiang  in  the  Northeast.  This  is  a   similar   finding   to   what   has   been   reported   previously   [78,   85,   86].   When   examined   at   the   prefecture   scale,   it   is   easier   to   see   patches   of   high   crop   residue   production   such   as   in   Chongqing,  southern  Henan  and  Hebei,  northern  Anhui  and  Jiangsu,  and  western  Heilongjiang   and   Jilin.   The   provincial   scale   also   doesn’t   tell   the   whole   story,   particularly   for   the   large   provinces.   For   example,   Inner   Mongolia   in   northern   China   has   patches   of   high   crop   residue   production,  particularly  in  the  east  where  it  connects  with  Heilongjiang  and  Jilin.  However,  on   the  provincial  scale  its  overall  yield  is  fairly  low.    Crop   residue   density,   when   expressed   in   terms   of   the   total   prefecture   or   province   land  area,  is  unsurprisingly  biased  toward  locations  with  a  high  crop  residue  yield  and  smaller   total   land   area,   such   as   Shandong,   Henan,   Anhui,   and   Jiangsu   (Figure   2.2   -­‐   C,   D).   Reporting   27         Figure   2.2:   Estimated   crop   residue   yields   in   China   in   2006:   (A,   B)   Total   yields   for   each   6 prefecture  and  province  (10  Mg);  (C,  D)  Residue  yield  per  total  land  area  (Mg/ha);  (E,  F)  Per   capita   crop   residue   yield   (kg/person).   Residues   included   cereals,   legumes,   tubers,   cotton,   and   oilseed   crops,   and   the   amounts   were   estimated   using   the   average   harvest   indices   for   the   respective  crops  reported  by  Xie  and  colleagues  [74,  75].   28       crop   residue   production   in   terms   of   the   total   land   area   probably   doesn’t   have   much   practical   value   as   the   locations   of   crop   production   are   likely   to   be   more   localized   and   regions   with   significant   amounts   of   forest,   mountains,   and   deserts   will   be   negatively   impacted   in   the   assessments.   Expressing   crop   residue   production   per   capita   (Figure   2.2   -­‐   E,   F)   tends   to   bias   towards   locations   that   have   high   crop   production   and   a   small   population,   such   as Heilongjiang,   Jilin,   Inner   Mongolia,   and   Xinjiang   [80,   89];   locations   with   more   large-­‐scale   mechanization  [70].    Per  capita  yield  of  crop  residues  is  important  if  it  is  desirable  for  reasons   of  transportation  costs  and  logistics  to  keep  all  stages  of  bioenergy  production  through  end   use   at   a   local   scale.   This   would   probably   have   greater   impact   when   crop   residues   are   used   as   a   feedstock   for   individual   or   community   biodigesters   or   home   heating   and   cooking   units   as   opposed  to  a  power  plant  or  a  biorefinery  which  would  require  significantly  larger  amounts  of   crop  residues.       2.3.2.  Distribution  of  crops  within  Mainland  China   The   different   crops   have   decidedly   different   crop   residue   distributions   within   Mainland  China.  In  terms  of  total  mass  yields,  wheat  production  is  focused  in  the  Huang-­‐Huai-­‐   Hai   region   (Shandong,   Hebei,   Henan),   with   some   production   in   Xinjiang   (Figure   2.3   A).   The   largest  amount  of  rice  production  is  located  in  the  Southern  region,  largely  in  Hunan,  Jiangxi,   and   Anhui;   however,   there   is   also   some   production   in   northeast   China   (Figure   2.3   B).   Corn   production   is   largely   in   northern   and   northeast   China   (Figure   2.3   C).   Cotton   production   is   almost  entirely  located  in  Xinjiang,  with  some  smaller  scale  cultivation  in  the  eastern  region   (Figure   2.3   D).   Oilseed   production,   primarily   canola,   is   located   in   a   broad   band,   stretching   29         6 Figure  2.3:  Spatial  distribution  of  total  residue  yields  (10  Mg)  for  different  types  of  crops  in   2006:  (A)  wheat,  (B)  rice,  (C)  corn,  (D)  cotton,  (E)  oilseeds,  and  (F)  legumes  and  tubers.       30         Figure  2.4:  Amount  and  proportion  of  the  different  types  of  crop  residues  for  each  province   in  Mainland  China.     across   central   China   (Figure   2.3   E),   and   legumes   and   tubers   are   primarily   grown   in   the   far   north  in  Heilongjiang  and  Inner  Mongolia  (Figure   2.3   F).  The  residue  density  maps  (Mg/ha)  for   each  crop  type  are  provided  in  the  supplemental  information  (Figure  A.1).     The  relative  proportions  of  the  different  crop  types  grown  in  each  province  are  shown   in   Figure   2.4.   The   regions   with   the   largest   production   of   crop   residues   are   unsurprisingly   areas   with   significant   amounts   of   double-­‐   and   triple-­‐cropping   [96].   Most   of   these   provinces   31       produce  large  amounts  of  either  wheat  straw  and  corn  stover  (Hebei,  Shandong,  and  Henan)   or   wheat   and   rice   straw   (Jiangsu,   Anhui,   Hubei,   and   Hunan).   Heilongjiang   and   Jilin   in   the   northeast   also   have   large   amounts   of   residues   from   corn,   rice,   legumes,   and   tubers,   and   Sichuan   in   the   west   has   a   large   amount   of   residues   that   evenly   distributed   among   the   different  categories.     Within  China  the  variation  in  the  distribution  of  different  crop  residues  could  have  an   impact   on   certain   bioenergy   production   scenarios.   The   type   of   crop   straw   can   be   a   very   important  consideration  as  different  straws  can  have  very  different  properties.  For  example,   rice  straw  has  a  very  high  ash  content,  particularly  silica,  and  practical  issues  that  would  need   to   be   considered   if   using   this   feedstock   include   rapid   equipment   wear   due   to   abrasion,   significant  fouling  of  combustion  boilers  unless  the  minerals  are  removed  by  leaching  [103],   and  ultimately  some  form  of  waste  disposal  or  end  use  [104].  In  order  to  avoid  these  issues,  it   may  be  desirable  to  construct  a  facility  in  a  location  that  is  dominated  by  a  less  problematic   feedstock.   To   give   another   example,   in   Chapter   3,   it   was   found   that   herbaceous   dicots   and   monocots  respond  differently  to  AFEX TM  pretreatment.  Cotton,  oilseed  straws,  legumes  and   TM tubers,   as   they   are   all   dicot   species   would   all   likely   respond   less   amenably   to   AFEX pretreatment   compared   to   the   cereal   crops,   which   are   monocots.   If   placing   an   AFEX TM     pretreatment  technology,  it  would  be  more  desirable  to  focus  on  the  major  cereal  producing   regions  such  as  the  rice-­‐producing  regions  of  Hunan,  Jiangxi,  Hubei  and  Anhui;  or  the  wheat   and  corn  producing  provinces  of  southern  Hebei,  Shandong,  and  Henan.  Chongqing  could  be  a   good  place  to  locate  a  biorefinery  as  they  have  a  fairly  high  population  [60],  all  petroleum  and   32       diesel   fuel   is   imported   to   the   municipality   [105],   and   there   is   a   high   mass   yield   of   all   three   cereal  residues  (Figure  2.3  –  A,  B,  C).  However,  the  residue  density  is  low  (Figure  A.1)  which   may   indicate   that   feedstock   collection   and   transportation   could   be   an   issue.   While   it   may   not   be  an  appropriate  location  for  a  large  biorefinery,  it  might  be  better  suited  for  a  distributed   biomass  processing  system  that  is  organized  at  a  more  local  scale  [10].     2.3.3.  Influence  of  residue-­‐to-­‐grain  ratio  on  spatial  distribution   As   very   little   data   is   presently   available   on   a   large   scale   for   the   generation   of   crop   residues,   values   are   typically   estimated   using   from   crop   yields   using   a   factor,   either   the   harvest  index  (HI)  which  is  the  ratio  of  the  yield  of  the  economic  product  (seed,  tuber,  etc…)   to  the  total  plant  weight,  or  the  residue-­‐to-­‐grain  ratio,  which  is  the  yield  of  the  crop  residue   to  the  yield  of  economic  product.  By  multiplying  the  economic  crop  yield  by  a  residue-­‐to-­‐grain   ratio  it  is  possible  to  estimate  the  residue  yield.  The  values  for  the  residue-­‐to-­‐grain  ratio  used   by   various   sources   for   estimating   crop   residue   amounts   in   China   varies   significantly   (Table   2.1),   however   many   papers   base   their   values   on   those   published   in   a   China   Ministry   of   Agriculture   (MOA)   and   U.S.   Department   of   Energy   (DOE)   joint   report   from   1998   [78].   The   main   issue   with   using   these   numbers   is   that   there   is   no   indication   as   to   where   they   were   derived   from,   whether   from   literature   or   experiments.   Additionally   it   is   possible   that   some   of   the  values  given  in  this  document  are  erroneous,  particularly  the  value  for  rice  which  is  lower   and   corn   which   is   higher   than   values   reported   elsewhere.   In   the   15   years   since   the   original   document   was   published,   advancements   in   crop   production,   either   through   improved   varieties  or  improved  production  practices  have  improved  wheat  and  corn  yields  per  hectare   33       [70].  As  crop  yields  increase  per  hectare,  generally  the  harvest  index,  decreases,  with  the  limit   being  around  0.4  -­‐  0.5  [106],  which  would  be  a  residue-­‐to-­‐grain  ratio  of  around  0.6  –  0.8.  This   is   most   likely   true   of   corn,   where   the   residue-­‐to-­‐grain   ratio   is   reported   was   2.0,   but   in   developed   countries,   the   residue-­‐to-­‐grain   ratio   is   typically   0.8   –   1.0   [106].   As   crop   production   and   crop   varieties   in   China   become   more   similar   to   those   in   the   West,   the   residue-­‐to-­‐grain   ratio  of  corn  would  be  expected  to  approach  these  values.       For  our  data  we  used  the  recent  values  from  two  papers  by  Xie  et  al.  that  surveyed  the   literature   to   determine   the   residue-­‐to-­‐grain   ratios   for   different   crops   in   different   parts   of   China   [74,   75].     These   values   are   fairly   conservative   but,   particularly   for   the   rice   straw   and   corn  stover  values,  they  seem  more  similar  to  those  that  have  been  set  forth  by  others  (Table   2.1).    When  we  compared  results  based  on  our  chosen  residue-­‐to-­‐grain  ratios  to  those  from   the  MOA  study,  apart  from  a  general  reduction  in  yields,  there  was  generally  little  apparent   difference   in   spatial   distribution   for   total   crop   residue   yields   either   at   the   prefecture   or   provincial   level.   The   biggest   differences   in   spatial   distribution   were   for   the   yields   per   cultivated   land   area   (Figure   2.5).   The   MOA   study,   which   had   a   comparatively   higher   corn   residue-­‐to-­‐grain  ratio  and  a  lower  rice  residue-­‐to-­‐grain  ratio  has  an  obvious  northerly  shift  in   crop  residue  density  compared  to  our  study,  with  comparatively  higher  crop  residue  densities   in  northern  China  and  lower  densities  in  southern  and  central  China.  The  choice  of  residue-­‐to-­‐ grain   ratio   could   lead   to   very   different   conclusions,   particularly   if   looking   at   the   coarser   provincial  scale  (Figure  2.5).  From  the  MOA  results  it  would  seem  that  the  highest  yields  are   in  the  northeast,  northern,  and  central  regions.  However,  for  our  results  the  highest  yields  are   in  eastern  and  central  China.  These  results  seem  more  logical  than  those  based  on  the  MOA   34         Figure  2.5:   Cultivated  land  as  the  proportion  of  total  land  area  in  a  prefecture  or  province,   and   estimated   2006   crop   residue   yields   per   prefecture   or   province   cultivated   land   area   (Mg/ha)   as   affected   by   the   chosen   harvest   indices.   (A,B)   Proportion   of   total   land   area   as   cultivated  land;  (C,  D)  Harvest  indices  from  Xie  et  al.  [74,  75];  and  (E,  F)  Harvest  indices  from   the  Ministry  of  Agriculture  report  [78].   35       study.  The  eastern  and  central  regions  of  China  have  a  high  prevalence  of  double-­‐  and  triple-­‐ cropping  [96],  which  would  result  in  greater  production  of  crop  residues  per  cultivated  land   area  compared  to  single-­‐cropped  land,  such  as  is  largely  present  in  northeast  China.     2.3.4.  Usable  amount  of  crop  residues   Traditionally  in  Mainland  China,  crop  straws  have  been  used  for  rural  energy  and  for   animal   feed.   However,   increases   large-­‐scale   animal   production   and   availability   of   cheap   commercial   energy   have   reduced   some   of   the   needs   for   these   materials   [107,   108].   This   is   especially  true  in  eastern  China,  which  has  developed  the  most  rapidly.  Straw  is  also  used  at  a   smaller   scale   as   a   raw   material   for   pulp   and   paper   production   and   as   a   substrate   for   mushroom  cultivation  [80].  Additionally,  a  portion  of  the  crop  straw  should  be  left  on  the  field   in   order   to   maintain   soil   organic   carbon   levels,   moderate   soil   temperatures   during   the   summer   growing   season,   limit   evaporation,   and   prevent   erosion.   However   it   is   possible   to   leave  too  much  crop  residue  on  the  field  [80],  particularly  in  no-­‐till  and  multi-­‐cropped  systems.   For  example,  the  presence  of  excess  crop  residues  from  one  crop  rotation  can  interfere  with   planting  and  seedling  germination  of  the  subsequent  rotation  [109,  110]  and  can  increase  the   prevalence   of   pests   and   pathogens.   Also,   due   to   the   high   prevalence   of   multi-­‐cropping   systems,  there  is  often  an  excess  of  crop  residues  that  are  too  abundant  to  be  left  on  the  field.   Often   there   is   either   insufficient   time   to   clear   the   field   before   the   next   planting   or   there   is   no   alternative  use  to  make  clearing  the  field  worthwhile,  so  much  of  these  residues  are  simply   burned   on   field,   despite   the   fact   that   this   practice   has   been   officially   banned   by   the   government   [111].   Besides   wasting   a   valuable   resource,   on-­‐field   combustion   of   crop   residues   36       Table  2.3:  Estimated  use  of  crop  residues  in  Mainland  China  from  various  sources.     Estimates  of  Residue  Use  in  Mainland  China   Reference   Year     This  Work     Li  et  al.  [78]   2006     1995     Cui  et  al.  [80]       2006     Zhang  et  al.  [36]   2007   millon   tons   %     millon   tons   %     millon   tons   %     millon   tons   %   Total     594   -­‐     604   -­‐     433   -­‐     755   -­‐   Rural  Energy   -­‐   -­‐     355   58     108   25     299   40   Returned  to  Field   297   50     91   15     130   30     113   15   Animal  Feed   154   26     145   24     79   18     208   28   Pulp  &  Paper   19   3     14   2     20   5     20   3   Mushroom   -­‐   -­‐     -­‐   -­‐     10   2     -­‐   -­‐   Burned   103   17     -­‐   -­‐     86   20     115   15     has   a   variety   of   negative   impacts   including   reducing   soil   organic   carbon,   harming   beneficial   soil  microorganisms,  increasing  air  pollution,  and  in  a  number  of  cases,  reducing  visibility  in   nearby  cities  to  the  point  of  grounding  air  travel  [107,  111-­‐113].   Compared  to  other  reported  values  for  the  amount  of  crop  residues  returned  to  the   field   [36,   78,   80]   and   the   amount   of   residues   used   for   animal   feed   [80]   our   numbers   are   higher   (Table   2.3).   This   is   expected   as   we   assumed   an   ideal   scenario   where   soil   health   is   adequately  protected  and  overgrazing  does  not  occur  on  pasture  and  rangeland.  It  is  almost   certain  that  these  values  are  higher  than  the  actual  situation  in  2006.  Li  et  al.  [78]  and  Zhang   et  al.  [36]  both  determined  the  amount  of  crop  residues  that  should  be  returned  to  the  field   using   a   set   percentage   of   the   total   residues   (15%).   However   this   amount   may   not   be   adequate   for   maintaining   soil   health   in   all   locations,   under   all   tillage   conditions,   and   for   all   cropping  systems  [114,  115].  For  some  situations  it  may  be  necessary  to  retain  as  much  as  70     37         Figure  2.6:  Estimated  ideal  use  of  crop  residues  in  each  province  in  Mainland  China  in  2006.   Values  for  the  amount  of  residues  used  as  animal  feed  and  returned  to  the  field  are  based  on   ideal   scenarios   without   overgrazing   or   adversely   effects   on   soil   health.   Red   circles   (at   the   same  scale  as  the  pie  graphs)  are  used  to  show  the  amount  of  additional  residues  needed  in   provinces  that  are  do  not  produce  the  amount  required  for  the  ideal  scenario.  Data  on  field   burning   of   crop   residue   are   from   [100].   A   table   listing   all   the   values   for   residue   use   and   on   field  burning  is  provided  in  the  supplemental  information  (Table  A.5).     –   80%   of   the   residue   on   the   field   in   order   to   maintain   soil   carbon   levels,   prevent   erosion,   etc.   [114-­‐116].   Our   values   represent   an   assumption   of   a   constant   amount   of   crop   residue   returned  per  ha  of  cultivated  land.  This  estimates  that,  nationally,  50%  of  the  residues  should   be   returned   to   the   field.   However,   the   actual   amount   that   should   be   retained   for   a   given   38       location   is   going   to   be   highly   dependent   on   climate,   topography,   cropping   systems,   and   management   (tillage,   irrigation,   fertilization).   Given   a   more   intensive   modeling   effort   that   incorporates  these  factors,  the  amount  of  residue  that  should  be  returned  to  the  field  across   the  country  would  be  improved.       A   number   of   provinces   did   not   generate   a   sufficient   quantity   of   crop   residues   to   sustain   livestock   production,   agriculture,   and   industry   (Figure   2.6).   Additionally,   all   of   these   provinces,   except   for   Guangxi   that   had   the   smallest   insufficiency,   had   very   little   on-­‐field   burning   of   crop   residues,   most   likely   due   to   the   high   demand   for   these   materials   for   other   applications  (Figure  2.6).  The  insufficiency  in  these  locations  is  not  surprising  as  a  number  of   these  provinces  are  known  to  have  issues  with  overgrazing  and  deterioration  of  agricultural   land   [95].   However,   while   these   locations   are   obviously   not   options   for   bioenergy   production   due   to   the   insufficient   quantities   of   crop   residues,   this   is   unfortunate   because   they   also   account   for   the   poorest   farming   populations   in   the   country   [117]   and   would   most   benefit   from  local  bioenergy  development.   We   were   unable   to   determine   the   provincial   values   of   the   amount   of   crop   straw   used   for  rural  energy  because,  although  this  is  reported  in  national  statistics,  it  is  generally  believed   to  be  significantly  overestimated  [97].  It  is  also  difficult  to  determine  the  amount  used  in  each   province   from   other   literature.   As   a   result,   the   amount   of   usable   residues   in   each   province   includes   the   amount   of   straw   needed   for   rural   energy   (Figure   2.6).   If   the   straw   was   combusted  at  a  power  plant  or  converted  to  biogas,  this  would  directly  replace  the  household   combustion  of  straw  for  fuel.  As  the  efficiency  of  these  conversion  processes  are  higher  [118],   more  energy  would  be  produced  for  the  same  amount  of  straw  without  negative  impacts  of   39       pollution  due  to  indoor  burning  of  straw  for  fuel  [119-­‐121].  If  this  material  was  used  for  liquid   fuel   production,   although   there   is   extra   electricity   produced   by   the   process   and   sent   to   the   grid  [12],  it  may  not  be  sufficient  to  replace  the  rural  energy.  Ideally  if  it  is  not  sufficient,  the   difference   would   be   replaced   by   some   other   renewable   source   of   electricity   and   not   by   kerosene  or  coal,  however  coal  is  the  most  likely  replacement.   One   possible   solution   to   the   demands   for   energy   and   animal   feed   is   ammonia   fiber   TM expansion   (AFEX )   pretreatment,   a   pressurized   ammonia   treatment   of   plant   materials.   In   addition   to   significantly   improving   the   enzymatic   digestibility   of   a   variety   of   plant   materials   for  the  ethanol  conversion  process,  this  method  has  also  been  shown  to  improve  their  quality   TM as   an   animal   feed.   AFEX -­‐treated   plant   materials   have   increased   digestibility   and   non-­‐ protein   nitrogen   content   and   have   been   shown   to   stimulate   milk   production   in   ruminants   [122,   123].   AFEX TM   technology   could   reduce   the   projected   dependence   on   animal   feed   imports  by  improving  grasses  and  crop  residues  for  animal  production,  thereby  increasing  the   potential   animal   feed   base   within   the   country.   The   north   region   of   China   (Beijing,   Tianjin,   Hebei,   Shanxi,   Shandong   and   Henan)   is   predicted   to   have   the   highest   future   production   of   ruminant  animals  (for  meat  and  milk)  and  the  highest  demand  for  animal  feed  [124].  By  our   estimates  Shandong,  Henan,  and  Sichuan  have  the  largest  number  of  non-­‐pastured  ruminants   (Table   A.2).   Henan   and   Shandong   are   also   the   largest   producers   of   crop   residues,   together   contributing  over  19%  of  the  total  national  production  of  crop  straw  (Table  A.5,  Figure  2.2).     Both  of  these  provinces  are  also  among  the  largest  potential  sources  of  usable  fuel  residues   and   the   largest   amount   of   on-­‐field   crop   residue   combustion.   Rural   development   is   also   40       important   to   consider   when   considering   bioenergy   applications.   Of   the   provinces   with   the   largest   amount   of   residues   available   for   energy   generation,   farmers   in   Henan,   Anhui,   and   Hunan  are  comparatively  less  well  off  and  would  perhaps  benefit  more  from  the  development   of   a   rural   bioenergy   compared   to   Shandong,   Jiangsu,   and   Jilin   [117].   Given   all   of   the   different   considerations,   Henan   in   particular,   and   the   central,   eastern,   and   northeastern   regions   of   China  in  general,  seem  like  the  best  locations  for  generation  of  bioenergy  from  crop  residues.       2.4.  Conclusion   Total   crop   residue   production   from   the   main   crops   in   Mainland   China   in   2006   was   estimated   at   593.5   million   dry   metric   tons.   The   largest   total   amounts   of   crop   straw   were   produced  in  Henan,  Shandong,  Heilongjiang,  Jiangsu,  and  Anhui,  while  the  densest  production   (Mg/ha  cultivated  land)  was  in  the  central  and  eastern  regions  where  there  is  intensive  multi-­‐ cropping.  When  different  residue-­‐to-­‐straw  ratios  were  used  which  were  based  on  the  Chinese   Ministry   of   Agriculture   evaluation,   the   regions   of   highest   density   straw   production   shifted   north   in   conjunction   with   the   relatively   higher   corn   and   lower   rice   residue-­‐to-­‐straw   ratios.   The  yields  of  all  the  crop  straws  were  largely  distributed  in  different  areas  of  the  country,  and   this   distribution   and   the   type   of   crop   straw   available   at   a   given   location   could   impact   the   choice  of  location  for  a  bioenergy  facility.  The  amount  of  usable  residues  for  bioenergy  was   estimated  to  be  124.7  million  metric  tons.  Given  the  location  of  usable  residues,  the  potential   TM for  rural  development,  and,  if  using  AFEX ,  the  production  of  animal  feed  as  a  co-­‐product,   Henan   in   particular,   and   central,   eastern,   and   northeastern   regions   of   China   in   general,   appear  to  be  the  best  locations  for  the  deployment  of  bioenergy  systems.   41       CHAPTER  3 :       CLASSIFICATION  SCALE:  INFLUENCE  OF  VARIABLE  SPECIES  COMPOSITION  ON  THE   TM   SACCHARIFICATION  OF  AFEX  PRETREATED  BIOMASS  FROM  UNMANAGED  FIELDS  IN   COMPARISON  TO  CORN  STOVER   3.1.  Introduction   Feedstock   cost   is   predicted   to   be   the   largest   contributor   to   the   overall   cost   of   cellulosic   ethanol   production   [12],   and   the   relative   importance   of   the   feedstock   will   only   increase  as  the  liquid  biofuel  industry  matures.  The  success  of  the  cellulosic  biofuel  industry   will  be  highly  dependent  on  the  availability  of  diverse  sources  of  inexpensive,  highly  digestible   plant  materials.  Potential  biofuel  feedstocks  can  be  categorized  in  terms  of  their  energy  and   chemical   inputs,   and   diversity,   ranging   from   high-­‐input,   low-­‐diversity   (conventional   monoculture  crops  e.g.  corn  &  soybeans)  to  low-­‐input,  high-­‐diversity  (native  prairie  /  mixed-­‐ species  grasslands)  [125].  In  addition  to  native  prairie,  old  fields  are  another  type  of  low-­‐input   natural  mixed  species  ecosystem.  Old  fields  are  defined  as  agricultural  fields  that  have  been   abandoned   and   no   longer   undergo   reseeding   and   maintenance.   First   year   production   from   these  abandoned  fields  is  comprised  primarily  of  mixed-­‐species  annual  weeds,  which  in  later   years  typically  succeeds  into  perennial  grasses,  composites  and  legumes,  and  eventually  into   shrubs  and  trees  [126].  Mixed-­‐species  ecosystems  such  as  native  prairie,  and  to  some  extent   old  fields,  provide  higher  value  ecosystem  services  compared  to  conventional  monocultures,   including  wildlife  habitat  and  pollination  services  [127-­‐129],  water  quality  maintenance  [130],   nitrogen-­‐fixation  in  fields  containing  legumes  [131,  132],  improved  soil  carbon  fixation/lower   carbon   debt   [4,   132,   133]   and   decreased   global   warming   potential   and   release   of   fine   particulate  matter  [133-­‐135].  However,  from  the  perspective  of  the  biorefinery  the  inherent   42       heterogeneity   of   polycultures   increases   the   apparent   risk   associated   with   these   materials.   Because  the  processing  characteristics,  potential  yields,  and  digestibility  cannot  currently  be   predicted   or   controlled,   this   could   intensify   the   challenges   associated   with   determining   feedstock   value   and   appropriate   processing   conditions   compared   to   monoculture   feedstocks.   Additionally,   because   of   high   harvest   costs   associated   with   low   predicted   biomass   yields,   at   the   farm   scale   mixed-­‐species   fields   are   not   considered   to   be   economically   competitive   with   other   bioenergy   cropping   systems   [136]. As   a   result,   polycultures   are   often   believed   to   be   undesirable  feedstocks.     But   in   spite   of   these   issues,   energy   generation   from   mixed-­‐species   feedstocks   has   been   examined   experimentally   using   a   number   of   different   methods   including   biogas   production   [137],   supercritical   gasification   [138],   liquefaction   [139]   and   co-­‐combustion   with   coal   [140].   To   date   there   has   been   no   experimental   research   on   ethanol   production   via   pretreatment   and   saccharification   of   mixed-­‐species   feedstocks,   although   two   studies   have   reported   theoretical   ethanol   yields   [141,   142].   Tilman   et   al.   [141],   in   their   paper   on   low-­‐input   high-­‐diversity   grasslands,   used   a   generic   ethanol   yield   (0.255   L/kg   DM)   which   was   based   on   a   reported   value   for   corn   stover   [143].   Adler   et   al.   [142]   estimated   ethanol   yields   from   conservation   grasslands   based   on   composition   data   with   a   set   conversion   rate,   and   determined   fermentability   using   in   vitro   gas   production.   However,   predicting   ethanol   yields   solely   from   composition   data   and   then   drawing   comparisons   between   feedstocks   gives   no   indication   of   potential   differences   in   digestibility,   which   can   also   be   affected   by   differences   in   organization  of  components  within  the  cell  wall,  the  presence  of  inhibitory  compounds,  etc.   For  example,  woody  materials  often  have  higher  structural  sugar  contents  than  grasses  and   43       based  solely  on  composition  data  they  might  be  expected  to  perform  better,  but  actually  they   are  typically  less  digestible  and  give  lower  sugar  yields  [14].   Because   enzymatic   sugar   yields   directly   impact   subsequent   ethanol   yields,   it   is   important   to   determine   whether   there   are   general   characteristics   of   mixed-­‐species   feedstocks   that   impact   yields   and   subsequent   profitability   for   the   biorefinery.   One   characteristic  that  is  unique  to  mixed-­‐species  feedstocks  is  the  combination  of  species  from   different  botanical  classifications.  The  simplest  classification  that  is  often  used  for  ecology  and   forage   research   includes   the   grasses   and   relatives   (graminoids)   and   forbs   (herbaceous   non-­‐ graminoids)   [144].   There   is   already   evidence   that   there   are   distinct   differences   in   effectiveness  of  chemical  pretreatments  and  saccharification  efficiency  between  species  from   these   two   groups   [13,   145].   By   comparing   feedstocks   that   contain   varying   distributions   of   the   different  plant  classifications,  it  should  be  possible  to  observe  whether  there  are  classification   effects   on   pretreatment   efficiency   and   saccharification   yields.   This   information   might   then   be   used  to  better  manage  a  mixed-­‐species  ecosystem  to  maintain  most  of  the  ecological  benefits   of  an  unmanaged  system  while  preserving  most  of  the  downstream  economic  benefits  of  an   intensively  managed  grass  monoculture.   For   this   study,   we   evaluated   the   sugar   yields   following   ammonia   fiber   expansion   (AFEX TM )   pretreatment   and   enzymatic   hydrolysis   of   biomass   harvested   from   five   newly   abandoned  alfalfa  fields  (e.g.  early  successional  old  fields),  each  of  which  varied  in  its  mixture   of   annual   forb   and   grass   species.   This   study   was   then   broadened   to   include   a   single   mixed-­‐ species   sample   that   was   prepared   by   mixing   biomass   harvested   from   three   fields   that   had   been   abandoned   for   at   least   45   years   (e.g.   late   successional   old   fields)   and,   as   a   control,   a   44       TM separate  sample  of  corn  stover.  Samples  were  pretreated  using  AFEX  and  sugar  yields  were   measured  following  enzymatic  saccharification.  As  animal  feed  is  one  possible  co-­‐product  for   TM a   biorefinery,   the   AFEX -­‐treated   samples   were   also   evaluated   for   their   forage   quality   as   measured  by  a  standard  in  vitro  digestibility  assay.       3.2.  Materials  and  methods   3.2.1.  Sample  harvest  and  preparation   Corn  stover  including  cobs  (CS)  was  provided  by  the  Great  Lakes  Bioenergy  Research   Center   (GLBRC)   at   the   University   of   Wisconsin   –   Madison.   This   material   was   harvested   on   rd September   3 ,   2008.     The   corn   stover   was   milled   through   a   2   mm   screen   using   a   Retsch   centrifugal  mill  prior  to  composition  analysis  and  enzymatic  hydrolysis.  Old  field  samples  were   provided  from  the  five  replicates  (R1  –  R5,  treatment  G9)  of  the  GLBRC  intensive  experiment   site   at   W.K.   Kellogg   Biological   Station   (KBS)   of   Michigan   State   University   (MSU).   During   the   previous  year  (2007)  this  site  had  been  planted  with  alfalfa.  Individual  plots  were  40  x  28  m   and  had  received  no  maintenance  or  inputs  in  2008  other  than  initial  disking  in  May.  Three   st quadrats   of   2.0   x   0.5   m   were   harvested   within   each   plot   on   August   20-­‐21 ,   2008.     A   sixth   sample   from   the   Long-­‐Term   Ecological   Research   (LTER)   project   was   also   provided   from   KBS   and  consisted  of  a  mixed  sample  from  three  late-­‐successional  old  fields  (SF1  –  SF3)  that  had   been   abandoned   and   unmanaged   since   1964,   1948   and   1963   respectively   [146].     Five   th quadrats  of  2.0  x  0.5  m  were  harvested  on  August  13-­‐18 ,  2008.    For  the  GLBRC  and  the  LTER   45       plots,   all   plants   rooted   in   each   quadrat   were   clipped   at   ground   level,   bagged   and   dried   at   60°C  for  a  minimum  of  48  hr.    The  dry  weights  were  determined  for  each  species,  following   which   the   quadrats   for   each   plot   were   combined   and   milled   through   a   2   mm   screen.     The   species  composition  of  these  plots  is  listed  in  the  supporting  information  in  Table   B.1   (GLBRC)   2 and   Table   B.2   (LTER).     The   yield   (g/m )   for   each   plot   was   calculated   from   the   combined   2 2 quadrat  yields  (GLBRC:  g/3  m  or  LTER:  g/5  m ).     3.2.2.  Composition  analysis   Biomass   moisture   content   was   determined   using   a   moisture   analyzer   (A&D,   Model   MF-­‐50;   San   Jose,   CA).     The   dry   matter   (DM)   composition   of   each   sample   (extractives,   ash,   lignin,  glucan  and  xylan  content)  was  determined  based  on  the  NREL  standard  protocols  [147].   The   acid   insoluble   lignin   analysis   method   was   modified   to   use   47   mm,   0.22   μm   pore-­‐size,   mixed-­‐cellulose   ester   filter   discs   (Millipore   Corp.;   Bedford,   MA)   during   the   filtration   step   instead  of  fritted  crucibles.  Due  to  problems  with  burning,  these  discs  with  the  filtered  lignin   residue  were  dried  overnight  in  a  desiccator  prior  to  weighing  rather  than  in  a  vacuum  oven.   The   nitrogen   content   of   the   extracted   and   unextracted   samples   were   determined   via   the   combustion   method   for   nitrogen   determination [148] using   a   Skalar   Primacs   SN   Total   Nitrogen   Analyzer   (Breda,   The   Netherlands).   Nitrogen   values   were   multiplied   by   6.25   to   determine   the   crude   protein   content.   This   conversion   factor   assumes   that   16%   of   the   protein   is   nitrogen   and   that   there   is   negligible   non-­‐protein   nitrogen   present   in   the   biomass.   This   factor   varies   with   different   types   of   plant   samples   due   to   differences   in   protein   structure,   46       however   accurate   determination   of   this   factor   requires   a   complete   amino   acid   analysis   [149].   As  protein  content  does  not  directly  impact  our  results  and  conclusions,  we  assume  that  the   standard  factor  of  6.25  allows  for  a  reasonable  approximation.  The  protein  that  was  removed   during  extraction  steps  was  subtracted  from  the  total  extractives  content.         3.2.3.  Effect  of  pretreatment  conditions  on  hydrolysis  yields  from  early  successional  samples   TM AFEX   pretreatment   of   the   early   successional   old   field   samples   was   conducted   using   3.0   TM g  DM  of  sample  in  22  mL  reaction  vessels  as  outlined  by  Bals  et  al.  [150].  AFEX  conditions   for   the   pretreatment   optimization   experiments   were   chosen   using   a   three   level   Box-­‐Behnken   statistical  design  and  the  parameters  ranged  from  0.5  to  2.0  g  NH3:g  DM,  0.5  to  2.0  g  H2O:g   DM,  90  to  180°C,  and  5  to  30  min  residence  time.  The  central  design  point  was  conducted  in   triplicate  to  give  a  total  of  24  different  conditions  and  27  experiments  for  each  sample.  In  an   attempt   to   improve   the   fit   of   the   statistical   models,   three   additional   experimental   points   were   tested   (Table   B.3).   Enzymatic   hydrolysis   was   then   conducted   on   the   samples   as   described   in   Section   3.2.5.   and   the   enzymatic   hydrolysis   total   sugar   conversions   (g   oligomeric   and  monomeric  glucose  and  xylose  released  per  g  sugar  theoretically  present  in  the  untreated,   dry  biomass)  were  used  as  the  metric  of  pretreatment  efficacy.   TM The  following  polynomial  quadratic  equation  of  the  AFEX  pretreatment  conditions  was   fitted  to  the  enzymatic  hydrolysis  total  sugar  yield  data  (combined  monomeric  and  oligomeric   47       glucose   and   xylose   in   terms   of   the   theoretical   maximum)   using   Minitab15   Statistical   Software   (2006  Minitab  Inc,  Pennsylvania,  USA):     n   n n 2+ Y = a0 + ∑ ai xi + ∑ aii xi ∑ aij xi x j   i=1 i=1 i, j=1 i≠ j (3.1)   where  Y  is  the  sugar  yield;  a0  is  the  regression  constant;  ai  is  the  linear  regression  coefficient   for  the  ith  parameter;  aii  is  the  quadratic  regression  coefficient  for  the  ith  parameter;  aij  is  the   interaction  coefficient  for  the  ith  and  jth  parameters;  xi   and  xj  are  the  values  of  the  ith  and  jth   parameters;  and  n  is  the  number  of  factors,  which  in  this  case  is  4.  Coefficients  with  p  >  0.1   were   removed   stepwise   from   the   model,   beginning   with   the   largest   p-­‐value.   The   resulting   2 coefficients  and  adjusted  and  predictive  R  values  for  each  sample  are  reported  in  Table  B.4.   An   attempt   was   made   to   determine   the   optimum   pretreatment   conditions   for   each   early  successional  sample  using  the  polynomial  models.  However,  the  predictive  capability  of   2 the   models   was   very   poor,   as   indicated   by   the   low   predictive   R   values   (Table   B.4)   and   we   were   unable   to   determine   the   optimum   pretreatment   conditions.   So   for   the   later   experiments,   the   same   pretreatment   condition   was   chosen   for   all   of   the   feedstocks:   2.0   g   NH3:g  DM;  0.5  g  H2O:g  DM;  90°C;  30  min.  This  set  of  conditions  was  chosen  as  it  resulted  in   comparatively   high   sugar   yields   for   four   of   the   five   feedstocks   (the   highest   sugar   yields   for   E7   and  E87  and  among  the  top  five  highest  yields  for  E60  and  E83)  (Table   B.3,   Figure   B.1).  As  the   48       last   sample,   E21,   performed   better   when   pretreated   at   slightly   higher   temperatures,   we   raised  the  selected  temperature  slightly  to  100°C.     TM 3.2.4.  AFEX  pretreatment  for  comparison  of  early  successional  old  field  replicates  with  late   successional  old  field  corn  stover  samples     TM AFEX   was   performed   on   the   early   successional   samples,   the   late   successional   sample  (LSF)  and  corn  stover  (CS)  in  a  300  mL  stainless  steel  (#316)  Parr  reactor  according  to   the  method  detailed  in  Bals  et  al.  [123].  The  conditions  were  chosen  because  of  the  relatively   high   hydrolysis   yields   for   the   early   successional   feedstocks   as   determined   in   the   previous   section:  2.0  g  NH3:g  DM,  0.5  g  H2O:g  DM,  100°C,  30  min.         3.2.5.    Enzymatic  hydrolysis   Samples  for  pretreatment  characterization  were  hydrolyzed  in  20  mL  screw-­‐cap  vials   at   1.5%   total   sugar   (glucan   +   xylan)   loading   and   a   total   volume   of   15   mL.   For   the   feedstock   comparison   experiments,   samples   were   hydrolyzed   in   20   mL   screw-­‐cap   vials   at   3%   solids   loading   and   a   total   volume   of   10   mL.   Samples   were   adjusted   to   a   pH   of   4.8   by   1M   citrate   buffer   solution.   To   prevent   fungal   and   bacterial   contamination   during   enzymatic   hydrolysis,   cycloheximide   and   tetracycline   were   loaded   at   a   final   concentration   of   30   µg/mL   and   40   µg/mL,  respectively. 49       Accelerase®1000,  Multifect®  Xylanase,  and  Multifect®  Pectinase  (Genencor  Division   of   Danisco   US,   Inc.)   were   used   for   all   of   the   hydrolysis   experiments.   The   protein   content   of   each   of   the   enzymes   are   as   follows:   Accelerase®1000   (84   mg   protein/mL),   Multifect®   Xylanase  (32  mg  protein/mL),  and  Multifect®  Pectinase  (61  mg  protein/mL).  Enzyme  protein   content   was   determined   from   total   N   analysis   using   the   Dumas   method   for   combustion   of   nitrogen   to   NOx   [148]   following   trichloroacetic   acid   (TCA)   precipitation   to   remove   non-­‐ protein  nitrogen  [151]. For   the   pretreatment   characterization   experiments,   Accelerase®1000   was   added   at   26.8  mg  protein/g  glucan  in  the  untreated  biomass,  and  Multifect®  Xylanase  and  Multifect®   Pectinase   were   each   added   at   7.5   mg   protein/g   xylan   in   the   untreated   biomass.   For   the   feedstock   comparison   experiments,   enzymes   were   loaded   at   6.0   mg   protein/g   DM   sample   and  in  the  following  relative  percentages  by  protein  mass,  which  were  similar  to  those  found   effective   for   Alamo   switchgrass   (unpublished   data):   Accelerase®1000   (42%),   Multifect®   Xylanase  (24%),  and  Multifect®  Pectinase  (34%).         Enzymatic  hydrolysis  for  all  experiments  was  conducted  in  a  New  Brunswick  Scientific   (Edison,   NJ)   shaking   incubator   at   50°C   and   200   rpm.     Samples   were   taken   at   72   hours   of   enzymatic   hydrolysis   for   both   monomeric   and   oligomeric   sugar   analysis,   as   detailed   below.   Samples  taken  for  monomeric  sugar  analysis  were  heated  at  100°C  for  15-­‐20  min,  cooled  in   the  freezer,  and  then  centrifuged  at  15,000  × g  for  5  minutes.  The  supernatant  was  filtered   50       into   HPLC   shell   vials   using   a   25   mm,   0.2   μm   polyethersulfone   syringe   filter   (Whatman   Inc.   Florham  Park,  NJ),  then  stored  at  -­‐20°C  until  further  sugar  analysis.     3.2.6.  Oligomeric  sugar  analysis   Oligomeric   sugar   analysis   of   the   pre-­‐wash   liquid   and   hydrolysate   was   conducted   using   a   scaled-­‐down   version   of   the   standard   NREL   method   for   oligomeric   sugar   determination   of   liquid   streams   [152].   The   modified   method   was   identical   except   that   it   was   scaled   down   to   use  2  mL  of  sample  and  run  in  duplicate  in  10  mL  screw-­‐cap  culture  tubes.  Instead  of  being   autoclaved,  the  tubes  were  incubated  in  a  121°C  bench-­‐top  hot  plate  for  one  hour,  cooled  on   ice,   and   the   liquid   was   filtered   into   HPLC   vials.   The   oligomeric   sugar   concentration   was   determined   by   subtracting   the   monomeric   sugar   concentration   of   the   non-­‐hydrolyzed   samples  from  the  total  sugar  concentration  of  the  acid  hydrolyzed  samples.         3.2.7.  HPLC  analysis   Sugar   contents   of   all   composition   analysis,   wash   liquid,   and   hydrolysate   samples   were   determined   using   a   Bio-­‐Rad   (Hercules,   California,   USA)   Aminex   HPX-­‐87H   column   equipped   with   appropriate   guard   columns.   Degassed   5   mM   aqueous   H2SO4   was   used   as   the   mobile   phase   and   the   column   temperature   was   held   at   60°C. The   reported   total   sugar   (glucose   or   xylose)  concentration  was  recalculated  as  the  sum  of  the  average  monomeric  and  the  average   oligomeric   sugar   concentration   for   each   sample.   Because   of   the   presence   of   soluble   sugars   in   51       the  biomass,  the  glucose  percent  conversions  were  calculated  using  the  following  equation,   where  0.9  corrects  for  addition  of  the  water  molecule  upon  hydrolysis  of  glucan  to  glucose:   Glucose Conversion (%) = GluHY   Gln / 0.9 + Glu + Suc * (180.2 / 342.3) (3.2)   GluHY  =  hydrolysate  glucose  mass  yield  (g/kg  dry  biomass)   Gln  =  biomass  glucan  content  (g/kg  dry  biomass)   Glu  =  biomass  soluble  glucose  content  (g/kg  dry  biomass)   Suc  =  biomass  soluble  sucrose  content  (g/kg  dry  biomass)   180.2/342.3   =   correction   for   glucose   contribution   by   sucrose   (molecular   weight   of   glucose/molecular  weight  of  sucrose)     3.2.8.    In  vitro  rumen  digestibility  and  neutral  detergent  fiber  determination   In  vitro  rumen  digestibility  was  performed  in  triplicate  based  on  the  method  reported   in   Tilley   and   Terry   [153]   using   rumen   fluid   obtained   from   a   fistulated   dairy   cow.   Neutral   detergent   fiber   (NDF)   concentration   (without   amylase   digestion)   was   determined   for   each   sample  (treated  and  untreated)  after  0  h  and  48  h  of  incubation.  The  amount  of  NDF  digested   after  48  h  was  determined  as  the  difference  between  these  two  values.         3.2.9.  Statistical  analysis   All   statistical   analyses,   except   for   Pearson’s   correlation   coefficients,   which   were   calculated   using   Excel   (Microsoft®   Office   Excel®   2007),   were   performed   using   Minitab15   Statistical  Software  (2006  Minitab  Inc,  Pennsylvania,  USA).  This  included  the  response  surface   52       optimization   (as   described   previously   in   Section   3.2.3.   ),   linear   regressions,   and   Tukey’s   pairwise  comparisons.     3.3.  Results   3.3.1.  Feedstock  characteristics  and  plot  yields   The   species   mass   composition   of   the   old   field   replicates   (E7,   E21,   E60,   E83,   E87)   consisted  almost  entirely  of  annual  forbs  and  grasses  (Table   B.1).  These  samples  have  been   relabeled   for   ease   of   analysis   in   terms   of   their   %   grass   content   on   a   mass   basis.   There   were   7   to  14  species  in  each  replicate  (Table   3.1),  however  five  main  species  contributed  >  95%  of   the   total   mass   for   all   five   replicates.   The   late   successional   old   field   sample   (LSF)   was   composed  of  three  different  replicate  plots  with  20  to  45  species  in  each  replicate  (Table   3.1).   Seven  of  these  species  comprised  86%  of  the  combined  sample  mass  (Table  B.2).  There  was   2 no   distinct   trend   between   plant   classification   and   biomass   yield   (g/m )   for   the   five   early   successional   replicates.   The   late   successional   replicate   that   contained   mostly   grass   had   a   much  higher  biomass  yield  than  the  other  two  samples  that  were  composed  predominantly  of   woody  species.     The   dry   matter   composition   of   each   sample   is   shown   in   Table   3.2.     The   early   successional  replicates  that  had  lower  grass  contents  (E7  and  E21)  also  had  lower  structural   sugar  contents  compared  to  the  samples  that  had  higher  grass  contents  (E60,  E83,  E87,  and   LSF).   Corn   stover   had   a   significantly   higher   structural   sugar   content   compared   to   all   other   samples   (65.5%   of   the   total   dry   mass).   The   Klason   lignin   content   of   the   different   samples   ranged   from   13.6%   -­‐   18.3%   of   the   total   dry   biomass   and   was   inversely   correlated   with   the 53       Table  3.1:  Species  composition,  biomass  yield,  and  distribution  for  the  GLBRC  old-­‐field  and  LTER  replicates.             #  of  Species     2 Biomass  Yield  (g/m )     Mass  Distribution  (%)   Expt.   Field     Grass   Forb   Wood   Total     Grass   Forb   Wood   Total     Grass   Forb   Wood     Early  Successional  Old  Field  Treatment  Replicates     4   E7   R1   10   0   14   R4   E60   R5   E83   R3   E87     E21   R2   -­‐   727     7%   93%   -­‐   6   0   7     210   799   -­‐   1009     21%   79%   -­‐   4   0   8     516   347   -­‐   863     60%   40%   -­‐     3     4   8   0   11     753   158   -­‐   911     83%   17%   -­‐   4   0   8     576   88   -­‐   664     87%   13%   -­‐   20     213   47   -­‐   260     82%   18%   -­‐   28     7   1   23   31     21%   4%   45     <1   17   47   64     0%   27%   73%     SF3   673     1     4   Late  Successional  Old  Field  Replicates     SF1     6   14   0   LSF   SF2     5   10   13       54     3   25   17   54     75%     Table  3.2:  Composition  analysis  data  as  %  of  total  dry  matter  (DM).    The  standard  error  is  reported  in  parenthesis  and  represents   three  replicates.  LSF  =  late  successional  old  field  sample.     Structural  Carbohydrates*   Early  Successional  Old  Field   Feedstock   Gln   Xyl   d   f   c   e   c   Klason   Ash   Lignin   a         E21   26.4 15.1 (0.7)   (0.1)   2.1   (0.1)   43.7   (0.7)       16.3 (0.3)       E60   29.1 17.2 (0.8)   (0.6)   2.5   (0.2)   48.9 (1.0)       16.9 (0.3)   e 13.9 (0.1)   Total     Water  Extractives   2.1 39.0 (0.04)   (0.2)   E7   23.0 (0.1)   Ara   e b d b   b   b c   d bc   b       h     18.3 (0.1)   b b d   a   Crude   g Protein   1.8   (0.4)   d 14.2   (0.6)   8.8   5.5   (0.04)   (0.1)       2.0 (0.1)   3.2   (0.3)   3.3   (0.7)   c 11.8   (1.0)   9.0   (0.1)   4.6   (0.1)       1.7 (0.1)   2.5   (0.1)   3.1   (0.1)   c 13.5   (0.3)   a   bc b d e bc bc c   d f e     b c bc   c Other     1.8   (0.2)   11.2   (0.1)   b Suc   1.9   (0.1)   9.7 (0.2)   a   Acetyl   Glu   a b b   E83   29.6 (0.9)   18.7 (0.5)   2.1 (0.2)   50.5 (1.0)     14.0 (0.1)   8.5 (0.1)   4.9 (0.02)     1.8 (0.1)   2.4 5.8 (0.05)   (0.1)   11.2   (0.5)   E87   30.7   17.8 (0.7)   (1.4)   2.5   (0.1)   50.9   (1.6)     13.6   (0.5)   9.2   (0.1)   6.2   (0.2)     1.6   (0.1)   2.2   4.9   (0.05)   (0.2)   13.3   (2.0)   2.6   (0.1)   47.3   (0.8)       14.9 (0.1)   5.7   (0.1)   8.5   (0.1)       1.5 (0.1)   3.5   (0.1)   65.5   (0.6)   14.9     (0.1)   4.6   (0.5)   NM   2.5     (0.1)   b c cd   bc     LSF   26.5 18.2 (0.7)   (0.4)   a a Corn   36.8   25.0   Stover   (0.5)   (0.2)   b b a b c a d c c ab d e c b cd d a 1.2   (0.1)   0.2   21.6   (0.03)   (0.4)   e   e 0.4 0.3   8.0   (0.02)   (0.04)   (0.3)     Ethanol     Total   h Mass   Ext.     3.4     101.3   (0.01)   (0.8)     4.8     99.8   (0.3)   (1.5)     2.8     103.1   (0.2)     2.9   (0.2)     3.6   (1.4)     ND     2.6   (0.1)   (1.1)     102.0   (1.2)     105.5   (2.9)     100.9   (0.9)     98.6   (0.8)   *Gln=  glucan;  Xyl  =  xylan;  Ara  =  arabinan;  Glu  =  glucose;  Suc  =  sucrose;  Ext.  =  extractives   a-­‐f   Means  with  different  superscripts  in  each  column  are  significantly  different  at  the  95%  confidence  level  using  Tukey’s  pairwise   comparison.   g     h   Crude  protein  content  was  not  determined  for  the  corn  stover  sample   The  ethanol  extraction  was  performed,  but  not  quantified  for  the  SF  replicates  sample.  “Other  Water  Extractives”  also  includes  the   amount  of  ethanol  extractives. 55       sample  grass  content  (r  =  -­‐0.95,  p  =  0.01,  n  =  7)  when  measured  on  a  cell  wall  basis  (generalized   as  the  sum  of  the  structural  carbohydrates  and  Klason  lignin).  Ash  content  andsoluble  glucose   and   sucrose   content   were   higher   for   the   early   successional   samples   compared   to   the   late   successional  and  corn  stover  samples.     TM 3.3.2.  Relationship   of   AFEX   pretreatment   conditions   to   hydrolysis   yields   from   early   successional  samples   TM Each  of  the  five  early  successional  replicates  was  pretreated  with  AFEX  using  various   ammonia  and  water  loadings,  temperatures  and  residence  times.  Each  sample  was  tested  using   30  different  sets  of  pretreatment  conditions,  except  for  E87  where  one  result  was  omitted  from   the   analysis   due   to   a   high   residual   value   that   corresponded   to   abnormally   low   sugar   yields.)   Information   on   the   specific   conditions   examined   and   the   resulting   total   sugar   yields   are   reported   in   the   supplemental   information   (Table  B.3).   The   sugar   conversion   varied   significantly   for  each  feedstock  across  the  pretreatment  conditions  (Figure  3.1)  and  trends  appeared  to  be   associated   in   part   with   forb   vs.   grass-­‐dominated   samples.   E7   had   the   lowest   conversions   of   monomeric   and   oligomeric   sugars   (glucose:   46-­‐67%;   xylose:   46-­‐82%),   followed   by   E21   (glucose:   54-­‐80%;  xylose:  57-­‐80%).  E60,  E83,  and  E87  all  had  similar  ranges  for  xylose  release  (E60:  69-­‐ 102%;  E83:  67-­‐98%;  E87:  65-­‐102%).  The  yields  that  are  slightly  over  100%  are  likely  due  to  small   errors   that   occurred   during   compositional   analysis.   Glucose   release   tended   to   increase   from   E87  to  E83  to  E60  and  is  more  evident  from  Figure   3.1   than  from  the  range  of  glucose  yields   (E60:  65-­‐91%;  E83:  59-­‐84%;  E87:  56-­‐82%).  Pretreatment  improved  sugar  yields  in  all  cases     56         Figure   3.1:   Relationship   between   enzymatic   hydrolysis   glucose   and   xylose   yields   for   the   GLBRC   old   field   replicates.   (A)   Forb-­‐dominated   samples   (E7   and   E21).     (B)   Grass-­‐dominated   samples  (E60,  E83,  and  E87).    Yields  are  calculated  as  the  total  monomeric  and  oligomeric  sugar   solubilized  based  on  the  total  sugar  theoretically  available  in  the  untreated  dry  biomass.  For  the   E60,   E83,   and   E87   regressions,   (p   =   0.000).     Each   data   point   represents   one   of   30   different   pretreatment   conditions   (except   for   E87,   which   only   has   data   for   29   conditions   due   to   one   significant  outlier).       compared  to  the  untreated  feedstock,  for  which  sugar  yields  ranged  from  32.2%  for  E7  to  46.3%   for  E83  (data  not  shown).     Each  feedstock  showed  a  significant  linear  correlation  (p  <  0.05)  between  glucose  yield   and  xylose  yield  (Figure  3.1).  However,  there  was  a  much  larger  spread  for  the  data  points  for   the  two  feedstocks  that  had  the  lowest  grass  content  (<  20%  grass),  E7  and  E21  (r  ~  0.45).  The   correlation   between   glucose   and   xylose   release   was   both   highly   linear   (r   >   0.90)   and   highly   significant  (p  <  0.001)  for  the  feedstocks  with  the  higher  grass  content  (>  60%  grass),  E60,  E83,   and  E87.   57       Due  to  high  error  associated  with  the  constructed  response  surface  models  (Table  B.4),   the  optimum  pretreatment  conditions  could  not  be  accurately  determined  for  the  feedstocks.   Because   of   this,   the   raw   data   were   used   to   select   a   pretreatment   condition   that   gave   reasonably   high   yields   for   most   of   the   feedstocks   and   could   be   used   for   further   experiments.   The  pretreatment  condition  chosen  was  2.0  g  NH3:g  DM,  0.5  g  H2O:g  DM;  100°C  and  30  min   residence   time.   This   condition   was   chosen   as   a   similar   pretreatment   condition   at   90°C   resulted   in  the  highest  sugar  yields  for  E7  and  E87  and  among  the  top  five  sugar  yields  for  E60  and  E83   (Table  B.3,  Figure  B.1).  Of  those  materials  examined,  only  E21  did  not  obtain  high  yields  when   operated  at  these  conditions  (lower  than  the  optimum  condition  by  50  g  sugars  released  per  kg   untreated   dry   biomass),   and   it   tended   to   require   slightly   higher   temperatures   compared   to   the   other   feedstocks.   This   difference   appears   unrelated   to   the   differences   between   forbs   and   grasses.  For  example,  sample  E7,  the  other  forb  dominated  feedstock,  has  the  same  optimum   as  E87.  The  difference  may  instead  be  related  to  the  specific  species  that  were  present  in  the   mixture.   E21   contained   a   much   larger   percentage   of   Amaranthus   retroflexus   L.   (redroot   pigweed),  and  it  may  be  that  this  forb  is  more  indigestible  than  the  forb  species  present  in  the   E7  sample.     3.3.3.  In  vitro  digestibility     The   sale   of   pretreated   biomass   as   an   animal   feed   co-­‐product   has   the   potential   to   improve   the   process   economics   in   a   biorefinery   [10],   so   the   in   vitro   rumen   digestibility   was   determined   for   the   untreated   and   the   pretreated   old   field   replicates   (Table   3.3).   In   all   cases,   58       TM Table   3.3:   In   vitro   rumen   digestibility   of   untreated   and   AFEX -­‐treated   early   successional  old  field  samples.   Initial  NDF    (g  NDF/kg  DM)     Untreated     Average   E7   E21   E60   E83   E87   528   574   586   597   595     a TM AFEX -­‐Treated     d Average             SEM   6.6   3.0   1.9   4.9   2.7   442   475   526   509   491   a SEM   4.5   4.0   5.9   5.8   0.9   b c Difference   86   99   60   87   104   %  Increase   -­‐   -­‐   -­‐   -­‐   -­‐   Digested  (g  NDF/kg  DM)       Untreated     Average   E7   E21   E60   E83   E87   130   205   260   290   313   TM   a AFEX -­‐Treated     Average             SEM   9.7   7.9   2.4   15.2   6.6   133   212   327   313   323   a SEM   11.8   4.6   6.5   8.2   2.8   b c Difference   3   7   67   23   10   %  Increase   2%   4%   26%   8%   3%   e Total  Digested  (g  NDF/kg  DM)   a b c f   SEM   Difference   %  Increase   %  Digested   E7   E21   E60   218   311   387   12.7   3.7   3.3   89   107   127   68%   52%   49%   41%   54%   66%   E83   E87   a     Average   400   427   7.6   3.8   111   114   38%   37%   67%   72%   Standard  error  of  the  mean  based  on  three  replicates   b     c     d e TM Difference  between  AFEX Percent  Increase  in  treated  sample  over    untreated  sample      NDF  digested  during  48  h  in  vitro  rumen  digestion      Total  NDF  removed  for  AFEX TM f -­‐treated  and  untreated  samples   due  to  AFEX TM -­‐treated  samples,  including  both  the  amount  removed    and  the  amount  digested  during  in  vitro  rumen  digestion   TM     Total   amount   of   NDF   removed   due   to   AFEX percentage  of  untreated  NDF     and   during   in   vitro   rumen   digestion   as   a   59       TM AFEX   treatment   decreased   neutral   detergent   fiber   (NDF)   content   compared   to   untreated   controls,   a   difference   of   60   –   104   g   NDF/kg   DM,   indicating   an   increase   in   digestibility   of   pretreated  materials.  Except  for  the  E60  sample,  where  almost  26%  more  was  digested  for  the   TM AFEX TM -­‐treated   sample,   there   was   very   little   difference   between   the   untreated   and   AFEX -­‐ treated   samples   in   the   amount   of   NDF   digested   by   the   rumen   microbes.   The   percent   of   TM digested  NDF  for  the  AFEX -­‐treated  samples,  from  largest  to  smallest  was  E87  >  E83  >  E60  >   E21   >   E7,   which   corresponded   with   decreasing   grass   content   in   the   samples.   In   all   cases,   TM AFEX   improved   overall   NDF   digestion   compared   to   untreated   samples.     The   materials   with   TM the   lowest   grass   content   had   the   largest   improvement   in   digestibility   due   to   AFEX   pretreatment.   One   issue   with   NDF   determination   is     that   pectins,   which   are   present   in   much   larger  amounts  in  forbs,  are  digested  by  the  NDF  process,  resulting  in  a  lower  estimate  of  NDF   value  than  is  actually  the  case  [154].  As  pectins  can  be  utilized  by  ruminants,  this  may  result  in   underestimating  the  value  of  the  forb-­‐dominated  samples  as  an  animal  feed.     3.3.4.  Comparison  of  early  successional  old  field  replicates  to  late  successional  old  field  (LSF)  and   corn  stover  (CS)  samples     The  glucose,  xylose  and  total  sugar  hydrolysis  yields  (g/kg  dry  biomass)  are  reported  in   Figure  3.2.  All  of  the  total  sugar  yields  were  significantly  different  except  for  E60,  E83,  and  E87,   and  can  be  arranged  in  decreasing  order  from  CS  >  (E87,  E83,  E60)  >  LSF  >  E21  >  E7.  Both  LSF   and  CS  had  a  large  amount  of  xylo-­‐oligomers  remaining,  indicating  that  for  these  samples  the   60         Figure  3.2:  Comparative  monomeric  and  oligomeric  sugar  yields.    (A)  Glucose   yields.   (B)   Xylose   yields.   (C)   Total   sugar   (glucose   +   xylose)   yields.   Oligomeric   sugars   are   reported   in   monomeric   equivalents.   The   maximum   theoretical   sugar   yield   is   the   maximum   amount   of   glucose,   xylose   or   total   sugars   that   could   be   released   from   the   untreated   dry   biomass.     Total   (monomeric   and   oligomeric)  sugar  yields  with  different  letters  are  statistically  different  based   on   Tukey’s   test   (p   <   0.05).   E7,   E21,   E60,   E83,   E87   =   early   successional   old   field   replicates  from  the  GLBRC  intensive  site;  LSF  =  LTER  late-­‐successional  old  field   combined  sample;  CS  =  corn  stover.  Each  sample  was  subjected  to  the  same   pretreatment  and  enzymatic  hydrolysis  conditions.     61         Figure  3.3:  Correlation  of  glucose  (A)  and  xylose  (B)  percent   conversion   (g   sugar   released/g   theoretically   available   in   untreated   dry   biomass)   to   Klason   lignin   content   on   a   cell   wall   basis   for   the   early   successional   old   field   replicates,   late-­‐successional   old   field   sample,   and   corn   stover.   The   solid   line   represents   the   correlation   when   the   old   field   replicate   E60   (represented   by   an   open   circle)   is   included   in   the   analysis   and   the   dashed   line   indicates   the   correlation   when  sample  E60  is  removed  from  the  analysis.       62       enzyme   combination   used   was   not   adequate   to   convert   all   of   the   xylo-­‐oligomers   to   xylose.   None   of   the   samples   reached   the   theoretical   maximum   sugar   yield,   except   for   the   LSF   xylose   yield.   The   grass-­‐enriched   samples   (E60,   E83,   and   E87)   had   higher   sugar   mas   yields   and   conversion  efficiencies  compared  to  the  forb-­‐enriched  samples  (E7  and  E21).  Based  on  Tukey’s   test  (p  <  0.05),  the  different  feedstocks  can  be  divided  into  two  statistically  different  categories   based  on  the  total  sugar  yields  on  a  percent  basis,  the  low  grass  content  samples  (<  20%  grass):   E7  and  E21;  and  the  corn  stover  and  high  grass  content  samples  (>  60%  grass):  E60,  E83,  E87,   LSF  and  CS.   When   the   total   (monomeric   +   oligomeric)   sugar   yields   were   plotted   against   the   lignin   content   on   a   cell   wall   basis,   there   was   some   correlation   between   xylose   release   and   lignin   content  (Figure  3.3),  but  it  was  not  statistically  significant.  However,  there  was  a  stronger  and   more   significant   correlation   between   glucose   release   and   lignin   content,   which   increased   considerably  when  sample  E60  was  removed  from  the  analysis.       3.4.  Discussion   Old  field  mixed-­‐species  samples  are  considered  unsuitable  for  biofuel  production  due  to   low   predicted   yields   per   hectare   that   cause   the   cost   to   the   farmer   to   become   prohibitively   expensive  [136].  Additionally,  as  harvestable  yield  per  unit  area  decreases,  the  collection  radius   for  the  biorefinery  increases,  leading  to  significantly  higher  transportation  costs.  The  large-­‐scale   biomass  yields  from  the  old  field  samples  can  be  extrapolated  from  the  small-­‐scale  sample  data   to  estimate  the  biomass  yields  for  a  larger  harvested  area;  however  these  numbers  should  be   viewed   with   caution   due   to   the   potential   for   over-­‐   or   under-­‐estimation   of   yields.   The   early   63       successional   old   field   replicates   had   rather   high   extrapolated   yields   (6.6-­‐10.1   Mg/ha)   that   are   promising  but  would  probably  not  be  sustainable  over  multiple  years  without  additional  inputs.   Additionally,  the  yields  may  be  high  because  the  GLBRC  intensive  plot,  while  unfertilized,  was   previously  planted  to  alfalfa,  and  there  may  have  been  residual  nitrogen  remaining  in  the  soil   due   to   nitrogen   fixation.   The   key   species   that   grew   in   the   old   field   plots,   particularly   Amaranthus   retroflexus   L.   (redroot   pigweed)   and   Chenopodium   album   L.   (common   lambsquarters)   are   extremely   responsive   to   soil   nitrogen   levels   and   are   highly   efficient   in   nitrogen  uptake  [155].  Any  fixed  nitrogen  remaining  from  the  alfalfa  would  have  been  readily   utilized  by  these  plants,  increasing  production  of  biomass,  assuming  the  presence  of  adequate   amounts  of  other  potentially  limiting  nutrients.  The  late  successional  (SF)  replicates,  harvested   from  sites  which  have  been  abandoned  and  unmanaged  for  ~50  years,  may  give  a  more  realistic   estimate  of  long  term  yields  from  fields  which  contain  very  few  nitrogen-­‐fixing  species  (0.3-­‐2.6   Mg/ha,   extrapolated).   The   biomass   yields   from   all   three   late   successional   (SF)   replicates   have   continually   decreased   since   1993,   particularly   SF2   and   SF3   which   have   become   dominated   by   woody  species  (>70%  of  the  total  biomass)  [146].     For   mixed-­‐species   feedstocks,   the   relationship   between   glucose   and   xylose   yields   for   different   pretreatment   conditions,   the   sugar   conversion   efficiencies   (a   measure   of   feedstock   digestibility),  and  the  in  vitro  digestibility  were  related  to  the  grass  content  the  sample.  Based   on   the   results   shown,   the   old   field   replicates   can   be   categorized   into   either   the   grass-­‐ dominated   samples   (≥   60%   grass   species   on   a   mass   basis):   E60,   E83,   and   E87;   or   the   forb-­‐ dominated  samples  (≤  20%  grass  species  on  a  mass  basis):  E7  and  E21.  The  differences  observed   between   the   two   groups   could   be   explained   by   the   fact   that   grasses   and   their   relatives   64       (commelinids)  and  forbs  (non-­‐commelinids)  have  very  different  cell  wall  chemistries,  including   the   type,   structure,   and   localization   of   hemicellulose   as   well   as   lignin   monomer   content   and   composition  [17,  18,  27,  156-­‐158].  Release  of  xylose  from  the  cell  wall  of  grasses  [150]  or  grass-­‐ dominated   samples   is   strongly   correlated   to   increases   in   glucose   yields.   However,   the   relationship   was   not   as   strong   for   the   forb-­‐dominated   samples.   This   may   indicate   that   compared  to  the  grass  cell  wall,  the  release  of  glucose  from  the  forb  cell  wall  depends  more  on   other   factors   than   on   the   presence   or   absence   of   xylan.   This   is   supported   by   research   on   dilute   acid   pretreatment,   which   selectively   removes   xylan   from   the   cell   wall   [159].   Dien   et   al.   [13]   observed   that   a   dilute   acid   pretreated   forb   (alfalfa   stems)   had   a   lower   glucan   conversion   efficiency   that   was   not   dependent   on   the   release   of   non-­‐glucan   sugars,   compared   to   two   pretreated   grasses.   They   hypothesized   that   the   stem   cellulose   that   was   not   digested   may   be   more   closely   associated   with   lignin,   which   unlike   for   commelinids,   is   not   evenly   distributed   within  the  non-­‐commelinid  cell  wall  [160,  161].     It  is  well  documented  that  total  lignin  content  is  negatively  correlated  with  both  rumen   degradability  [162,  163]  and  enzymatic  saccharification  [164,  165].  However,  it  is  important  to   note  that  lignin  content,  while  correlated  with  digestibility,  is  in  itself  not  a  sufficient  indicator   for   biomass   digestibility,   particularly   if   one   is   interested   in   total   sugars.   Based   on   our   results,   the   cell   wall   lignin   content   was   strongly   negatively   correlated   with   glucose   digestibility,   but   not   xylose  digestibility,  and  the  correlation  was  stronger  when  E60  was  not  included  in  the  analysis.   This   is   interesting   because   there   are   indications   that   there   is   something   different   about   E60   compared   to   the   other   samples   that   were   tested.   E60,   which   is   the   most   digestible   sample   and   contains  nearly  60%  grass,  and  E21,  which  is  significantly  less  digestible  and  contains  only  21%   65       grass,   have   statistically   identical   Klason   lignin   contents   (Table   3.1).   The   large   difference   in   digestibility   between   these   samples   could   be   attributed   to   differences   between   forbs   and   grasses,   such   as   lignin   distribution   within   the   cell   wall,   or   there   might   be   some   unique   characteristic  about  E60  that  makes  it  more  responsive  to  pretreatment  compared  to  the  other   samples.  E60  had  the  largest  increase  in  glucose  release  between  the  untreated  material  and   the   optimally   pretreated   materials   (45%   increase   versus   34-­‐39%   for   the   other   old   field   replicates).  E60  is  also  the  only  sample  that  experienced  a  large  increase  in  rumen  digestibility   TM because   of   AFEX   pretreatment   (Table   3.3).   Another   possibility   is   that   there   are   different   types   and   quantities   of   covalent   linkages   between   cellulose   and   lignin   in   the   different   plants.   Recent   evidence   indicates   the   presence   of   oxygen-­‐containing   linkages   (ether   or   ester)   between   lignin  and  cellulose  in  corn  leaves  [166].  Ester  linkages  are  known  to  be  more  readily  cleaved   under  alkaline  conditions  [19],  so  if  more  of  these  were  present  linking  the  cellulose  and  lignin   in  the  E60  sample,  then  this  might  partially  explain  an  increase  in  digestibility.     At   the   same   pretreatment   conditions   and   enzyme   loading,   the   high   grass   content   samples,   CS,   E87,   E83,   E60,   and   LSF,   had   statistically   identical   sugar   conversions   (g   sugar   released/g  sugar  theoretically  available),  approximately  80%  of  the  total  sugars.  Woody  species   have  previously  been  shown  to  result  in  lower  yields  compared  to  cereal  residues  [14],  but  their   presence   in   the   LSF   sample   did   not   reduce   the   digestibility.   This   was   largely   because   a   high   xylose  yield  compensated  for  a  lower  glucose  yield  (Figure   3.2).  Glucose  yields  for  the  mixed-­‐ species  samples  ranged  from  170  –  300  g/kg  biomass  and  total  sugars  ranged  from  290  –  470   g/kg  biomass.  This  is  in  comparison  to  the  glucose  and  total  sugars  released  from  corn  stover:   340  and  580  g/kg  biomass,  respectively.  While  the  corn  stover  had  the  highest  sugar  yields,  this   66       was   not   because   it   was   significantly   more   digestible   than   the   other   materials,   but   rather   because  it  had  a  significantly  higher  cell  wall  sugar  content.  This  may  be  partly  related  to  the   relative   maturity   of   the   samples.   Even   though   they   were   harvested   around   the   same   time   of   the  year  (albeit  in  different  locations),  the  corn  stover  may  have  been  more  mature  than  the   old  field  replicates,  because  the  cell  wall  characteristics  (high  glucan,  xylan  and  lignin  with  low   ash  and  soluble  sugars)  were  typical  of  a  more  mature  plant  [158].  If  the  old  field  samples  had   been  allowed  to  mature  further  and  harvested  in  October  or  November,  it  is  possible  that  the   structural  sugar  content  and  subsequent  sugar  mass  yields  would  have  been  higher.     TM AFEX   pretreatment   has   been   examined   as   a   possible   treatment   for   forages   due   to   its   similarity   to   the   ammonia   treatment   that   has   been   used   by   farmers   for   decades   to   increase   TM digestibility  of  ruminant  feeds  [123].  The  increase  in  digestibility  we  observed  due  to  AFEX   pretreatment   of   the   mixed   species   feedstocks   is   quite   similar   to   that   observed   by   Bals   et   al.   for   the  more  readily  digestible  forages  [123].  As  they  concluded,  it  seems  unlikely  that  the  increase   in   digestibility   observed   for   these   types   of   materials   is   large   enough   to   warrant   the   use   of   pretreatment.   It   may   be   that   if   the   mixed   species   feedstocks   we   examined   were   allowed   to   mature   further   and   harvested   in   late   October   or   November,   the   recalcitrance   would   increase   and  there  would  be  more  of  an  effect  by  pretreatment  on  digestibility.   From   the   biorefinery   standpoint,   grass-­‐dominated   feedstocks   will   likely   be   more   profitable  because  of  their  higher  mass  sugar  yields  that  are  directly  related  to  a  higher  cell  wall   sugar  content  and  greater  digestibility.  C4  grasses  have  previously  been  predicted  to  produce   the   greatest   amount   of   ethanol   per   hectare   [142].   It   might   be   possible   to   manage   mixed-­‐ 67       species  sites  to  have  a  majority  of  grass  species,  which  could  have  a  beneficial  impact  on  the   biomass   yield,   but   this   would   likely   come   with   some   cost   to   the   species   diversity   [142].   Additionally,  although  a  farmer  may  not  wish  to  replace  their  corn  crop  with  a  mixed-­‐species   feedstock  based  on  the  economic  considerations,  it  may  be  worthwhile  to  harvest  one  of  their   abandoned  fields  or  convert  it  to  a  mixed-­‐species  stand,  such  as  native  prairie  or  mixed  native   grasses,  for  use  as  a  biofuel  feedstock.  The  addition  of  a  small  amount  of  fertilizer  to  the  fields   and/or   the   incorporation   of   legumes   in   the   species   mixture   may   also   increase   yields   to   the   extent   that   the   fields   become   more   profitable   for   the   farmer,   while   still   limiting   the   cost   associated  with  inputs  and  fertilizer-­‐related  environmental  impacts.   Mixed-­‐species   feedstocks   have   been   considered   unsuitable   for   biofuel   production   because  of  their  inherently  heterogeneous  nature.  However,  no  lignocellulosic  feedstock  will  be   homogenous,  even  monocultures  of  intensively  selected  varieties.  This  is  because  the  feedstock   characteristics   and   digestibility   are   highly   dependent   on   the   fraction   of   the   plant   considered   (leaf,  stem,  etc.),  maturity,  location,  and  environmental  conditions  experienced  during  growth   [29,  150,  167].  Large  biorefineries,  because  of  transportation  costs  and  supply  limitations,  will   need  to  accept  a  wide  variety  of  feedstocks  and  have  a  method  to  quickly  determine  their  value.   However,   it   may   be   feasible   to   process   most   feedstocks   at   similar   conditions   to   obtain   the   highest   yields.   Four   of   the   five   feedstocks   gave   fairly   high   yields   at   the   chosen   operating   condition.    It  is  very  likely,  given  the  example  of  E21,  that  there  will  be  some  feedstocks  that  do   not   process   as   well   at   the   chosen   condition.   The   operator   will   need   to   decide   which   is   more   cost   effective:   changing   operating   conditions   for   different   materials   to   maximize   yields,   operating   at   the   same   conditions   for   all   feedstocks   and   potentially   losing   yields   for   some   of   68       them,   or   perhaps   blending   feedstocks   to   make   up   for   deficiencies   in   low-­‐yielding   materials   and   improve  overall  process  stability. While   dedicated,   managed   monocultures   will   likely   give   higher   yields   and   be   more   reliable   once   established,   polycultures   can   be   equally   digestible,   and   could   be   used   as   a   supplemental   feedstock.   By   implementing   this   approach,   the   biorefinery   can   diversify   their   feedstock   supply   and   increase   source   security,   while   simultaneously   decreasing   the   collection   area   and   reducing   transportation   costs.   At   the   same   time,   mixed-­‐species   feedstocks   provide   more   valuable   ecological   services   compared   to   monocultures   and   have   much   to   offer   the   lignocellulosic  biorefinery  and  surrounding  communities  and  landscapes.     69       CHAPTER  4 :       TM   SPECIES  SCALE:  OPTIMIZATION  OF  AFEX  PRETREATMENT  CONDITIONS  AND  ENZYME   MIXTURES  TO  MAXIMIZE  SUGAR  RELEASE  FROM  UPLAND  AND  LOWLAND  SWITCHGRASS   4.1.  Introduction   Switchgrass  (Panicum  virgatum  L.)  is  a  perennial  C4  grass,  native  to  the  Great  Plains  of   North  America.  Experimental  research  on  switchgrass  breeding  has  been  conducted  since  the   1970s,   with   most   currently   available   cultivars   developed   for   improved   forage   qualities   [168].   After   evaluating   35   potential   herbaceous   crops,   the   US   Department   of   Energy   (DOE)   chose   switchgrass  as  one  of  the  promising  species  for  bioenergy  production  because  of  its  potential   for   high   yields,   wide   range   of   distribution,   and   beneficial   environmental   characteristics   [169,   170].   Switchgrass   cultivars   can   be   broadly   classified   according   to   their   cytotype   (upland   vs.   lowland)   and   latitude-­‐of-­‐origin   (southern   vs.   northern).   Lowland   varieties   are   tetraploid   and   tend   to   be   taller,   coarser,   and   have   thicker   stems   and   wider   leaves   compared   to   upland   varieties,   which   are   either   tetraploid   or   octaploid   [168,   171].   Typically,   upland   cultivars   have   higher   biomass   yields   and   survivability   at   northern   latitudes,   while   lowland   cultivars   perform   better   at   more   southern   latitudes   [172,   173].   Cytotype,   latitude-­‐of-­‐origin,   and   location   significantly  impact  biomass  yield,  survivability,  cell  wall  composition,  and  in  vitro  digestibility  of   different  switchgrass  varieties  [172,  174].     One   method   to   convert   switchgrass   to   a   usable   transportation   fuel   is   through   biochemical  conversion  of  the  biomass  structural  sugars  to  a  liquid  fuel,  such  as  ethanol.  This   process   uses   a   mechanical   or   thermochemical   pretreatment   to   disrupt   biomass   structure,   followed  by  enzymatic  hydrolysis  of  the  structural  carbohydrates  to  fermentable  sugars.  While   there   have   been   many   recent   publications   covering   a   range   of   pretreatment   options   for   70       biochemical  conversion  of  switchgrass,  only  three  papers  have  compared  multiple  switchgrass   cultivars   and/or   cytotypes   [150,   175,   176].   It   is   possible   that   different   cultivars   could   have   different  optimal  pretreatment  parameters  and  enzyme  combinations  for  hydrolysis.  However,   because   of   the   strong   effect   of   location   [172]   and   harvest   timing   [177]   on   switchgrass   phenotype,   it   is   necessary   to   grow   and   harvest   the   different   varieties   under   as   similar   conditions   as   possible   to   accurately   determine   which   differences   are   specifically   tied   to   the   cultivar,  and  not  differences  in  environment  or  harvest  timing.   One   way   to   improve   the   optimization   of   biofuel   processing   parameters   is   by   using   statistical   design   of   experiments.   The   benefit   of   statistical   methods   compared   to   the   one-­‐ factor-­‐at-­‐a-­‐time  approach  is  that,  for  a  comparable  number  of  experiments,  statistical  methods   provide   more   detailed   information   including   the   interactions   between   process   variables.   Examples   of   statistical   methods   include   factorial   designs,   which   are   best   suited   for   determining   which  factors  significantly  impact  the  process  output;  response  surface  optimization,  which  is   useful  for  predicting  the  optimum  process  design  points  given  a  target  endpoint;  and  mixture   optimization,  which  is  useful  for  optimizing  the  ratio  of  components  in  a  mixture  given  a  desired   output  [178].  Statistical  optimization  methods  have  been  used  to  characterize  various  steps  of   the   bioconversion   process   including   optimization   of   pretreatment   parameters   [179,   180],   enzyme  combinations  [181,  182],  enzymatic  hydrolysis  and  fermentation  parameters  [183],  and   fermentation  media  formulations  [184].   Our   objective   for   this   project   was   to   use   a   response   surface   optimization   method   to   TM determine   and   compare   the   optimal   ammonia   fiber   expansion   (AFEX )   pretreatment   parameters   for   two   switchgrass   varieties,   one   from   each   cytotype,   Alamo   (lowland)   and   71       Shawnee   (upland)   that   had   the   same   harvest   timing   (December)   in   the   same   region   (central   Oklahoma).   The   optimally   pretreated   switchgrass   was   then   used   to   determine   the   optimum   combination  of  commercial  enzymes  (Spezyme®  CP,  Novozyme®  188,  Multifect®  Xylanase  and   Multifect®  Pectinase)  for  each  variety  using  a  mixture  optimization  design.       4.2.  Materials  and  methods   4.2.1.  Feedstock   Two  switchgrass  varieties,  Alamo,  an  upland  ecotype,  and  Shawnee,  a  lowland  ecotype,   were   provided   by   Ceres,   Inc.   (Thousand   Oaks,   CA).   Both   varieties   were   planted   in   June   2005;   the  Alamo  switchgrass  in  Ardmore,  OK  (34°N,  Elev.  870  ft.)  and  the  Shawnee  in  Stillwater,  OK   (36°N,   Elev.   960   ft.),   and   both   were   harvested   in   December   2006.   Following   harvest,   the   switchgrass   was   air-­‐dried   to   less   than   10%   moisture   and   then   milled   through   a   2   mm   screen   using  a  standard  Wiley  mill  (Thomas  Scientific,  Swedesboro,  NJ).  Samples  were  stored  at  room   temperature  until  composition  analysis,  pre-­‐washing,  or  pretreatment  were  performed.     4.2.2.  Pre-­‐wash   A   pre-­‐wash   step   was   performed   to   remove   any   soluble   sugars   that   could   mask   the   solubilization   of   cell   wall   sugars.   100   g   of   switchgrass   was   soaked   in   1   L   of   80-­‐90°C   distilled   water  for  10-­‐15  min.    The  switchgrass  slurry  was  vacuum-­‐filtered  through  Whatman  No.  1  filter   paper   (Whatman   Ltd.).   This   process   was   repeated   three   times   and   after   each   wash   step   a   72       portion  of  the  filtrate  was  retained  for  oligomeric  sugar  analysis.  The  washed  solids  were  dried   in  a  45°C  oven.  The  extracted  weight  loss  of  the  switchgrass  was  determined  by  subtracting  the   dry  weight  of  the  washed  switchgrass  and  the  dry  mass  loss  to  the  filter  paper  from  the  initial   dry   weight.   Washed   switchgrass   was   used   for   the   hydrolysis   rate   determination   and   enzyme   mixture  experiments,  but  not  for  the  pretreatment  response  surface  experiments.     4.2.3.  Composition  analysis   Biomass  moisture  content  was  determined  using  a  moisture  analyzer  (A&D,  Model  MF-­‐ 50;  San  Jose,  CA).  The  composition  of  each  sample  (extractives,  ash,  lignin,  glucan,  and  xylan)   was   determined   according   to   the   standard   National   Renewable   Energy   Laboratory   (NREL)   protocol  that  uses  a  two-­‐stage  extraction  followed  by  two-­‐step  acid  hydrolysis  [147].  The  acid-­‐ insoluble  lignin  analysis  method  was  modified  to  use  47  mm,  0.22  μm  pore-­‐size  mixed-­‐cellulose   ester   filter   disks   (Millipore   Corp.,   Bedford,   MA)   during   the   filtration   step   instead   of   fritted   crucibles.  The  filtered  lignin  residues  were  dried  overnight  in  a  desiccator  prior  to  weighing.  The   nitrogen   content   of   the   extracted   and   unextracted   samples   were   determined   via   the   combustion  method  for  nitrogen  determination  [148]  using  a  Skalar  Primacs  SN  Total  Nitrogen   Analyzer   (Breda,   The   Netherlands).   Nitrogen   values   were   multiplied   by   6.25   to   determine   the   crude  protein  content.  This  conversion  factor  assumes  that  16%  of  the  protein  is  nitrogen  and   that   there   is   negligible   non-­‐protein   nitrogen   present   in   the   biomass.   This   factor   varies   with   different   types   of   plant   samples   due   to   differences   in   protein   structure,   however   accurate   determination  of  this  factor  requires  a  complete  amino  acid  analysis  [149].  As  protein  content   does   not   directly   impact   our   results,   we   assume   that   the   standard   factor   of   6.25   allows   for   a   73       reasonable   approximation.   Protein   that   was   removed   during   the   extraction   steps   was   subtracted   from   the   total   extractives   content.   The   composition   data   for   unwashed   and   washed   Alamo  and  Shawnee  switchgrass  are  listed  in  Table  4.1.     Table  4.1:  Composition  analysis  data  for  untreated  Alamo  and  Shawnee  switchgrass   (%  of  total  dry  biomass).  Washed  switchgrass  samples  had  been  sequentially  washed   three  times  with  80-­‐90°C  water  in  order  to  remove  the  majority  of  the  soluble  sugars.   Values  with  different  superscripts  in  each  row  were  statistically  different  based  on   Tukey’s  pair-­‐wise  comparisons  with  α  =  0.05.             Alamo   b 21.3    ±  0.7     c 3.1    ±  0.1     d 17.4    ±  0.1     a a   6.4    ±  0.2     b 3.3    ±  0.2     a 2.0    ±  0.1   b 15.5    ±  0.9   Alamo   1.9    ±  0.2       Xylan   21.2    ±  0.3     Arabinan   3.2    ±  0.1     Klason  Lignin   15.4    ±  0.1     Total  Ash   5.3    ±  0.1     Protein   6.5  ±  0.2   Soluble  Glucose   2.6    ±  0.2   Sucrose   5.2    ±  0.4   Acetyl   Other  Extractives   c Washed   5.3    ±  0.1     27.3    ±  0.5   Extractives     30.2    ±  0.5     Glucan   Total   Shawnee   d       Unwashed   b c c Shawnee   b 34.8    ±  0.3   a 26.3    ±  0.3   a 4.6    ±  0.1   b a 37.1    ±  1.2   a 25.0    ±  0.8   b 4.2    ±  0.1   a   18.9    ±  0.3   20.8  ±  0.3   n.d.   n.d.   6.0    ±  0.1   4.7    ±  0.1   a 0.7    ±  0.0*   0.8    ±  0.0*   b n.d.   n.d   2.0    ±  0.0     2.2    ±  0.1   2.0    ±  0.0   13.0    ±  0.4     0   0   103.9         93.1   93.9   a a b 104.2       b a c b n.d.  =  not  determined;    *Soluble  monomeric  glucose  in  the  acid-­‐hydrolyzed  water   extract  –  includes  both  soluble  glucose  and  glucose  contributed  by  sucrose.     74       4.2.4.  Design  of  experiments   4.2.4.1.  Response  surface  optimization  of  pretreatment  conditions   TM AFEX   conditions   for   the   pretreatment   optimization   experiments   were   chosen   using   a   Box-­‐Behnken  statistical  design  and  the  parameters  ranged  from  0.5  to  2.0  g  NH3:g  dry  matter   (DM),  0.5  to  2.0  g  H2O:g  DM,  90  to  180°C  and  5  to  30  min  residence  time.  The  central  design   point   was   conducted   in   triplicate   to   give   a   total   of   27   experiments   for   both   the   Alamo   and   Shawnee.  In  an  attempt  to  improve  the  fit  of  the  statistical  models,  five  additional  experiment   points  were  tested  for  Alamo  and  three  for  Shawnee  (Table   C.1).  Enzymatic  hydrolysis  was  then   conducted  on  the  samples  as  described  in  Section  4.2.6.    The  enzymatic  hydrolysis  sugar  yields   (g   sugar   released   per   g   sugar   theoretically   present   in   the   untreated,   dry   biomass)   were   used   as   the  metric  of  pretreatment  efficacy.   TM The   following   polynomial   quadratic   equation   of   the   AFEX   pretreatment   conditions   was   fitted   to   the   enzymatic   hydrolysis   total   sugar   yield   data   (combined   monomeric   and   oligomeric   glucose   and   xylose   in   terms   of   the   theoretical   maximum)   using   Minitab15   Statistical   Software  (2006  Minitab  Inc.,  Pennsylvania,  USA):   n   n n 2+ Y = a0 + ∑ ai xi + ∑ aii xi ∑ aij xi x j   i=1 i=1 i, j=1 i≠ j (4.1)   where  Y  is  the  sugar  yield;  a0  is  the  regression  constant;  ai  is  the  linear  regression  coefficient  for   the   ith   parameter;   aii   is   the   quadratic   regression   coefficient   for   the   ith   parameter;   aij   is   the   75       interaction  coefficient  for  the  ith  and  jth  parameters;  xi  and  xj  are  the  values  of  the  ith  and  jth   parameters;   and   n   is   the   number   of   factors,   which   in   this   case   is   4.   Coefficients   with   p   >   0.1   were   removed   stepwise   from   the   model,   beginning   with   the   largest   p-­‐value.   The   model   was   TM then  optimized   for   the   AFEX   conditions  that  gave   the   highest   composite   desirability   for   both   glucose   and   xylose   (monomeric   +   oligomeric)   sugar   yields   following   enzymatic   hydrolysis.   Contour   plots   were   generated   that   show   the   effect   of   pairs   of   pretreatment   parameters   on   glucose   and   xylose   yields,   with   the   remaining   two   parameters   in   each   figure   held   at   their   optimum  levels.     4.2.4.2.  Mixture  optimization  of  hydrolysis  enzymes   Minitab   was   also   used   to   create   and   analyze   a   mixture   optimization   experiment   to   determine   the   optimal   combination   of   β-­‐glucosidase   (Novozyme®   188,   Novozymes   Corp.,   Bagsværd,   Denmark),   Spezyme®   CP,   Multifect®   Xylanase,   and   Multifect®   Pectinase   (Genencor   Division   of   Danisco   US   Inc.,   New   York,   USA)   for   release   of   sugars   from   optimally   pretreated   Alamo  and  Shawnee  switchgrass.  For  this  experiment,  an  extreme  vertices  design  with  a  design   degree   of   three   was   generated,   which   included   the   four   enzymes   and   two   enzyme   loadings   (15   and   30   mg   total   protein   per   g   glucan).   The   constraints   on   the   relative   enzyme   proportions   in   terms   of   total   protein   were:   Spezyme®   CP   ≥   20%,   Novozyme®   188   ≤   50%,   Multifect®   Xylanase   ≤   80%,   and   Multifect®   Pectinase   ≤   80%.   For   this   design,   only   vertices   (type   1)   and   edge   midpoints   (type   2)   were   replicated,   resulting   in   84   total   design   points,   including   replicates.   A   76       regression  model  was  generated  for  each  sugar  yield  (monomeric  and  total  glucose,  xylose,  and   glucose  +  xylose)  in  Minitab  from  the  hydrolysis  yield  data:   n n n Y = ∑ ai xi + ∑ aij xi x j + ∑ aijk xi x j xk i=1 i, j=1 i, j,k=1 i≠ j i≠ j≠k # &   % ( n n n % ( + z % b0 + ∑ bi xi + ∑ bij xi x j + ∑ bijk xi x j xk ( % ( i=1 i, j=1 i, j,k=1 % ( i≠ j i≠ j≠k $ '  (4.2)   where  Y  is  the  sugar  yield;  ai  is  the  linear  regression  coefficient  for  the  ith  component;  aij   is  the   quadratic   interaction   coefficient   for   the   ith   and   jth   components;   aijk   is   the   cubic   interaction   coefficient  for  the  ith,    jth,  and  kth  components;  xi,  xj,  and  xk  are  the  values  of  the  ith,  jth  and   kth  parameters;  n  is  the  number  of  components,  in  this  case  4;  z  is  the  enzyme  loading;  b0  is   the   enzyme   loading   coefficient;   and   bi,   bij,   bijk   are   the   linear,   quadratic,   and   cubic   enzyme   loading   interaction   regression   coefficients,   respectively.     The   coefficients   in   the   regression   models  were  selected  stepwise,  beginning  with  the  four  linear  enzyme  terms  and  sequentially   adding  terms  with  α  <  0.05  and  removing  terms  with  α  >  0.05.  The  regression  model  was  then   used   to   predict   the   optimum   mixture   composition   for   each   switchgrass   variety   and   generate   contour  plots  showing  the  effect  of  enzyme  combinations  on  sugar  yields.     77       TM 4.2.5.  AFEX  pretreatment   TM AFEX   pretreatment   of   the   Alamo   and   Shawnee   switchgrass   for   the   pretreatment   optimization   experiments   was   conducted   in   22   mL   reactors   as   outlined   by   Bals   et   al.   [150].   Conditions   used   for   these   experiments   are   detailed   earlier   in   the   design   of   experiments   section   and   listed   in   the   supplemental   information   (Table   C.1).   For   the   enzyme   mixture   optimization   TM and   hydrolysis   rate   determination   experiments,   AFEX   was   performed   on   Alamo   and   Shawnee   switchgrass   in   a   300   mL   reactor   as   detailed   by   Kim   et   al.   [175].   The   pretreatment   conditions   used   for   each   switchgrass   variety   were   chosen   based   on   the   results   of   the   pretreatment  response  surface  optimization:  Alamo:  1.5  g  NH3:g  DM,  2.0  g  H2O:g  DM,  140°C,   20  min;  Shawnee:  1.5  g  NH3:g  DM,  2.0  g  H2O:g  DM,  150°C,  30  min.       4.2.6.  Enzymatic  hydrolysis   Samples   for   pretreatment   optimization   and   enzyme   mixture   optimization   were   hydrolyzed  in  20  mL  screw-­‐cap  vials  at  1%  glucan  loading  and  a  total  volume  of  15  mL.  For  the   rate  determination  experiments,  samples  were  hydrolyzed  in  250  mL  Erlenmeyer  flasks  at  1%   glucan  loading  and  a  total  volume  of  100  mL.  All  samples  were  adjusted  to  a  pH  of  4.8  using  1  M   citrate   buffer   solution.   To   prevent   fungal   and   bacterial   contamination   during   enzymatic   hydrolysis,  cycloheximide  and  tetracycline  were  loaded  at  a  final  concentration  of  30  µg/mL  and   40  µg/mL,  respectively.     78       For  the  pretreatment  optimization  experiments,  the  standard  enzyme  loading  was  used:   15   filter   paper   units   (FPU)   Spezyme®   CP   and   30   cellobiase   units   (CBU)   β-­‐glucosidase   per   g   glucan   in   the   untreated   biomass,   or   27.33   mg   total   protein   per   g   glucan.   For   the   enzyme   mixture   optimization,   all   four   enzymes   were   loaded   in   combinations   as   determined   by   the   experimental  design  at  two  different  protein  loadings  (15  mg  protein  per  g  glucan  and  30  mg   protein  per  g  glucan).  For  the  rate  determination  experiments,  two  different  enzyme  mixtures   were   compared:   15   FPU   Spezyme®   CP   and   30   CBU   Novozyme®   188   per   g   glucan   in   the   untreated   substrate,   and   the   optimal   enzyme   loading   determined   from   enzyme   mixture   optimization  loaded  at  27  mg  protein  per  g  glucan.  The  protein  content  and  enzyme  activity  for   each   of   the   enzymes   are   as   follows,   where   known:   Spezyme®   CP   (82   mg   protein/mL,   50   FPU/mL),   Novozyme®   188   (67   mg   protein/mL,   735   CBU/mL),   Multifect®   Xylanase   (27   mg   protein/mL)   and   Multifect®   Pectinase   (52   mg   protein/mL).   Protein   content   was   determined   from   total   N   analysis   using   the   Dumas   method   for   combustion   of   nitrogen   to   NOx   [148]   following  trichloroacetic  acid  (TCA)  precipitation  to  remove  non-­‐protein  nitrogen  [151].   Enzymatic  hydrolysis  for  all  experiments  was  conducted  in  a  shaking  incubator  at  50°C   and  200  rpm.  For  the  optimization  experiments,  samples  were  taken  at  72  hours  of  enzymatic   hydrolysis   for   both   monomeric   and   oligomeric   sugar   analysis,   as   detailed   below.   For   the   rate   determination  experiments,  samples  were  taken  at  1  h,  24  h,  and  168  h  for  monomeric  sugar   analysis.   Samples   taken   for   monomeric   sugar   analysis   were   heated   at   100°C   for   15-­‐20   min,   cooled   in   the   freezer,   and   then   centrifuged   at   15,000g   for   5   min.     The   supernatant   was   filtered   79       into   HPLC   shell   vials   using   a   25   mm,   0.2   μm   polyethersulfone   syringe   filter   (Whatman   Inc.   Florham  Park,  NJ)  then  stored  at  -­‐20°C  until  further  sugar  analysis.       4.2.7.  Soluble  total  and  oligomeric  sugar  analysis   Oligomeric   sugar   analysis   of   the   pre-­‐wash   liquid   and   hydrolysate   was   conducted   using   a   scaled-­‐down   version   of   the   standard   NREL   method   for   oligomeric   sugar   determination   of   liquid   streams  [152].  The  modified  method  was  identical  except  that  it  was  scaled  down  to  use  2  mL   of   sample,   which   were   run   in   duplicate   in   10   mL   screw-­‐cap   culture   tubes.   Instead   of   being   autoclaved,  the  tubes  were  incubated  in  a  121°C  bench-­‐top  hot  plate  for  one  hour,  cooled  on   ice,   and   the   liquid   was   filtered   into   HPLC   vials.   The   oligomeric   sugar   concentration   was   determined  by  subtracting  the  monomeric  sugar  concentration  of  the  non-­‐hydrolyzed  samples   from  the  total  sugar  concentration  of  the  acid  hydrolyzed  samples.     4.2.8.  HPLC  analysis   Sugar   contents   of   all   composition   analysis,   wash   liquid,   and   hydrolysate   samples   were   determined   using   a   Bio-­‐Rad   (Hercules,   California,   USA)   Aminex   HPX-­‐87H   column   equipped   with   appropriate   guard   columns.     Degassed   5   mM   H2SO4   was   used   as   the   mobile   phase   and   the   column   temperature   was   held   at   60°C.     Glucose,   xylose   (plus   galactose   and   mannose),   and   arabinose   concentrations   were   determined   for   each   liquid   stream.     Because   the   xylose,   galactose,   and   mannose   peaks   cannot   be   separated   using   the   HPX-­‐87H   column   [185],   any   results  reported  for  xylose  also  includes  mannose  and  galactose.  For  grasses  the  galactose  and   80       mannose  contents  tend  to  be  very  low  –  in  sum  less  than  1.5%  of  the  total  biomass  [186].  Sugar   yields   were   calculated   as   reported   previously   [30],   taking   into   account   soluble   glucose   and   sucrose  present  in  the  untreated  switchgrass.       4.3.  Results  and  discussion   4.3.1.  Switchgrass  characteristics   The   Alamo   and   Shawnee   switchgrass   used   for   these   experiments   did   not   exhibit   large   differences   in   composition   (Table   4.1).   Shawnee   had   a   statistically   higher   glucan   content   and   higher   lignin   content   compared   to   the   Alamo,   and   a   statistically   lower   total   extractives   content   (data  not  shown),  which  may  indicate  that  it  is  a  slightly  more  mature  sample  [158].  However   xylan,  ash,  and  protein  were  all  statistically  identical,  so  this  conclusion  is  not  certain.  Following   washing  of  the  switchgrass  to  remove  potentially  interfering  soluble  sugars,  the  overall  cell  wall   content   increased   as   expected,   and   the   composition   data   corresponded   to   expected   values   based   on   mass   balance   calculations   around   the   washing   step   (data   not   shown).   The   relative   differences  in  the  cell  wall  content  between  Alamo  and  Shawnee  did  not  change  to  any  large   extent.     A   previous   study   on   AFEX TM   pretreatment   of   switchgrass   cultivars   used   lowland   switchgrass   (Alamo)   that   was   harvested   in   Alabama,   and   upland   switchgrass   (Cave-­‐in-­‐Rock)   that   was   harvested   in   Michigan   [150].   This   difference   in   latitude   made   it   difficult   to   accurately   determine   whether   the   differences   in   the   results   were   due   to   the   difference   in   cytotype,   harvest  timing,  or  location.  For  our  experiments,  Alamo  and  Shawnee  were  both  planted  and   harvested   at   the   same   time   in   the   same   year   and   within   2°   latitude,   which   reduces   the   81       environmental   impacts   on   digestibility   compared   to   the   previous   work  [150].   Other   research   has   shown   that   when   grown   at   the   same   latitude   in   the   south   central   U.S.,   upland   varieties   tended   to   have   lower   cellulose   content   and   consistently   produced   lower   biomass   yields   compared   to   the   lowland   varieties   [174].   The   2°   higher   latitude   of   the   Shawnee   switchgrass   could  be  one  reason  for  Shawnee’s  higher  glucan  content  compared  to  the  Alamo.       Figure   4.1:   Relationship   between   enzymatic   hydrolysis   glucose   and   xylose   yields   for   Alamo   and  Shawnee  switchgrass.  (A)  Monomeric  sugar  yields.    (B)  Total  monomeric  +  oligomeric  sugar   yields.     Yields   are   calculated   as   the   total   sugar   solubilized   based   on   the   total   sugar   theoretically   available  in  the  untreated  dry  biomass.  Each  data  point  represents  one  of  30  (Shawnee)  or  32   (Alamo)  different  pretreatment  experiments.     4.3.2.  Pretreatment  parameter  optimization   The  monomeric  and  total  monomeric  +  oligomeric  sugars  for  each  of  the  pretreatment   experiments   are   shown   graphically   in   Figure   4.1.   Total   monomeric   +   oligomeric   glucose   and   xylose  yields  from  untreated  Shawnee  were  23.5%  and  19.8%,  respectively,  and  from  untreated   82       Alamo   were   25.3%   and   25.4%,   respectively.   In   all   cases,   pretreatment   increased   sugar   yields   compared   to   the   untreated   samples.   In   general   Alamo   tended   to   have   higher   sugar   yields   (%   sugar  released  of  sugar  present  in  untreated  biomass)  compared  to  the  Shawnee  switchgrass,   indicating  greater  digestibility.    In  some  cases,  oligomeric  xylose  accounted  for  as  much  as  30-­‐ 35%   of   the   total   xylose   solubilized   from   the   pretreated   biomass   during   enzymatic   hydrolysis.   TM For   AFEX   pretreated   grasses,   often   there   is   a   statistically   significant   linear   relationship   between  glucose  yields  and  xylose  yields.  As  increasing  amounts  of  xylose  and  xylo-­‐oligomers   are  released  from  the  biomass,  increasing  amounts  of  glucose  are  also  released  [150].  For  other   pretreatments   that   selectively   remove   xylan   from   the   plant   cell   wall,   such   as   dilute   acid,  glucan   conversion  in  grasses  can  also  be  dependent  on  release  of  hemicellulose  sugars  [13].   For   the   pretreatment   optimization,   regression   models   of   total   monomeric   and   oligomeric  glucose  and  xylose  yields  were  evaluated  in  terms  of  the  pretreatment  parameters   and  only  coefficients  with  α  <  0.10  were  included  in  the  final  model  (Table  4.2).    This  value  was   chosen   because   using   a   lower   value   such   as   0.05   resulted   in   the   removal   of   too   many   terms   from  the  regression  models  and  extremely  poor  representation  of  the  data.  Of  the  terms  in  the   models,   moisture   content   for   Alamo   switchgrass,   the   interaction   between   ammonia   and   residence   time   for   Shawnee,   and   the   residence   time   quadratic   term   for   Shawnee   were   less   2 significant  with  α  >  0.05.  The  models  adequately  describe  the  data  with  adjusted  R -­‐values  of   2 2 around   85%   for   both   regressions.   Unlike   the   standard   R   value,   adjusted   R -­‐values   take   into   2 account   the   number   of   terms   in   the   model   and   may   decrease   compared   to   the   standard   R -­‐   value   if   there   are   more   terms   than   necessary   to   describe   the   data.   The   deviation   of   values   83       predicted  by  the  model  from  the  actual  yield  values  for  each  of  the  pretreatment  experiments   is   shown   in   Figure   4.2.   The   histograms   show   the   number   of   experiments   that   deviated   by   a   certain  percent  yield.  Negative  values  indicate  that  the  model  underpredicted  the  sugar  yields   and  values  greater  than  one  indicate  that  the  model  overpredicted  the  sugar  yields.  In  general   the   Alamo   model   tended   to   underpredict,   while   the   Shawnee   model   tended   to   overpredict   yields,   however   both   curves   were   fairly   normal   in   distribution   with   a   couple   of   outliers.   For   Table  4.2:  Response  surface  optimization  of  pretreatment  parameters  in  terms  of  total   monomeric  and  oligomeric  glucose  and  xylose  release  following  enzymatic  hydrolysis.     A  =  ammonia  loading  (g  NH3:g  DM),  B  =  water  loading  (g  H2O:g  DM),  C  =  temperature  (°C),     D  =  residence  time  (min).     Alamo   a Term   Constant   A   B   C   D   2   b a Shawnee   b T   P     Coef.   T   P   Coef.   SE   SE   -­‐88.7919   19.517   -­‐4.550   0.000     -­‐54.1452   12.787   -­‐4.234   0.000   26.5272   6.081   4.362   0.000     18.5639   4.549   4.081   0.001   -­‐13.6733   7.568   -­‐1.807   0.085     3.3716   1.053   3.201   0.004   1.6561   0.211   7.865   0.000     1.1577   0.160   7.230   0.000   3.6793   0.578   6.363   0.000     1.2562   0.321   3.919   0.001   A   -­‐4.4631   1.540   -­‐2.897   0.009     -­‐3.2350   1.139   -­‐2.842   0.010   2 C   -­‐0.0057   0.001   -­‐7.443   0.000     -­‐0.0039   0.001   -­‐6.729   0.000   2 D   AD   BC   CD   -­‐0.0270   0.010   -­‐2.699   0.013     -­‐0.0130   0.007   -­‐1.751   0.094   0.180   -­‐2.253   0.035     0.054   2.290   0.032     0.004   -­‐3.753   0.001         89.22%           84.08%       -­‐0.2468   -­‐   -­‐   0.140   -­‐1.763   0.092   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   68.97%     -­‐0.4064   0.1239   -­‐0.0132   2 R     2 c R  (adj)   2   a d R (pred)         b 90.29%         86.59%         78.65%         c 2 2 d 2   Coef.  =  regression  model  coefficient;   SE  =  standard  error;   R  (adj)  =  adjusted  R -­‐value;   R 2 (pred)  =  predictive  R -­‐value   84         Figure   4.2:   Histograms   and   scatter   plots   showing   deviation   of   predicted   sugar   yields   from   actual   sugar   yields   for   the   pretreatment   regression   models.  (A)  Shawnee  histogram.  (B)  Alamo  histogram.  (C)  Shawnee  scatter   plot.  (D)  Alamo  scatter  plot.     both  switchgrass  samples,  over  50%  of  the  predicted  values  were  within  ±2%  of  the  actual  value   and   94%   of   the   predicted   yields   were   within   ±4%   and   ±5%   of   the   actual   sugar   yields   from   Shawnee  and  Alamo,  respectively.     The   regression   models   were   used   to   determine   the   optimum   pretreatment   conditions   for   both   Shawnee   and   Alamo   switchgrass   using   the   Minitab   response   surface   optimizer   for   maximizing  total  monomeric  +  oligomeric  yields  from  both  glucose  and  xylose.    The  optimum   condition   for   Shawnee   was   initially   determined   to   be   1.75   g   NH3:g   DM,   2.0   g   H2O:g   DM,   146°C   85       and  30  min  residence  time  and  for  Alamo  was  2.49  g  NH3:g  DM,  2.0  g  H2O:g  DM,  152°C  and   12.3   min   residence   time.   However,   due   to   constraints   on   the   300   mL   reactor   operation,   the   optimal   conditions   were   constrained   to   1.5   g   NH3:g   DM   and   the   other   parameters   were   adjusted  to  obtain  yields  as  close  as  possible  to  the  original  optimum  (Table  4.3).  The  optimal   pretreatment  conditions  for  Alamo  and  Shawnee  were  fairly  similar,  with  Alamo’s  optimum  at  a   slightly   lower   temperature   and   shorter   residence   time,   while   simultaneously   obtaining   higher   sugar   yields.   Compared   to   previously   determined   optima   for   AFEX TM   pretreatment   of   switchgrass,  the  values  determined  here  seem  closest  to  those  proposed  by  Bals  et  al.  [150]  for   Alamo   switchgrass.   The   optimal   values   found   by   Alizadeh   et   al.   [187]   were   quite   different   from   our  findings,  and  while  the  information  on  the  harvest  date  and  variety  for  those  experiments   were   not   provided,   the   optimum   pretreatment   conditions   were   similar   to   the   mild   optimum   pretreatment  conditions  for  early  harvest  Cave-­‐in-­‐Rock  [150].  This  may  indicate  that  the  initial   TM work  on  AFEX -­‐treated  switchgrass  also  used  a  fairly  immature  sample.  Dacotah  switchgrass,   which  was  used  for  the  CAFI  III  project  [30,  175,  188],  was  harvested  in  South  Dakota  in  May   following   over-­‐wintering   on   the   field.   This   sample   had   consistently   lower   glucose   yields   compared  to  Alamo  and  Shawnee  for  all  pretreatment  methods  tested  [175].    Contour   plots   of   monomeric   glucose   and   xylose   yields   (Figure  C.1  and  Figure  C.2)   were   nearly  identical  to  the  contour  plots  of  total  monomeric  +  oligomeric  glucose  and  xylose  yields   (Figure  4.3  and  Figure  4.4),  and  have  been  included  in  the  online  supplemental  information  for   reference.  The  contour  plots  show  that  the  optimum  pretreatment  conditions  for  total  xylose   release  corresponded  roughly  with  the  optimum  values  for  total  glucose  release.  The  optimum   86       TM Table  4.3:  Comparison  of  literature  on  optimal  AFEX  pretreatment  conditions  and  sugar  yields  from  switchgrass.  Predicted   yields  are  sugar  yields  as  predicted  by  the  response  surface  regression  model.  Values  in  parenthesis  represent  monomeric  +   oligomeric  sugars.  Pred.  =  predicted   TM       AFEX  Pretreatment   Switchgrass   Sugar  Yields   Conditions         Temp   Time     Glucose   Xylose   NH3   H2O       Harvest   Variety   Cytotype   Location   Timing       Pred.   Actual   Pred.   Actual   (g:   (g:   Ref.   (°C)   (min)   g  DM)   g  DM)   (%)   (%)   (%)   (%)   68.0   (75.4)   63.9     (67.3)   68.7   (93.3)   63.0   (82.1)   -­‐   69.3   (75.3)   62.0   (66.7)   52.7   (54.5)   75.2   -­‐   67.7   (97.9)   59.5   (79.1)   58.0   (79.9)   44.5     -­‐   44.6   -­‐   35.5   30     -­‐   51.8   -­‐   34.4   150   25     -­‐   57.3   -­‐   38.3   100   5     -­‐   ~80   -­‐   ~65   Alamo   Lowland   Oklahoma   Dec.     1.5   2.0   140   20   Shawnee   Upland   Oklahoma   Dec.     1.5   2.0   150   30   [188]   Dacotah   Upland   South   Dakota   May     1.5   2.0   150   30     -­‐   [150]   Cave-­‐in-­‐ Rock   Upland   Michigan   Jul.     0.9   0.4   80   20     Oct.     2.0   0.4   130   30   Alamo   Lowland   Alabama   Jul.     1.6   2.0   160   Oct.     2.0   2.0   -­‐     1.0   0.8   This   Study           [187] a -­‐   -­‐   -­‐   Same  enzymatic  hydrolysis  conditions  as  this  paper.     b c -­‐   d   Same  enzymatic  hydrolysis  conditions  except  biomass  loaded  at  3%  solids  and  3.2  FPU  Accelerase  per  g  DM  (~10  FPU  per  g  glucan).   Enzymatic  hydrolysis  using  15  FPU  Spezyme  CP  per  g  glucan  and  40  CBU  Sigma  β-­‐glucosidase  per  g  glucan.   d Included  a  30  minute  heat-­‐up. 87         TM Figure   4.3:   Contour   plots   showing   the   interactive   effect   of   pairs   of   AFEX   pretreatment   parameters   on   monomeric   +   oligomeric   glucose   yields   from   (A)   Alamo   and   (B)   Shawnee   switchgrass.  The  two  pretreatment  parameters  not  shown  in  each  sub-­‐figure  were  held  at  the   optimal  level.  Hydrolysis  was  conducted  at  50°C,  200  rpm,  and  1%  glucan  loading  using  30  FPU   Spezyme®  CP  and  15  CBU  Novozyme®  188  per  g  glucan,  with  72  h  sampling.   88         Figure   4.4:   Contour   plots   showing   the   interactive   effect   of   pairs   of   AFEX TM   pretreatment   parameters   on   monomeric   +   oligomeric   xylose   yields   from   (A)   Alamo   and   (B)   Shawnee   switchgrass.  The  two  pretreatment  parameters  not  shown  in  each  sub-­‐figure  were  held  at  the   optimal   level.   Hydrolysis   was   conducted   at   50°C,   200   rpm,   and   1%   glucan   loading   using   30   FPU   Spezyme®  CP  and  15  CBU  Novozyme®  188  per  g  glucan,  with  72  h  sampling.     89       ammonia   loading   tended   to   be   slightly   lower   for   xylose   release   compared   to   glucose   release,   which   in   most   cases   was   outside   the   charted   range.   Of   the   pretreatment   parameters,   water   loading  had  the  smallest  impact  on  both  Alamo  glucose  and  xylose  yields  and  Shawnee  xylose   yields,   although   all   pretreatments   showed   a   strong   interaction   between   water   loading   and   temperature,   with   the   highest   yields   at   moderate   temperatures   and   high   water   loading.   A   moderate   residence   time   (~20-­‐25   min)   resulted   in   higher   glucose   and   xylose   release   from   Alamo   switchgrass;   however,   the   optimum   for   Shawnee   may   actually   be   higher   than   30   min,   which  was  the  limit  of  the  parameters  tested.  However,  it  may  not  be  economically  desirable  to   operate   for   a   longer   residence   time,   as   this   increases   the   capital   cost   associated   with   the   pretreatment   reactor   for   the   same   amount   of   throughput   and   can   significantly   increase   the   minimum  ethanol  selling  price  (MESP)  [189].   While   it   is   apparent   that   solubilization   of   hemicellulose   is   important   for   increasing   TM glucose   yields   from   AFEX   treated   switchgrass,   even   when   operated   at   optimal   conditions,   the   pretreatment   was   still   insufficient   to   solubilize   100%   of   the   hemicellulose   or   obtain   greater   than  75%  glucose  yields  from  both  varieties  of  switchgrass.  While  hemicellulose  and  lignin  are   known   to   be   extracted   from   the   biomass   and   redeposited   on   the   cell   wall   surface   during   TM AFEX   pretreatment   [188,   190],   much   of   the   lignin   remains   insoluble   even   following   enzymatic   hydrolysis   [30].   It   seems   likely   that   this   lignin   will   still   have   a   portion   of   the   hemicellulose   associated   with   it   following   pretreatment,   which   would   be   rendered   inaccessible   to  enzymes.  This  lignin  may  also  impede  access  to  the  cellulose.  In  addition  to  lignin  content,   reduction   in   cellulose   degree   of   polymerization   is   also   important   for   improving   cellulose   90       TM conversion.  As  AFEX   is  known  to  not  influence  this  parameter  significantly  [191],  the  lack  of   cellulose  reducing  ends  may  also  hinder  enzymatic  hydrolysis.  One  possibility  that  is  currently   being   explored   is   to   pretreat   biomass   using   liquid   ammonia   with   very   little   water,   which   has   been  shown  to  generate  the  cellulose  IIII  crystalline  allomorph  [192].  This  crystalline  structure  is   more  readily  converted  by  enzymes  compared  to  native  cellulose  I  [193],  and  using  this  method   it  may  be  possible  to  further  increase  glucan  conversion  from  switchgrass.     4.3.3.  Commercial  enzyme  mixture  optimization   Sugar  yield  data  from  the  mixture  optimization  experiments  was  fitted  to  a  regression   model   for   both   Alamo   (Table  4.4)   and   Shawnee   (Table  4.5)   switchgrass.   The   only   terms   initially   included   in   the   model   were   the   four   base   commercial   enzyme   mixtures:   Spezyme®   CP   (S),   Novozyme®   188   (B),   Multifect®   Xylanase   (X)   and   Multifect®   Pectinase   (P).   New   terms   were   sequentially   added   (α   <   0.05)   or   removed   from   the   model   (α   >   0.05).   The   models   adequately   describe   the   data   with   adjusted   R2-­‐values   of   around   95%   for   all   sugar   yields   except   monomeric   +   oligomeric   xylose   for   which   adjusted   R2-­‐values   were   around   80%.   The   low   R2-­‐value   for   the   monomeric  +  oligomeric  xylose  yields  is  likely  due  to  the  smaller  spread  in  the  total  xylose  sugar   yields  between  all  the  experiments  (~12-­‐14%)  compared  to  the  other  sugars  (~30-­‐35%).     In  addition  to  the  base  enzymes,  the  amount  of  enzyme  was  also  significant  for  all  sugar   yields   for   both   switchgrass   varieties.   Other   terms   which   were   in   all   eight   models   include   the   interactions   between:   Spezyme   x   Xylanase,   Spezyme   x   Pectinase,   and   Spezyme   x   Xylanase   x   Pectinase.   β-­‐glucosidase   had   a   significant   impact   on   sugar   yields,   however   in   all   cases   the   91       highest  sugar  yields  were  obtained  when  this  enzyme  was  not  included  in  the  enzyme  mixture   (Table   4.6).   Both   Multifect®   Pectinase,   and   to   a   lesser   extent,   Spezyme®   CP   contain     Table  4.4:  Mixture  regression  of  enzymes  and  total  protein  loading  in  terms  of  sugar  release   following  enzymatic  hydrolysis  of  Alamo  switchgrass.  S  =  Spezyme®  CP;  B  =  Novozyme®  188   (β-­‐glucosidase);  X  =  Multifect®  Xylanase;  P  =  Multifect®  Pectinase;  Amt  =  Enzyme  loading  in   terms  of  total  protein.     Monomeric   Glucose     a Term     a P     *   *   *   *   0.000   0.000   0.000   0.000   -­‐   0.001   0.000   -­‐   0.000   -­‐   0.006   -­‐   -­‐                                     Coef.   50.14   26.42   55.43   61.92   49.96   27.77   55.62   78.54   36.02   109.33   201.82   -­‐   7.28   16.37   -­‐   -­‐   -­‐77.19   S   B   X   P   SB   SX   SP   BX   BP   SBP   SXP   BXP   Amt   SB*Amt   SX*Amt   BX*Amt   SBP*Amt   Coef.   57.57   7.85   43.12   40.19   101.69   39.55   65.63   112.7   -­‐   129.66   268.50   -­‐   13.93   -­‐   18.44   -­‐   -­‐   R     95.73%         *   *   *   *   0.000   0.000   0.000   0.000   0.032   0.030   0.000   -­‐   0.000   0.001   -­‐   -­‐   0.010                                     63.79   23.39   44.11   41.26   60.70   42.26   50.29   81.56   -­‐   100.50   247.00   -­‐   13.92   -­‐   -­‐   34.17   -­‐   97.19%   R  (adj)   95.08%       R  (pred)   93.95%       2 2 2 a P   Monomeric  +   Oligomeric   Glucose   a P   Coef.   Monomeric   Xylose   b c       85.26   81.43   83.77   80.52   -­‐   14.15   11.80   -­‐   -­‐   -­‐   44.48   97.16   4.02   -­‐   -­‐   -­‐   -­‐   *   *   *   *   -­‐   0.001   0.008   -­‐   -­‐   -­‐   0.009   0.008   0.000   -­‐   -­‐   -­‐   -­‐       95.80%       81.61%     96.67%       95.16%       79.59%     96.10%       94.05%       76.25%     b 2 2 *   *   *   *   0.000   0.000   0.000   0.000   -­‐   0.007   0.000   -­‐   0.000   -­‐   -­‐   0.005   -­‐   c 2                                   Monomeric  +   Oligomeric   Xylose   a P   Coef.   2 Coef.  =  regression  model  coefficient;   R  (adj)  =  adjusted  R -­‐value;   R  (pred)  =  predictive  R -­‐ value   92       Table  4.5:  Mixture  regression  of  enzymes  and  total  protein  loading  in  terms  of  sugar  release   following  enzymatic  hydrolysis  of  Shawnee  switchgrass.  S  =  Spezyme®  CP;  B  =  Novozyme®   188  (β-­‐glucosidase);  X  =  Multifect®  Xylanase;  P  =  Multifect®  Pectinase;  Amt  =  Enzyme  loading   in  terms  of  total  protein.     Monomeric   Glucose     a Term     P     *   *   *   *   0.000   0.000   0.000   0.000   0.000   -­‐   -­‐   0.000   -­‐   0.000   -­‐   -­‐                                   a S   B   X   P   SB   SX   SP   BX   BP   SBX   SBP   SXP   BXP   Amt   P*Amt   BX*Amt   Coef.   52.56   8.36   40.36   32.41   96.12   31.29   67.68   94.58   41.20   -­‐   -­‐   202.77   -­‐   12.58   -­‐   -­‐   R     96.34%     *   *   *   *   0.004   0.010   0.000   -­‐   -­‐   0.000   0.000   0.000   0.008   0.000   0.032   -­‐                                   59.68   52.49   43.24   34.33   -­‐   26.30   49.97   -­‐   -­‐   125.95   -­‐   185.18   -­‐   11.68   -­‐   21.93       95.34%   R (adj)   95.84%   R (pred)   95.41%   2 2   2   a P   Monomeric  +   Oligomeric   Glucose   a P   Coef.   Monomeric   Xylose   b c Coef.   50.94   38.46   60.37   60.98   31.22   13.52   50.15   -­‐   -­‐   190.75   144.95   175.80   129.67   7.50   -­‐3.03   -­‐         84.73   81.37   83.12   77.85   -­‐   8.00   16.74   -­‐   -­‐   40.34   -­‐   95.82   79.06   3.74   -­‐   -­‐   *   *   *   *   -­‐   0.038   0.000   -­‐   -­‐   0.018   -­‐   0.000   0.020   0.000   -­‐   -­‐       96.24%       84.52%         94.57%       95.79%       82.59%         93.82%       95.32%   b 2   2 *   *   *   *   -­‐   0.000   0.000   -­‐   -­‐   0.000   -­‐   0.000   -­‐   0.000   -­‐   0.021                                   Monomeric  +   Oligomeric   Xylose   a P   Coef.       79.87%   c 2     2 Coef.  =  regression  model  coefficient;   R (adj)  =  adjusted  R -­‐value;   R (pred)  =  predictive  R -­‐ value     β-­‐glucosidase   activity   [194].   Between   these   two   enzymes,   there   was   sufficient   activity   to   eliminate   the   need   for   Novozyme®   188   in   the   enzyme   mixture.   In   fact,   the   optimal   mixture,   when   loaded   at   the   same   total   protein   loading,   had   higher   estimated   β-­‐glucosidase   activity   93       compared   to   the   standard   enzyme   loading   of   15   FPU   Spezyme®   CP   and   30   CBU   Novozyme®   TM   188  per  g  glucan  (Table   4.6).  Other  research  on  AFEX and  dilute  acid  pretreatment  has  found   that  when  Multifect®  Xylanase  and  Multifect®  Pectinase  are  included  in  the  enzyme  mixture,   Novozyme®   188   is   not   necessary   for   optimal   sugar   yields   [150,   181].   In   terms   of   the   regression   models,   this   indicates   that   those   terms   that   include   β-­‐glucosidase   can   be   neglected   as   insignificant   and   the   model   can   be   thought   of   as   a   ternary   mixture   of   the   other   three   enzymes.   Once  the  β-­‐glucosidase  terms  are  neglected  from  the  model,  the  amount  of  enzyme  has  little   significant   interaction   with   the   specific   enzymes   used,   only   with   Pectinase   for   Shawnee   monomeric   xylose   yields   and   with   Spezyme   x   Xylanase   for   Alamo   monomeric   glucose   yields.   Even  so,  when  contour  plots  of  the  two  enzyme  loadings  (15  mg  protein  per  g  glucan  and  30  mg   protein  per  g  glucan)  were  compared;  there  was  almost  no  difference  in  the  shape  of  the  curves   (data  not  shown).  At  the  two  enzyme  loadings,  glucose  release  is  impacted  more  significantly   with   the   increase   in   enzymes   compared   to   the   xylose   release.   This   may   indicate   either   that   the   xylo-­‐oligomers   and   hemicellulose   are   competing   substrates   for   the   cellulase   enzymes,   or   that   the  small  amount  of  hemicellulases  is  limiting  the  capacity  to  break  down  hemicelluloses  that   block  enzyme  access  to  the  cellulose.   When   comparing   the   models   between   the   two   switchgrass   varieties,   most   of   the   coefficients   for   the   same   sugar   yield   are   very   similar.   This   indicates   that   the   response   curves   of   the  two  varieties  should  also  be  very  similar,  which  is  indeed  the  case  (Figure   4.5).  (The  total   glucose   contour   plot   is   almost   identical   to   the   monomeric   glucose   contour   plot   and   so   was   not   included.)   As   can   be   seen,   the   monomeric   glucose   and   xylose   yield   contour   plots   are   nearly   94       Table   4.6:   Base   case   and   optimized   commercial   enzyme   mixtures   for   monomeric   and   monomeric  +  oligomeric  (total)  sugar  yields  (both  glucose  +  xylose)  from  Alamo  and  Shawnee   switchgrass.    Enzyme  proportions  are  expressed  as  a  percentage  of  total  protein  in  the  enzyme   mixture.   Sugar   yields   are   presented   as   %   of   sugars   theoretically   available   in   untreated,   dry   biomass.  Sugar  yields  in  parentheses  represent  total  monomeric  +  oligomeric  sugars,  and  those   not   in   parentheses   represent   monomeric   sugar   yields.   Enzyme   activities   for   each   enzyme   mixture  are  estimated  based  on  activities  per  mL  reported  by  Dien  et  al.  [194]  for  each  of  the   commercial  enzyme  preparations.               Alamo   Base   Case     Maximize     Monomers   Enzyme  Mixture  (%  of  total  protein)     Spezyme  CP   90%     β-­‐Glucosidase   10%     Xylanase   -­‐     Pectinase   -­‐   46%   -­‐   20%   34%   Total   54%   -­‐   23%   23%           Shawnee   Maximize     Base   Case   Monomers   Total   90%   10%   -­‐   -­‐   47%   -­‐   20%   33%   57%   -­‐   20%   23%   Predicted  Sugar  Yields  (g/g  sugar  theoretically  present  in  untreated  biomass)   15  mg/g  glucan     Glucose   54.8   (58.2)   47.9   (82.9)     Xylose   63.3     (64.7)   69.1     (85.9)   62.9   (65.7)   66.5   (86.4)   50.5   (53.1)   48.8     (82.5)     56.9     (58.2)   67.4   (86.6)   57.0   (59.1)   64.5   (86.6)   30  mg/g  glucan   68.7   79.0   79.1   63.1   69.5     (72.2)   (78.6)   (79.6)   (64.8)   (69.8)     56.6   76.4   73.8   56.3   73.9   Xylose     (86.9)   (89.7)   (90.4)   (86.3)   (90.3)   Estimated  Activity  of  the  Mixture  Shown  Above  (activity  per  10  mg  protein  in  mixture)   Glucose   a   Cellulase  (FPU )     β-­‐Glucosidase   6.5   3.6   4.1     6.5   3.7   4.3   24   32   27     24   32   27     Xylanase  (OSX )   290     α-­‐Arabinofuranosidase   3     β-­‐Xylosidase   1     α-­‐Galactosidase   1.8     Feruloyl  esterase   0.0   2100   2400     290   2100   2100   120   14   2.3   0.6   85   11   1.6   0.4   3   1   1.8   0.0   120   14   2.2   0.6   85   10   1.6   0.4   b a 69.5   (70.8)   71.3   (90.3)   b FPU  =  Filter  paper  units;   OSX  =  Oat  spelt  xylan.   95                 Figure   4.5:   Ternary   plots   based   on   the   enzyme   mixture   regression   model   showing   the   interactive   effect   of   Spezyme®   CP,   Multifect®   Xylanase   and   Multifect®   Pectinase   on   sugar   yields  from  AFEX TM   pretreated  Alamo  (A-­‐C)  and  Shawnee  (D-­‐F)  switchgrass.  (A,  D)  monomeric   glucose;   (B,   E)   monomeric   xylose;   (C,   F)   monomeric   +   oligomeric   xylose.   Percentages   of   each   enzyme   are   with   respect   to   total   protein   loading,   which   was   held   at   30   mg   per   g   glucan   for   these   figures.   β-­‐glucosidase   was   held   at   0%   in   all   plots.   Enzyme   loadings   were   constrained   to   Spezyme   CP   >   20%,   Multifect   Xylanase   <   80%   and   Multifect   Pectinase   <   80%   of   total   enzyme   protein  added  to  hydrolysis.     identical  between  the  two  varieties  and  the  optimal  enzyme  mixtures  are  also  almost  identical   (Table   4.6).  There  is  a  more  obvious  difference  in  the  shape  of  the  total  xylose  contour  plots  for   the   two   varieties,   with   Alamo   shifted   toward   a   higher   xylanase   content   compared   to   the   Shawnee.  However,  the  range  in  total  xylose  yields  for  these  figures  are  very  small  (~4-­‐5%  for   96       the   entire   plot),   and   any   number   of   points   within   the   center   of   these   diagrams   should   give   almost   identical   yields.   Even   the   base   case   of   Spezyme®   CP   and   Novozyme®   188   only   has   about   3-­‐4%   less   total   monomeric   +   oligomeric   xylose   release   compared   to   the   optimized   enzyme  mixtures  (Table  4.6).  What  this  indicates  is  that  pretreatment  has  the  greatest  impact   on   effective   removal   and   solubilization   of   hemicellulose   from   the   cell   wall,   and   most   of   the   impact   by   the   xylanase   and   pectinase   is   on   conversion   of   glucan   and   soluble   hemicellulose   oligomers   to   monomers.   Accessory   enzymes   are   known   to   be   necessary   to   increase   xylose   TM yields   from   AFEX   treated   grasses   [195].   Other   work   on   AFEXTM-­‐treated   switchgrass   also   found  that  the  optimal  enzyme  mixture  contained  both  xylanase  and  pectinase  [150].  Xylanase   is  necessary  for  its  high  xylanase  content  that  breaks  apart  the  xylan  backbone,  while  pectinase   is   necessary   for   its   high   accessory   enzyme   activity   (including   β-­‐glucosidase   and   β-­‐xylosidase)   [194].   If   it   is   possible   to   directly   use   the   oligomeric   sugars,   perhaps   by   employing   a   microorganism  that  can  consume  or  ferment  oligosaccharides,  then  the  proportion  of  pectinase   in  the  optimized  mixture  could  be  decreased.   While   the   optimized   enzyme   mixtures   were   theoretically   expected   to   increase   monomeric   glucose   release   compared   to   the   base   case   (Table   4.6),   this   was   not   observed   experimentally   (Figure   4.6),   with   only   a   slight   increase   in   glucose   yields   due   to   mixture   optimization.  Alamo  had  higher  glucose  yields  compared  to  Shawnee,  although  in  terms  of  total   sugar   release   per   kg   biomass,   the   yields   were   almost   identical.   The   primary   benefit   of   the   optimized  enzyme  mixture  is  in  increasing  the  initial  rate  of  xylose  release  through  more  rapid   conversion   of   the   soluble   oligomers.   For   both   switchgrass   varieties   the   initial   rate   of   xylose 97         Figure   4.6:   Monomeric   glucose   (A)   and   xylose   (B)   release   from   optimally   pretreated   Alamo   and  Shawnee  switchgrass  hydrolyzed  with  base  enzyme  loading  (15  FPU  Spezyme®  CP  and  30   CBU  Novozyme®  188  per  g  glucan)  compared  to  with  our  optimized  enzyme  mixture  for  each   variety   containing   Spezyme®   CP,   Multifect®   Xylanase   and   Multifect®   Pectinase.   Enzyme   loading  was  27  mg  protein  per  g  glucan  and  samples  were  taken  at  1  h,  24  h  and  168  h.     hydrolysis  at  1  h  increases  significantly  for  the  optimized  mixture  compared  to  the  base  case.   This  again  indicates  that  much  of  what  is  hindering  xylose  yields  is  lack  of  appropriate  xylanases   and   accessory   enzymes   in   the   standard   enzyme   cocktail.   Interestingly,   the   optimum   enzyme   cocktail  is  able  to  achieve  identical  glucose  yields  compared  to  the  base  case,  although  the  base   case  mixture  contains  nearly  double  the  cellulase  activity  (Table   4.6).  As  stated  earlier,  this  may   indicate   that   the   xylanase   and   pectinase   enzymes   either   improve   access   to   the   cellulose   or   remove  a  competing  substrate.     The   cost   of   enzymes,   either   purchased   or   produced   in-­‐house,   is   one   of   the   major   contributors   to   the   total   costs   for   the   production   of   cellulosic   ethanol.   A   recent   study   by   the   National   Renewable   Energy   Laboratory   (NREL,   Golden,   Colorado)   found   that   the   cost   of   98       enzymes   produced   in-­‐house   accounted   for   almost   16%   of   the   minimum   ethanol   selling   price   (MESP),   nearly   half   of   the   cost   of   the   feedstock   [196].   This   increased   to   20%   of   the   MESP   if   the   enzymes   were   purchased   from   external   sources.   One   way   to   reduce   the   costs   of   producing   cellulosic   ethanol   is   to   make   improvements   to   enzyme   efficiency   that   allow   for   reductions   in   either   the   enzyme   loading   required   to   maintain   a   set   level   of   conversion   (leading   to   reduced   enzyme   costs),   and/or   the   residence   time   required   for   a   set   level   of   conversion   (leading   to   reduced  capital  and  operating  costs).  For  our  work,  the  optimized  mixture  gave  higher  xylose   yields  but  lower  glucose  yields  at  the  lower  enzyme  loading  (15  mg/g  glucan)  compared  to  the   base  case  at  the  higher  enzyme  loading  (30  mg/g  glucan)  (Table  4.6).  By  optimizing  the  enzyme   mixture,  it  was  not  possible  to  maintain  glucan-­‐to-­‐glucose  conversion  while  reducing  the  total   protein   loading   by   50%,   a   finding   that   has   been   observed   in   other   studies   [181].   However,   xylan-­‐to-­‐xylose   conversions   were   increased   and   as   a   result,   the   total   monomeric   sugar   conversion  was  slightly  higher  for  the  optimized  enzyme  loading  at  15  mg/g  glucan  compared   to  the  base  case  at  30  mg/g  glucan  for  both  switchgrass  varieties.     One   final   point   to   note   is   that   while   this   lab-­‐scale   analysis   provides   a   baseline   for   optimum  pretreatment  conditions  and  enzymatic  hydrolysis  enzyme  combinations,  it  is  entirely   possible  that  the  optimum  parameters  and  combinations  would  change  due  to  reactor  scale-­‐up   and  increased  enzymatic  hydrolysis  solids  loading.  Further  work  would  need  to  be  conducted  to   determine  these  new  optima,  however  the  results  presented  here  provide  a  starting  point  for   further  analysis.     99       4.4.  Conclusions   When  grown  in  similar  locations  and  harvested  at  the  same  time  of  the  year,  there  was   little  difference  in  optimal  pretreatment  conditions  and  optimal  enzyme  loadings  required  for   conversion  of  Alamo  (lowland)  and  Shawnee  (upland)  switchgrass.  Additionally,  the  total  sugar   yields  on  a  mass  basis  were  almost  identical  for  both  varieties.    Inclusion  of  hemicellulases  in   the  enzyme  mixture  primarily  functioned  to  increase  monomeric  xylose  release  and  allowed  for   a  50%  reduction  in  enzyme  loading  while  maintaining  total  sugar  yields.  The  biorefinery  should   be  able  to  effectively  process  both  switchgrass  varieties  using  the  same  pretreatment   conditions  and  obtain  relatively  high  yields.     100       CHAPTER  5 :       COMPONENT  SCALE:  OPTIMIZING  HARVEST  OF  CORN  STOVER  FRACTIONS  BASED  ON   TM OVERALL  SUGAR  YIELDS  FOLLOWING  AFEX    PRETREATMENT  AND  ENZYMATIC  HYDROLYSIS   5.1.  Introduction   Corn   stover,   the   aboveground,   vegetative   portion   of   maize   (Zea   mays   L.),   makes   up   roughly  80%  of  all  agricultural  residues  produced  in  the  USA  [116].  Data  on  annual  corn  stover   production   in   the   USA   are   not   readily   available;   so   various   sources   have   independently   estimated  that  anywhere  from  200  to  250  million  dry  tons  of  corn  stover  are  produced  per  year   [114,   116,   197,   198].   Sustainably   harvested   corn   stover   could   be   used   as   a   feedstock   for   a   variety  of  applications,  including  lignocellulosic  ethanol  production.  It  has  been  estimated  that   38.4   billion   liters   of   ethanol   per   year   could   be   produced   from   North   American   corn   stover,   assuming  that  40%  of  the  stover  is  collected  [76].  It  is  widely  acknowledged  that  a  percentage   of   the   produced   corn   stover   should   be   retained   on   the   field   following   harvest   in   order   to   prevent   soil   erosion   and   maintain   soil   organic   carbon   (SOC)   levels.   The   amount   that   can   be   sustainably   harvested   is   highly   debated   and   depends   heavily   on   cropping   practices,   climate,   topography,  and  soil  type  [114,  199-­‐201].  Estimates  on  the  amount  of  corn  stover  that  can  be   sustainably   harvested   vary   widely   because   of   these   factors,   anywhere   from   20-­‐80%   [76,   116,   199].         Lignocellulosic   feedstocks,   such   as   corn   stover,   derive   their   name   from   the   three   primary   components   of   the   plant   cell   wall:   cellulose,   hemicellulose   and   lignin.   The   complex   polysaccharides,   cellulose   and   hemicellulose,   must   be   broken   down   into   monomeric   form   (primarily   glucose   and   xylose)   prior   to   microbial   fermentation   into   ethanol   or   other   valuable   products.   High   sugar   yields   require   a   two-­‐step   process:   generally   a   chemical   and/or   physical   101       pretreatment  step  followed  by  enzymatic  hydrolysis  of  the  polysaccharides.  Previous  work  has   TM shown  that  ammonia  fiber  expansion  (AFEX )  is  a  promising  pretreatment  that  can  be  used  in   the  process  of  converting  corn  stover  polysaccharides  into  ethanol  as  a  liquid  fuel  source  [202-­‐ TM 205].   AFEX   pretreatment   uses   concentrated   ammonia-­‐water   mixtures   under   moderate   temperatures  (60-­‐180°C)  and  high  pressures  (200-­‐1000  psi)  to  disrupt  the  cellular  structure  of   the  plant  material  by  decrystallizing  the  cellulose;  partially  depolymerizing  and  solubilizing  the   hemicellulose;  and  altering  the  form,  location,  and  structure  of  lignin  [203,  204].     The   structure   and   composition   of   the   plant   cell   wall   depends   on   a   number   of   factors   including:   developmental   stage   at   harvest,   geographical   origin,   type   of   tissue,   and   other   external  factors  including  season  of  harvest  and  environmental  conditions  experienced  during   growth  [206].  Corn  stover,  like  most  grasses,  experiences  considerable  compositional  changes   throughout   the   yearly   growth   period,   as   well   as   significant   variation   between   the   various   fractions   of   the   plant   (that   is,   leaf   versus   stem)   [158,   207,   208].   Largely   because   of   these   differences   in   composition,   stover   fractions   have   been   shown   to   respond   differently   to   pretreatment   and   enzymatic   hydrolysis,   resulting   in   different   sugar   yields   [209-­‐211].   It   is   reasonable  to  assume  that  differences  in  composition,  due  largely  to  differences  in  morphology   and  cell  and  tissue  organization,  could  cause  different  stover  fractions  to  have  different  optimal   pretreatment   conditions   for   maximizing   sugar   yields.   For   example,   wheat   straw   leaves,   when   treated   with   dilute   NaOH,   required   less   severe   pretreatment   conditions   to   optimize   glucan   yields   than   stem   internodes   and   nodes   [212].   The   same   might   be   true   for   corn   stover   TM pretreated  with  ammonia  (or  AFEX ).  Maximum  sugar  yields  from  individual  fractions  would   102       be  one  criterion  for  determining  which  fractions  should  be  left  on  the  field  following  harvest.   Assuming  that  there  are  no  other  constraining  factors,  it  would  be  most  logical  to  harvest  the   least  recalcitrant  biomass  and  leave  the  remainder  for  erosion  control  and  soil  organic  carbon   maintenance   [213].   Crofcheck   and   Montross   recommended,   based   on   glucose   yields   from   fractionated   corn   stover,   a   roughly   30%   corn   stover   harvest   scenario   where   the   selectively   harvested  corn  stover  (SHCS)  was  composed  of  all  of  the  available  cobs  and  74%  of  the  leaves   and  husks,  leaving  the  most  recalcitrant  stalks  on  the  field  [209].   For  our  experiment,  AFEX TM   followed  by  enzymatic  hydrolysis  was  performed  on  four   corn  stover  fractions  (stem,  leaf,  husk,  and  cob)  from  September  (early)  and  November  (late)   harvests.   The   objectives   of   this   project   were:   (1)   to   determine   whether   individual   stover   TM fractions   have   different   optimal   AFEX   conditions   and   whether   this   is   different   from   previously   optimized   values   for   homogeneously   milled   corn   stover   [203,   204];   (2)   to   discover   which  fractions  give  the  highest  glucose  and  xylose  yields  at  optimal  pretreatment  conditions;   and   (3)   to   model   optimal   harvest   scenarios,   assuming   30%   and   70%   collection   of   total   available   dry   corn   stover,   based   on   the   maximum   monomeric   glucose   and   xylose   yields   from   each   fraction.       5.2.  Materials  and  methods   5.2.1.  Harvest  and  milling   Corn  stover,  from  a  variety  intended  for  grain  production,  was  manually  harvested  from   the   Michigan   State   University   Agronomy   Center   in   East   Lansing,   Michigan,   USA   in   September   103       (early   harvest)   and   November   (late   harvest)   of   2006.   The   early   and   late   stover   harvests   were   hand-­‐sorted  into  four  individual  fractions:  stems,  leaves  with  leaf  sheaths,  cobs,  and  husks.  The   early  husk  and  early  cob  fractions  were  not  used  due  to  spoilage  of  the  material  prior  to  use.  All   other  fractions  were  air-­‐dried,  with  stems  split  lengthwise  in  order  to  increase  the  drying  rate.   Fractions   were   then   milled   using   a   Fitzpatrick   JT-­‐6   Homoloid   mill   (Continental   Process   Systems,   Inc.,   Westmont,   Illinois,   USA),   with   leaf,   husk,   and   cob   fractions   passing   through   a   4.763   mm   (3/16  in)  mesh  screen,  and  stem  fractions  passing  through  a  3.175  mm  (1/8  in)  mesh  screen.       5.2.2.  Composition  analysis   Biomass  moisture  content  was  determined  using  a  moisture  analyzer  (A&D,  Model  MF-­‐ 50;  California,  USA).  The  composition  of  each  corn  stover  fraction  (ash,  lignin,  glucan  and  xylan   content)   was   determined   using   the   National   Renewable   Energy   Laboratory   (NREL,   Colorado,   USA)  standard  protocols  for  ash  analysis,  removal  of  extractives,  and  structural  carbohydrates   and  lignin  [214-­‐216].  The  acid  insoluble  lignin  analysis  method  was  modified  to  use  47  mm,  0.22   μm  pore-­‐size,  mixed-­‐cellulose  ester  filter  discs  (Millipore  Corp,  Massachusetts,  USA)  during  the   filtration   step   instead   of   fritted   crucibles.   Due   to   problems   with   burning,   these   discs,   with   their   filtered   lignin   residue,   could   not   be   dried   in   the   vacuum   oven   and   were   therefore   dried   overnight   in   a   desiccator   prior   to   weighing.   Soluble   sugars   could   not   be   quantified   after   extraction   due   to   difficulties   in   resolving   distinct   peaks   using   high-­‐performance   liquid   chromatography  (HPLC)  and  were  therefore  not  included  in  the  composition.     104       TM 5.2.3.  AFEX  treatment   A  small-­‐scale  bench  top  reactor  system,  consisting  of  four  separate  22  mL  stainless  steel   (No.   316)   reaction   vessels,   was   used   for   the   pretreatment   process.     Prior   to   its   loading,   the   biomass   was   adjusted   to   the   appropriate   moisture   content   with   deionized   water,   after   which   3.0   g   dry   weight   (dwb)   of   biomass   was   added   to   each   reaction   vessel.   A   metal   screen   was   placed  over  the  biomass  inside  each  vessel,  to  prevent  escape  of  biomass  during  venting.  The   loaded  reactor  units  were  weighed  and  attached  to  the  reactor  manifold,  and  any  air  within  the   reactor  vessels  was  then  removed  using  a  rotary  vacuum  pump.  Liquid  anhydrous  ammonia  was   dispensed   into   the   manifold   via   Swagelok   screw   valves   (Swagelok   Co,   Ohio,   USA)   and   then   added  to  the  reactor  vessels.  The  reactors  were  weighed  in  order  to  determine  the  amount  of   ammonia  added  and  they  were  then  vented  slightly  to  reach  the  appropriate  ammonia  loading.   A   heating   mantle   was   used   to   raise   the   reactors   to   the   desired   temperature   and   maintain   it   for   the  set  residence  time.  On  completion  of  the  residence  time,  the  reactor  pressure  was  released   and   the   reactor   was   simultaneously   cooled.   The   pretreated   biomass   was   removed   from   the   vessel  and  left  in  the  fume  hood  overnight  to  allow  the  residual  ammonia  to  evaporate.     5.2.4.  Enzymatic  hydrolysis   NREL  protocol  (LAP  009)  [152]  was  followed  for  the  enzymatic  hydrolysis  of  pretreated   and   untreated   (control)   samples.   All   samples   were   hydrolyzed   in   20mL   screw-­‐cap   vials   at   1%   glucan  loading  and  a  total  volume  of  15mL.  Samples  were  adjusted  to  a  pH  of  4.8  by  1M  citrate   buffer  solution.  Spezyme®  CP  (Genencor  Division  of  Danisco  US,  Inc.,  New  York,  USA)  cellulase   105       Table  5.1:    Enzymatic  hydrolysis  xylanase  loading  in  terms  of  xylan  content  of  each  fraction.     Xylanase  loading   -­‐1 mg  xylanase  g  xylan   Leaves   4.78   Early   Stem   5.73   Leaves   5.03   Stem   4.97   Late   Husk   4.57   Cob   2.64   *Oat  spelt  xylan,  based  on  activity  numbers  from  Dien  et  al.  [194]   Corn  stover  fraction   Xylanase  activity   -­‐1 OSX*  g  xylan   2891   3456   3030   2997   2754   1593     -­‐1 at   15   FPU   g   glucan   (31.3   mg   protein   g-­‐1   glucan)   and   β-­‐glucosidase   (Novozyme®   188,   -­‐1   -­‐1   Novozymes   Corp.,   Bagsværd,   Denmark)   at   64   p-­‐NPGU   g glucan   (41.3   mg   protein   g glucan)   -­‐1   were   added   to   each   vial   with   a   total   protein   content   of   72.6   mg   protein   g glucan.   In   addition,   certain   samples   were   also   hydrolyzed   using   xylanase   (Multifect®   Xylanase,   Genencor   Division   -­‐1   of  Danisco  US  Inc.)  at  10%  of  total  cellulase  protein  (1871  OSX  (oat  spelt  xylan)  g glucan  or  3.1   -­‐1   -­‐1   mg  protein  g glucan),  giving  a  total  protein  content  of  75.7  mg  protein  g glucan.    The  data   for  the  xylanase  activity  are  based  on  the  activity  per  mL  provided  by  Dien  et  al.  [194]  and  the   activity,  in  terms  of  the  xylan  content  of  each  sample,  is  included  in  Table  5.1.  Enzyme  loading   throughout  the  paper  is  referred  to  in  terms  of  protein  loading,  as  opposed  to  activity,  because   of   the   probable   relationship   between   protein   and   enzyme   cost   to   the   biorefinery   [217].   Samples   were   placed   in   a   New   Brunswick   Scientific   (New   Jersey,   USA)   incubator   shaker   and   hydrolyzed   at   50°C   and   150   rpm   for   72   h.   The   hydrolysates   were   sampled   at   24   h   and   72   h,   106       following  which  samples  were  heated  at  90°C  for  15  min,  cooled  and  centrifuged  at  15K  for  5   min.  The  supernatant  was  filtered  into  HPLC  shell  vials  using  a  25  mm,  0.2  μm  polyethersulfone   syringe  filter  (Whatman  Inc.,  New  Jersey,  USA)  after  which  samples  were  stored  at  -­‐20°C  until   further  sugar  analysis.               5.2.5.  Sugar  analysis   An   HPLC   system   was   used   to   determine   the   sample   monomeric   glucose   and   xylose   concentrations   following   enzymatic   hydrolysis.   The   HPLC   system   consisted   of   a   Waters   (Massachusetts,   USA)   pump,   auto-­‐sampler   and   Waters   410   refractive   index   detector,   equipped   with  a  Bio-­‐Rad  (Hercules,  California,  USA)  Aminex  HPX-­‐87P  carbohydrate  analysis  column  with   attached  deashing  guard  column.  Degassed  HPLC  grade  water  was  used  as  the  mobile  phase,  at   -­‐1 0.6   mL   min ,   with   the   column   temperature   set   at   85°C.   Injection   volume   was   10   μL   with   a   run   time   of   20   min   per   sample.   Mixed   sugar   standards   were   used   to   quantify   the   amount   of   monomeric   glucose   and   xylose   in   each   hydrolysate   sample.     All   sugar   yields   are   from   the   enzymatic  hydrolysate  and  are  reported  in  terms  of  the  untreated  dry  biomass.     5.2.6.  Statistical  analysis   Monomeric   glucose   and   xylose   yields   following   enzymatic   hydrolysis   were   analyzed   using   MANOVA   in   Minitab15   Statistical   Software   (2006   Minitab   Inc.,   Pennsylvania,   USA).     The   interactive   effects   plot   which   compares   the   harvest   period   and   the   stover   fraction   with   each   107       TM other,   the   four   AFEX   pretreatment   parameters   (moisture   content,   ammonia   loading,   temperature  and  residence  time)  and  the  xylanase  addition  was  also  constructed  using  Minitab.         5.2.7.  Empirical  modeling  of  harvest  scenarios   TM For  this  analysis,  the  sugar  yields  used  were  from  the  72  h  hydrolysis  of  AFEX   -­‐treated   late-­‐harvest  corn  stover.  The  option  of  an  early  harvest  was  not  analyzed  because  of  the  lack  of   data  for  husk  and  cob  fractions.  Scenarios  were  analyzed  with  regard  to  the  effect  of  increasing   -­‐1   ammonia   loading   from   1.0   to   1.5   (g   NH3   g biomass)   and   for   the   maximized   sugar   yield,   either   glucose  or  xylose.    This  gave  four  potential  scenarios  (1.0  +  glucose,  1.5  +  glucose,  1.0  +  xylose   and   1.5   +   xylose).   All   other   AFEX TM   and   hydrolysis   conditions   were   held   constant   (60%   dwb   moisture,   90°C,   5   min   residence   time   +   10%   xylanase   addition).   As   the   glucose   yields   were   consistently   higher   than   the   xylose   yields,   the   harvest   conditions   used   to   obtain   maximum   glucose  yields  for  all  of  the  scenarios  also  corresponded  with  the  maximum  total  sugar  yields.       5.3.  Results   5.3.1.  Composition  analysis   The  composition  of  each  of  the  corn  stover  fractions  from  each  harvest  is  listed  in  Table   5.2.   The   value   of   the   ‘other’   column   was   determined   by   the   difference   of   the   total   of   the   other   columns  from  100%.  The  standard  deviation  is  representative  of  three  replicates.  Statistically,   the  early  and  late  stem  and  the  late  leaves  and  husk  had  the  highest  glucan  content,  while  the   108       Table   5.2:   Corn   stover   composition   for   early   and   late   harvest   stover   fractions.   Values   with   different  superscripts  in  a  column  were  statistically  different  using  Tukey’s  pairwise  comparison   with  α  =  0.05.    The  ‘other’  column  determined  by  difference  from  100%.     Corn  stover   fraction   Corn  stover  fraction  composition  (%  dry  biomass)   Acid-­‐insoluble   Glucan   Xylan   Ash   lignin   Leaves   17.8  ±  1.7   13.2  ±  0.7   7.3  ±  0.13   a 34.2   35.1  ±  2.6   19.0  ±  1.1   14.9  ±  0.2   c 27.6   35.3  ±  1.2   21.8  ±  0.6   bc 3.4  ±  0.10   Leaves   13.6  ±  1.7   6.0  ±  0.25   b 23.3   Stem   37.8  ±  0.9   b 2.4  ±  0.08   d 19.3   Husk   Late   27.5  ±  3.2   Stem   Early   39.0  ±  2.2   26.5  ±  1.5   11.6    ±  0.3   c 2.1  ±  0.11   e 20.8   27.5  ±  1.1   32.3  ±  1.3   25.8    ±  2.6   f 1.1    ±  0.02   13.3   Cob   b a a a a b e de cd bc 23.6  ±  0.4   b a bc Other   bc 16.9    ±  0.5   a   early   leaves   and   late   cob   had   the   lowest   glucan   content.   The   xylan   content   of   the   late   fractions   was  significantly  higher  than  their  early  counterparts  and  tended  to  decrease  from  late  cob  >   late   husk   >   late   stem   >   late   leaves   >   early   stem   >   early   leaves.   The   acid-­‐insoluble   lignin   content   was  similar  for  all  fractions,  except  for  the  cob,  which  had  the  highest  lignin  content,  and  the   late  husk,  which  had  statistically  less  lignin  than  the  late  stem.  The  ash  content  of  all  fractions   were   statistically   different   and   decreased   from   early   leaves   >   late   leaves   >   early   stem   >   late   stem  >  late  husk  >  late  cob.           TM 5.3.2.  AFEX  pretreatment  and  hydrolysis   TM Pretreatment   conditions   for   AFEX -­‐treated   corn   stover   have   been   previously   -­‐1   optimized  at  1.0  (g  NH3  g dry  biomass),  60%  moisture  content  (dry-­‐weight  basis;  dwb),  90°C   and  5  min  residence  time  [203,  204].  These  conditions  were  treated  as  the  ‘base  case’  for  the   109       analysis   of   pretreatment   conditions.   The   effect   of   pretreatment   conditions   on   monomeric   glucose   and   xylose   yields   following   hydrolysis   was   tested   by   varying   one   process   parameter   (temperature,   ammonia   loading,   moisture   content   or   residence   time)   at   a   time   (for   example,   raising   the   temperature   from   90°C   to   100°C).   Once   the   preliminary   data   had   been   gathered,   the   untreated   control,   base   case   and   best   case   were   supplemented   with   xylanase   during   hydrolysis  to  observe  the  effect  on  sugar  yields.     Figure   5.1   shows   the   monomeric   glucose   and   xylose   yields   for   a   variety   of   conditions   TM with   particular   comparisons   between   untreated   and   AFEX -­‐treated   materials   at   a   range   of   ammonia   loadings.   The   effect   of   xylanase   addition   to   the   enzyme   cocktail   can   also   be   observed   in  Figure   5.1.  Error  bars  in  all  figures  represent  the  mean  ±1  standard  deviation.  From  Figure   TM 5.1,   it   can   be   seen   that   AFEX   substantially   improves   both   glucose   and   xylose   monomeric   sugar   yields   for   all   harvest   periods   and   corn   stover   fractions   when   compared   to   untreated   materials.       -­‐1 The   increase   in   ammonia   loading   from   0.5   to   1.5   (g   NH3   g   biomass)   had   different   effects   on   early   harvest   and   late   harvest   corn   stover   fractions.   For   the   early   harvest   stover   -­‐1 without   xylanase   addition,   glucose   yields   peak   at   1.0   (g   NH3   g   biomass).   This   optimum   is   TM similar   to   what   has   been   seen   previously   with   AFEX -­‐treated   corn   stover   [203,   204],   which   may   indicate   that   that   material   was   from   an   earlier   harvest.   The   xylose   yields   are   relatively   -­‐1 unaffected  by  any  further  increase  above  1.0  (g  NH3  g  biomass).  However,  when  performing   110       the  same  experiment  with  the  late  harvest  corn  stover,  there  is  an  increase  in  both  glucose  and   -­‐1 xylose  yields  for  all  fractions  when  increasing  from  1.0  to  1.5  (g  NH3  g  biomass).         Xylanase  addition  had  little  effect  on  the  increase  of  either  glucose  or  xylose  sugar  yields     TM Figure  5.1:   Effect  of  ammonia  fiber  expansion  (AFEX   )  pretreatment  ammonia  loading  and   xylanase   addition   on   enzymatic   hydrolysis   monomeric   sugar   yields.     Glucose   yields   are   TM reported   in   part   A   and   xylose   yields   are   in   part   B.     All   AFEX   runs   were   kept   at   constant   moisture   content   (60%   dry-­‐weight   basis),   temperature   (90°C)   and   residence   time   (5   min).     Yields  are  in  terms  of  sugar  available  in  untreated  dry  biomass.   111       TM in  untreated  corn  stover  fractions.  For  AFEX -­‐1   xylanase   at   1.0   (g   NH3   g -­‐treated  early  harvest  fractions,  the  addition  of   biomass)   had   no   effect   on   monomeric   xylose   yields   and   it   slightly   -­‐1   lowered  glucose  yields.    At  1.5  (g  NH3  g biomass),  all  fractions  and  harvests  experienced  an   increase  in  both  the  monomeric  xylose  and  glucose  yields  with  the  addition  of  xylanase.         The   leaf   and   stem,   for   both   early   and   late   harvests,   have   similar   glucose   yields   at   1.5     -­‐1 (g  NH3  g  biomass)  ammonia  loading.  However,  the  leaf  glucan  is  more  digestible,  as  seen  by   the   greater   yield   (percent   of   maximum   theoretical   glucan   available).   The   late   harvest   husk     TM Figure   5.2:   Effect   of   ammonia   fiber   expansion   (AFEX   )   pretreatment   temperature   on   TM   runs   were   kept   at   a   constant   -­‐1 moisture   content   (60%   dry-­‐weight   basis),   ammonia   loading   (1.0   g   NH3   g   dry   biomass)   and   enzymatic   hydrolysis   monomeric   sugar   yields.     All   AFEX residence  time  (5  min).    Yields  are  in  terms  of  sugar  available  in  untreated  dry  biomass.    Glu  =   glucose,  Xyl  =  xylose,  MTSY  =  maximum  theoretical  sugar  yield.       112       -­‐1 approaches  theoretical  glucose  yields  at  the  optimal  condition  of  1.5  (g  NH3  g  biomass).  As  a   result   of   this,   the   addition   of   xylanase   for   this   pretreatment   condition   increases   husk   xylose   yields  slightly  but  not  the  glucose  yields,  as  is  seen  in  the  other  fractions.  With  the  addition  of   -­‐1 xylanase  at  1.5  (g  NH3  g  biomass),  the  cob  and  leaf  also  approach  theoretical  glucose  yields.         Figure   5.2  shows  the  effect  of  pretreatment  temperature  on  glucose  and  xylose  yields   from   corn   stover   fractions.   Altering   the   temperature   by   10°C   from   the   base   case   had   little   effect   on   glucose   and   xylose   yields.   There   is   a   definite   peak   in   glucose   yields   at   90°C   for   the   TM Figure   5.3:  Effect   of   ammonia   fiber   expansion   (AFEX   )   moisture   content   and   residence   time   TM   conditions:    moisture  content   -­‐1 (60%  dry-­‐weight  basis),  ammonia  loading  (1.0  g  NH3  g  dry  biomass),  temperature  (90°C)  and   on  enzymatic  hydrolysis  monomeric  sugar  yields.    Base  AFEX residence  time  (5  min).    Yields  are  in  terms  of  sugar  available  in  untreated  dry  biomass.    MC  =   moisture  content,  RT  =  residence  time,  Glu  =  glucose,  Xyl  =  xylose,  MTSY  =  maximum  theoretical   sugar  yield.   113       early   harvest   but   the   late   harvest   has   no   apparent   difference   in   yields   for   80°,   90°   or   100°C.   In   a  previous  work  [203],  raising  the  temperature  above  90°C  had  a  negative  impact  on  ethanol   yields  from  simultaneous  saccharification  and  fermentation.     Decreasing  the  moisture  content  to  40%  (dwb)  and  eliminating  the  residence  time  (the   time   for   which   the   reactor   was   held   at   the   set   temperature   following   heat-­‐up)   each   had   a   negative  impact  on  glucose  and  xylose  yields  for  all  fractions  (Figure  5.3).  For  all  stover  fractions,   except  the  late  husk,  it  was  more  detrimental  in  terms  of  sugar  yields  to  decrease  the  residence   time  rather  than  the  moisture  content.         Table  5.3:  Analysis  of  variance  for  factors  influencing  sugar  yields.     Factor   Harvest  date   Corn  stover  fraction   Ammonia  loading   Temperature   Moisture  content   Residence  time   Xylanase  addition   Harvest  x  ammonia   Harvest  x  temperature   Harvest  x  moisture     Harvest  x  residence  time   Harvest  x  xylanase   Fraction  x  ammonia   Fraction  x  temperature   Fraction  x  moisture   Fraction  x  residence  time   Fraction  x  xylanase   24  h  Glucose   0.775   0.000*   0.000*   0.082   0.018*   0.001*   0.001*   0.007*   0.918   0.687   0.829   0.919   0.288   0.746   0.278   0.916   0.711   p-­‐value   72  h  Glucose   24  h  Xylose   0.437   0.000*   0.006*   0.526   0.000*   0.000*   0.161   0.000*   0.002*   0.000*   0.000*   0.003*   0.002*   0.000*   0.001*   0.001*   0.932   0.824   0.762   0.943   0.719   0.377   0.760   0.111   0.080   0.416   0.684   0.588   0.075   0.163   0.859   0.715   0.300   0.008*   *Significant  at  α  =  0.05.   114     72  h  Xylose   0.000*   0.528   0.000*   0.022*   0.000*   0.000*   0.000*   0.002*   0.392   0.424   0.317   0.063   0.152   0.400   0.109   0.542   0.030*     TM Figure  5.4:  Interaction  effect  plot  of  AFEX     parameters,  stover  fraction  and  harvest  period   on   monomeric   (A)   glucose   and   (B)   xylose   yields   (g   sugar   per   kg   untreated   dry   biomass)   following  72  h  enzymatic  hydrolysis.    MC  =  moisture  content,  RT  =  residence  time,  DWB  =  dry-­‐ weight  basis,  N  =  no  xylanase  added,  Y  =  xylanase  added  (10%  of  total  cellulase  protein).     5.3.3.  Statistical  analysis   Multivariate   analysis   of   variance   (MANOVA)   was   conducted   in   order   to   determine   the   TM significance  of  harvest  date,  corn  stover  fraction,  AFEX   parameters  and  xylanase  addition  on   both   the   24   hour   and   72   hour   monomeric   glucose   and   xylose   yields.   Interactive   effects   were   also  examined  between  harvest  date  and  stover  fraction  and  each  of  the  other  parameters.  As   115       the  conclusions  regarding  significance  were  the  same  for  24  hour  and  72  hour  yields  for  both   glucose  and  xylose  (Table   5.3),  only  the  72  hour  yields  were  used  for  the  interactive  effects  plot   TM (Figure   5.4).   Glucose   yields   were   significantly   affected   by   three   of   the   AFEX   pretreatment   conditions:   ammonia   loading,   moisture   content   and   residence   time,   but   not   by   temperature.   Glucose   yields   were   also   dependent   on   the   corn   stover   fraction   and   whether   xylanase   was   added   to   the   hydrolysis   cocktail.   Of   the   interactive   effects   analyzed,   only   harvest   date   x   ammonia   loading   had   any   significant   affect   on   monomeric   glucose   yields.     If   the   α-­‐value   is   increased  to  0.1,  the  fraction  x  ammonia  and  fraction  x  moisture  also  significantly  affect  72  h   glucose  yields.  However,  compared  to  the  majority  of  the  other  significant  parameters  (except   the   moisture   content   and   harvest   x   ammonia   effect   on   24-­‐hour   glucose   yields),   which   are   significant  at  α  <  0.005,  the  effect  of  these  two  interactions  on  the  glucose  yield  seems  minimal.     TM Xylose   yields   were   significantly   affected   by   all   four   AFEX   pretreatment   conditions,   including   temperature.   Unlike   the   case   for   glucose   yields,   xylose   yields   were   not   significantly   affected   by   corn   stover   fraction,   but   they   were   affected   by   both   the   harvest   date   and   the   addition   of   xylanase   to   the   hydrolysis   cocktail.   There   were   also   interactive   effects   on   xylose   yields  from  harvest  date  x  ammonia  loading  and  corn  stover  fraction  x  xylanase  addition.     When   analyzing   the   interactive   effects   plot   (Figure   5.4),   significant   interactive   effects   will  have  very  different  slopes  for  the  different  lines  in  that  portion  of  the  graph.  For  example,   when  observing  the  interactive  effect  of  harvest  x  ammonia  on  xylose  yields,  the  slope  of  the   early  and  late  harvest  lines  are  roughly  the  same  when  the  ammonia  loading  is  increased  from   -­‐1 0.5  to  1.0  (g  NH3  g  biomass).  However,  when  the  ammonia  loading  is  increased  from  1.0  to  1.5   116       -­‐1 (g  NH3  g  biomass),  the  slope  of  the  late  harvest  line  is  significantly  steeper  than  the  slope  of   the   early   harvest   line.   This   difference   in   slope   signifies   that   most   of   the   impact   of   ammonia   loading   on   this   interaction   is   due   to   the   second,   not   the   first   increase.   This   implies   that   the   higher  ammonia  loading  has  a  greater  effect  on  the  late  harvest  than  the  early  harvest.           Figure  5.5:  Estimated  dry  matter  distribution  for  70%  and  30%  (dry-­‐weight  basis)  harvest  of   late   harvest   corn   stover.     Percentages   of   the   individual   fractions   harvested   are   based   on   the   total  amount  of  each  fraction  available.   117       Table   5.4:   Optimized   harvest   scenarios   based   on   desired   sugar   and   ammonia  fiber  expansion  ammonia  loading.   Harvest  scenario   Optimized  sugar   Ammonia  loading   A   Glucose/total   B   Xylose   C   Xylose   1.0,  1.5   1.0   1.5   Husk   Leaf   Stem   Cob   Husk   Stem   Cob   Leaf   Cob   Husk   Stem   Leaf   -­‐1   (g  NH3  g dry  SHCS)   Best  fraction       Worst  fraction     5.3.4.  Optimization  of  harvest  scenarios   The  conditions  selected  resulted  in  three  scenarios  for  selectively  harvesting  corn  stover   (Table   5.4)   because   the   harvest   scenario   to   maximize   glucose   yields   was   the   same   for   both   ammonia   loadings.   The   relative   amounts   of   harvested   fractions   for   each   scenario   are   represented  in  Figure  5.5  for  both  the  70%  and  30%  harvests.    A  comparison  of  Table  5.5  and   Table   5.6   reveals   that   the   amount   of   corn   stover   harvested   has   the   largest   impact   on   theoretical   ethanol   yield   per   hectare.   Decreasing   stover   collection   from   70%   of   available   material   to   30%,   with   the   same   harvest   scenario,   decreased   theoretical   ethanol   yields   by   852   –   -­‐1 -­‐1 1139  L  ha .  Decreasing  the  ammonia  loading  from  1.5  to  1.0  (g  ammonia  g  biomass)  for  the   -­‐1 same  harvest  scenario  caused  a  decrease  in  the  theoretical  ethanol  yield  of  150  –  462  L  ha ,   while  switching  desired  sugars  from  glucose  to  xylose  (that  is,  changing  harvest  scenarios  but   TM keeping  stover  collection  and  AFEX   and  enzymatic  hydrolysis  conditions  constant)  caused  a   -­‐1 decrease   in   the   theoretical   ethanol   yield   of   29   –   64   L   ha .   In   order   to   determine   the   sensitivity   118       Table  5.5:  Estimated  yields  for  70%  collection  of  selectively  harvested  corn  stover  (SHCS)   TM following  AFEX ,  enzymatic  hydrolysis  and  fermentation.   -1 -1 1.0 g NH3 g dry SHCS 1.5 g NH3 g dry SHCS Scenario Scenario Worst Scenario Scenario Worst A B Case A C case Yield Glucose 273.7 254.2 240.9 331.5 310.8 303.1 Xylose 150.0 153.1 146.6 195.7 206.8 203.2 Total 423.6 407.3 387.5 527.2 517.5 506.3 L kg SHCS Theoretical ethanol 0.274 0.263 0.250 0.341 0.335 0.327 -1 Theoretical ethanol 1648 1585 1508 2051 2014 1970 g sugar -1 kg SHCS -1 L ha   Table  5.6:  Estimated  yields  for  30%  collection  of  selectively  harvested  corn  stover  (SHCS)   TM following  AFEX ,  enzymatic  hydrolysis  and  fermentation.   -1 -1 1.0 g NH3 g dry SHCS 1.5 g NH3 g dry SHCS Scenario Scenario Worst Scenario Scenario Worst A B Case A C case Yield Glucose 305.1 278.2 228.7 354.4 311.3 288.2 Xylose 151.8 161.3 144.0 192.7 215.4 210.2 Total 456.9 439.5 372.6 547.1 526.7 498.3 L kg SHCS Theoretical ethanol 0.295 0.284 0.241 0.354 0.340 0.322 -1 Theoretical ethanol 762 733 621 912 878 831 g sugar -1 kg SHCS -1 L ha   of  changing  the  harvest  scenario,  the  model  was  also  run  for  a  worst  case  scenario,  where  the   biomass  was  harvested  in  a  manner  that  would  give  the  worst  possible  sugar  yields.  The  worst   case  scenario  led  to  a  decrease  in  the  theoretical  ethanol  yields  per  hectare  ranging  from  81  –   119       -­‐1 TM 141  L  ha .  As  expected,  when  comparing  untreated  corn  stover  to  the  AFEX -­‐treated  cases   (data  not  shown),   the   theoretical  ethanol   yield   was   substantially   lower   for   the   untreated   cases:   -­‐1 -­‐1 a  decrease  of  1234  –  1695  L  ha  for  the  70%  harvest  and  527  –  719  L  ha  for  the  30%  harvest.         5.4.  Discussion   5.4.1.  Composition  analysis   Fractions  from  the  late  harvest  tended  to  have  a  slightly  higher  percentage  of  cell  wall   components  and  slightly  lower  percentage  of  ash  compared  to  their  early  harvest  counterparts.   For  corn  stover,  the  increase  in  lignin  and  cellulose  and  the  decrease  in  ash  have  been  observed   elsewhere   [218,   219].   There   is   also   a   general   increase   in   all   cell   wall   components   with   a   decrease  in  soluble  solids  and  non-­‐structural  carbohydrates  and  an  increase  in  lignin  and  xylan   with   increasing   maturity   [207,   220].   This   observed   increase   in   the   cellulose   (glucan),   hemicellulose   (glucan   and   xylan)   and   lignin   content   is   due   to   the   secondary   thickening   of   the   plant  cell  wall  that  continues  to  occur  for  as  long  as  the  plant  matures.  During  this  time  there  is   also   a   decrease   in   ash   content   [158].     However,   while   there   is   a   continual   change   in   the   dry   matter   composition   until   late   in   the   season,   there   tend   to   be   very   small   changes   during   the   grain  harvest  period  [197,  207],  the  time  during  which  our  samples  were  harvested.     TM 5.4.2.  AFEX  pretreatment   TM Based  on  the  final  total  sugar  yields,  the  optimal  AFEX   pretreatment  conditions  were   observed  to  be  consistent  for  all  fractions,  for  both  early  and  late  harvest  corn  stover:  1.5:1  (g   120       -­‐1 NH3  g  biomass),  60%  moisture  content  (dwb),  90°C,  5  min  residence  time  and  10%  xylanase   -­‐1   addition   (mg   xylanase   protein   mg cellulase   protein),   in   addition   to   the   standard   enzyme   mixture   used   during   enzymatic   hydrolysis.   For   AFEX TM -­‐treated   early   harvest   fractions,   the   -­‐1 addition  of  xylanase  at  1.0  (g  NH3  g  biomass)  had  no  effect  on  monomeric  xylose  yields  and   slightly  lowered  the  glucose  yields.  This  drop  in  glucose  yields  could  be  due  to  the  competition   for   binding   sites   on   the   cellulose   chains   between   enzymes   in   the   xylanase   and   cellulase   mixtures.   The   fact   that   there   is   no   increase   in   xylose   yields   with   the   addition   of   xylanase   supports   this   conclusion.     If   the   xylanase,   which   has   a   much   lower   cellulase   activity   [194],   is   competitively   binding   to   the   cellulose   instead   of   the   xylan,   this   could   result   in   a   decrease   in   glucose  yields  with  no  significant  change  in  xylose  yields.     The   higher   optimal   ammonia   loading   for   the   late   harvest   fractions   compared   with   the   early  harvest  could  be  due  to  a  number  of  reasons.  AFEX TM ,  by  the  ammoniation  of  the  active   methoxyl   sites   of   lignin   [221],   may   be   preventing   the   lignin   from   binding   to   the   hydrolysis   -­‐1 enzymes.  This  may  be  one  of  the  main  reasons  for  the  increase  of  0.5  to  1.0  (g  NH3  g   biomass).   However,  if  this  were  the  reason  for  the  difference  in  optimum  ammonia  loading  between  the   early  and  late  harvests,  then  the  lignin  content  of  the  later  harvest  should  be  greater.  This  is  not   the  case  however,  as  statistically  the  lignin  contents  of  the  early  and  late  fractions  are  identical.     Another  possibility  is  that,  although  they  have  identical  lignin  contents,  the  later  harvest  may   contain   a   greater   quantity   of   methoxyl   sites   compared   to   the   early   harvest,   which   may   make   the  later  harvest  lignin  more  reactive  with  ammonia.  In  both  hardwoods  and  grasses  there  has   121       been   an   observed   increase   in   the   amount   of   syringyl   residues   and/or   the   S:G   ratio   as   they   mature  [162,  222]  which  would  increase  the  relative  methoxyl  content  of  the  lignin  (Figure   1.4).   It  is  also  possible  that  the  difference  in  optimal  ammonia  loading  could  be  due  to  the  increase   in  xylan  content  and  possibly  increased  cross-­‐linking  between  hemicellulose  and  lignin  from  the   early   to   late   harvest.   Ferulate   cross-­‐linking   occurs   between   lignin   and   arabinoxylan   in   the   plant   cell   wall,   with   the   ferulates   ether-­‐linked   to   lignin   and   ester-­‐linked   to   the   arabinoxylan   [20].   Ammonolysis   of   the   ferulate   ester   linkages   to   arabinoxylan   side-­‐chains   is   believed   to   be   one   TM major  reaction  occurring  during  the  AFEX   process  [223].  These  mechanisms  may  be  opening   up   the   cell   wall   ultrastructure   more   effectively   at   the   higher   ammonia   loading,   allowing   the   enzymes   greater   access   to   cellulose.   Also,   by   increasing   access   to   the   substrate,   the   xylanase   enzymes   would   have   more   potential   xylan   binding   sites   and   therefore   be   less   likely   to   bind   competitively   to   the   cellulose   chains.   This   could   also   explain   the   increase   in   glucose   yields   with   -­‐1   the  addition  of  xylanase  for  1.5  (g  NH3  g biomass).     These   hypotheses   are   supported   by   the   fact   that   the   husk,   the   material   with   the   lowest   lignin   content,   while   having   the   second-­‐highest   xylan   content,   is   least   affected   by   the   -­‐1   combination  of  increased  ammonia  loading  and  xylanase  addition.  At  1.5  (g  NH3   g biomass),   the  xylose  yield  only  increases  by  6.1%  with  the  addition  of  xylanase  to  the  hydrolysis  cocktail.   The   late   cob,   which   has   a   significantly   higher   lignin   and   xylan   content   than   all   of   the   other   materials,  experiences  the  largest  impact  on  xylose  yields  due  to  the  combination  of  increased   ammonia   loading   and   addition   of   xylanase   -­‐   a   22.5%   increase.   The   higher   ammonia   loading   would   cleave   more   linkages   between   the   hemicellulose   and   lignin,   solubilizing   more   oligomeric   122       and   monomeric   xylose   and,   perhaps,   some   lignin   as   well.   These   exposed,   solubilized   sugars   would  be  much  easier  to  hydrolyze  with  the  xylanase.  It  might  be  possible,  given  the  very  high   xylan   content   of   the   cob,   that   more   xylanase   would   be   needed   to   achieve   near   complete   monomeric   xylose   yields.   As   the   xylanase   loading   was   based   on   a   percentage   of   the   cellulase   loading  (and  therefore  the  glucan  content),  and  because  the  glucan-­‐to-­‐xylan  ratio  for  the  cob   -­‐1   -­‐1   (0.85   g   glucan   g xylan)   is   much   lower   than   the   other   fractions   (1.47   –   1.85   g   glucan   g xylan),   -­‐1   the  xylanase  loading  in  terms  of  the  xylan  content  (mg  xylanase  g xylan)  is  much  lower  for  the   cob  fraction  (Table  5.1).  This  may  be  one  reason  for  the  much  lower  xylose  yield  relative  to  the   maximum  theoretical  xylose  yield  of  the  late  cob  fraction.         5.4.3.  Statistical  analysis   TM All   AFEX   parameters  had  significant  impacts  on  sugar  yields,  except  for  temperature,   which  had  no  significant  effect  on  glucose  yields.  Based  on  least  squares  means  analysis  (data   not  shown),  the  temperature  effect  on  xylose  yield  is  likely  due  to  a  greater  yield  increase  as   the  temperature  is  raised  from  80°  to  90°C  rather  than  the  decrease  in  yield  when  temperature   is  raised  from  90°  to  100°C.  For  the  range  of  conditions  tested,  optimizing  the  ammonia  loading,   moisture   content,   and   residence   time   are   more   important   for   maximizing   sugar   yields   from   corn   stover.   However,   this   conclusion   may   change   for   a   different   range   of   temperatures   and   should  not  be  extrapolated  to  other  conditions.     123       Harvest  date  had  a  significant  impact  on  xylose  yields  but  not  on  glucose  yields.  This  is   largely  due  to  the  fact  that  the  xylan  content  of  the  late  fractions  was  greater  than  the  xylan   content   of   the   early   fractions,   whereas   the   glucan   content   was   not   significantly   different   between  harvests.  The  corn  stover  fractions  tested  had  a  significant  effect  on  glucose  yields  but   not   on   xylose   yields.   The   late   stem,   husk,   leaf,   and   early   stem   fractions   had   no   significant   statistical   difference   in   their   glucan   contents,   so   their   relative   recalcitrance,   in   terms   of   glucose   yields,   can   be   inferred   from   Figure   5.4.     As   the   husk   has   the   highest   glucose   yield,   it   can   be   considered   the   least   recalcitrant,   followed   by   the   leaf   and   then   the   stem.   Inferences   cannot   be   made   regarding   the   cob   because   its   glucan   content   is   statistically   lower   than   the   other   three   fractions.  However,  because  the  cob  approaches  theoretical  glucan  yields  at  optimal  conditions   while  the  stem  does  not  (Figure  5.1),  it  may  be  less  recalcitrant  in  terms  of  glucan  conversion.     For  interactive  effects,  only  two  were  significant:  harvest  date  X  ammonia  loading  and   corn  stover  fraction  x  xylanase  addition.  The  harvest  x  ammonia  interaction  was  significant  for   -­‐1   both   glucose   and   xylose   yields.   The   increase   in   ammonia   loading   from   1.0   to   1.5   (g   NH3   g biomass)  appears  to  have  a  greater  effect  on  the  late  harvest  than  the  early  harvest,  but  this   may  be  due  to  the  lack  of  data  for  the  early  harvest  cob.  As  the  cob  is  the  fraction  most  affected   by   the   increase   in   ammonia   loading,   particularly   for   xylose   yield,   the   lack   of   early   cob   data   may   lead  to  an  apparent  difference  in  effects  that  is  not  actually  present  between  harvests.  None  of   TM the   other   AFEX   parameters   show   this   relationship   with   either   harvest   date   or   corn   stover   fraction,   which   indicates   that   the   same   pretreatment   conditions   (moisture,   temperature   and   124       residence   time)   can   be   used   to   maximize   glucose   release,   regardless   of   the   fractional   composition  of  the  corn  stover  or  the  harvest  date.     The  second  significant  interactive  effect  was  for  corn  stover  fraction  x  xylanase  addition,   but  only  for  xylose  yields.  The  main  reason  for  this  effect,  as  can  be  observed  from  Figure   5.4,  is   due  to  the  cob  fraction  that  was  much  more  strongly  affected  by  the  addition  of  xylanase  than   all  of  the  other  fractions,  whose  responses  were  fairly  similar.  This  conclusion  is  supported  by   the   fact   that   when   the   data   for   the   late   cob   was   removed   from   the   analysis,   the   fraction   x   xylanase   interaction   became   non-­‐significant   (data   not   shown).   Taken   together,   these   results   indicate   that   of   all   the   corn   stover   components,   the   cob   reacts   more   differently   during   enzymatic   hydrolysis.   As   mentioned   previously,   the   cob   may   require   a   much   higher   xylanase   loading   than   the   other   fractions   in   order   to   release   xylose   remaining   in   the   biomass   or   to   TM convert  the  AFEX  solubilized  xylo-­‐oligomers.     5.4.4.  Empirical  modeling  of  harvest  scenarios   As  a  result  of  the  wide  range  of  opinions  on  how  much  corn  stover  can  be  sustainably   harvested   and   because   the   amount   will   likely   change   for   a   given   field   depending   on   environmental  conditions  and  agricultural  practices  [114,  199-­‐201],  we  have  modeled  a  number   of  corn  stover  harvest  scenarios  for  both  a  liberal  harvest  estimate  (70%  of  available  corn  stover)   and  a  conservative  harvest  estimate  (30%  of  available  corn  stover).  The  goal  was  to  determine   which  combination  of  fractions  provides  the  most  benefit  to  the  biorefinery  in  terms  of  sugar   yields,  and  to  determine  the  preferential  order  in  which  fractions  should  harvested.   125         Crofcheck  and  Montross  [209]  found  that  the  weighted  sum  of  the  glucose  yields  from   individual   pretreated   fractions   was   not   statistically   different   from   the   glucose   yield   from   whole   pretreated   corn   stover.   This   means   that   glucose   yields   for   SHCS   could   be   predicted   using   glucose  yields  from  individual  fractions.  Our  estimate  of  the  late  harvest  dry  matter  distribution   of   corn   stover   (Figure  5.5),   was   based   on   published   data   from   four   sources   [207-­‐209,   218],   and   is  similar  to  standard  estimates  of  corn  stover  dry  matter  distribution  near  harvest  [224].  Corn   stover   dry   matter   yields,   particularly   of   the   husk   and   leaf,   tend   to   decrease   rapidly   due   to   weathering  over  the  course  of  the  harvest  season  [207,  208,  213,  218-­‐220,  225].  This  estimate   attempts   to   account   for   both   the   effects   of   the   late   harvest   date   as   well   as   our   inclusion   of   the   leaf  sheath  with  the  leaf  fraction  instead  of  the  stem  fraction,  as  is  often  the  case  [207,  208].         The  estimated  whole  corn  stover  dry  matter  distribution  was  used  to  predict  monomeric   glucose  and  xylose  yields  from  the  three  different  harvest  scenarios  and  the  worst  case  scenario   (where   the   least   digestible   fractions   were   harvested)   for   both   a   70%   (Table   5.5)   and   a   30%   (Table  5.6)   harvest   of   on-­‐field   corn   stover   using   weighted   averaging   of   individual   fraction   sugar   yields.   It   is   important   to   note   that   values   given   in   these   tables   do   not   attempt   to   take   into   account  the  ability  or  inability  to  harvest  the  specific  fractions  or  any  losses  due  to  inefficiencies   in   harvest,   transport   and   storage   of   corn   stover,   which   can   be   significant   depending   on   the   methods   used.   A   recent   study   found   that   the   maximum   amount   of   corn   stover   was   available   at   -­‐1 grain   physiological   maturity   (15.6   t   ha )   and   steadily   decreased   over   the   harvest   period   to   a   -­‐1   minimum   of   8.6   t   ha [225].   As   this   value   takes   into   account   the   late   season   of   harvest,   and   because   it   is   within   the   range   of   most   estimates   of   corn   stover   yields   reported   in   the   published   126       -­‐1 -­‐1 literature   (7.8   –   8.8   t   ha )   [116,   197,   226],   8.6   t   ha   of   available   corn   stover   was   chosen   to   estimate   the   total   sugars   that   could   be   produced   per   hectare   for   the   given   harvest   scenario.   -­‐1 The  standard  value  of  0.51  (theoretical  g  ethanol  produced  g   sugar  consumed)  was  used  to   determine   the   theoretical   ethanol   production   from   both   a   kilogram   of   SHCS   and   a   hectare   of   harvested  SHCS  and  does  not  take  into  account  inefficiencies  of  fermentation.     Harvest  scenario  A,  which  selectively  harvests  the  husk  followed  by  the  leaf,  stem  and,   lastly,  the  cob,  obtained  the  highest  sugar  and  ethanol  yields  of  all  the  scenarios  and,  as  a  result,   was  chosen  as  the  optimal  harvest  scenario  for  AFEX TM -­‐treated  corn  stover.  Harvest  scenario  A   was   also   preferable   to   scenarios   B   and   C   for   a   number   of   other   reasons.   First,   optimizing   the   collection   for   maximum   glucose   yields   is   preferable   because   most   current   and   relevant   microbial  strains  selectively  utilize  hexoses  over  pentoses  as  a  carbon  source  during  ethanolic   fermentation  [205,  227].  Second,  harvest  scenario  A  selectively  leaves  behind  the  more  lignified   fractions   on   the   field   which   may   prove   more   valuable   for   improving   SOC   levels   due   to   the   longer   half-­‐life   of   lignin   compared   to   cellulose   and   hemicellulose   [114,   228].   Lastly,   harvest   scenario  A  seems  to  be  the  most  feasible  option  from  a  technical  viewpoint.           Selective   harvesting   of   corn   stover   fractions   will   involve   either   returning   the   cob   and/or   husk  to  the  field  following  the  removal  of  the  grain  from  the  ear,  and/or  raising  the  header  on   the  combine  to  increase  the  stover  cut  height  [197,  229,  230].  As  a  result  of  the  association  of   the  leaves  with  the  stem,  at  higher  cut  heights  it  would  be  almost  impossible  to  remove  all  of   the  leaves  while  leaving  the  entire  stem  behind.  Taking  these  factors  into  consideration,  of  all   of   the   scenarios,   the   most   feasible   from   a   technical   aspect   would   be:   scenario   A   (70%   harvest),   127       where   all   of   the   cob   and   a   portion   of   the   lower   stem   is   returned   to   the   field;   and   scenario   C   (30%   harvest),   where   only   the   stover   associated   with   the   ear   (husk   and   cob)   is   retained.   Scenario  A  (30%  harvest)  could  also  be  feasible  if  we  replaced  the  percentage  associated  with   the   leaf   material   with   a   mixture   of   the   upper-­‐most   portion   of   the   corn   plant   (leaf   and   stem).   This   might   be   a   reasonable   option,   because   the   upper   portion   of   the   stem   tends   to   be   more   easily  digestible  than  the  lower  portion  of  the  stem  and  also  has  a  higher  sugar  content  than   the   leaf   [221,   229].   So,   harvesting   the   upper   portion   of   the   corn   plant   could   hypothetically   give   higher  yields  than  harvesting  the  leaf  alone.  Unfortunately,  this  cannot  be  modeled  because  for   this  study,  only  the  entire,  homogenized  corn  stem  was  tested.           Crofcheck  and  Montross  [209]  recommended,  based  on  glucose  yields  from  fractionated   corn  stover,  a  roughly  30%  corn  stover  harvest  scenario  where  the  SHCS  was  composed  of  all  of   the  available  cobs  and  74%  of  the  leaves  and  husks,  leaving  the  most  recalcitrant  stalks  on  the   field.  The  difference  between  their  optimal  harvest  scenario  and  ours  is  most  probably  due  to   their   experimental   methods   for   pretreatment   and   the   subsequent   analysis.   Pretreatment   of   lignocellulosic  biomass  using  dilute  sodium  hydroxide  solubilizes  much  of  the  lignin  and  some  of   the   hemicellulose   into   the   liquid   pretreatment   stream   [231,   232].   It   is   therefore   unlikely   that   glucan   content   of   the   pretreated   corn   stover   corresponds   to   glucan   content   of   the   untreated   corn  stover.  For  similar  pretreatment  conditions  of  corn  stover,  Varga  et  al.  found  a  41.9%  mass   loss   from   the   untreated   dry   corn   stover   to   the   pretreated   solids   and   the   composition   of   the   pretreated  material  shifted  in  favor  of  a  higher  glucan  content  [232].  The  cob  has  a  significantly   higher  xylan  and  lignin  content  than  the  other  fractions  of  the  corn  plant  and,  therefore,  it  is   reasonable   to   assume   that   it   will   lose   a   greater   proportion   of   its   mass   following   dilute   alkali   128       pretreatment.   As   this   mass   loss   was   not   taken   into   account   [209],   the   amount   of   glucan   that   could   be   obtained   on   a   mass   basis   from   the   untreated   fractions   was   over-­‐exaggerated,   particularly   from   the   xylan-­‐   and   lignin-­‐rich   cob.   If   the   mass   loss   had   been   taken   into   account,   it   TM is  likely  that  their  choice  of  optimal  fractions  for  harvest  would  have  been  different.  As  AFEX   is  a  dry-­‐to-­‐dry  process  with  insignificant  mass  loss  during  pretreatment,  the  glucan  content  of   the  pretreated  material  can  be  assumed  to  be  the  same  as  the  glucan  content  of  the  untreated   material   [204].   It   is   feasible   because   of   differences   in   reaction   chemistries,   that   other   pretreatment  methods  would  give  different  results  for  the  selective  harvest  ratio  of  corn  stover   TM fractions   compared   to   those   for   AFEX .   However,   because   Crofcheck   and   Montross   did   not   take   into   account   the   mass   losses   which   occurred   during   their   pretreatment   and   as   we   therefore   do   not   know   their   sugar   yields   based   on   the   untreated   stover   fractions,   we   cannot   attribute   the   difference   between   our   results   and   theirs   to   differences   between   the   pretreatment  methods.    Rather  the  difference  may  be  due  to  errors  in  their  analysis.     Shinners  et  al.  [229]  analyzed  the  effect  of  cut  height  of  corn  stover  (a  harvest  scenario   that   leaves   a   portion   of   the   lower   stem   and   leaves   behind)   on   predicted   ethanol   yields   and   found   that   the   amount   of   ethanol   produced   was   only   ~3%   greater   (L   Mg -­‐1   DM)   for   the   low   cut   compared   to   the   high   cut.   If   one   were   to   assume   that   the   amount   of   material   harvested   per   hectare  was  constant,  focusing  only  on  the  composition  differences  in  the  harvested  material,   this  result  would  indicate  that  the  fraction  harvested  has  little  impact  on  the  theoretical  ethanol   production,  which  is  similar  to  our  results.    However,  when  they  analyzed  their  results  based  on   the   ethanol   yield   per   hectare,   the   increase   in   total   dry   matter   harvested   with   the   lower   cut   129       height  increased  the  predicted  ethanol  yield  by  52%  compared  to  the  higher  cut  [229],  which   indicates   that   the   amount   of   material   harvested   has   a   significant   impact   on   theoretical   ethanol   yields  and  corresponds  to  our  findings.     Based   on   these   results,   optimizing   the   fractions   collected   during   harvest   has   a   much   smaller   impact   on   potential   yields   than   optimizing   pretreatment   and   hydrolysis   conditions,   even   if   the   worst   case   scenario   occurs   and   the   least   digestible   materials   are   preferentially   harvested.   However,   the   amount   of   stover   harvested   has   the   greatest   impact   on   theoretical   ethanol   production   per   hectare.   It   will   be   very   important,   in   terms   of   maximizing   ethanol   production,   to   develop   methods   to   efficiently   maximize   harvest   of   corn   stover,   while   still   maintaining  soil  productivity  and  preventing  erosion.             5.5.  Conclusions   TM Based  on  monomeric  glucose  and  xylose  yields,  the  optimal  AFEX   conditions,  for  all   stover  fractions  (leaf,  stem,  husk  and  cob)  regardless  of  harvest  period,  were  found  to  be  1.5  (g   -­‐1 NH3  g   biomass),  60%  moisture  content  (dwb),  90°C  and  5  min  residence  time;  with  enzyme   loading   during   hydrolysis   of   31.3   mg   of   cellulase   (Spezyme®   CP),   41.3   mg   of   β-­‐glucosidase   -­‐1 (Novozyme®  188)  and  3.1  mg  xylanase,  g   glucan.  These  conditions  are  different  from  those   presented   in   previous   analyses   [203,   204]   largely   due   to   the   inclusion   of   xylanase   in   the   hydrolysis  cocktail.  The  addition  of  xylanase  was  necessary  in  order  to  achieve  high  xylose  yields   130       TM at   moderate   cellulase   loadings   and   moderate   AFEX   conditions,   particularly   with   respect   to   the  more  recalcitrant  cob  and  stem  fractions.     The  optimal  harvest  scenario  for  the  collection  of  SHCS  would  harvest  the  husk  followed   by   the   leaves,   then   the   stem,   and   lastly,   the   cob.   This   harvest   scenario   was   independent   of   ammonia  loading  during  AFEX TM   pretreatment  and  maximized  glucose  and  ethanol  yield  from   TM SHCS.   This   scenario,   combined   with   the   optimal   AFEX gave  a  theoretical  ethanol  yield  of  2051  L  ha -­‐1   pretreatment   conditions   for   SHCS,     for  the  70%  dry  matter  harvest  and  912  L  ha-­‐1   for   the   30%   dry   matter   harvest.   Decreasing   the   stover   collection   from   70%   to   30%   dropped   the   -­‐1 ethanol  yield  by  852  –  1139  L  ha ,  depending  on  harvest  scenario  and  pretreatment  conditions.   Maximizing  stover  collection  while  protecting  soil  health  will  be  the  most  important  factor  for   maximizing  ethanol  yields  from  corn  stover.     Optimizing   the   collection   of   corn   stover   fractions   has   little   impact   on   the   theoretical   -­‐1 ethanol  yield  (29  –  141  L  ha ),  especially  compared  to  optimizing  pretreatment  and  hydrolysis   -­‐1 conditions   (150–   462   L   ha ).   The   dry   matter   distribution   of   collected   corn   stover   fractions   is   less   important   than   the   optimization   of   the   ethanol   production   process.   However,   it   is   still   something  that  needs  to  be  taken  into  account  because  harvesting  the  worst  fractions  can  still   decrease  ethanol  yields,  especially  when  a  smaller  percentage  of  the  stover  is  collected.     131       TM MOLECULAR  SCALE:  AFEX CHAPTER  6 :        PRETREATMENT  OF  POPLAR  MODIFIED  FOR  LIGNIN  CONTENT   AND  COMPOSITION     6.1.  Introduction   Lignin   is   one   of   the   three   main   components   present   in   the   higher   plant   cell   wall,   comprising   between   12   –   30%   of   the   total   mass,   depending   on   the   type   of   plant   [18,   233].   Lignin  is  important  for  plant  growth  and  survival  as  it  provides  strength  and  rigidity  to  the  plant   structure  that  is  necessary  for  hydraulic  transport,  as  well  as  passive  defense  against  pest  and   pathogen   attack.   Lignin   is   also   of   interest   to   biofuel   researchers   as   it   represents   both   a   significant   hindrance   to   biological   processing   of   plant   structural   carbohydrates   to   liquid   fuels   [233-­‐235],  and  also,  by  providing  carbon-­‐neutral  energy  and  steam  for  biorefinery  operations,   an  economic  and  environmental  asset  [12,  235].     With   advances   in   plant   transgenic   techniques,   it   has   become   possible   to   tweak   individual  cell  wall  components  by  altering  genes  that  encode  enzymes  in  molecular  pathways.   The   total   lignin   content   and   monomer   composition   can   be   altered   by   either   up-­‐   or   down-­‐ regulating   or   knocking-­‐out   different   enzymes   within   the   synthesis   pathway   (Figure   1.4).   A   number  of  reviews  have  been  published  comparing  the  different  lignin  transgenics  and  mutants   that   have   been   researched   in   recent   years   and   the   corresponding   changes   in   lignin   content,   structure   and   composition   due   to   these   modifications   [22,   25,   27,   236,   237].   By   comparing   modified  plant  materials  with  their  unmodified  counterparts,  it  is  possible  to  examine  changes   in  interactions  due  to  altering  a  single  parameter  in  the  plant  biochemistry.  It  is  also  possible  to   determine  which  modifications  are  most  significant  in  terms  of  improving  plant  digestibility  and   132       reducing   process   costs   either   by   allowing   use   of   a   milder   pretreatment   or   reducing   the   enzyme   loading  required  during  enzymatic  hydrolysis.  Previous  experiments  on  AFEX TM   pretreatment   of   hardwoods   (poplar   and   black   locust)   found   that   high   ammonia   and   water   loadings,   temperatures   (180°C),   and   enzyme   loadings   during   hydrolysis   are   necessary   to   achieve   significant   saccharification   yields   [14,   31].   At   these   conditions,   it   is   unlikely   that   an   AFEX TM -­‐ based   biorefinery   could   profitably   use   hardwoods.   However,   by   genetically   modifying   the   plant   lignin  content  or  composition,  it  may  be  possible  to  increase  yields  sufficiently  to  be  profitable.   It   is   well-­‐known   that   reductions   in   total   lignin   content   can   improve   enzymatic   saccharification   [164,   234,   238,   239],   however   there   is   no   clear   evidence   of   the   effect   of   alterations   to   monomer   composition   [21,   164,   234].   One   study   which   tested   the   saccharification   potential   of   cinnamyl   alcohol   dehydrogenase   (CAD)   down-­‐regulated   switchgrass  with  altered  lignin  content  and  composition  found  a  negative  correlation  between   the   syringyl:guaiacyl   (S:G)   ratio   and   total   sugar   release,   however,   the   authors   concluded   that   the  total  lignin  content  of  the  samples  was  the  determining  factor  for  increased  saccharification   efficiency   [239].   Another   study   on   caffeic   acid   o-­‐methyltransferase   (comt)   downregulated   alfalfa  with  significantly  reduced  S-­‐lignin  content  showed  an  increase  in  sugar  yields  following   dilute-­‐ammonia  pretreatment  compared  to  the  control  [240].  In  contrast  to  these  two  studies,   Studer   et   al.   [241]   found   for   a   large   group   of   naturally   occurring   poplar,   samples   with   S:G   greater   than   2.0   released   more   sugar   compared   to   those   with   S:G   less   than   2.0.   And   finally,   some   studies   find   no   relationship   between   the   lignin   composition   and   sugar   yields.   When   model   plant   cell   walls   were   constructed   with   varying   ratios   of   lignin   subunits,   there   was   no   133       subunit  effect  on  degradation  of  these  materials  by  fungal  enzymes,  or  extraction  of  lignin  by   NaOH   [242],   which   led   the   authors   to   conclude   that   improvements   in   digestibility   that   were   previously   related   to   changes   in   lignin   subunits   were   actually   the   result   of   other   changes   in   cell   wall   chemistry   or   architecture.     However,   as   the   materials   they   used   were   model   primary   plant   cell   walls,   it   is   uncertain   how   well   they   represent   naturally   occurring   plant   materials,   which   are   more  complex  systems.     For   this   project   we   examined   the   interaction   of   three   different   lignin   modifications   in   poplar:   up-­‐regulated   ferulate-­‐5-­‐hydroxylase   (F5H),   down-­‐regulated   4-­‐coumarate:CoA   ligase   TM (4CL),   and   down-­‐regulated   cinnamoyl-­‐CoA   reductase   (CCR)),   with   AFEX   pretreatment   conditions  and  the  subsequent  effect  on  sugar  yields.  We  also  assessed  whether  modifications   to  lignin  content  (CCR-­‐downregulation)  or  composition  and  structure  (C4H::F5H  up-­‐regulation)   would  allow  for  reductions  in  the  severity  of  pretreatment,  by  testing  these  materials  at  three   TM different   sets   of   AFEX   conditions:   low   temperature   –   long   time,   moderate   temperature   –   moderate   time,   high   temperature   –   short   time.   We   also   used   an   AFEX TM -­‐pretreated   poplar   hybrid,   NM-­‐6   (Populus   maximowiczii   x   nigra),   to   determine   an   optimum   combination   of   commercial   enzymes   for   hydrolysis   of   AFEXTM-­‐treated   hardwoods   consisting   of   Accellerase®   1000,  Accellerase®  XY,  and  Multifect®  Pectinase.     134       6.2.  Materials  and  methods   6.2.1.  Feedstock   One   control   and   two   genetically   engineered   lines   of   poplar   that   had   been   down-­‐ regulated  for  cinnamoyl-­‐CoA  reductase  expression  (ΔCCR  5-­‐2-­‐3  and  ΔCCR  5-­‐2-­‐40)  were  grown  in   a  greenhouse  at  VIB-­‐Ghent.  Three  different  progeny  from  each  line  were  examined  separately,   for   a   total   of   nine   samples.   One   control   and   one   genetically   engineered   line   of   hybrid   poplar   (Populus  tremula  x  Populus  alba)  that  had  been  up-­‐regulated  for  ferulate-­‐5-­‐hydroxylase  (F5H)   expression   (C4H::F5H   construct)   were   grown   in   a   greenhouse   at   the   University   of   British   Columbia   [243].   One   control   and   nine   genetically   engineered   lines   of   poplar   that   had   been   down-­‐regulated  for  4-­‐coumarate:CoA  ligase  (4CL)  expression  (35s::antisense  Pt4CL1a  construct)   were  grown  in  the  greenhouse  at  Michigan  State  University.  The  lines  were  selected  for  level  of   suppression  of  protein  expression:  weak  (1,  2  and  3),  medium  (4,  22,  and  32),  and  strong  (7,  12,   and  16);  and  three  progeny  from  each  control  and  transgenic  line  were  selected  and  tested.  For   the   enzyme   mixture   optimization   experiments,   year-­‐old   coppiced   hybrid   poplar   Populus   maximowiczii  x  nigra  (NM-­‐6)  was  harvested  from  the  Michigan  State  University  Tree  Research   Center.   All   of   the   samples   were   manually   debarked,   after   which   the   CCR   and   F5H   poplar   samples   were   milled   twice   through   a   5.56   mm   (7/32”)   screen   and   4CL   and   NM-­‐6   samples   were   milled  once  through  a  6.35  mm  (1/4”)  screen  using  a  Fitzpatrick  JT-­‐6  Homoloid  mill  (Continental   Process   Systems,   Inc,   Westmont,   Illinois,   USA).   All   samples   were   then   milled   through   a   2   mm   screen  using  a  FOSS  Cyclotec  Mill  (FOSS,  Hillerød,  Denmark)  and  stored  at  room  temperature.       135       6.2.2.  Composition  analysis   Biomass   composition   was   determined   by   the   Great   Lakes   Bioenergy   Research   Center   (GLBRC)   cell   wall   analytical   platform.   Samples   were   initially   ground   into   a   powder   by   robot   and   then   sequentially   treated   with   70%   ethanol   followed   by   an   extraction   using   1:1   (v/v)   chloroform/methanol   solution,   in   order   to   remove   soluble   materials   [244].   Any   starch   in   the   sample   was   then   removed   via   amylase   treatment.   The   resulting   cell   wall   material   was   then   analyzed   for   hemicellulose   sugars,   crystalline   cellulose,   acetyl   bromide   lignin   [245]   and   lignin   composition  via  thioacidolysis  [246].  The  hemicellulose  sugar  composition  was  determined  by   treating   the   extracted   samples   with   trifluoroacetic   acid   and   then   deriviatizing   the   solubilized   monosaccharides   into   alditol-­‐acetates,   which   were   then   separated   and   quantified   by   GC-­‐MS   [247].  Crystalline  cellulose  was  isolated  from  the  cell  wall  residue  using  Updegraff  reagent  and   then   hydrolyzed   with   sulfuric   acid   to   generate   glucose   that   was   quantified   using   an   anthrone   colorimetric  assay  [248].  The  composition  of  the  NM-­‐6,  CCR,  F5H,  and  4CL  samples  are  reported   in  the  supplemental  information  (Table  D.1).       6.2.3.  Acidic  and  alkaline  digestibility  assay   Alkaline  and  acidic  pretreatment  of  the  F5H  and  control  poplar,  followed  by  enzymatic   digestibility   assays,   were   conducted   by   the   GLBRC   cell   wall   digestibility   platform   using   their   micro-­‐scale   method   for   analyzing   biomass   digestibility   [249].   Untreated   samples   were   compared  to  those  that  had  been  pretreated  using  2%  H2SO4  at  120°C;  2.0%,  0.2%,  and  0.02%   136       NaOH  at  90°C;  and  hot  water  at  90°C.  All  samples  were  hydrolyzed  using  Accellerase®  1000.   Glucose  release  was  determined  using  an  enzymatic  assay  kit  (K-­‐GLUC,  Megazyme,  Ireland).     TM 6.2.4.  AFEX  pretreatment   TM AFEX   pretreatment  was  conducted  in  22  mL  reactors  as  outlined  by  Bals  et  al.  [150].   Pretreatment  conditions  were  chosen  for  their  similarity  to  previously  determined  optima  for   TM AFEX   pretreatment  of  hardwoods  [14,  31],  however  water  loading  was  reduced  in  order  to   minimize   potential   soluble   mass   losses   during   the   venting   and   unloading   steps.   Four   batches   of   NM-­‐6   poplar   and   two   batches   of   each   4CL   sample   were   pretreated   at   1.0   g   NH3:g   DM,   1.0   g   H2O:g   DM,   180°C,   and   20   min   residence   time.   The   batches   for   each   sample   were   then   combined  to  homogenize  batch  differences  prior  to  enzymatic  hydrolysis.  Two  batches  of  each   CCR  and  F5H  sample  were  pretreated  using  1.0  g  NH3:g  DM,  1.0  g  H2O:g  DM,  and  because  of   the   significant   impact   of   temperature   on   yields   from   hardwoods   when   operated   at   the   same   residence   time   [14,   31],   we   chose   to   test   three   different   temperature   and   residence   time   combinations:   180°C   for   20   min,   120°C   for   60   min,   and   60°C   for   240   min.   These   conditions   were   chosen   to   compare   the   optimum   high-­‐temperature,   short-­‐time   to   a   low-­‐temperature,   long-­‐time  pretreatment  that  may  be  suitable  for  a  regional  processing  center  [250].     137       Untreated  and  pretreated  biomass  moisture  content  was  determined  using  a  moisture   analyzer  (A&D,  Model  MF-­‐50;  San  Jose,  CA).  Prior  to  microplate  enzymatic  hydrolysis,  untreated   and  pretreated  samples  were  milled  through  a  0.5  mm  screen  using  the  FOSS  Cyclotec  Mill.       6.2.5.  Enzyme  optimization  -­‐  enzymatic  hydrolysis  and  sugar  analysis   Enzyme  protein  content  was  determined  using  trichloroacetic  acid  (TCA)  precipitation  to   remove   non-­‐protein   nitrogen   [151]   follwed   by   total   N   analysis   using   the   Dumas   method   for   combustion   of   nitrogen   to   NOx   [148].   Enzyme   protein   contents   were:   Accellerase®   1500   (67   mg   protein/mL),   Accellerase®   XY   (29   mg   protein/mL),   and   Multifect®   Pectinase   (72   mg   protein/mL)  (Genencor  Division  of  Danisco  US,  Inc.,  New  York,  USA).     The   pretreated   NM-­‐6   poplar   was   used   to   optimize   the   enzyme   cocktail   that   was   then   used   for   the   transgenic   poplar   experiments.   Enzymatic   hydrolysis   was   conducted   using   the   microplate   technique   [251].   Samples   were   loaded   at   0.2%   glucan   loading   with   one   of   two   different  enzyme  loadings:  15  mg  total  protein  per  g  glucan  or  30  mg  total  protein  per  g  glucan.   Four  replicates  were  run  for  each  enzyme  combination.  Glucose  and  xylose  release  were  both   analyzed   using   bio-­‐enzymatic   assay   kits   (glucose:   R-­‐Biopharm,   Inc.,   Marshall,   MI;   xylose:   K-­‐ XYLOSE,  Megazyme,  Ireland),  however  due  to  extremely  high  error  associated  with  the  xylose   readings,   those   results   were   inconclusive.   It   is   unclear   why   the   xylose   error   was   large   for   the   enzyme   optimization   experiments,   but   not   for   the   poplar   comparison   experiments.   It   seems   most  likely  that  it  was  either  due  to  the  specific  test  kits  that  were  used  or  to  operator  error.     138       6.2.6.  Poplar  comparison  -­‐  enzymatic  hydrolysis  and  sugar  analysis   Samples  were  hydrolyzed  in  20  mL  screw-­‐cap  vials  at  0.0125  g  cell  wall  sugars  (cellulose,   glucan,   xylan,   arabinan,   mannan   and   galactan)   per   mL   (1.25%   total   sugar   loading)   and   15   mL   final   hydrolysis   volume.   The   pH   was   adjusted   to   4.8   using   1   M   citrate   buffer,   and   to   limit   microbial   contamination,   cycloheximide   and   tetracycline   were   loaded   at   final   concentrations   of   30  and  40  µg/mL,  respectively.  Enzymes  were  loaded  at  24.00  mg  total  protein  per  g  cell  wall   sugars   and   in   the   optimum   proportions   by   mass   protein   as   were   determined   for   the   NM-­‐6   poplar:   Accellerase®   1500   (80%),   Accellerase®   XY   (10%),   and   Multifect®   Pectinase   (10%).   Vials   were  placed  in  a  New  Brunswick  Scientific  (Edison,  NJ)  incubator  shaker  and  hydrolyzed  at  50°C   and   200   rpm   for   168   h.   Samples   were   then   taken   from   each   hydrolysate   at   24   h   and   168   h,   filtered  in  microplates,  and  the  glucose  and  xylose  contents  were  determined  using  microplate   enzyme  assay  kits  (glucose:  R-­‐Biopharm,  Inc.,  Marshall,  MI;  and  xylose:  K-­‐XYLOSE,  Megazyme,   Ireland)  [249,  251].     6.2.7.  Statistical  analysis   Except  for  the  95%  confidence  intervals  on  the  S:G  ratio  that  were  conducted  in  Excel,   all  other  statistical  analyses  including  Pearson  correlation  coefficients  and  p-­‐values,  box  plots,   fully-­‐nested   ANOVAs,   and   Tukey’s   pairwise   comparisons   (95%   CI)   were   conducted   using   Minitab16  Statistical  Software  (2010  Minitab  Inc,  Pennsylvania,  USA).  Pearson  coefficients  and   p-­‐values  were  determined  for  the  correlations  between  different  cell  wall  components  in  the   untreated   samples   and   for   the   correlations   between   cell   wall   components   and   sugar   yields   139       from   untreated   and   pretreated   samples.   For   the   4CL   samples,   the   fully   nested   ANOVAs   and   general  linear  models  for  Tukey’s  comparisons  were  conducted  by  nesting  sample  pool  within   the  line,  and  the  line  within  the  strength  of  downregulation.  For  the  CCR  samples,  the  pool  was   nested  within  the  parent  line.  For  Tukey’s  comparisons  on  sugar  yields  from  the  CCR  and  F5H   samples,   the   interactions   between   pretreatment   x   line   and   pretreatment   x   pool   were   also   included  in  the  general  linear  model.  Letters  were  used  to  indicate  statistically  different  sugar   yields  and  cell  wall  component  values  based  on  Tukey’s  pairwise  comparisons  (α  <  0.05).       Figure  6.1:  Ternary  diagrams  for  enzyme  optimization  experiments  on  AFEX TM -­‐treated  poplar   (NM-­‐6)  using  either  (A)  15  or  (B)  30  mg  total  protein  per  g  glucan.  Hydrolysis  performed  was  in   microplates  at  0.2%  glucan  loading  with  quadruplicates  of  each  enzyme  combination.  Glucose   yields  are  reported  as  the  percentage  of  the  theoretically  available  glucan  (hemicellulose  glucan   and  cellulose)  in  untreated,  dry  biomass  that  was  released.     140       6.3.  Results  and  discussion   6.3.1.  Enzyme  optimization   TM For   previous   work   where   AFEX -­‐treated   poplar   was   hydrolyzed   at   a   constant   cellulase   loading,   supplementation   with   xylanase   significantly   increased   glucan   and   xylan   conversion,   TM which   is   expected   given   that   AFEX -­‐pretreated   poplar   retains   all   of   the   xylan   going   into   enzymatic   hydrolysis   [252].   However,   no   work   was   done   to   determine   an   optimum   enzyme   mixture   for   a   constant   enzyme   loading,   which   is   what   we   attempted   to   do   for   our   experiments   using  NM-­‐6  hybrid  poplar.  Glucose  yields  for  the  various  enzyme  combinations  are  reported  in   Figure   6.1.   Xylose   yields   were   also   measured   using   an   enzyme-­‐based   assay,   however   the   replicate  error  in  most  cases  was  greater  than  the  difference  between  enzyme  combinations,   and   no   conclusions   could   be   made.   For   our   experiments,   the   trends   in   yields   were   similar   regardless   of   the   enzyme   loading   used   (15   mg   total   protein   per   g   glucan   or   30   mg   total   protein   per   g   glucan).   This   similarity   in   optimum   enzyme   combinations   for   different   enzyme   loadings   was  also  observed  for  AFEXTM-­‐treated  switchgrass  (refer  to  Chapter  4).  In  all  cases,  the  xylanase   (Accellerase®  XY)  and  pectinase  (Multifect®  Pectinase)  by  themselves  gave  the  lowest  yields;   however,  some  supplementation  with  the  accessory  enzymes  was  necessary  to  give  the  highest   yields.   Compared   to   the   optimum   enzyme   combination   determined   for   switchgrass,   around   50%   cellulase,   20%   xylanase   and   30%   pectinase   (refer   to   Chapter   4);   the   optimal   enzyme   mixture   for   poplar   required   a   much   higher   cellulase   content,   as   much   as   80%   of   the   enzyme   mixture.  Poplar  glucan  tends  to  be  much  more  indigestible  compared  to  grass  glucan  [14]  and   this   higher   cellulase   loading   may   be   necessary   to   compensate.   For   further   experiments,   the   141       optimal   enzyme   mixture   was   set   at   80%   Accellerase®   1500,   10%   Accellerase®   XY,   and   10%   Multifect®  Pectinase.     6.3.2.  Influence  of  4CL  downregulation  on  cell  wall  composition   4-­‐coumarate:CoA  ligase  (4CL)  occurs  fairly  early  in  the  lignin  synthesis  pathway  (Figure   1.4)   and   catalyzes   the   conversion   of   p-­‐coumaric   acid   to   the   thioester,   p-­‐coumaryl-­‐Co-­‐A.   4CL   downregulation  in  plants  typically  results  in  decreased  total  lignin  content  [253-­‐256].  This  was   observed  for  our  samples,  with  all  but  line  1  containing  less  lignin  than  the  control  line  (Figure   6.2).   There   was   also   a   decreasing   trend   in   lignin   content   as   the   strength   of   downregulation   increased   (Figure   6.3).   However   there   was   no   statistical   difference   in   lignin   content   between   the  medium  and  strong  downregulation.  This  is  in  contrast  to  findings  by  Voelker  et  al.  [257]; although   thioacidolysis   results   indicated   a   reduced   total   lignin   content   for   the   4CL   transgenic   lines  compared  to  the  control,  there  was  very  little  reduction  in  acetyl  bromide  lignin  content   (~10%).  They  attributed  this  to  the  presence  of  flavonoids  that  were  unable  to  be  extracted  by   the   sample   preparation   methods   and   interfered   with   the   UV   absorbance   readings.   There   is   the   possibility  that  this  may  have  occurred  with  our  samples,  which  would  result  in  overestimates   of   the   transgenics’   lignin   content,   and   mean   an   even   stronger   decrease   in   lignin   content   for   these  samples  compared  to  the  control.     In   some   4CL   downregulated   plants,   the   decrease   in   total   lignin   content   is   sometimes   concurrent  with  an  increase  in  the  percentage  of  cellulose  [253-­‐255],  however  it  is  unclear  from   reported  data  whether  the  apparent  increase  is  due  to  the  plant  producing  more  cellulose  or   less  biomass.  In  other  studies  there  is  no  significant  change  in  structural  sugars  [256].  For  our   142         Figure  6.2:  Structural  carbohydrate  and  acetyl  bromide  lignin  content  of  the  control  and  4CL   downregulated   transgenics,   arranged   by   strength   of   downregulation   (weak,   medium,   or   strong)   and   parent   line.   Each  data  point  represents  one  pool.    The  (+)  symbols  represent  the   average   structural   carbohydrate   content   across   pools   for   each   transgenic   line,   and   the   (x)   symbol   represents   the   same   for   acetyl-­‐bromide   lignin.   Values   with   different   letters   for   the   average   line   structural   carbohydrates   or   lignin   are   statistically   different   based   on   Tukey’s   pairwise  comparisons  (95%  CI)  (p  <  0.05).   143         Figure   6.3:   Boxplot   of   lignin,   total   glucose   (from   hemicellulose   glucan   and   crystalline   cellulose),   and   xylose   within   the   cell   wall   for   the   different   strengths   of   4CL   downregulation   compared  to  the  control.  Boxplots  with  different  letters  within  each  component  are  statistically   different  based  on  Tukey’s  pairwise  comparisons  (95%  CI)  (P  <  0.05).     samples   there   was   very   little   statistical   difference   in   total   sugar   content   between   the   lines   (Figure   6.2).   Of   the   parent   lines,   line   2   had   the   highest   structural   carbohydrate   content,   and   lines   22   and   12   had   the   lowest,   and   none   of   the   lines   were   statistically   different   from   the   control.  Total  glucose  contents  for  the  different  strengths  of  downregulation  were  statistically   identical   to   the   control   (Figure   6.3),   but   xylose   contents   were   statistically   greater.   While   this   could   indicate   that   some   of   the   lines   responded   to   the   downregulation   by   increasing   xylan   deposition  within  the  cell  wall,  there  was  no  statistical  correlation  between  the  lignin  content   144       and   xylose   content   for   the   samples   (Table  6.1).   This   indicates   that   a   reduction   in   lignin   content   in  a  specific  sample  did  not  necessarily  correspond  to  an  increase  in  xylan.     The   acetyl   bromide   lignin   content   was   most   influenced   by   the   strength   of   the   downregulation,   which   explained   52%   of   the   between   sample   variance   (Table   D.2).     For   the   structural   sugar   content,   the   biggest   influence   on   differences   between   samples   was   from   between   pool   effects,   as   opposed   to   differences   between   the   parent   lines   or   the   strength   of   downregulation.   This   indicates   that   the   downregulation   itself   did   not   contribute   the   most   significant   effects   to   differences   in   sugar   content   between   the   different   samples,   but   differences  were  mostly  due  to  natural  variability.  Three  of  the  sugars,  mannose,  glucose,  and   crystalline  cellulose  were  most  influenced  by  error  in  the  method.  Glucomannans,  particularly   those   with   low   galactose   substitution   such   as   are   present   in   hardwoods,   can   be   tightly   associated  with  the  cellulose  microfibril  [258].  It  is  likely  that  small  errors  in  sample  preparation   may   not   solubilize   the   most   tightly   associated   glucomannans.   There   is   a   small   set   of   samples   with  decidedly  lower  glucose  and  mannose  hemicellulose  sugar  contents,  so  It  seems  likely  that   for   these   samples   the   glucomannans   were   not   effectively   solubilized   during   composition   processing   (Figure   D.1).   The   hemicellulose   glucose,   and   perhaps   the   mannose,   would   be   measured  as  cellulose  in  the  colorimetric  assay  used  for  cellulose  quantification,  and  we  do  see   that   those   samples   with   an   abnormally   low   glucose   content   also   had   an   abnormally   high   cellulose   content.   This   error   should   not   have   strongly   impacted   the   enzymatic   hydrolysis,   as   samples  were  loaded  on  total  cell  wall  sugars  basis  rather  than  a  cellulose  basis,  and  mannose   content  for  the  samples  were  comparatively  low.     145       Table   6.1:   Pearson   coefficients   for   4CL   poplar   sample   cell   wall   components.   Data   were   analyzed   as   the   combined   wildtype   and   transgenic  samples.     Total   Ara   Xyl   Man   Gal   Glc   Cry   Sugars   a   Xyl   0.50     Man   -­‐0.20   -­‐0.04     Gal   0.19   Glc           a                             -­‐0.04   -­‐0.001   -­‐0.19   0.68 -­‐0.23 0.04   0.96 0.57 Cry   -­‐0.11   -­‐0.48 -­‐0.14   0.23 Lignin   0.05   -­‐0.04   0.25 c   a   a   c   a   c   c   0.22 a c   -­‐0.25 0.18   b c Influence  was  statistically  significant  with   p  =  0.000,   p  <  0.01,   p  <0.05.   Ara  =  arabinose;  Xyl  =  xylose;  Man  =  mannose;  Gal  =  galactose;  Glc  =   glucose;  Cry  =  crystalline  cellulose;     Of  the  cell  wall  sugars,  there  was  a  high  level  of  positive  correlation  (p  =  0.00)  between   xylose   and   arabinose   (R   =   0.50);   mannose,   galactose   and   glucose   (R~0.57-­‐0.96);   and   a   strong   negative   correlation   between   crystalline   cellulose   and   xylose   content   (R   =   -­‐0.48),   which   is   more   difficult  to  interpret  (Table   6.1).  The  strong  correlation  between  glucose  and  mannose  (R  =  0.96)   relates   to   the   sample   glucomannan   content,   the   second   most   abundant   hardwood   hemicellulose  after  4-­‐O-­‐methyl-­‐glucuronoxylan  [18,  259,  260].  The  small  amount  of  arabinose   and   galactose,   and   their   weaker   correlations   to   xylose   and   mannose/glucose,   respectively   (R   ~0.5   –   0.7)   could   be   related   to   glucuronoarabinoxylan   and   galactoglucomannan   which   can   be   present   in   very   small   amounts   in   dicot   cell   walls   [15].   However,   most   of   the   rhamnose,   arabinose  and  galactose  are  likely  derived  from  pectins,  which  are  a  major  component  of  the   dicot   primary   cell   wall   [16].   Of   the   sugars,   mannose   and   galactose   have   a   slight   positive   correlation  to  lignin  content  (p  <  0.05,  R  =  0.25).  In  work  examining  lignin-­‐carbohydrate  linkages   146       it   was   found   that   galactose   and   xylose   residues   are   the   most   common   residues   linked   to   beech   residual   lignin,   and   mannose   and   galactose   residues   are   most   commonly   linked   to   spruce   residual  lignin  [261].  The  observed  correlation  may  indicate  that  there  is  a  tendency  for  reduced   linkages  between  the  hemicellulose  sugars  and  lignin  with  decreased  lignin  content.   The  S:G  ratio  for  all  the  4CL  samples  were  between  1.8  and  2.2  (except  for  one  sample   at   2.4),   which   is   a   similar   range   to   what   has   been   reported   previously   [253,   257].   There   are   conflicting   reports   on   the   effect   of   4CL   downregulation   on   the   ratio   of   syringyl   to   guaiacyl   monomers  (S:G  ratio)  in  angiosperms,  in  some  cases  showing  no  apparent  change  compared  to   the   control   [253,   254],   and   in   others   a   slight   increase   for   some   of   the   lines   [257].   Six   out   of   the         Figure   6.4:   95%   confidence   intervals   around   the   mean   S:G   ratio   for   each   4CL   control   and   transgenic   line.   UCL   and   LCL   lines   represent   the   upper   and   lower   confidence   limit   for   all   the   control   samples.   Stars   represent   lines   that   were   statistically   different   from   all   the   control   samples.  Ctrl  =  control.     147       27  transgenic  lines  showed  statistically  higher  S:G  ratios  compared  to  the  control  (Figure   6.4),   however,   there   appeared   to   be   no   relationship   with   either   strength   of   downregulation   or   acetyl-­‐bromide   lignin   content.   In   a   population   of   1,100   Populus   trichocarpa   samples,   the   average   S:G   ratio   was   2.0,   but   ranged   from   1.0   to   3.0   [241].   It   is   entirely   likely   that   the   range   in   S:G  ratios  observed  for  our  samples  fall  within  the  natural  variation  for  the  population.       6.3.3.  Influence  of  CCR  downregulation  on  cell  wall  composition     Cinnamoyl-­‐CoA   reductase   (CCR)   occurs   later   in   the   lignin   synthesis   pathway   than   4CL   and   catalyzes   the   transformation   of   feruloyl   and   p-­‐coumaryl   thioesters   to   their   respective   aldehydes   (Figure   1.4).   Like   4CL   downregulation,   CCR   downregulation   typically   results   in   decreased   total   lignin   content   [238,   262-­‐266].   For   our   samples,   the   control   line   had   the   highest   lignin  content,  followed  by  5-­‐2-­‐40,  and  then  5-­‐2-­‐3  (Table  D.1).  Of  the  samples,  only  5-­‐2-­‐40  pool   3   had   statistically   identical   lignin   content   to   the   controls.   The   lignin   contents   of   all   other   transgenic  lines  were  statistically  identical  to  each  other  and  lower  than  the  controls.  Like  4CL   downregulation,   there   are   also   conflicting   reports   on   the   effect   of   CCR   downregulation   on   structural   carbohydrates.   One   experiment   saw   an   increase   in   proportion   of   structural   carbohydrates   following   CCR   downregulation   [265].   Another   found   an   increase   in   cellulose   content,  but  a  decrease  in  hemicellulose  sugars  [266].  Arabinose  was  statistically  lower  in  the   transgenics;   galactose   was   higher   in   5-­‐2-­‐3;   and   xylose,   mannose,   hemicellulose   glucose   and   crystalline   cellulose   were   statistically   identical   for   all   the   lines   (Table  D.1).   Because   glucose   and   xylose   contents   are   statistically   identical,   when   the   enzymatic   hydrolysis   xylose   and   glucose   sugar   yields   (%   of   theoretical)   are   compared   between   samples,   they   should   show   a   similar   148       Table  6.2:  Pearson  coefficients  for  CCR  poplar  sample  cell  wall  components.  Data  were   analyzed  as  the  combined  wildtype  and  transgenic  samples.   Total     Ara   Xyl   Man   Gal   Glc   Cry   Lignin   S   G   Sugars   c Xyl   0.44                     Man   -­‐0.17   Gal   Glc   Cry   c -­‐0.42     c 0.15   b   0.48   b   0.23   a                             -­‐0.50 -­‐0.56   0.90 0.13               -­‐0.07   0.51   0.01   0.25   -­‐0.01             0.13   0.07   -­‐0.24   -­‐0.21   -­‐0.02   0.03         a Lignin   0.73   S:G   -­‐0.40 0.05   -­‐0.17   -­‐0.01   0.09   0.26   0.21   -­‐0.58 b       S   -­‐0.27   -­‐0.04   -­‐0.09   -­‐0.08   0.13   0.19   0.13   c     G   c -­‐0.43   0.46   -­‐0.09   0.19   0.64   -­‐0.91   a   -­‐0.23   0.26   -­‐0.20   a -­‐0.63   0.25   H   c   b a -­‐0.01   -­‐0.08   -­‐0.29   -­‐0.25   -­‐0.15   0.22   -­‐0.15   0.09   0.16   a b c Influence  was  statistically  significant  with   p  =  0.000,   p  <  0.01,   p  <0.05.   Ara  =  arabinose;  Xyl  =  xylose;  Man  =  mannose;  Gal  =  galactose;  Glc  =  glucose;  Cry  =   crystalline  cellulose;  S  =  syringyl  lignin;  G  =  guaiacyl  lignin;  H  =  p-­‐hydroxyphenyl  lignin     relationship   to   the   mass   sugar   yields.   As   with   the   4CL   samples,   there   was   a   strong   positive   correlation  between  hemicellulose  glucose  and  mannose  contents  (R  =  0.90,  p  =  0.000)  (Table   6.2)  indicative  of  the  presence  of  glucomannan  in  the  samples.       The   largest   influence   on   differences   in   structural   sugar   contents   between   the   CCR   samples  was  due  to  between  pool  effects  (Table   D.2).  This  indicates  that  the  downregulation   did   not   contribute   the   most   significant   effects   to   differences   in   sugar   content   between   the   different  samples.  Conversely  the  lignin  content  was  most  strongly  influenced  by  the  transgenic   modification.  Like  the  sugars,  the  variability  in  the  syringyl  and  guaiacyl  lignin  monomer  content,   was   mostly   explained   by   differences   between   pools   (natural   variation).   For   CCR   down-­‐regulaed   angiosperms,  generally  there  is  a  decrease  in  total  monomer  yield  and  an  increase  in  the  S:G   149       ratio  [238,  263-­‐265],  though  in  some  cases  there  is  no  significant  difference  [266].  In  our  case,   the  S:G  ratio  was  negatively  correlated  with  lignin  content  (R  =  -­‐0.58,  p  <  0.01)  (Table  6.2)  and   three   of   the   six   transgenic   lines   showed   a   statistically   significant   increase   in   the   S:G   ratio   compared   to   the   control   samples   (Figure   6.5).   Because   thioacidolysis   selectively   cleaves   the   most   reactive   β-­‐O-­‐4   ether   linkages,   the   increase   in   the   thioacidolysis   S:G   ratio   from   CCR   downregulated  materials  has  been  attributed  to  a  reduction  in  β-­‐O-­‐4  linked  G  units  [263,  264].       Figure  6.5:  95%  confidence  intervals  around  the   mean   S:G   ratio   for   each   CCR   control   and   transgenic  line.  UCL  and  LCL  lines  represent  the   upper   and   lower   confidence   limit   for   all   the   control  samples.  Stars  represent  lines  that  were   statistically   different   from   all   the   control   samples.     6.3.4.  Influence  of  composition  and  pretreatment  on  enzymatic  digestibility  –  F5H  upregulation   Deposition   of   syringyl   and   guaiacyl   lignin   in   the   plant   cell   wall   can   be   controlled   by   manipulating   the   ferulate-­‐5-­‐hydroxylase   (F5H)   enzyme,   which   controls   the   flux   of   150       coniferaldehyde   as   a   precursor   to   syringyl   lignin   (Figure   1.4).   By   knocking   out   the   gene   it   is   possible   to   produce   plants   with   mostly   guaiacyl   residues,   while   up-­‐regulating   using   a   cinnamate-­‐4-­‐hydroxylase   promoter   (C4H)   results   in   mostly   syringyl   residues   [243,   267,   268].   The   lignin   chains   resulting   from   such   up-­‐regulations   are   highly   linear   and   lower   in   molecular   weight  compared  to  the  control  [243],  and  it  is  believed  that  these  structural  changes  increase   the  ability  to  extract  lignin  polymers  from  the  plant  cell  wall  [243,  268].    Samples  of  C4H::F5H   upregulated  and  control  materials  were  initially  examined  for  structural  carbohydrate  and  total   lignin   content   (Table   D.1).   Statistically   there   was   no   difference   in   any   of   the   major   cell   wall   carbohydrates   between   the   control   and   transgenic   samples   (p   <   0.05).   Total   acetyl   bromide       Figure   6.6:   Pretreatment   and   enzymatic   digestibility   assay   for   control   and   C4H::F5H   poplar  samples.   Green   boxplots   represent   no   pretreatment,   red   represents   acidic,   and   blue   represents   alkaline.   Boxplots   with   different   letters   have   statistically   different   glucose  release  based  on  Tukey’s  pairwise  comparisons  (95%  CI),  (p  <  0.05).   151       lignin   content   was   higher   for   the   control   compared   to   the   transgenic,   although   the   statistical   significance  could  not  be  determined.    The  lignin  contents  are  similar  to  those  reported  for  acid   insoluble  lignin  of  these  materials  by  Stewart  et  al.  [243].  In  their  case,  the  total  amount  of  acid   insoluble  plus  acid  soluble  lignin  was  the  same  for  both  samples.     When  the  samples  were  screened  under  a  variety  of  different  pretreatments  followed   by   enzymatic   hydrolysis   (Figure   6.6),   there   was   no   difference   in   glucose   release   between   the   control  and  C4H::F5H  transgenic  for  the  untreated  and  acidic  pretreatments,  although  in  both   cases  the  dilute  acid  pretreatment  was  more  effective  than  both  the  untreated  and  hot  water   pretreatments.  In  contrast,  the  C4H::F5H  transgenic  released  statistically  greater  quantities  of   sugars   compared   to   the   control   under   alkaline   pretreatment   conditions   (p   <   0.05).   Sodium   hydroxide,   unlike   sulfuric   acid   or   other   acidic   pretreatments,   selectively   removes   lignin   from   the   plant   cell   wall   [231,   269].   This   supports   the   hypothesis   that   the   C4H::F5H   lignin   is   more   extractable   than   control   lignin   under   alkaline   conditions,   increasing   the   susceptibility   of   the   poplar  carbohydrates  to  enzymatic  conversion.  As  the  samples  for  the  digestibility  assays  were   loaded  on  a  solids  basis,  there  is  the  potential  for  differences  in  digestibility  to  be  observed  due   to   differences   in   total   glucan   content   of   the   biomass.   However,   because   the   structural   carbohydrate  contents  of  the  two  samples  were  statistically  identical,  the  observed  increase  in   digestibility   for   the   transgenic   sample   should   be   due   primarily   to   the   difference   in   lignin   structure.   The   same   materials   were   also   pretreated   using   AFEX TM   pretreatment  at  three  different   conditions:   low-­‐temperature,   long-­‐time;   moderate-­‐temperature   moderate-­‐time;   and   high-­‐   152         Figure   6.7:   (A)   Glucose   and   (B)   xylose   yields   from   control   and   C4H::F5H   poplar   samples   for   TM   different   AFEX pretreatment   conditions.   Sugar   yields   (168   h)   with   different   letters   within   each   subplot   are   statistically   different   based   on   Tukey’s   pairwise   comparisons   (95%   CI),   (p   <   0.05).   All   samples   were   pretreated   using   1:1   g   NH3:g   DM   and   1:1   g   H2O:g   DM.   Enzymatic   hydrolysis   was   conducted   at   200   rpm   and   1.25%   total   sugar   loading   with   24   mg   total   protein   per  g  cell  wall  sugars  (80%  Accellerase®  1500,  10%  Accellerase®  XY,  10%  Multifect®  Pectinase).   Ctrl  =  Control;  F5H  =  C4H::F5H.     temperature,  short-­‐time.  The  glucose  and  xylose  yields  from  the  untreated  transgenic  samples   were  identical  to  the  control  (Figure  6.7),  as  was  observed  for  the  digestibility  assay.  The  xylose   yields   for   the   pretreated   samples   from   both   the   control   and   the   transgenic   were   statistically   identical   at   all   pretreatment   conditions.   Also,   there   is   no   statistical   effect   of   pretreatment   condition  on  xylose  yields  from  either  sample.  For  the  control  there  was  a  lower  glucose  release   for  the  180°C,  20  min  pretreatment  that  was  not  observed  for  the  C4H::F5H  sample.  The  glass   transition   temperature   in   lignin   is   often   reported   as   being   between   130-­‐150°C   [270]   and   it   is   possible   that   the   lower   molecular   weight   lignin   in   the   C4H::F5H   becomes   more   fluidic   at   the   higher   temperatures   compared   to   the   control,   facilitating   its   removal   from   the   cell   wall.   Interestingly   there   is   not   a   large   difference   in   yields   across   pretreatment   conditions.   This   153       indicates   that   it   may   be   possible   to   operate   at   a   lower   temperature   for   a   longer   residence   time   and   obtain   similar   yields   from   AFEX TM -­‐treated   poplar.   While   there   wasn’t   a   significant   TM improvement  in  yields  for  these  novel  transgenic  materials,  conventional  AFEX   also  does  not   remove   lignin   as   effectively   as   other   alkaline   pretreatments   [30],   this   is   at   least   partly   due   to   the   low   liquid   to   solid   ratio   of   the   pretreatment   method.   For   AFEX TM -­‐treated   hardwoods   there   also  seems  to  be  limitations  on  accessibility  to  the  cellulose  [31].  More  interesting  results  may   be   obtained   by   using   a   pretreatment   method   currently   under   development   which   uses   liquid   ammonia   to   simultaneously   delignify   the   biomass   [271]   and   generate   the   more   readily   digestible  cellulose  IIII  allomorph  [272].     6.3.5.  Influence   of   composition   and     pretreatment   on   enzymatic   digestibility   –   4CL   and   CCR   downregulation   Two   studies   have   looked   at   4CL   and   CCR   downregulated   transgenics   for   biofuel   production.   CCR   down-­‐regulated   alfalfa   showed   improved   in   vitro   dry   matter   digestibility   and   saccharification   efficiency   following   dilute   acid   pretreatment,   which   was   related   to   the   amount   of   total   lignin   reduction   [238].     In   the   study   by   Voelker   et   al.   on   poplar   [257],   except   for   the   most   severely   suppressed   lines   which   had   reduced   yields,   4CL   downregulated   transgenics   showed   no   difference   in   total   sugar   released   (glucose   +   xylose)   following   hot   water   pretreatment  and  enzymatic  hydrolysis  compared  to  the  control  (around  0.55  g  sugar  per  g  DM).   However   they   also   loaded   72.5   mg   of   enzyme   per   g   of   biomass.   For   their   samples   this   was   equivalent   to   around   150   mg   of   enzyme   per   g   glucan,   which   is   roughly   4.5   to   5   times   the   154       amount   of   enzyme   used   in   this   study   and   most   likely   significantly   overloading   the   amount   required   for   conversion.   This   is   particularly   true   for   the   small   particle   size   used   for   their   microplate   hydrolysis,   where   enzyme   mass   transfer   limitations   are   less   significant.   Saturating   with  enzymes  can  mask  differences  in  sugar  yields  between  feedstocks.     The  actual  sugar  yields  from  untreated  and  pretreated  control  and  transgenic  4CL  and   CCR  poplar  for  each  line  are  reported  in  the  supporting  information  (Figure   D.2   -­‐   Figure   D.4).   For  the  pretreated  4CL  materials,  there  was  no  difference  in  24  h  glucose  yields  between  lines   compared   to   the   control.   Except   for   line   22   (medium   downregulation),   all   other   samples   had   higher  24  h  xylose  yields  than  the  control,  with  the  highest  from  lines  7,  32  and  3.  Only  lines  22   and  3  had  higher  168  h  glucose  yields  compared  to  the  control,  and  lines  7,  32,  3,  2,  and  12  had   the   highest   xylose   168   h   xylose   yields,   all   higher   than   the   control.   For   the   pretreated   CCR   samples,   5-­‐2-­‐3   consistently   had   higher   glucose   and   xylose   yields   compared   to   5-­‐2-­‐40   and   the   control.   While   xylose   yields   from   5-­‐2-­‐40   were   lower   than   the   control.   Glucose   yields   from   untreated  poplar  were  between  20-­‐40%  for  all  samples.  Studer  et  al.  also  observed  a  high  yield   for  untreated  poplar  [241].  In  some  cases  their  glucose  release  was  as  high  as  0.36  g  per  g  dry   biomass.   If   their   samples   contained   500   mg   glucose   per   g   dry   biomass,   then   this   would   equal   a   70%  glucose  yield.  Quite  a  few  of  their  samples  yielded  between  0.1  and  0.2  g  glucose  per  g  dry   biomass,  which  would  be  a  20-­‐40%  yield  given  the  hypothetical  glucose  content.   Glucose  release  from  the  4CL  untreated  poplar  samples  was  negatively  correlated  with   xylose   content   (R   =   -­‐0.54,   p   =   0.000)   and   was   not   correlated   with   lignin   content   (Table   6.3,   Figure   6.8).   Once   pretreated,   the   correlation   with   xylose   content   disappears.   As   the   TM pretreatment   yields   indicate,   AFEX   pretreatment   is   very   effective   at   solubilizing   and   155       releasing  all  of  the  4CL  poplar  xylan  (Figure   6.8,   Figure   D.3)  and  this  is  particularly  interesting   given  the  low  release  of  xylose  from  untreated  poplar  (Figure  6.8,  Figure  D.2),  even  given  the   inclusion  of  xylanases  in  the  enzyme  cocktail.  Removal  of  the  interfering  xylan  from  the  cell  wall   improves   access   to   the   cellulose   and   removes   the   negative   relationship   between   xylan   content   and  glucan  conversion.  There  is  no  negative  relationship  between  xylose  content  and  glucose   yields  from  the  untreated  CCR  poplar  samples  (Table  6.3),  which,  although  they  have  a  higher   xylose  content,  have  less  difference  in  xylose  content  compared  to  the  4CL  samples.    In  contrast     Table  6.3:  Pearson  coefficients  for  24  h  and  168  h  enzymatic  hydrolysis  sugar  yields  from   untreated  and  pretreated  4CL  poplar  samples.  Data  were  analyzed  as  the  conglomerate   TM of  all  control  and  transgenic  samples.  AFEX -­‐pretreatment  conditions  were  1:1  g  NH3:g   DM;  1:1  g  H2O:g  DM;  180°C  and  20  min.           Cell  Wall  Composition   Total   Total   Xylose   Lignin   Sugars   Glucose   Untreated     24  Glc   -­‐0.06     24  Xyl   -­‐0.20   0.10   -­‐0.16     168  Glc   -­‐0.08   0.08     168  Xyl   -­‐0.29c   -­‐0.24   Pretreated     24  Glc   -­‐0.57a   -­‐0.59a     24  Xyl   0.29c   0.29c     168  Glc   -­‐0.60a   -­‐0.64a     168  Xyl   0.18   0.14   a -­‐0.53   -­‐0.01     S  %   0.13   -­‐0.15   a -­‐0.49   -­‐0.21   -­‐0.54   0.08   -­‐0.16   0.03   -­‐0.38   -­‐0.18   a b b Sugar  Yields   24   24   168   Glc   Xyl   Glc         c   -­‐0.30       0.97a   -­‐0.23     -­‐0.22   c     0.38   -­‐0.15   -­‐0.28 -­‐0.13   b     a 0.86   -­‐0.10   -­‐0.43   0.30 c   a 0.46   0.08             0.20       c     a -­‐0.17   -­‐0.25 0.75   0.02     b a b -­‐0.37   0.25     0.21   0.69   0.41   a b c Influence  was  statistically  significant  with   p  =  0.000,   p  <  0.01,   p  <0.05.   Glc  =  glucose;  Xyl  =  xylose;  S  %  =  percentage  of  lignin  monomers  as  syringyl  lignin;       156       to  the  glucose  yields,  xylose  yields  from  untreated  4CL  samples  were  negatively  correlated  with   lignin   content.   However,   the   yields   are   so   low   that   there   is   not   much   difference   between   samples.   CCR   untreated   poplar   samples   show   different   relationships   compared   to   the   4CL   samples  in  that  both  glucose  and  xylose  yields  were  negatively  correlated  with  lignin  content   (Table  6.3),  though  the  impact  is  slight  (Figure  6.8).     Although  the  transgenic  modifications  effectively  altered  lignin  content  in  many  of  the   samples  compared  to  the  controls,  the  modification  did  not  have  any  impact  on  glucose  yields     (except   for   the   untreated   and   120°C/60   min   pretreated   CCR   (Table   6.4)),   which   is   somewhat   surprising.  Many  other  studies,  including  our  own  on  AFEX TM -­‐treated  mixed-­‐species  feedstocks   (refer  to  Chapter  3)  have  observed  a  strong  impact  of  lignin  content  on  glucose  yields  from  a   variety   of   different   materials   [164,   234,   238,   239,   241],   so   it   is   unclear   why   there   is   no   relationship   in   this   case.   Completely   opposite   to   our   earlier   findings   we   find   a   negative   correlation  between  lignin  content  and  xylose  yields  for  both  CCR  and  4CL  materials  (Table  6.3,   Table   6.4,  Figure   6.8).  When  the  CCR  transgenics  are  grouped  with  the  control  samples,  there  is   no   apparent   relationship   between   lignin   content   and   xylose   release   (Figure   6.8   -­‐   K   and   L,   dashed  line).  However,  when  they  are  grouped  separately  (Figure  6.8  -­‐  K  and  L,  solid  red  lines),   a  definite  effect  of  lignin  content  on  xylose  release  from  the  transgenics  is  observed.   Studer   et   al.   [241]   reported   that   poplar   S:G   ratios   greater   than   2.0     gave   higher   sugar   release  (glucose  and  xylose)  compared  to  samples  with  S:G  ratios  less  than  2.0.    They  state  that   at  S:G  ratios  >  2.0  there  is  no  impact  by  overall  lignin  content  on  sugar  yields  (glucose  &  xylose).   We  found  very  little  impact  of  S:G  ratio  on  sugar  yields,  perhaps  because  of  the  much  smaller   157         Figure  6.8:  Correlations  between  sugar  yields  and  cell  wall  composition  (g/g  DM)  for  4CL  (A-­‐H)   and   CCR   (I-­‐L)   poplar   samples   across   all   pretreatment   conditions.   The   top   row   shows   correlation  of  sample  xylose  content  (A-­‐B)  and  glucose  content  (C-­‐D)  on  glucose  release  from   4CL   poplar.   The   effect   of   lignin   content   on   glucose   yields   (E-­‐F,   I-­‐J)   and   xylose   yields   (G-­‐H,   K-­‐L)   is   shown   for   4CL   and   CCR   poplar.   Untreated   poplar   samples   are   shown   in   black   and   pretreated   samples  are  shown  in  red.  Sugar  yields  are  expressed  in  terms  of  the  total  sugar  theoretically   present  in  the  untreated  dry  biomass.    Lines  represent  the  linear  regression  for  the  respective   parameters.   The   dashed   lines   in   (K)   and   (L)   represent   all   of   the   pretreated   samples.   The   two   solid  lines  represent  the  separate  regression  for  transgenic  samples  and  control  samples.       158       Table  6.4:  Pearson  coefficients  for  24  h  and  168  h  enzymatic  hydrolysis  sugar  yields  from   untreated  and  pretreated  CCR  poplar  samples.  Data  were  analyzed  as  the  conglomerate   TM of  all  control  and  transgenic  samples.  AFEX -­‐pretreatment  conditions  were  1:1  g  NH3:g   DM;  1:1  g  H2O:g  DM  for  all  pretreated  samples.           Total   Sugars   Cell  Wall  Composition   Total   Xylose   Lignin   Glucose     S  %   Sugar  Yields   24   24   168   Glc   Xyl   Glc   Untreated     24  Glc   -­‐0.04     24  Xyl   -­‐0.38   -­‐0.01   -­‐0.06   -­‐0.83   0.12   -­‐0.26   -­‐0.35   -­‐0.50   -­‐0.26     168  Glc   0.24   0.30   0.05   -­‐0.89   0.35     168  Xyl   -­‐0.24   -­‐0.02   -­‐0.42   -­‐0.66 -­‐0.05   60°C  –  240  min     24  Glc   -­‐0.11   -­‐0.37   0.31   -­‐0.43   0.23   b c -­‐0.46   -­‐0.28   0.23   -­‐0.16   0.34   -­‐0.53   0.41         b   0.64       0.86a   0.51c   -­‐0.38   -­‐0.44   -­‐0.19     0.44   -­‐0.04   0.08   -­‐0.55   0.60   -­‐0.27   -­‐0.65     24  Xyl   -­‐0.61   -­‐0.50     168  Glc   0.06     168  Xyl   -­‐0.51   c 120°C  –  60  min     24  Glc   0.01     c 24  Xyl   -­‐0.54   a c a b   c 0.36     0.91a   0.38       0.56c   0.88a   0.49c   b 168  Xyl   -­‐0.47   -­‐0.18   -­‐0.66   180°C  –  20  min   c   24  Glc   0.56   0.41   0.51   c 0.03   0.27     24  Xyl   -­‐0.05   0.18   -­‐0.33   0.11   0.54     168  Glc   0.63   b 0.35   0.71   -­‐0.34   0.01   b -­‐0.71     a c 0.83   0.57       b           0.51c       0.30   -­‐0.45   0.01   b     0.24   -­‐0.08           c b -­‐0.31   0.53     0.63     b a a c -­‐0.61   0.76     0.87   0.56   c c a -­‐0.26   0.60     0.51   0.89   168  Glc   0.16   168  Xyl   -­‐0.42     c             0.56c     c -­‐0.03   a 0.15   b b   0.62   0.10   c Influence  was  statistically  significant  with   P  =  0.000,   P  <  0.01,   P  <0.05.   Glc  =  glucose;  Xyl  =  xylose;  S  %  =  percentage  of  lignin  monomers  as  syringyl  lignin;         159         c 0.58       -­‐0.36     range   of   S:G   ratios   present   in   our   samples.   Glucose   yields   from   pretreated   4CL   had   a   slight   negative  correlation  to  the  percentage  of  S-­‐lignin  in  the  samples,  and  the  xylose  content  had  a   slight   positive   correlation.   Of   the   CCR   treatments,   only   the   120°C   and   180°C   treatments   showed   any   relationship   to   monomer   composition.   The   samples   pretreated   at   120°C   had   a   positive  correlation  between  the  percent  S-­‐lignin  and  24  h  and  168  h  xylose,  and  168  h  glucose.   There   may   be   some   impact   of   the   S:G   ratio   on   sugar   yields,   but   the   correlations   observed   were   not   consistent   across   the   different   transgenics,   nor   strong   enough   to   make   any   definite   statements.       Figure  6.9:  Influence  of  intial  rate  of  hydrolysis  at  24  h  on  168  hr  glucose   and   xylose   yields   from   the   control   and   transgenic   (A)   4CL   and   (B)   CCR   control   poplar.   Xylose  yields  are  shown  in  red  and  glucose  yields  in  black.   Solid   ines   represent   regressions   on   the   given   sugar   yields   for   separate   untreated   and   pretreated   groupings.   Dashed   lines   in   each   figure   represent  the  linear  regression  for  untreated  and  pretreated  samples  as   one  group.     160       Of  all  the  correlations,  whether  4CL  or  CCR,  untreated  or  pretreated  (except  the  180°C   condition),   the   24   h   glucose   yields   had   the   largest,   most   significant   relationship   with   168   h   glucose   yields,   and   the   same   for   24   h   xylose   yields   with   168   h   xylose   yields   (R   ≥   0.83   (except   pretreated  4CL),  p  =  0.000)    (Table  6.3,  Table  6.4,  Figure  6.9).  While  it  may  be  a  stretch  to  call   24   h   sugar   release   an   initial   rate,   the   sugar   release   at   this   point   is   a   deciding   factor   for   the   amount   of   sugars   released   from   the   biomass   over   longer   periods   of   time.   It   also   appears   to   have   a   greater   impact   on   168   h   sugar   yields   than   most   of   the   other   factors   examined.   It   is   difficult   to   release   glucose   from   AFEX TM -­‐treated   hardwoods,   particularly   at   larger   particle   sizes   [31].  One  issue  that  may  be  hindering  enzymatic  breakdown  of  poplar  samples,  even  more  so   than  lignin  content,  is  cellulose  crystallinity.  Although  poplar  and  corn  stover  they  have  similar   TM crystallinity  indices,  poplar  crystallinity  is  much  less  affected  by  AFEX .  Cellulose  crystallinity  is   also   a   major   factor   determining   the   initial   rate   of   hydrolysis   [273-­‐275].   The   increase   in   initial   TM rate   for   ionic   liquid   pretreatment   compared   to   AFEX   is   likely   one   reason   for   the   comparatively   higher   yields   from   this   method   [276].   After   a   certain   amount   of   cellulose   decrystallization   occurs,   the   cellulose   crystallinity   plays   a   less   important   role   in   limiting   hydrolysis  [274].  By  168  h,  xylose  should  no  longer  be  a  major  hindrance  to  cellulose  conversion,   as   most   of   the   xylan   has   been   solubilized.   However,   it   is   possible   that   even   after   168   h   of   conversion,   hardwood   cellulose   crystallinity   may   still   be   a   major   limiter   of   yields.   Zhu   et   al.   also   found   that   for   long   hydrolysis   periods   crystallinity   is   more   important   for   yields   when   lignin   TM content   is   higher   [275].   As   AFEX   does   little   to   remove   lignin,   it   is   likely   that   the   issue   of   161       poplar   crystallinity   is   exacerbated.   The   addition   of   more   enzymes   can   compensate   for   the   hindrance   of   cellulose   crystallinity   [273,   275]   however   this   is   not   a   practical   solution   from   an   industrial   perspective.   Another   possibility   that   might   be   effective   is   to   completely   alter   the   cellulose  crystallinity  by  using  an  extractive  ammonia  pretreatment,  such  as  is  currently  being   developed  in  our  laboratory  and  was  mentioned  previously  [272].     In   addition   to   the   cellulose   crystallinity,   another   factor   that   may   be   hindering   glucose   yields   is   the   presence   of   hemicellulose   still   retained   within   the   biomass.   Although   xylans   are   TM effectively   released   from   AFEX   treated   hardwoods   in   the   presence   of   hemicellulases,   it   is   unknown   what   is   happening   with   the   other   compounds,   particularly   glucomannnans.   Glucomannans  can  be  tightly  associated  with  cellulose  [258,  277],  and  it  is  unknown  how  they   TM are   affected   by   AFEX   pretreatment.   If   the   glucomannans   are   still   associated   with   the   cellulose  microfibrils  following  pretreatment,  and  if  sufficient  mannan  degrading  enzymes  are   not  present  in  the  enzyme  cocktail,  these  compounds  could  present  a  significant  hindrance  to   conversion   of   hardwood   cell   walls,   and   even   more   so   for   softwoods   which   have   a   higher   glucomannan  content.     6.4.  Conclusions   TM The  optimum  commercial  enzyme  cocktail  for  conversion  of  AFEX   treated  hardwoods   contained   primarily   cellulase   (Accellerase®   1500   –   80%)   with   supplemental   amounts   of   hemicellulases  (Accellerase®  XY  –  10%  and  Multifect®  Pectinase  –  10%),  indicating  that  when   162       TM compared   to   the   grass   optimal   enzyme   mixture   that   contains   around   50%   cellulase,   AFEX -­‐ treated  hardwood  cellulose  is  comparatively  more  difficult  to  break  down.     C4H::F5H   poplar   is   a   promising   feedstock   for   biofuel   production   and   performs   better   TM   with  alkaline  pretreatments  than  acidic  due  to  its  more  readily  extractible  lignin.  For  AFEX – pretreated  materials  however,  there  was  very  little  difference  between  the  transgenic  and  the   TM control,   indicative   of   the   small   amount   of   lignin   AFEX   typically  removes  from  the  biomass.  A   pretreatment   with   a   higher   liquid:solid   ratio   may   perform   better,   improving   extraction   of   biomass   components.   4CL   and   CCR   downregulations   both   resulted   in   transgenic   lines   with   reduced   lignin   contents,   and   in   some   cases,   increased   S:G   ratios.   However,   the   reduction   in   lignin  content  did  not  serve  to  improve  glucose  yields,  for  which  there  was  no  relationship  with   overall  lignin  content.  Instead  reductions  in  lignin  content  were  positively  correlated  to  xylose   yields   both   untreated   and   pretreated   poplar   samples.   However,   the   general   impact   was   still   fairly  low,  less  than  20%  increase  in  yields  for  reductions  in  lignin  content  of  60  mg  per  g  dry   biomass.  The  impact  of  S:G  ratio  on  sugar  yields  was  inconclusive  though  in  some  cases  there   was  some  evidence  for  improvements  in  sugar  yields  due  to  an  increased  S:G  ratio.  Of  the  lines   examined,  some  had  higher  yields  compared  to  the  control,  but  the  differences  were  not  very   pronounced,   particularly   with   respect   to   glucose   yields.   In   general   there   was   little   difference   in   TM the   AFEX   pretreatment   conditions   tested,   indicating   that   high   temperature-­‐short   time   pretreatments   and   low   temperature-­‐long   time   pretreatments   may   be   interchangeable   in   terms   of  biomass  conversion  efficiencies.     163       TM AFEX   is  most  successful  at  solubilizing  xylan  from  hardwood  materials,  achieving  60-­‐ TM   70%  xylan  yields  within  24  hours.  In  most  cases  AFEX improved  release  of  glucose,  however   in   almost   all   cases,   168   hr   hydrolysis   yields   were   less   than   60%.   Two   factors   may   be   limiting   conversion   of   hardwood   cellulose:   cellulose   crystallinity,   which   is   relatively   unchanged   by   TM AFEX   pretreatment,   and   the   potential   association   of   residual   glucomannan   with   cellulose   microfibrils,  hindering  enzyme  access.     164       CHAPTER  7 :     CONCLUSIONS  AND  RECOMMENDATIONS     7.1.  Conclusions   The  characteristics  of  a  given  feedstock  and  its  interactions  with  bioenergy  conversion   processes   influence   the   decisions   that   are   made   with   respect   to   bioenergy   conversion   at   a   variety   of   different   scales.   At   the   landscape   scale,   the   distribution   and   availability   of   the   feedstock   is   one   factor   that   impacts   the   decision   on   where   to   locate   a   new   biorefinery.   In   Mainland   China,   the   majority   of   crop   residues   are   produced   in   the   central,   eastern   and   northeastern  regions,  with  the  highest  yields  in  areas  with  a  high  incidence  of  multi-­‐cropping.   Of   all   the   provinces,   Henan   appears   to   be   the   location   most   suited   for   construction   of   a   biorefinery  due  to  the  large  amount  of  crop  residues  available  for  bioenergy,  the  potential  for   TM rural   development,   and   a   potential   market   for   an   animal   feed   co-­‐product   if   using   an   AFEX   platform   for   ethanol   production.   The   central,   eastern,   and   northeastern   regions   of   China   all   appear  to  be  potential  locations  for  a  biorefinery  due  to  the  large  amount  of  crop  residues  that   are  available  for  bioenergy.   Of   the   scales   examined   for   interactions   between   feedstock   and   biomass   processing,   the   classification  scale  and  the  component  scale  showed  the  largest  differences  between  different   materials.   Comparatively,   there   was   little   difference   in   optimal   pretreatment   and   hydrolysis   conditions   and   resulting   yields   at   the   species   scale   for   two   different   varieties   of   switchgrass.   Across   the   scales,   Klason   lignin   content   was   a   major   inhibitor   of   glucose   yields.   This   was   observed  for  both  the  mixed-­‐species  feedstocks  and  the  corn  stover  fractions.  Interestingly  this   was   not   observed   for   the   poplar   samples   that   had   been   modified   for   reduced   lignin   content,   as   165       this   alteration   resulted   in   no   significant   improvement   to   glucose   yields   following   AFEX TM   pretreatment.   In   contrast,   extreme   changes   to   lignin   composition   showed   some   increase   in   sugar  yields.  The  C4H::F5H  transgenic  poplar  sample  that  had  a  high  proportion  of  syringyl  units   was   slightly   more   digestible   than   the   control   sample.   This   lignin   is   more   extractable   than   the   TM   control   poplar   lignin,   however   because   conventional   AFEX does   little   to   extract   cell   wall   components,   the   difference   in   yields   between   the   two   samples   was   not   very   large.   More   distinct   results   may   be   obtained   using   a   pretreatment   that   is   able   to   more   effectively   solubilize   and  remove  lignin  from  the  plant  cell  wall.   Mixed-­‐species  feedstocks  are  an  ecologically  desirable  bioenergy  feedstock,  but  due  to   their  inherent  heterogeneity,  they  are  often  viewed  negatively  from  a  processing  perspective.     However,   this   seems   to   be   an   unfair   assessment   as   mixed-­‐species   with   a   high   proportion   of   grass   species   were   as   digestible   as   the   conventional   biofuel   feedstock,   corn   stover.   It   may   be   possible   to   increase   yields   of   both   biomass   and   of   soluble   sugars   following   processing   by   managing  mixed-­‐species  stands  for  a  higher  grass  content.  While  this  may  counteract  some  of   the  environmental  benefits  of  these  systems,  with  some  management  they  could  still  be  more   environmentally   beneficial   than   a   corn   monoculture,   and   have   better   production   and   processing  characteristics  than  a  naturally  occurring  system.   Within   the   respective   experiments   on   mixed-­‐species   feedstocks,   switchgrass   varieties,   and   corn   stover   fractions,   it   was   found   that   similar   processing   conditions   could   be   used   to   obtain   high   sugar   yields   from   most   of   the   materials.   It   may   be   possible   to   process   samples   harvested  from  a  similar  location  and  at  the  same  maturity  using  the  same  pretreatment  and   166       hydrolysis   conditions.   But   because   there   will   be   some   samples   where   the   chosen   processing   condition   is   far   from   the   optimum,   it   will   be   up   to   the   operator   to   determine   whether   it   is   preferable   to   maintain   set   operating   conditions   and   sacrifice   some   of   the   potential   yields,   alter   operating   conditions   to   maximize   the   yields,   or   blend   feedstocks   in   order   to   make-­‐up   for   deficiencies  in  low-­‐yielding  materials  and  improve  process  stability.     When  harvesting  plant  materials,  without  considering  other  factors,  it  is  most  desirable   to   harvest   the   portions   of   the   plant   that   give   the   highest   yields.   For   corn   stover,   this   is   both   technically  feasible  and  the  best  option  for  increasing  soil  organic  carbon  levels.  The  optimum   order  of  selective  harvest  was  also  the  order  of  decreasing  lignin  content:  husk  >  leaf  >  stem  >   cob.   This   order   of   harvest   can   be   readily   accomplished   by   raising   the   header   on   the   combine   and  by  ejecting  the  cob  back  onto  the  field  following  removal  of  the  grain.  Additionally,  because   of  the  longer  half-­‐life  of  lignin  compared  to  the  other  cell  wall  components,  by  leaving  behind   the  fractions  with  higher  lignin  content,  soil  organic  carbon  levels  are  more  likely  to  increase.     However,   this   selective   harvest   of   corn   stover   may   not   provide   sufficient   ground   cover   to   prevent  wind  and  water  erosion  and  would  need  to  be  examined  in  greater  detail.       7.2.  Recommendations  for  future  research   There   are   broad   opportunities   to   expand   on   the   research   that   was   presented   here.   In   the   chapter   on   distribution   of   crop   residues   within   Mainland   China,   potential   locations   for   a   biorefinery   were   determined   based   on   the   available   feedstock   supply.   This   information   provides  an  opportunity  for  future  case  studies  that  are  based  in  one  or  more  specific  locations   and   that   investigate   the   feasibility   and   the   impacts   of   constructing   bioenergy   facilities,   167       potentially  looking  at  the  farm-­‐scale  economics,  biorefinery  economics,  environmental  impacts,   and   rural   development   opportunities.   Additionally   it   would   be   worthwhile   to   compare   the   feasibility   and   impacts   of   constructing   and   operating   a   single   centralized   biorefinery   versus   a   decentralized   system   with   a   single   biorefinery   and   local   biomass   processing   depots.   The   decentralized   system   holds   a   great   deal   of   promise   because   farming   in   China,   unlike   the   U.S.,   is   still  largely  operated  at  the  small-­‐scale.     Another  area  of  potential  future  research  could  be  in  development  of  crop  models  that   examine   the   impacts   of   climate,   topography,   and   management   practices   on   cropland,   the   environment,   and   biomass   yields.   These   types   of   models   would   allow   better   estimates   of   the   amount   of   crop   residues   that   should   be   left   on   field.   As   farming   in   China   tends   to   be   more   intensive   than   in   the   U.S.,   models   and   assumptions   that   apply   here   may   not   be   applicable   within   a   Chinese   context   and   a   China-­‐specific   model   would   be   a   very   valuable   tool   for   furthering  the  biofuel  industry.   At   the   classification   scale   there   is   a   need   for   research   that   compares   the   differences   between   representative   materials   and   potential   biofuel   feedstocks   from   each   of   the   different   TM biofuel   classes.   Additionally,   although   AFEX   pretreatment   facilitates   the   removal   of   xylans   from   the   cell   walls   of   grasses,   herbaceous   dicots,   and   hardwoods,   it   is   not   well   known   how   TM AFEX   pretreatment   impacts   the   other   types   of   hemicelluloses,   particularly   mannans   that   could   be   more   strongly   associated   with   cellulose   within   the   plant   cell   wall.   Future   research   TM could   compare   the   impact   of   AFEX   on   solubilization   and   redistribution   of   the   different   classes   of   hemicelluloses   within   the   cell   walls   of   different   classes   of   plant   materials.   It   is   also   168       likely   that   the   necessary   mannases   are   either   not   present   or   are   not   present   in   sufficient   quantities  in  the  enzyme  cocktails  during  enzymatic  hydrolysis.  Future  work  could  also  examine   the   impact   of   adding   these   classes   of   enzymes   to   enzymatic   hydrolysis   of   hardwoods   and   softwoods   to   determine   whether   the   recalcitrance   of   these   materials   is   entirely   due   to   the   ineffectiveness   of   AFEXTM,   or   whether   the   addition   of   appropriate   enzymes   is   sufficient   to   alleviate  this  recalcitrance.     Another   area   of   research   that   would   benefit   from   further   research   is   the   nature   cellulose   recalcitrance   within   the   hardwood   cell   wall.   It   will   be   very   important   to   test   these   TM materials   using   the   extractive   AFEX   pretreatment   that   is   currently   under   development.   It   may  be  that  the  combination  of  lignin  removal  and  the  conversion  of  cellulose  I  to  cellulose  III  is   sufficient  to  increase  digestibility  of  these  materials.   Little  difference  was  found  between  the  processing  conditions  for  the  two  varieties  of   switchgrass  that  were  tested.  However,  it  is  important  to  test  a  larger  number  of  switchgrass   varieties   in   order   to   determine   whether   this   is   a   generalization   that   can   be   made   across   the   board.  It  is  also  important  to  determine  the  impact  of  harvest  date  on  processing  yields  from   switchgrass.  There  are  indications  that  harvest  date  has  a  much  more  significant  effect  on  yields   than   the   differences   between   varieties,   and   this   should   be   examined   in   greater   depth,   particularly   with   respect   to   the   impact   of   maturation   and   over-­‐wintering   on   the   biomass   structure  and  the  subsequent  impact  on  processing  yields.   169                                         APPENDICES   170           APPENDIX  A :  SUPPLEMENTARY  INFORMATION  FOR  CHAPTER  2     Table   A.1:   Paper   and   paperboard   production   (million   metric   tons)   in   2006.   Values   for   provinces  with  greater  than  1  million  metric  tons  of  production  (and  Jiangxi)  are  from  [93].  All   other  values  are  estimated  from  [92].   Paper  and  Paperboard  Production   (million  metric  tons)   East     Shandong     Zhejiang     Guangdong     Jiangsu     Hebei     Fujian     Shanghai     Heilongjiang     Liaoning     Jilin     Tianjin     Beijing     48.80   11.66   10.44   9.69   7.58   3.70   2.54   0.96   0.68   0.68   0.48   0.26   0.15                             Central     Henan     Hunan     Anhui     Guangxi     Hubei     Jiangxi     Inner  Mongolia     Shanxi     Hainan                 171     12.96   6.23   1.80   1.20   1.15   1.13   0.91   0.26   0.26   0.00                                   West     Sichuan     Ningxia     Shaanxi     Xinjiang     Yunnan     Chongqing     Gansu     Guizhou     Tibet     Qinghai           3.24   1.25   0.51   0.51   0.26   0.26   0.26   0.15   0.05   0.00   0.00         Table  A.2:  Total  and  non-­‐pastured  ruminant  animals  (million  head),  grazing  and  pastureland   area  (million  ha),  and  carrying  capacity  per  province  in  2006.  Data  on  ruminant  production  and   pastureland   area   are   from   [70],   and   carrying   capacity   data   are   from   [95],   except   for   Heilongjiang,  Jilin,  Sichuan,  Guizhou  and  Gansu,  which  were  estimated.         National   Beijing   Tianjin   Hebei   Shanxi   Inner  Mongolia   Liaoning   Jilin   Heilongjiang   Shanghai   Jiangsu   Zhejiang   Anhui   Fujian   Jiangxi   Shandong   Henan   Hubei   Hunan   Guangdong   Guangxi   Hainan   Chongqing   Sichuan   Guizhou   Yunnan   Tibet   Shaanxi   Gansu   Qinghai   Ningxia   Xinjiang   Cattle  and   Sheep  and   Buffalo   Goats   million  head   105.9   0.2   0.3   4.8   1.1   6.1   3.3   5.4   5.2   0.1   0.3   0.2   1.4   0.6   2.2   5.7   10.3   3.1   4.1   2.2   4.0   0.8   0.9   9.9   5.1   7.3   6.2   1.7   4.2   4.5   1.0   3.8   285.6   0.8   0.4   15.8   7.5   50.6   6.8   4.6   8.2   0.1   4.0   1.1   5.4   0.8   0.6   23.4   19.4   3.0   5.0   0.4   1.6   0.6   1.2   17.1   2.2   8.3   17.1   6.7   15.9   15.0   3.9   38.4     Grazing  and   Land  Carrying   Non-­‐pastured   Pasture  Area   Capacity     Ruminants       million  ha                                                                   172     261.9   0.0   0.0   0.8   0.7   65.6   0.3   1.0   2.2   0.0   0.0   0.0   0.0   0.0   0.0   0.0   0.0   0.0   0.1   0.0   0.7   0.0   0.2   13.7   1.6   0.8   64.4   3.1   12.6   40.4   2.3   51.1   animal  unit   per  ha   -­‐   -­‐   -­‐   -­‐   -­‐   0.198   -­‐   0.198   0.198   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   0.143   0.143   -­‐   0.143   0.416   0.138   0.184   0.114   0.134   million     animal  units   120.5   0.4   0.3   7.9   2.6   3.3   4.7   6.1   6.4   0.1   1.1   0.4   2.5   0.8   2.3   10.4   14.2   3.7   5.1   2.3   4.3   0.9   1.2   11.3   5.3   8.9   0.5   1.7   5.6   0.0   1.5   4.6     Table  A.3:  Estimated  amount  of  vegetable  sown  area  cropped  only  with  vegetables,  either   single-­‐cropped  or  triple-­‐cropped  (not  multi-­‐cropped  with  other  types  of  crops)  for  different   regions  in  Mainland  China.  Regions  defined  based  on  Qiu  et  al.  [96].   Region   Percentage  of  Vegetable  Sown   Area  Cropped  Only  with  Vegetables   Provinces   Heilongjiang,  Inner  Mongolia,   Jilin,  Liaoning   Gansu,  Xinjiang,  Ningxia,   Northwest   Qinghai,  Shaanxi,  Tibet   Beijing,  Hebei,  Henan,   North  China  Plain   Shandong,  Shanxi,  Tianjin   Chongqing,  Guizhou,  Sichuan,   West   Yunnan   Middle  &  Lower   Hubei,  Hunan,  Jiangsu,  Jiangxi,   Yangtze     Shanghai,  Zhejiang   Fujian,  Guangdong,  Guangxi,   South   Hainan   Northeast/North     173     90%   70%   50%   30%   10%   5%     Table   A.4:   National   and   provincial   sown,   cultivated,   and   fallow   land   area   of   cereals,   legumes,   tubers,  oilseeds  and  cotton  crops  in  Mainland  China  in  2006.     Land  Area     (million  ha)       National   Beijing   Tianjin   Hebei   Shanxi   Inner  Mongolia   Liaoning   Jilin   Heilongjiang   Shanghai   Jiangsu   Zhejiang   Anhui   Fujian   Jiangxi   Shandong   Henan   Hubei   Hunan   Guangdong   Guangxi   Hainan   Chongqing   Sichuan   Guizhou   Yunnan   Tibet   Shaanxi   Gansu   Qinghai   Ningxia   Xinjiang     Reported  Sown   Land  Area     124.63   0.23   0.38   7.36   3.46   5.12   3.32   4.61   9.45   0.19   6.15   1.78   8.04   1.52   4.19   8.60   11.65   5.82   5.89   3.08   3.65   0.49   2.75   7.68   3.67   4.50   0.20   3.65   3.00   0.42   0.88   2.92     Calculated   Cultivated     Land  Area     112.62   0.19   0.38   5.73   3.93   6.86   3.75   5.31   11.35   0.24   4.64   1.83   5.63   1.22   2.73   6.60   6.90   4.47   3.51   2.59   3.29   0.65   2.05   5.44   4.13   5.22   0.35   3.77   4.41   0.52   1.06   3.87   174     Estimated   Fallow  Land   9.60   -­‐   -­‐   -­‐   0.47   1.73   0.43   0.70   1.89   0.05   -­‐   0.05   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   0.16   -­‐   -­‐   0.45   0.73   0.15   0.12   1.41   0.10   0.18   0.95   Estimated   Cultivated  Land   Area  of  Crops  of   Interest     103.02   0.19   0.38   5.73   3.46   5.12   3.32   4.61   9.45   0.19   4.64   1.78   5.63   1.22   2.73   6.60   6.90   4.47   3.51   2.59   3.29   0.49   2.05   5.44   3.67   4.50   0.20   3.65   3.00   0.42   0.88   2.92     Table   A.5:   Crop   residue   production   and   use   by   province.   The   amount   that   is   usable   as   fuel   includes   the   amount   used   for   rural   energy   and   assumes   this   amount   is   either   replaced   by   combustion  of  crop  straw  for  electricity  or  that  this  energy  is  replaced  by  some  other  means.   Values  for  the  amount  of  crop  residues  burned  on-­‐field  are  from  Wang  and  Zhang  [100].   Crop  Residues     (million  metric  tons)       Total   National   Beijing   Tianjin   Hebei   Shanxi   Inner  Mongolia   Liaoning   Jilin   Heilongjiang   Shanghai   Jiangsu   Zhejiang   Anhui   Fujian   Jiangxi   Shandong   Henan   Hubei   Hunan   Guangdong   Guangxi   Hainan   Chongqing   Sichuan   Guizhou   Yunnan   Tibet   Shaanxi   Gansu   Qinghai   Ningxia   Xinjiang   593.53   1.20   1.90   32.70   12.25   20.08   18.44   29.64   38.11   1.28   37.86   10.13   37.46   6.99   20.38   50.97   64.36   30.34   31.07   14.06   15.49   1.82   9.62   33.54   13.11   15.97   0.87   12.94   9.11   1.58   3.35   16.91   Pulp  and   Paper   19.35   0.04   0.08   1.10   0.08   0.08   0.20   0.14   0.20   0.29   2.26   3.11   0.36   0.76   0.27   3.47   1.85   0.34   0.54   2.88   0.34   0.00   0.08   0.37   0.01   0.08   0.00   0.15   0.04   0.00   0.15   0.08   Animal   Feed   153.52   0.50   0.44   10.09   3.31   4.15   5.95   7.77   8.20   0.12   1.46   0.55   3.19   1.01   2.96   13.24   18.08   4.77   6.47   2.91   5.45   1.14   1.51   14.42   6.81   11.35   0.58   2.19   7.13   0.05   1.88   5.84   175     Returned   to  Field   Usable  as   Fuel   296.45   0.57   1.14   17.19   10.37   15.37   7.63   10.61   21.74   0.57   13.93   5.34   16.88   3.65   8.20   19.81   20.70   13.40   10.52   7.77   9.87   1.46   6.16   16.32   11.02   13.49   0.45   10.94   9.00   0.97   2.63   8.75   124.69   0.09   0.25   4.32   -­‐1.51   0.48   4.66   11.11   7.96   0.30   20.22   1.14   17.03   1.58   8.95   14.45   23.73   11.83   13.55   0.49   -­‐0.18   -­‐0.77   1.88   2.43   -­‐4.74   -­‐8.95   -­‐0.16   -­‐0.34   -­‐7.07   0.57   -­‐1.32   2.24   Burned   On-­‐Field   103.34   0.25   0.34   5.84   2.26   2.69   2.73   4.53   4.56   0.31   9.53   2.61   9.74   1.99   1.76   9.04   10.04   2.65   8.53   4.25   4.76   0.55   1.03   3.29   1.41   1.76   0.06   2.17   1.37   0.18   0.56   2.56       Figure  A.1:  Residue  density  based  on  prefecture  cultivated  area  (Mg/ha)  for  different  crops  in   2006:  (A)  wheat,  (B)  rice,  (C)  corn,  (D)  cotton,  (E)  oilseeds,  and  (F)  legumes  and  tubers.     176       APPENDIX  B :  SUPPLEMENTARY  INFORMATION  FOR  CHAPTER  3     Table  B.1:  Species  composition  of  GLBRC  early  successional  old  field  replicates.   Scientific  Name   Common  Name   Grasses   Digitaria  sanguinalis  (L.)  Scop.     Echinochloa  crus-­‐galli  (L.)  Beauv.     Panicum  dichotomiflorum  Michx.     Poa  compressa  L.     Setaria  faberi  Herrm.     Setaria  viridis  (L.)  Beauv.     Hairy  Crabgrass   Barnyardgrass   Fall  Panicgrass   Canada  Bluegrass   Giant  Foxtail   Green  Foxtail   Forbs   Abutilon  theophrasti  Medikus     Amaranthus  retroflexus  L.     Capsella  bursa-­‐pastoris  (L.)  Medicus     Chenopodium  album  L.     Lamium  purpureum  L.     Phytolacca  americana  L.     Silene  alba  (Mill.)  E.H.L.Krause     Stellaria  media  (L.)  Vill.     Taraxacum  officinale  Weber   Trifolium  pratense  L.     Trifolium  repens  L.     Veronica  sp.   Unknown  Dicots   Velvetleaf   Redroot  Pigweed   Shepherd's  Purse   Lambsquarters   Purple  Deadnettle   American  Pokeweed   White  Campion   Common  Chickweed   Common  Dandelion   Red  Clover   White  Clover   Speedwell   -­‐   2 ANPP  (g/3m )**   Growth   Habit*   R1  (E7)   A   A   A   P   A   A   128.1   35.0   -­‐   0.0   -­‐   -­‐   1040.5   683.4   0.8   -­‐   -­‐   1.9   540.0   1699.9   -­‐   -­‐   -­‐   19.8   630.7   -­‐   -­‐   -­‐   -­‐   -­‐   1515.8   0.99   -­‐   -­‐   18.35   14.2   A   A   A   A   A   P   B/P   A/P   P   B/P   P   A/P   -­‐   38.9   248.7   285.3   1378.7   0.1   2.3   -­‐   60.8   2.0   -­‐   0.7   -­‐   0.1   46.7   -­‐   5.6   207.8   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   3.5   -­‐   27.1   60.5   4.1   379.5   -­‐   1.6   -­‐   0.1   -­‐   -­‐   -­‐   0.3   0.0   40.9   1219.5   0.5   1134.5   -­‐   -­‐   1.1   0.1   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   319.8   0.1   720.5   -­‐   -­‐   -­‐   -­‐   -­‐   0.0   -­‐   -­‐   -­‐   *A  =  Annual,  B  =  Biennial,  P  =  Perennial;  **ANPP  =  annual  net  primary  productivity   177     R2  (E87)   R3  (E83)   R4  (E21)   R5  (E60)     Table  B.2  :  Species  composition  of  LTER  late  successional  old  field  replicates.   Scientific  Name   Grasses   Arrhenatherum  elatius  (L.)  Beauv.  ex  J.  &  C.  Presl     Dactylis  glomerata  L.     Danthonia  spicata  (L.)  Beauv.  ex  R.  &  S.     Elytrigia  repens  (L.)  Nevski     Panicum  sp.   Phleum  pratense  L.     Poa  compressa  L.     Poa  pratensis  L.     Unknown  Grasses   Forbs   Achillea  millefolium  L.     Alliaria  petiolata  (Bieb.)  Cavara  &  Grande     Antennaria  neglecta  Greene     Apocynum  cannabinum  L.     Asplenium  platyneuron  (L.)  Oakes   Aster  pilosus  Willd.     Barbarea  vulgaris  R.  Br.     Circaea  lutetiana  L.   Daucus  carota  L.     Dianthus  armeria  L.     Desmodium  sp.   Erigeron  annuus  (L.)  Pers.     Common  Name   Tall  Oatgrass   Orchardgrass   Poverty  Oatgrass   Quackgrass   -­‐   Timothy   Canada  Bluegrass   Kentucky  Bluegrass   -­‐   Common  Yarrow   Garlic  Mustard   Field  Pussytoes   Indianhemp   Ebony  Spleenwort   Hairy  White  Oldfield  Aster   Garden  Yellowrocket   Broadleaf  Enchanter's  Nightshade   Queen  Anne's  Lace   Deptford  Pink   Ticktrefoil   Eastern  Daisy  Fleabane   2 ANPP  (g/5m )**   Growth   Habit*   SF1   SF2   SF3   P   P   P   P   P   P   P   P   -­‐   421.4   5.9   -­‐   286.9   -­‐   209.6   0.6   142.8   -­‐   -­‐   -­‐   0.9   -­‐   1.5   0.7   29.3   -­‐   0.0   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   0.2   0.1   0.0   P   A/B   P   P   P(F)   P   B   P   B   A/B   P   A   17.7   -­‐   -­‐   -­‐   0.3   3.3   0.0   -­‐   7.3   0.2   -­‐   -­‐   -­‐   2.9   0.5   0.3   -­‐   -­‐   -­‐   -­‐   -­‐   0.3   -­‐   -­‐   0.0   12.7   -­‐   *   3.5   0.2   *   2.5   -­‐   0.1   1.4   0.1   *A  =  Annual,  B  =  Biennial,  P  =  Perennial,  P(F)  =  Perennial  Fern;  **ANPP  =  annual  net  primary  productivity   178       Table  B.2  (cont’d):  Species  composition  of  LTER  late  successional  old  field  replicates.   Scientific  Name   Common  Name   Forbs   Euphorbia  corollata  L.     Euthamia  graminifolia  (L.)  Nutt.     Geum  laciniatum  Murray   Geum  sp.     Hieracium  sp.   Lactuca  canadensis  L.     Oxalis  stricta  L.     Parthenocissus  quinquefolia  (L.)  Planch.     Phytolacca  americana  L.     Polygonum  convolvulus  L.     Potentilla  recta  L.     Rumex  acetosella  L.     Rumex  obtusifolius  L.     Silene  alba  (Mill.)  E.H.L.Krause     Solidago  canadensis  L.     Solidago  nemoralis  Ait.     Taraxacum  officinale  F.H.  Wigg.   Torilis  japonica  (Houtt.)  DC.     Trifolium  pratense  L.     Trifolium  repens  L.     Verbena  urticifolia  L.     Veronica  chamaedrys  L.     Unknown  Dicots   Flowering  Spurge   Flat-­‐Top  Goldenrod   Rough  Avens   Avens   Hawkweed   Canada  Lettuce   Common  Yellow  Oxalis   Virginia  Creeper   American  Pokeweed   Black  Bindweed   Sulphur  Cinquefoil   Common  Sheep  Sorrel   Bitter  Dock   White  Campion   Canada  Goldenrod   Gray  Goldenrod   Common  Dandelion   Erect  Hedgeparsley   Red  Clover   White  Clover   White  Vervain   Germander  Speedwell   -­‐   Growth   Habit*   P   P   P   P   P   A/B   P   P   P   A   P   P   P   B/P   P   P   P   A   B/P   P   P   P   -­‐   2 ANPP  (g/5m )**   SF1   SF2   SF3   -­‐   0.3   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   0.0   0.4   3.7   -­‐   0.1   200.9   -­‐   -­‐   -­‐   -­‐   0.1   -­‐   -­‐   0.1     *A  =  Annual,  B  =  Biennial,  P  =  Perennial,  P(F)  =  Perennial  Fern;  **ANPP  =  annual  net  primary  productivity   179     -­‐   -­‐   -­‐   0.5   0.2   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   1.6   -­‐   -­‐   -­‐   0.5   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   0.2   2.0   -­‐   4.9   34.6   -­‐   0.8   0.1   5.7   0.4   -­‐   -­‐   -­‐   0.0   -­‐   -­‐   -­‐   0.0   6.0   0.9   -­‐   2.5   1.8   7.0     Table  B.2  (cont’d):  Species  composition  of  LTER  late  successional  old  field  replicates.   Scientific  Name   Common  Name   Woody   Acer  spp.     Celastrus  orbiculatus  Thunb.   Crataegus  spp.     Elaeagnus  umbellata  Thunb.     Lonicera  spp.     Populus  sp.   Prunus  serotina  Ehrh.   Quercus  spp.     Rhamnus  cathartica  L.     Rhamnus  frangula  L.     Rosa  sp.   Rubus  allegheniensis  T.C.  Porter   Rubus  occidentalis  L.   Rubus  sp.   Sassafras  albidum  (Nutt.)  Nees     Toxicodendron  radicans  (L.)  Ktze.     Unknown  Woody   Maple   Oriental  Bittersweet   Hawthorn   Autumn  Olive   Honeysuckle   Cottonwood/Poplar   Black  Cherry   Oak   Common  Buckthorn   Glossy  Buckthorn   Rose   Allegheny  Blackberry   Black  Raspberry   Blackberry   Sassafras   Eastern  Poison  Ivy   -­‐   2 ANPP  (g/5m )**   Growth   Habit*   SF1   P   P   P   P   P   P   P   P   P   P   P   P   P   P   P   P   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐     *A  =  Annual,  B  =  Biennial,  P  =  Perennial,  P(F)  =  Perennial  Fern;  **ANPP  =  annual  net  primary  productivity 180     SF2   0.3   31.3   8.6   0.3   -­‐   -­‐   0.8   0.3   0.8   61.5   -­‐   10.3   -­‐   0.1   2.5   0.0   0.3   SF3   4.3   169.6   2.7   8.3   9.5   1.0   0.7   0.3   8.2   1.5   8.1   12.3   1.7   -­‐   -­‐   1.7   3.9       Table  B.3  :  Pretreatment  conditions  and  total  sugar  yields  for  each  design  point  for  the  early   successional   samples.   Three   extra   design   points   were   included   in   an   attempt   to   improve   the   model  fit.  The  center  design  point  (red)  was  replicated  three  times.  The  pretreatment  condition   chosen   for   further   experiments   is   highlighted   in   teal.     The   top   five   sugar   yields   for   each   sample   and   the   highest   total   sugar   yield   for   each   sample   are   highlighted   in   orange   and   yellow,   respectively.  One  sample  of  E87  (gray)  was  excluded  from  the  analysis  as  a  statistical  outlier.   Pretreatment  Conditions   Ammonia   g:g  DM   Moisture   g:g  DM   0.5   0.5   0.5   1.25   0.5   1.25   0.5   1.25   0.5   1.25   0.5   2.0   1.25   0.5   1.25   0.5   1.25   0.5   1.25   0.5   1.25   1.25   1.25   1.25   1.25   1.25   1.25   1.25   1.25   1.25   1.25   1.25   1.25   1.25   1.25   2.0   1.25   2.0   1.25   2.0   1.25   2.0   2.0   0.5   2.0   1.25   2.0   1.25   2.0   1.25   2.0   1.25   2.0   2.0   Extra  Design  Points   1.25   2.0   2.0   0.5   2.0   2.0   Temp   17.5   17.5   30   5   17.5   17.5   17.5   30   5   17.5   5   30   17.5   17.5   17.5   30   5   17.5   5   30   17.5   17.5   17.5   5   30   17.5   17.5                                                           Total  Mono  &  Oligomeric  Sugar  Yields   (g  glucose  +  xylose  released/   g  untreated,  dry  biomass)   E7   E21   E60   E83   E87   267.8   344.4   432.0   482.7   441.8   250.7   358.3   436.5   457.4   416.9   267.2   302.6   468.8   395.5   480.7   260.3   367.7   403.2   431.1   446.0   217.5   287.7   374.8   493.1   399.5   268.3   346.0   518.9   482.4   471.2   284.3   353.8   457.0   471.5   471.5   266.8   368.6   464.1   527.1   482.3   271.6   307.9   404.1   460.9   411.6   225.2   291.8   398.4   532.5   349.4   303.4   342.7   424.7   479.0   511.3   271.8   359.5   492.3   488.7   428.2   285.2   369.7   469.9   490.2   482.9   297.6   377.6   460.2   386.9   500.4   289.7   362.6   483.8   491.4   484.0   220.7   345.3   393.6   492.8   424.4   280.4   302.7   486.0   460.5   385.8   266.3   343.1   447.8   428.0   463.8   271.2   347.7   442.6   490.6   455.7   277.5   343.5   486.1   466.7   501.0   278.4   328.4   506.7   475.0   417.2   301.4   361.6   472.6   434.8   484.6   305.7   398.8   446.3   417.2   474.2   290.8   361.0   457.3   386.6   295.9   256.9   362.0   478.7   412.1   481.7   273.6   293.4   436.9   369.7   417.5   286.6   342.3   529.9   406.5   493.4   17.5   30   30         269.7   310.1   290.8   °C   Time   min   135   90   135   135   180   135   90   135   135   180   90   90   135   135   135   180   180   90   135   135   180   135   90   135   135   180   135   135   90   90       181     343.6   341.1   354.7   474.4   523.1   518.1   481.3   501.8   469.2   474.2   517.0   488.5     Table   B.4:   Response   surface   model   coefficients   for   pretreatment   optimization   of   the   early   successional   samples   in   terms   of   total   monomeric   and   oligomeric   glucose   and   xylose   release   following   enzymatic   hydrolysis.   A  =  ammonia  loading  (g:g  DM),  B  =  water  loading   2 (g:g   DM),   C   =   temperature   (°C),   D   =   residence   time   (min).     Pred.   R   =   2 2 2 predictive  R  value;  Adj.  R  =  adjusted  R  value.     Response  Surface  Model  Coefficients     A   E7   126.699   53.740   -­‐67.235   1.836   5.070   -­‐14.117   E21   -­‐80.802   163.313   19.372   4.971   4.454   -­‐27.168   E60   153.556   25.934   -­‐76.597   3.255   10.235   -­‐   E83   -­‐15.0175   99.8024   67.6287   5.435   0.9736   -­‐   E87   172.089   18.705   -­‐54.109   4.757   1.810   -­‐   B   2 -­‐   -­‐36.357   -­‐   -­‐   -­‐   C   2 -­‐0.008   -­‐0.021   -­‐0.012   -­‐0.022   -­‐0.023   2 D   AB   BC   CD     Pred.  R   -­‐   -­‐   0.532   -­‐0.037     60.1%   -­‐0.096   -­‐   0.623   -­‐     0.0%   -­‐   -­‐   0.807   -­‐0.067     45.5%   -­‐   -­‐47.747   -­‐   -­‐     66.5%   -­‐   -­‐   0.510   -­‐     78.6%   Adj.  R     72.5%   43.5%   62.4%   72.5%   83.4%   Constant   A   B   C   D   2 2 2 182         Figure  B.1:  Range  of  total  sugar  yields  and  location  of  the   chosen   pretreatment   condition   within   the   range   for   each   early   successional   old   field   sample.   The   red   circle   represents  the  location  of  the  pretreatment  condition  that   was   chosen   as   a   basis   for   further   experimentation   within   the  range  of  the  sugar  yield  raw  data.   183         APPENDIX  C :  SUPPLEMENTARY  INFORMATION  FOR  CHAPTER  4   TM Figure   C.1:   Contour   plots   showing   the   interactive   effect   of   pairs   of   AFEX     pretreatment   parameters  on  monomeric  glucose  yields  from  (A)  Alamo  and  (B)  Shawnee  switchgrass.  The   two   pretreatment   parameters   not   shown   in   each   sub-­‐figure   were   held   at   the   optimal   level.   Enzymatic   hydrolysis   was   conducted   at   50°C,   200   rpm,   and   1%   glucan   loading   using   30   FPU   Spezyme®  CP  and  15  CBU  Novozyme®  188  per  g  glucan,  with  72  h  sampling.   184       TM Figure   C.2:   Contour   plots   showing   the   interactive   effect   of   pairs   of   AFEX     pretreatment   parameters   on   monomeric   xylose   yields   from   (A)   Alamo   and   (B)   Shawnee   switchgrass.   The   two   pretreatment   parameters   not   shown   in   each   sub-­‐figure   were   held   at   the   optimal   level.   Enzymatic   hydrolysis   was   conducted   at   50°C,   200   rpm,   and   1%   glucan   loading   using   30   FPU   Spezyme®  CP  and  15  CBU  Novozyme®  188  per  g  glucan,  with  72  h  sampling.     185       Table   C.1:   Experimental   levels   and   additional   design   points   for   TM Box-­‐Behnken   response   surface   optimization   of   AFEX pretreatment  conditions.     NH3  Loading   H2O  Loading   Temp     Time     (g:g  DM)   (g:g  DM)   (°C)   (min)   +1   2.0   2.0   180   30   0   1.25   1.25   135   17.5   -­‐1   0.5   0.5   90   5   Parameter  Levels     Additional  Design  Points   1.5   2.0   150   30       2.0   2.0   1.25   2.0   150   130   15   30     2.0   2.0   150   30     3.0   2.0   150   15   Shawnee   1.5   2.0   150   30     2.0   2.0   150   30     3.0   2.0   150   15     Alamo         186         Table  C.2:  ANOVA  of  pretreatment  optimization  regression  model  for   total  sugar  yields  (monomers  +  oligomers  of  glucose  and  xylose)  from   Alamo  and  Shawnee  switchgrass.   a b c d Source   DF   Seq  SS   Adj  SS   Adj  MS   Alamo     Regression       Linear       Square       Interaction     Residual     Total   Shawnee     Regression       Linear       Square       Interaction     Residual     Total     10   4   3   3   21   31     8   4   3   1   21   29   a     2616.21   2616.21   1220.96   1429.43   1060.66   853.98   334.59   334.59   316.25   316.25   2932.45         1625.05   1625.05   1215.21   589.42   383.97   402.20   25.88   25.88   174.77   174.77   1799.82       261.62   357.36   284.66   111.53   15.06       203.13   147.36   134.07   25.88   8.32     F   P     17.37   23.73   18.90   7.41         24.41   17.71   16.11   3.11         0.000   0.000   0.000   0.001         0.000   0.000   0.000   0.092       b c DF  =  degrees  of  freedom;   Seq  SS  =  sequential  sum  of  squares;   Adj   d SS  =  adjusted  sum  of  squares;   Adj  MS  =  adjusted  mean  square   187       Table  C.3:  ANOVA  of  Alamo  switchgrass  enzyme  mixture  regression.  Total  glucose  and   total  xylose  refer  total  monomeric  +  oligomeric  sugar  yields.   a Source   Mono  Glucose     Regression       Linear       Quadratic       Cubic       Amount       Comp*Amt  Quadratic     Residual       Lack-­‐of-­‐Fit       Pure  Error   Mono  Xylose     Regression       Linear       Quadratic       Cubic       Amount       Comp*Amt  Quadratic       Comp*Amt  Cubic     Residual       Lack-­‐of-­‐Fit       Pure  Error   Total  Glucose     Regression       Linear       Quadratic       Cubic       Amount       Comp*Amt  Quadratic     Residual       Lack-­‐of-­‐Fit       Pure  Error   a b DF   Seq  SS   b c Adj  SS   d Adj  MS   F   P   11   3   4   2   1   1   72   42   30   6913.44   6913.44   280.72   958.23   1158.78   1048.65   613.66   613.66   4826.39   2111.01   33.89   33.89   308.15   308.15   163.91   163.91   144.24   144.24   628.49   146.85   319.41   74.63   262.16   61.25   306.83   71.69   2111.01   493.24   33.89   7.92   4.28     3.90   0.81   4.81     0.000   0.000   0.000   0.000   0.000   0.006     0.737     13   3   5   2   1   1   1   69   39   30   4659.56   4659.56   2334.00   173.74   767.27   571.12   313.52   321.19   1211.47   547.33   19.77   24.70   13.53   13.53   134.53   134.53   105.76   105.76   28.77   28.77   358.43   183.84   57.91   29.70   114.22   58.59   160.60   82.37   547.33   280.73   24.70   12.67   13.53   6.94   1.95     2.71   2.83   0.96     0.000   0.000   0.000   0.000   0.000   0.001   0.010     0.002     11   3   4   2   1   1   72   42   30   6649.99   6649.99   738.02   1344.03   768.96   675.22   504.46   504.46   4604.54   2741.82   34.00   34.00   291.72   291.72   161.24   161.24   130.48   130.48   604.54   149.21   448.01   110.57   168.80   41.66   252.23   62.25   2741.82   676.72   34.00   8.39   4.05     3.84   0.88   4.35     0.000   0.000   0.000   0.000   0.000   0.005     0.651     c DF  =  degrees  of  freedom;   Seq  SS  =  sequential  sum  of  squares;   Adj  SS  =   d adjusted  sum  of  squares;   Adj  MS  =  adjusted  mean  square   188       Table   C.3   (cont’d):   ANOVA   of   Alamo   switchgrass   enzyme   mixture   regression.   Total   glucose  and  total  xylose  refer  total  monomeric  +  oligomeric  sugar  yields.   Source   DF   Seq  SS   a b                 8   3   2   2   1   73   45   28   525.11   98.65   48.31   48.45   329.70   118.36   99.22   19.14   a Regression     Linear     Quadratic     Cubic     Amount   Residual     Lack-­‐of-­‐Fit     Pure  Error   d c Adj  SS   Adj  MS   525.11   38.80   31.04   50.64   329.70   118.36   99.22   19.14   b 65.64   12.94   15.52   25.32   329.70   1.62   2.21   0.68   F   P   40.48   7.98   9.57   15.62   203.34     3.23     0.000   0.000   0.000   0.000   0.000     0.001     c DF  =  degrees  of  freedom;   Seq  SS  =  sequential  sum  of  squares;   Adj  SS  =   d adjusted  sum  of  squares;   Adj  MS  =  adjusted  mean  square       Table   C.4:   ANOVA   of   Shawnee   switchgrass   enzyme   mixture   regression.   Total   glucose  and  total  xylose  refer  to  the  total  monomeric  +  oligomeric  sugar  yields.   a b Source   DF   Seq  SS   Mono  Glucose     Regression       Linear       Quadratic       Cubic       Amount     Residual       Lack-­‐of-­‐Fit       Pure  Error   10   3   5   1   1   73   43   30   a b c Adj  SS   4935.22   4935.22   401.09   765.82   900.73   787.55   311.10   311.10   3322.30   3322.30   187.28   187.28   121.88   121.88   65.40   65.40   d Adj  MS   d 189     P   493.52   192.37   0.000   255.27   99.50   0.000   157.51   61.40   0.000   311.10   121.27   0.000   3322.30   1295.02   0.000   2.57       2.83   1.30   0.227   2.18       c DF  =  degrees  of  freedom;   Seq  SS  =  sequential  sum  of  squares;   Adj  SS  =   adjusted  sum  of  squares;   Adj  MS  =  adjusted  mean  square F     Table  C.4  (cont’d):  ANOVA  of  Shawnee  switchgrass  enzyme  mixture  regression.  Total   glucose  and  total  xylose  refer  total  monomeric  +  oligomeric  sugar  yields.   Mono  Xylose     Regression       Linear       Quadratic       Cubic       Amount       Comp*Amt  Linear     Residual       Lack-­‐of-­‐Fit       Pure  Error   Total  Glucose     Regression       Linear       Quadratic       Cubic       Amount       Comp*Amt  Quadratic     Residual       Lack-­‐of-­‐Fit       Pure  Error   Total  Xylose     Regression       Linear       Quadratic       Cubic       Amount     Residual       Lack-­‐of-­‐Fit       Pure  Error   a 12   3   3   4   1   1   71   41   30   3642.71   3642.71   1642.84   197.57   490.23   273.68   529.92   529.92   967.77   627.71   11.95   11.95   177.49   17.49   123.34   123.34   54.15   54.15   303.56   65.86   91.23   132.48   627.71   11.95   2.50   3.01   1.81   121.43   26.35   36.49   53.00   251.11   4.78     1.67     0.000   0.000   0.000   0.000   0.000   0.032     0.074     9   3   2   2   1   1   74   44   30   4754.14   4754.14   803.90   1366.97   442.26   304.34   344.40   344.40   3149.58   1929.27   14.01   14.01   185.49   185.49   121.86   121.86   63.63   63.63   528.24   455.66   152.17   172.20   1929.27   14.01   2.51   2.77   2.12   210.74   181.78   60.71   68.70   769.67   5.59     1.31     0.000   0.000   0.000   0.000   0.000   0.021     0.223     60.11   24.64   16.46   39.14   286.24   1.38   1.45   1.259   43.69   17.91   11.96   28.45   208.07     1.15     0.000   0.000   0.000   0.000   0.000     0.351     9   3   2   3   1   72   44   28   540.98   83.95   50.78   120.01   286.24   99.05   63.80   35.25   540.98   73.91   32.91   117.42   286.24   99.05   63.80   35.25   b c DF  =  degrees  of  freedom;   Seq  SS  =  sequential  sum  of  squares;   Adj  SS  =   d adjusted  sum  of  squares;   Adj  MS  =  adjusted  mean  square 190       APPENDIX  D :  SUPPLEMENTARY  INFORMATION  FOR  CHAPTER  6       Table  D.1:  Composition  data  for  the  NM-­‐6  poplar,  and  F5H,  CCR,  and  4CL  control  and  transgenic  lines.         Expt         Line         Pool   1.8   (0.1)   NM-­‐6   1.6   (0.1)   100.7       4.1   (3.2)   (0.2)   4.0   16.8       508.7       (0.2)   (1.0)   (16.9)   1.7   (0.1)   3.4ab   (0.3)   1.5   (-­‐)   3.9a   (0.3)   95.8       (3.8)   230.3ab   (5.7)   4.3   (0.1)   6.1b   (0.4)   a F5H   Ctrl     Rha   2.0   (0.3)   Cell  Wall  Composition  (mg/g  alcohol  insoluble  residue)   Structural  Carbohydrates   Hemicellulose  Sugars   Crystalline   ABSL   Cellulose   Ara   Xyl   Man   Gal   Glc   2.9   106.6     11.1     5.8   38.6   415.7   n.d.   (0.1)   (5.8)   (0.8)   (0.3)   (4.6)   (14.3)   C4H::F5H   CCR   Ctrl   1   a abc a b abc a b b     2   3.4   3.6   210.5   (0.5)   (0.7)   (21.6)       3   3.6   (0.2)   a a 3.9   (0.3)   ab a a 4.1   (0.2)   7.1ab   (0.5)   ab b a ab b a 18.5       (2.1)   43.5cd   (3.9)   ab b a 527.5       (19.3)   408.5a   (11.6)   a 7.5   6.3   50.7   393.0   (0.5)   (0.9)   (1.7)   (30.6)   ab ab bc a 237.7   7.5   6.7   48.2   437.5   (12.4)   (0.1)   (0.5)   (1.7)   (31.0)   Lignin  Composition  (%)*   Monomers   S   G   H   n.d.   n.d.   n.d.   191.4   65.5   34.5   -­‐   157.0   93.4   6.6   -­‐   207.5ab   71.5bc   28.2ab   0.3c   (5.4)   (0.2)   (0.2)   (0.1)   ab   cd a c 208.2 (5.6)   70.7   (0.3)   29.0   (0.2)   0.3   (0.1)   217.1   (5.7)   70.1   (0.5)   29.6   (0.4)   0.3   (0.0)   a cd a c *S  and  G  values  for  the  F5H  poplar  samples  are  thioacidolysis  values  from  Stewart  et  al.  [243].   For  the  different  experiments,  values  in  each  column  with  different  superscript  letters  are  statistically  different  based  on   Tukey’s  pairwise  comparisons  (95%  CI),  P  <  0.05.  Ctrl:  control;  Ara:  arabinose;  Xyl:  xylose;  Man:  mannose;  Gal:  galactose;   Glc:   glucose;   n.d.:   not   determined;   ABSL:   acetyl   bromide   soluble   lignin;   S:   syringyl   lignin;   G:   guaiacyl   lignin;   H:   p-­‐ hydroxyphenyl  lignin;         191       Table  D.1  cont’d.:  Composition  data  for  the  NM-­‐6  poplar,  and  the  CCR,  4CL,  and  F5H  control  and  transgenic  lines.         Expt         Line         Pool   Rha   abc Cell  Wall  Composition  (mg/g  alcohol  insoluble  residue)   Structural  Carbohydrates   Hemicellulose  Sugars   Crystalline   Cellulose   Ara   Xyl   Man   Gal   Glc   abcd ab ab ab bc ab ab a b a Lignin  Composition  (%)   Monomers   ABSL   d S   ab G   cd H   c CCR   5-­‐2-­‐3   1   3.1   (0.3)     2   2.8   (0.1)   2.9   (0.1)   233.3   (13.1)   6.9   6.9   (0.1)   (0.3)   48.1   432.2   (0.7)   (30.7)   186.3   (2.7)   74.2   (0.6)   25.6   (0.6)   0.2   (0.1)   3   3.1   (0.2)   3.2   (0.1)   222.3   (8.6)   7.7   7.5   (0.1)   (0.2)   51.5   406.4   (0.7)   (18.6)   189.4   (2.5)   70.0   28.9   (0.9)   (0.8)   1.2   (0.1)   55.3   (0.8)   452.6   (27.1)   195.7   (3.7)   71.3   27.4   1.4   (0.0)   (0.1)   (0.1)     bc abc 3.1   (0.3)   235.5   (12.6)   7.3   6.8   (0.4)   (0.6)   49.2   (3.0)   441.6   (4.0)   189.4   (2.9)   73.1   26.6   0.2   (0.3)   (0.3)   (0.0)   bcd abcd ab b bc ab a a d d a cd d a c b       5-­‐2-­‐40   1   2.8   (0.1)   2.5   (0.1)   217.4   (15.3)   7.9   (0.2)       2   2.8   (0.0)   2.8   (0.1)   225.6   (10.7)   7.6   6.8   (0.1)   (0.1)   50.2   407.4   (1.4)   (43.5)   191.2   (0.8)   69.9   28.6   1.5   (0.3)   (0.2)   (0.2)       3   4CL   Ctrl   1       2       3   3.3   (0.1)   3.7abc   (0.2)   2.9abcd   (0.1)   3.1abcd   (0.1)   3.7   (0.1)   2.7cdefg   (0.2)   3.1bcdef   (0.9)   2.5efg   (0.1)   263.2   (8.3)   122.2defg   (3.7)   116.9efg   (2.3)   108.8g   (4.5)   6.3   (0.2)   10.2a   (0.9)   8.2ab   (0.3)   9.0ab   (0.5)   38.7   (1.2)   44.2a   (3.7)   32.0ab   (0.3)   33.4ab   (0.7)   201.4   (2.5)   213.3a   (19.5)   198.2abc     (3.3)   197.0abc     (2.8)   69.6   (1.2)   63.9   (0.1)   63.5   (0.3)   64.3   (0.1)   c bc abc d cd ab b b a a ab c ab 6.3   (0.2)   ab ab 6.9   (0.2)   7.1abcd   (0.3)   7.2abc   (0.1)   7.4ab   (0.4)   a ab d a a a 435.6   (16.6)   561.9abcde   (22.0)   538.8bcdefg   (9.6)   528.6bcdefg   (27.0)   cd cd bc cd bc cd ab b ab a d b 28.4   2.0   (0.9)   (0.3)   35.7   0.4   (0.1)   (0.0)   36.1   0.4   (0.3)   (0.0)   35.3   0.4   (0.1)   (0.0)   For   the   different   experiments,   values   in   each   column   with   different   superscript   letters   are   statistically   different   based   on   Tukey’s   pairwise  comparisons  (95%  CI),  P  <  0.05.  Ctrl:  control;  Ara:  arabinose;  Xyl:  xylose;  Man:  mannose;  Gal:  galactose;  Glc:  glucose;  n.d.:   not  determined;  ABSL:  acetyl  bromide  soluble  lignin;  S:  syringyl  lignin;  G:  guaiacyl  lignin;  H:  p-­‐hydroxyphenyl  lignin;     192       Table  D.1  cont’d.:  Composition  data  for  the  NM-­‐6  poplar,  and  the  CCR,  4CL,  and  F5H  control  and  transgenic  lines.   Lignin  Composition   (%)   Pool   Line   Strength   Expt   Cell  Wall  Composition  (mg/g  alcohol  insoluble  residue)   Rha   abcd Structural  Carbohydrates   Hemicellulose  Sugars   Xyl   Man   Gal   Ara   defg bcdefg 1   3.3   2.5   (0.1)   (0.2)     2   2.4   (0.6)     3   3.4   2.7 (0.2)   (0.1)   2   1   3.0   3.0 (0.6)   (0.0)     126.3 (7.9)     2   2.7   2.4   (0.2)   (0.1)   111.6   (1.5)     3   3.4   2.5   (0.4)   (0.3)   129.8 (7.3)   3   1   3.2   2.8 (0.3)   (0.3)     136.4 (15.6)     2   3.6   (0.1)     144.7 (4.1)   3   3.0   2.3   (0.1)   (0.1)   Weak   4CL     fg cdefg abcdefg abcd bcdef abcd fg abcd efg abcd cdefg abc abcd cdefg 2.7 (0.1)   fg cdefg   g 111.7   (2.4)     a abcdefg   ab abc ab abcdef ab fg ab abcde defg b abc ab bcdefg ab ab ab abcd ab abcdef ab bcdefg ab efg ab abcdefg   29.9   492.2 (2.3)   (3.5)   5.7   5.9 (3.4)   (1.1)       9.2   7.0 (0.5)   (0.1)       6.2   6.6 (0.3)   (0.2)   6.4   5.4   (0.4)   (0.2)   8.0   7.0 (0.3)   (0.0)   G   H   199.6 (0.3)   67.7   31.9   0.4   (0.8)   (0.8)   (0.1)       67.7   32.0   0.3   (0.1)   (0.1)   (0.0)   abcd     67.6   32.1   0.3   (0.2)   (0.2)   (0.0)       68.8   30.9   0.4   (0.5)   (0.5)   (0.0)   176.6 (6.2)   bcdefg   67.6   32.0   0.4   (0.5)   (0.5)   (0.0)   ab 69.9   29.6   0.5   (0.2)   (0.2)   (0.1)   abc   28.1   603.5   (14.3)   (2.7)     31.8   568.8 (2.6)   (10.8)       197.3 (6.9)     38.2   550.6 (2.5)   (7.2)   191.4 (10.5)   167.2 (15.3)     39.3   555.9 (4.4)   (10.6)       199.8     (4.6)   27.2   535.6 (1.1)   (22.6)     189.8 (4.9)       66.3   33.0   0.7   (0.5)   (0.6)   (0.1)   186.0 (7.0)       64.2   35.2   0.6   (0.5)   (0.5)   (0.0)   29.5   482.6 (2.2)   (6.7)     S   abc 18.6   595.2   (17.5)   (20.8)   9.3   7.2   (1.0)   (0.5)   Monomers   ABSL   defghi 7.9   6.0 (0.5)   (0.6)   8.1   6.4 (0.5)   (0.2)   abcdefg abcde abcde Glc   ab 7.9   7.5   (3.1)   (0.8)   g bcdefg ab ab   136.0 (2.3)   abcd efg 2.5   (0.1)   cdefg ab 113.9   (15.6)   d   cdefg ab 128.1 (16.1)   1   ab Crystalline   Cellulose   34.1   554.9 (1.1)   (36.8)     abcde bcdef i   145.1     (7.0)   66.8   32.7   0.5   (0.6)   (0.6)   (0.0)   For   the   different   experiments,   values   in   each   column   with   different   superscript   letters   are   statistically   different   based   on   Tukey’s   pairwise   comparisons   (95%   CI),   P   <   0.05.   Ara:   arabinose;   Xyl:   xylose;   Man:   mannose;   Gal:   galactose;   Glc:   glucose;   n.d.:   not   determined;  ABSL:  acetyl  bromide  soluble  lignin;  S:  syringyl  lignin;  G:  guaiacyl  lignin;  H:  p-­‐hydroxyphenyl  lignin;   193       Table  D.1  cont’d.:  Composition  data  for  the  NM-­‐6  poplar,  and  the  CCR,  4CL,  and  F5H  control  and  transgenic  lines.   Lignin   Composition  (%)   Pool   Line   Strength   Expt   Cell  Wall  Composition  (mg/g  alcohol  insoluble  residue)   Rha   abcd Structural  Carbohydrates   Hemicellulose  Sugars   Xyl   Man   Gal   Ara   g defg 122.5 (6.5)   2   3.8   (0.2)   3   3.5   2.7 (0.1)   (0.1)   4   1   2.6   (0.7)     120.5 (5.9)     2   3.2   3.1   (0.6)   (0.1)     3   3.3   3.0 (0.2)   (0.3)   ab abcd bcd abcd abcd bcd cdefg cdefg 2.6 (0.2)   bcde bcdef ab ab bcdefg ab abcdef ab abcdefg ab abcde ab abcdefg ab abcd   30.7   529.6 (0.6)   (13.6)   ab   abcdefg 136.0 (14.5)   bcdefg 26.6   505.0 (2.5)   (11.2)   ab 6.3   6.1 (4.0)   (0.9)         6.6   6.6 (2.5)   (0.5)     32   1   3.8   (0.8)   3.7   (0.1)   139.6 (9.6)       6.9   7.0 (2.5)   (0.6)     2   3.5   2.6 (0.2)   (0.0)     134.6 (0.8)       7.4   6.2 (0.4)   (0.1)   3   3.7   (0.6)     abcd abc cdefg a 4.2   (0.3)   abcdef abcdefg abcd 146.3 (5.4)   g 6.7     5.4     (0.5)   (0.3)       7.5   6.3 (0.2)   (0.5)   ab   23.7 (0.4)     130.3 (6.7)   ab ghi 5.6   5.2     (0.1)   (0.2)   ab defg fg 7.4   6.0 (0.0)   (0.1)   a 3.3   (0.1)   g ab   155.4   (18.2)   a   cdefg     6.3   7.1   (2.2)   (0.4)     Monomers   ABSL   bcdefg Glc   ab 160.7   (5.0)     Medium   2.9   2.2   (0.0)   (0.1)     4CL   22   1   Crystalline   Cellulose     ab a ab abcd ab defg ab bcdefg ab defg ab bcdefg   29.7   489.5 (0.7)   (44.2)   22.1   534.5 (10.9)   (15.2)   G   H   64.8   34.6   0.6   (0.4)   (0.5)   (0.1)       64.4   34.9   0.6   (0.3)   (0.3)   (0.0)   bcdefg     65.3   34.2   0.5   (0.1)   (0.1)   (0.0)   175.8 (8.7)   ghi 159.6 (3.3)       68.7   31.1   0.3   (0.1)   (0.1)   (0.0)     158.0 (6.5)   ghi     65.1   34.4   0.5   (1.9)   (1.8)   (0.1)     156.2 (8.8)   ghi     68.5   31.0   0.5   (1.4)   (1.4)   (0.0)   25.7   549.0 (12.5)   (29.6)     30.4   489.8 (2.8)   (6.0)       165.8 (3.2)   20.5   637.7   (19.1)   (26.3)   26.4   573.7 (13.8)   (3.5)     159.8 (4.2)   efghi 468.0   (12.7)   defg S     bcdefg   178.9 (7.3)   ghi     65.6   33.7   0.7   (1.4)   (1.3)   (0.1)       67.7   31.9   0.5   (0.1)   (0.1)   (0.0)   defgh   67.8   31.6   0.6   (0.2)   (0.2)   (0.0)   160.6 (5.9)     171.1 (5.2)   For   the   different   experiments,   values   in   each   column   with   different   superscript   letters   are   statistically   different   based   on   Tukey’s   pairwise   comparisons   (95%   CI),   P   <   0.05.   Ara:   arabinose;   Xyl:   xylose;   Man:   mannose;   Gal:   galactose;   Glc:   glucose;   n.d.:   not   determined;  ABSL:  acetyl  bromide  soluble  lignin;  S:  syringyl  lignin;  G:  guaiacyl  lignin;  H:  p-­‐hydroxyphenyl  lignin;   194         Table  D.1  cont’d.:  Composition  data  for  the  NM-­‐6  poplar,  and  the  CCR,  4CL,  and  F5H  control  and  transgenic  lines.   Lignin   Composition  (%)   2     3   12   1   2.9   2.3   (0.1)   (0.2)   125.5 (4.6)     2   3.4   2.7 (0.0)   (0.1)     141.1 (4.2)     3   3.7   (0.1)   3.3   (0.1)   152.4 (2.2)   16   1   3.1   2.4   (0.0)   (0.2)   123.3 (7.3)     2   3.2   2.3   (0.2)   (0.1)     Strong   1     Expt   7   4CL   Pool   Structural  Carbohydrates   Hemicellulose  Sugars   Rha   Ara   Xyl   Man   Gal   3.0abcd   2.4efg   115.9fg   8.0ab   6.6abcdef   (0.1)   (0.1)    (3.1)   (0.3)   (0.3)   2.8abcd   2.6cdefg   124.3defg     6.5ab   6.1bcdefg   (0.2)   (0.1)   (5.5)   (0.3)   (0.2)   2.7abcd   2.6cdefg   119.0defg   5.1b   5.9cdefg   (0.7)   (0.0)   (18.6)   (3.0)   (0.5)   Line   Strength   Cell  Wall  Composition  (mg/g  alcohol  insoluble  residue)   3   2.5   (0.4)   2.     (0.3)   abcd fg abcd cdefg abc abcd abcd cd bc efg g efg cdefg abcdef abc   ab abcdefg ab abcdefg ab   defg ab fg ab abcdefg Crystalline   Cellulose   Glc   37.3ab   552.9abcdefg   (1.1)   (22.4)   29.2ab   559.9abcde   (0.8)   (17.5)   18.5b   607.7ab   (16.2)   (18.0)   ab bcdefg ab cdefg ab fg ab defg ab bcdefg 7.8   6.4 (0.2)   (0.5)     30.2   548.5 (0.7)   (7.9)     6.9   6.4 (0.2)   (0.1)     30.8   512.0 (1.2)   (4.2)   5.5   5.8   (0.2)   (0.1)   7.3   5.6   (0.2)   (0.0)   32.7   662.5 (1.2)   (28.8)   124.2 (4.2)     7.4   6.2 (0.4)   (0.2)     29.4   703.6 (2.3)   (46.8)   122.3 (2.7)     4.8   (2.3)   defg defg defg b efg 5.7   (0.2)   b 17.4   (12.6)   ab 759.0   (77.7)   S   66.9   (0.4)   66.2   (0.2)   66.7   (0.3)   66.5   (0.8)   156.7ghi   (12.7)   155.4ghi   (6.1)   150.0ghi     (3.7)   ghi 157.4 (12.1)       fghi G   32.4   (0.3)   33.1   (0.2)   32.7   (0.2)   33.0   (0.9)   H   0.7   (0.1)   0.7   (0.0)   0.6   (0.0)   0.5   (0.2)       64.1   35.3   0.6   (0.0)   (0.0)   (0.0)   cdefg       66.0   33.1   0.9   (0.9)   (1.0)   (0.1)   163.1 (3.3)   21.8   470.7   (0.9)   (54.5)       Monomers   ABSL   175.0 (1.2)   fghi   165.1 (1.9)       65.1   34.3   0.6   (0.2)   (0.2)   (0.0)   168.8 (3.9)   defghi   63.6   35.9   0.5   (0.4)   (0.4)   (0.0)   168.4 (7.1)   defghi   64.8   34.8   0.5   (0.1)   (0.1)   (0.0)   For   the   different   experiments,   values   in   each   column   with   different   superscript   letters   are   statistically   different   based   on   Tukey’s   pairwise   comparisons   (95%   CI),   P   <   0.05.   Ara:   arabinose;   Xyl:   xylose;   Man:   mannose;   Gal:   galactose;   Glc:   glucose;   n.d.:   not   determined;  ABSL:  acetyl  bromide  soluble  lignin;  S:  syringyl  lignin;  G:  guaiacyl  lignin;  H:  p-­‐hydroxyphenyl  lignin;     195       Table  D.2:  Variance  within  the  fully  nested  ANOVA  for  each  line,  related  to  downregulation  strength  (4CL  only),  parent   line,  or  sample  pool  for  the  4CL  and  CCR  samples.  An  ANOVA  was  unable  to  be  performed  for  the  4CL  lignin  monomers  as   information  on  the  individual  data  replicates  was  not  provided.         Hemicellulose  Sugars     Ara   Xyl   %  of  Total  Variance     Strength   13.10   14.69   4CL     Line   5.68   10.96     Pool   58.71   43.08     Error   22.52   31.27   %  of  Total  Variance   39.23   0.00   CCR     Line     Pool   34.57   57.97     Error   26.20   42.03   Man   20.45   0.00   13.32   66.23   Gal   Glc   Cry   Total  Glc   Total  Sugar   Lignin   Lignin  Monomers   S   G   -­‐   -­‐   -­‐   -­‐   H   14.12   13.12   0.00   21.86   0.00   13.27   31.86   22.16   60.07   32.15   64.71   26.65   0.00   17.66   57.66   24.68   0.00   52.15   6.28   7.97   56.68   25.59   37.04   14.29   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   -­‐   0.00   23.70   0.00   0.00   76.88   14.24   87.87   24.41   23.12   62.07   12.13   75.59   0.00   26.52   73.48   0.00   81.57   21.81   26.56   75.58   23.51   8.37   66.45   60.46   21.28   76.49   10.06   11.74   12.98   3.14   Ara  =  arabinose;  Xyl  =  xylose;  Man  =  mannose;  Gal  =  galactose;  Glc  =  glucose;  Cry  =  crystalline  cellulose;  S  =  syringyl  lignin;     G  =  guaiacyl  lignin;  H  =  p-­‐hydroxyphenyl  lignin   196           Figure   D.1:   Hemicellulose   glucose   vs.   mannose   content   for   4CL   poplar   samples.   The   equation   represents  the  linear  regression  of  the  data.    Glucose   is   derived   from   hemicellulose   and   does   not   include   cellulose-­‐derived  or  soluble  glucose.     197         Figure   D.2:   Untreated   wildtype   and   4CL   downregulated   poplar   (A)   glucose   and   (B)   xylose   yields.  Sugar  yields  with  different  letters  within  each  subplot  are  statistically  different  based  on   Tukey’s  pairwise  comparisons  (95%  CI),  (p  <  0.05)  and  are  not  comparable  between  24  h  and   168   h   data.   Enzymatic   hydrolysis   was   conducted   at   200   rpm   and   1.25%   total   sugar   loading   with   24  mg  total  protein  per  g  cell  wall  sugars  (80%  Accellerase®  1500,  10%  Accellerase®  XY,  10%   Multifect®  Pectinase).   198         Figure   D.3:   Pretreated   wildtype   and   4CL   downregulated   poplar   (A)   glucose   and   (B)   xylose   yields.  Sugar  yields  with  different  letters  within  each  subplot  are  statistically  different  based  on   Tukey’s  pairwise  comparisons  (95%  CI),  (p  <  0.05)  and  are  not  comparable  between  24  h  and   168  h  data.  All  samples  were  pretreated  using  1:1  g  NH3:g  DM;  1:1  g  H2O:g  DM;  180°C  for  20   min.   Enzymatic   hydrolysis   was   conducted   at   200   rpm   and   1.25%   total   sugar   loading   with   24   mg   total   protein   per   g   cell   wall   sugars   (80%   Accellerase®   1500,   10%   Accellerase®   XY,   10%   Multifect®  Pectinase).     199         Figure   D.4:   Control   and   CCR   downregulated   poplar   (A)   glucose   and   (B)   xylose   yields.   Sugar   yields   with   different   letters   within   each   subplot   are   statistically   different   based   on   Tukey’s   pairwise  comparisons  (95%  CI),  (p  <  0.05)  and  are  not  comparable  between  24  h  and  168  h  data.   Samples   were   pretreated   using   1:1   g   NH3:DM   and   1:1   g   H2O:g   DM.   Enzymatic   hydrolysis   was   conducted   at   200   rpm   and   1.25%   total   sugar   loading   with   24   mg   total   protein   per   g   cell   wall   sugars  (80%  Accellerase®  1500,  10%  Accellerase®  XY,  10%  Multifect®  Pectinase).   200         REFERENCES   201        1.   REFERENCES       Cassman,   K.G.   and   A.J.   Liska,   Food   and   fuel   for   all:   realistic   or   foolish?   Biofuel.   Bioprod.   Bior.,  2007.  1(1):  18-­‐23.   2.   Pimentel,  D.,  et  al.,  Food  versus  biofuels:  Environmental  and  economic  costs.  Hum.  Ecol.,   2009.  37(1):  1-­‐12.   3.   Rosillo-­‐Calle,   F.   and   D.O.   Hall,   Brazilian   alcohol:   Food   versus   fuel?   Biomass,   1987.   12(2):   97-­‐128.   4.   Fargione,   J.,   et   al.,   Land   clearing   and   the   biofuel   carbon   debt.   Science,   2008.   319(5867):   1235-­‐1238.   5.   Searchinger,   T.,   et   al.,   Use   of   U.S.   croplands   for   biofuels   increases   greenhouse   gases   through  emissions  from  land-­‐use  change.  Science,  2008.  319(5867):  1238-­‐1240.   6.   Mathews,   J.A.   and   H.   Tan,   Biofuels   and   indirect   land   use   change   effects:   the   debate   continues.  Biofuel.  Bioprod.  Bior.,  2009.  3(3):  305-­‐317.   7.   Bacovsky,   D.,   How   close   are   second-­‐generation   biofuels?   Biofuel.   Bioprod.   Bior.,   2010.   4(3):  249-­‐252.   8.   Ragauskas,   A.J.,   et   al.,   The   path   forward   for   bofuels   and   biomaterials.   Science,   2006.   311(5760):  484-­‐489.   9.   Haveren,   J.v.,   E.L.   Scott,   and   J.   Sanders,   Bulk   chemicals   from   biomass.   Biofuel.   Bioprod.   Bior.,  2008.  2(1):  41-­‐57.   10.   Carolan,   J.E.,   S.V.   Joshi,   and   B.E.   Dale,   Technical   and   financial   feasibility   analysis   of   distributed  bioprocessing  using  regional  biomass  pre-­‐processing  centers.  J.  Agr.  Food  Ind.   Organ.,  2007.  5(2):  Article  10.   11.   Lora,   J.H.   and   W.G.   Glasser,   Recent   industrial   applications   of   lignin:   A   sustainable   alternative  to  nonrenewable  materials.  J.  Polym.  Environ.,  2002.  10(1):  39-­‐48.   12.   Aden,  A.  and  T.  Foust,  Technoeconomic  analysis  of  the  dilute  sulfuric  acid  and  enzymatic   hydrolysis   process   for   the   conversion   of   corn   stover   to   ethanol.   Cellulose,   2009.   16(4):   535-­‐545.   13.   Dien,   B.S.,   et   al.,   Chemical   composition   and   response   to   dilute-­‐acid   pretreatment   and   enzymatic   saccharification   of   alfalfa,   reed   canarygrass,   and   switchgrass.   Biomass   Bioenerg.,  2006.  30(10):  880-­‐891.   202       14.   Balan,  V.,  et  al.,  Enzymatic  digestibility  and  pretreatment  degradation  products  of  AFEX-­‐ treated  hardwoods  (Populus  nigra).  Biotechnol.  Progr.,  2009.  25(2):  365-­‐375.   15.   Scheller,  H.V.  and  P.  Ulvskov,  Hemicelluloses.  Annu.  Rev.  Plant  Biol.,  2010.  61(1):  263-­‐289.   16.   Harris,   P.J.   and   B.G.   Smith,   Plant   cell   walls   and   cell-­‐wall   polysaccharides:   structures,   properties  and  uses  in  food  products.  Int.  J.  Food  Sci.  Tech.,  2006.  41(s2):  129-­‐143.   17.   McCann,   M.C.   and   N.C.   Carpita,   Designing   the   deconstruction   of   plant   cell   walls.   Curr.   Opin.  Plant  Biol.,  2008.  11(3):  314-­‐320.   18.   Pauly,  M.  and  K.  Keegstra,  Cell-­‐wall  carbohydrates  and  their  modification  as  a  resource  for   biofuels.  Plant  J.,  2008.  54(4):  559-­‐568.   19.   Iiyama,   K.,   T.B.T.   Lam,   and   B.A.   Stone,   Phenolic   acid   bridges   between   polysaccharides   and   lignin  in  wheat  internodes.  Phytochemistry,  1990.  29(3):  733-­‐737.   20.   Hatfield,  R.D.,  J.  Ralph,  and  J.H.  Grabber,  Cell  wall  cross-­‐linking  by  ferulates  and  diferulates   in  grasses.  J.  Sci.  Food  Agr.,  1999.  79(3):  403-­‐407.   21.   Grabber,  J.H.,  et  al.,  Genetic  and  molecular  basis  of  grass  cell-­‐wall  degradability.  I.  Lignin-­‐ cell  wall  matrix  interactions.  C.  R.  Biol.,  2004.  327(5):  455-­‐465.   22.   Vanholme,  R.,  et  al.,  Lignin  engineering.  Curr.  Opin.  Plant  Biol.,  2008.  11(3):  278-­‐285.   23.   Vanholme,  R.,  et  al.,  Lignin  biosynthesis  and  structure.  Plant  Physiol.,  2010.  153(3):  895-­‐ 905.   24.   Lapierre,   C.,   et   al.,   Structural   alterations   of   lignins   in   transgenic   poplars   with   depressed   cinnamyl   alcohol   dehydrogenase   or   caffeic   acid   O-­‐methyltransferase   activity   have   an   opposite  impact  on  the  efficiency  of  industrial  kraft  pulping.  Plant  Physiol.,  1999.  119(1):   153-­‐164.   25.   Baucher,  M.,  et  al.,  Lignin:  Genetic  engineering  and  impact  on  pulping.  Crit.  Rev.  Biochem.   Mol.,  2003.  38:  305-­‐350.   26.   Grabber,  J.H.,  How  do  lignin  composition,  structure,  and  cross-­‐linking  affect  degradability?   A  review  of  cell  wall  model  studies.  Crop  Sci.,  2005.  45(3):  820-­‐831.   27.   Boerjan,   W.,   J.   Ralph,   and   M.   Baucher,   Lignin   biosynthesis.   Annu.   Rev.   Plant   Biol.,   2003.   54(1):  519-­‐546.   28.   Garlock,  R.J.,  V.  Balan,  and  B.E.  Dale,  Optimization  of  AFEX™  pretreatment  conditions  and   enzyme   mixtures   to   maximize   sugar   release   from   upland   and   lowland   switchgrass.   Bioresource  Technol.,  (0).   203       29.   Garlock,  R.,  et  al.,  Optimizing  harvest  of  corn  stover  fractions  based  on  overall  sugar  yields   following   ammonia   fiber   expansion   pretreatment   and   enzymatic   hydrolysis.   Biotechnol.   Biofuels,  2009.  2(1):  29.   30.   Garlock,  R.J.,  et  al.,  Comparative  material  balances  around  pretreatment  technologies  for   the   conversion   of   switchgrass   to   soluble   sugars.   Bioresource   Technol.,   2011.   102(24):   11063-­‐11071.   31.   Garlock,  R.,  et  al.,  AFEX  pretreatment  and  enzymatic  conversion  of  black  locust  (Robinia   pseudoacacia  L.)  to  soluble  sugars.  Bioenergy  Res.:  1-­‐13.   32.   International   Energy   Statistics,   U.S.   Energy   Information   Administration.   Available   from:   http://www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm.   33.   Recent  Quarterly  and  Annual  World  Petroleum  (Oil)  Demand,  2005-­‐2009.  [Excel  File].  Last   Updated:   17   February   2010   2010.   Date   Accessed:   18   February   2010.   Available   from:   http://www.eia.doe.gov/ipm/demand.html.   34.   Medium  and  long-­‐term  development  plan  for  renewable  energy  in  China.  2007,  National   Development  and  Reform  Commission  (NDRC).   35.   Wu,   C.Z.,   et   al.,   The   development   of   bioenergy   technology   in   China.   Energy,   2010.   35(11):   4445-­‐4450.   36.   Zhang,   P.D.,   et   al.,   Bioenergy   industries   development   in   China:   Dilemma   and   solution.   Renew.  Sust.  Energ.  Rev.,  2009.  13(9):  2571-­‐2579.   37.   Kumar,  A.  and  S.  Sokhansanj,  Switchgrass  (Panicum  vigratum,  L.)  delivery  to  a  biorefinery   using  integrated  biomass  supply  analysis  and  logistics  (IBSAL)  model.  Bioresource  Technol.,   2007.  98(5):  1033-­‐1044.   38.   Liu,  J.W.,  et  al.,  Life  cycle  assessment  on  straw  directly  combustion  for  power  generation   system.  Renewable  Energy  Resources,  2009.  27(5):  102-­‐106.  (In  Chinese).   39.   Beijing  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   40.   Tianjin  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  English  and  Chinese).   41.   Hebei  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   42.   Shanxi  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   43.   Inner  Mongolia  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   44.   Liaoning  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   45.   Jilin  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   204       46.   Heilongjiang  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   47.   Shanghai  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   48.   Jiangsu  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   49.   Zhejiang  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   50.   Anhui  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   51.   Fujian  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   52.   Jiangxi  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   53.   Shandong  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   54.   Henan  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   55.   Hubei  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   56.   Hunan  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   57.   Guangdong  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   58.   Guangxi  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   59.   Hainan  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   60.   Chongqing  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   61.   Sichuan  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   62.   Guizhou  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   63.   Yunnan  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   64.   Tibet  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   65.   Shaanxi  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   66.   Gansu  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   67.   Qinghai  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   68.   Ningxia  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   69.   Xinjiang  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   205       70.   China  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   71.   China  Statistical  Yearbook  for  Regional  Economy.  2007:  China  Statistics  Press.  (In  Chinese   and  English).   72.   China   Data   Online,   China   Data   Center,   University   of   Michigan.   Access   Date:   11   September   2011.  Available  from:  http://chinadataonline.org.   73.   China  Rural  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese).   74.   Xie,  G.H.,  et  al.,  Harvest  index  and  residue  factor  of  cereal  crops  in  China.  Journal  of  China   Agricultural  University,  2011.  16(1):  1-­‐8.  (In  Chinese).   75.   Xie,  G.H.,  et  al.,  Harvest  index  and  residue  factor  of  non-­‐cereal  crops  in  China.  Journal  of   China  Agricultural  University,  2011.  16(1):  9-­‐17.  (In  Chinese).   76.   Kim,   S.   and   B.E.   Dale,   Global   potential   bioethanol   production   from   wasted   crops   and   crop   residues.  Biomass  Bioenerg.,  2004.  26(4):  361-­‐375.   77.   Zhang,   F.C.   and   Z.H.   Zhu,   Harvest   index   for   various   crops   in   China.   Scientia   Agricultura   Sinica,  1990.  23(2):  83-­‐87.  (In  Chinese).   78.   Li,   J.J.,   J.M.   Bai,   and   R.   Overend,   Assessment   of   Biomass   Resource   Availability   in   China.   1998:  Beijing.   79.   Xie,   G.H.,   X.Y.   Wang,   and   L.T.   Ren,   China's   crop   residues   resources   evaluation.   Chin.   J.   Biotechnol.,  2010.  26(7):  855-­‐863.   80.   Cui,  M.,  et  al.,  Analysis  and  evaluation  on  energy  utilization  of  main  crop  straw  resources   in   China.   Transactions   of   the   Chinese   Society   of   Agricultural   Engineering,   2008.   24(12):   291-­‐296.  (In  Chinese).   81.   Gao,  L.W.,  et  al.,  Analysis  on  the  quantities  and  utilization  of  crop  straw  and  its  nutrient  in   Huang-­‐Huai-­‐Hai   Region.   Chinese   Agricultural   Science   Bulletin,   2009.   25(11):   186-­‐193.   (In   Chinese).   82.   Gao,  L.W.,  et  al.,  Analysis  on  crop  straw  resources  and  its  current  status  of  utilization  in   Hebei   province.   Journal   of   Anhui   Agricultural   Sciences,   2009.   37(23):   11079-­‐11083.   (In   Chinese).   83.   Huang,   Y.,   et   al.,   Net   primary   production   of   Chinese   croplands   from   1950   to   1999.   Ecol.   Appl.,  2007.  17(3):  692-­‐701.   84.   Liu,  H.,  et  al.,  Distribution,  utilization  structure  and  potential  of  biomass  resources  in  rural   China:   With   special   references   of   crop   residues.   Renew.   Sust.   Energ.   Rev.,   2008.   12(5):   1402-­‐1418.   206       85.   Liao,   C.P.,   et   al.,   Study   on   the   distribution   and   quantity   of   biomass   residues   resource   in   China.  Biomass  Bioenerg.,  2004.  27(2):  111-­‐117.   86.   Chen,   L.J.,   L.   Xing,   and   L.J.   Han,   Renewable   energy   from   agro-­‐residues   in   China:   Solid   biofuels  and  biomass  briquetting  technology.  Renew.  Sust.  Energ.  Rev.,  2009.  13(9):  2689-­‐ 2695.   87.   Zeng,  X.Y.,  Y.T.  Ma,  and  L.R.  Ma,  Utilization  of  straw  in  biomass  energy  in  China.  Renew.   Sust.  Energ.  Rev.,  2007.  11(5):  976-­‐987.   88.   Zhou,   X.P.,   et   al.,   Assessment   of   sustainable   biomass   resource   for   energy   use   in   China.   Biomass  Bioenerg.,  2011.  35(1):  1-­‐11.   89.   Yang,   Y.L.,   et   al.,   Quantitative   appraisal   and   potential   analysis   for   primary   biomass   resources  for  energy  utilization  in  China.  Renew.  Sust.  Energ.  Rev.,  2010.  14(9):  3050-­‐3058.   90.   DIVA-­‐GIS.   Last.   Date   Accessed:   10   September   2011.   Available   from:   http://www.diva-­‐ gis.org/.   91.   China  Paper  Industry  Report  2006.  2007,  China  Paper  Association.   92.   China  Paper  Industry  Report  2005.  2006,  China  Paper  Association.   93.   China  Paper  Industry  Report  2007.  2008,  China  Paper  Association.   94.   He,  L.J.,  et  al.,  Straw  resources  and  their  utilization  in  China.  Transactions  of  the  Chinese   Society  of  Agricultural  Engineering,  2002.  18(3):  87-­‐91.  (In  Chinese).   95.   Shen,   Y.C.,   Natural   Resources   and   their   Utilization   in   the   Drylands   of   China,   in   Desertification   and   Its   Control   in   China,   L.   Ci   and   X.   Yang,   Editors.   2010,   Springer   Berlin   Heidelberg.  p.  101-­‐176.   96.   Qiu,  J.J.,  et  al.,  Mapping  single-­‐,  double-­‐,  and  triple-­‐crop  agriculture  in  China  at  0.5°  ×  0.5°   by   combining   county-­‐scale   census   data   with   a   remote   sensing-­‐derived   land   cover   map.   Geocarto  International,  2003.  18(2):  3-­‐13.   97.   Yan,  X.Y.,  T.  Ohara,  and  H.  Akimoto,  Bottom-­‐up  estimate  of  biomass  burning  in  mainland   China.  Atmos.  Environ.,  2006.  40(27):  5262-­‐5273.   98.   Gao,  X.Z.,  et  al.,  Analysis  on  the  current  status  of  utilization  of  crop  straw  in  China.  Journal   of  Huazhong  Agricultural  University,  2002.  21(3):  242-­‐247.  (In  Chinese).   99.   Huang,  X.,  et  al.,  Mercury  Emissions  from  Biomass  Burning  in  China.  Environ.  Sci.  Technol.,   2011.  45(21):  9442-­‐9448.   207       100.   Wang,   S.X.   and   C.Y.   Zhang,   Spatial   and   temporal   distribution   of   air   pollutant   emissions   from  open  burning  of  crop  residues  in  China.  Sciencepaper  Online,  2008.  3(5):  329-­‐333.   101.   Wang,   Y.J.,   Y.Y.   Bi,   and   C.Y.   Gao,   The   assessment   and   utilization   of   straw   resources   in   China.  Agr.  Sci.  China,  2010.  9(12):  1807-­‐1815.   102.   Li,   H.Q.,   L.M.   Wang,   and   L.   Shen,   Potential   CO2   emission   reduction   by   development   of   non-­‐grain-­‐based  bioethanol  in  China.  Environ.  Manage.,  2010.  46(4):  555-­‐564.   103.   Jenkins,   B.M.,   et   al.,   Combustion   properties   of   biomass.   Fuel   Process.   Technol.,   1998.   54(1-­‐3):  17-­‐46.   104.   Xiao,  J.,  et  al.,  Integrated  analysis  of  energy,  economic,  and  environmental  performance  of   biomethanol  from  rice  straw  in  China.  Ind.  Eng.  Chem.  Res.,  2009.  48(22):  9999-­‐10007.   105.   China  Energy  Statistical  Yearbook.  2007:  China  Statistics  Press.  (In  Chinese  and  English).   106.   Fischer,   R.A.   and   G.O.   Edmeades,   Breeding   and   cereal   yield   progress.   Crop   Sci.,   2010.   50(Supplement_1):  S-­‐85-­‐S-­‐98.   107.   Cao,   G.L.,   et   al.,   Estimating   the   quantity   of   crop   residues   burnt   in   open   field   in   China.   Resources  Science,  2006.  28(1):  9-­‐13.  (In  Chinese).   108.   Zhou,   Z.R.,   et   al.,   Analysis   of   changes   in   the   structure   of   rural   household   energy   consumption  in  northern  China:  A  case  study.  Renew.  Sust.  Energ.  Rev.,  2009.  13(1):  187-­‐ 193.   109.   Wuest,   S.B.,   S.L.   Albrecht,   and   K.W.   Skirvin,   Crop   residue   position   and   interference   with   wheat  seedling  development.  Soil  Till.  Res.,  2000.  55(3-­‐4):  175-­‐182.   110.   Morris,   N.L.,   et   al.,   The   effect   of   wheat   straw   residue   on   the   emergence   and   early   growth   of  sugar  beet  (Beta  vulgaris)  and  oilseed  rape  (Brassica  napus).  Eur.  J.  Agron.,  2009.  30(3):   151-­‐162.   111.   Cao,  G.L.,  et  al.,  Estimation  of  emissions  from  field  burning  of  crop  straw  in  China.  Chin.  Sci.   Bull.,  2008.  53(5):  784-­‐790.   112.   Li,   H.,   R.B.   Yang,   and   L.   Chen,   Utilization   of   crop   straws   in   development   of   energy.   Journal   of  Agro-­‐Environment  Science,  2007.  26:  628-­‐631.   113.   Li,  W.J.,  L.Y.  Shao,  and  P.R.  Buseck,  Haze  types  in  Beijing  and  the  influence  of  agricultural   biomass  burning.  Atmos.  Chem.  Phys.,  2010.  10(17):  8119-­‐8130.  (In  English).   114.   Wilhelm,   W.W.,   et   al.,   Crop   and   soil   productivity   response   to   corn   residue   removal:   A   literature  review.  Agron.  J.,  2004.  96(1):  1-­‐17.   208       115.   Gregg,  J.S.  and  R.C.  Izaurralde,  Effect  of  crop  residue  harvest  on  long-­‐term  crop  yield,  soil   erosion   and   nutrient   balance:   trade-­‐offs   for   a   sustainable   bioenergy   feedstock.   Biofuels,   2010.  1(1):  69-­‐83.   116.   Kadam,   K.L.   and   J.D.   McMillan,   Availability   of   corn   stover   as   a   sustainable   feedstock   for   bioethanol  production.  Bioresource  Technol.,  2003.  88(1):  17-­‐25.   117.   China  Agricultural  Statistical  Yearbook.  2008:  China  Statistics  Press.  (In  Chinese).   118.   Lu,   W.   and   T.Z.   Zhang,   Life-­‐cycle   implications   of   using   crop   residues   for   various   energy   demands  in  China.  Environ.  Sci.  Technol.,  2010.  44(10):  4026-­‐4032.   119.   Zhang,  J.,  et  al.,  Greenhouse  gases  and  other  airborne  pollutants  from  household  stoves  in   China:  a  database  for  emission  factors.  Atmos.  Environ.,  2000.  34(26):  4537-­‐4549.   120.   Mestl,   H.E.S.,   et   al.,   Urban   and   rural   exposure   to   indoor   air   pollution   from   domestic   biomass  and  coal  burning  across  China.  Sci.  Total  Environ.,  2007.  377(1):  12-­‐26.   121.   Jiang,  R.T.  and  M.L.  Bell,  A  comparison  of  particulate  matter  from  biomass-­‐burning  rural   and   non-­‐biomass-­‐burning   urban   households   in   northeastern   China.   Environ.   Health.   Perspect.,  2008.  116(7):  907-­‐914.   122.   Weimer,  P.J.,  et  al.,  FIBEX-­‐treated  rice  straw  as  a  feed  ingredient  for  lactating  dairy  cows.   Anim.  Feed  Sci.  Tech.,  2003.  103(1-­‐4):  41-­‐50.   123.   Bals,   B.,   et   al.,   Ammonia   fiber   expansion   (AFEX)   treatment   of   eleven   different   forages:   Improvements  to  fiber  digestibility  in  vitro.  Anim.  Feed  Sci.  Tech.,  2010.  155(2-­‐4):  147-­‐155.   124.   Fischer,  G.,  et  al.  China's  agricultural  prospects  and  challenges.  Last  Updated:  December   2007.   Date   Accessed:   16   February   2010.   Available   from:   http://www.iiasa.ac.at/Research/LUC/External-­‐Chinagro/SOW-­‐VU-­‐CHINAGRO-­‐ prospects_challenges.pdf.   125.   Donohue,   T.   Building   the   Next   Generation   of   Biofuels.   Conversations   in   Science   [Presentation].  Last  Updated:  8  Jan  2009  2009.  Date  Accessed:  4  Feb  2010.  Available  from:   http://www.greatlakesbioenergy.org/wp-­‐ content/uploads/2009/01/donohue_conversations.pdf.   126.   Bazzaz,   F.A.,   Succession   on   abandoned   fields   in   the   Shawnee   Hills,   Southern   Illinois.   Ecology,  1968.  49(5):  924-­‐936.   127.   Fargione,   J.E.,   et   al.,   Bioenergy   and   wildlife:   Threats   and   opportunities   for   grassland   conservation.  Bioscience,  2009.  59:  767-­‐777.   128.   Haddad,   N.M.,   et   al.,   Contrasting   effects   of   plant   richness   and   composition   on   insect   communities:  A  field  experiment.  Am.  Nat.,  2001.  158(1):  17.   209       129.   Öckinger,   E.   and   H.G.   Smith,   Semi-­‐natural   grasslands   as   population   sources   for   pollinating   insects  in  agricultural  landscapes.  Journal  of  Applied  Ecology,  2007.  44:  50-­‐59.   130.   Schilling,   K.E.   and   P.   Jacobson,   Groundwater   conditions   under   a   reconstructed   prairie   chronosequence.  Agr.  Ecosyst.  Environ.,  2010.  135(1-­‐2):  81-­‐89.   131.   Dybzinski,   R.,   et   al.,   Soil   fertility   increases   with   plant   species   diversity   in   a   long-­‐term   biodiversity  experiment.  Oecologia,  2008.  158(1):  85-­‐93.   132.   Fornara,  D.A.  and  D.  Tilman,  Plant  functional  composition  influences  rates  of  soil  carbon   and  nitrogen  accumulation.  J.  Ecol.,  2008.  96(2):  314-­‐322.   133.   Piñeiro,   G.,   et   al.,   Set-­‐asides   can   be   better   climate   investment   than   corn   ethanol.   Ecol.   Appl.,  2009.  19(2):  277-­‐282.   134.   Robertson,  G.P.,  E.A.  Paul,  and  R.R.  Harwood,  Greenhouse  gases  in  intensive  agriculture:   Contributions  of  individual  gases  to  the  radiative  forcing  of  the  atmosphere.  Science,  2000.   289(5486):  1922-­‐1925.   135.   Hill,   J.,   et   al.,   Climate   change   and   health   costs   of   air   emissions   from   biofuels   and   gasoline.   P.  Natl.  Acad.  Sci.  USA,  2009.  106(6):  2077-­‐2082.   136.   James,   L.A.,   S.M.   Swinton,   and   K.D.   Thelen,   Profitability   analysis   of   cellulosic   energy   crops   compared  to  corn.  Agron.  J.,  2010.  102(2):  675-­‐687.   137.   Pakarinen,   O.,   et   al.,   Storing   energy   crops   for   methane   production:   Effects   of   solids   content  and  biological  additive.  Bioresource  Technol.,  2008.  99(15):  7074-­‐7082.   138.   Zhang,   B.,   et   al.,   Treatment   variable   effects   on   supercritical   gasification   of   high-­‐diversity   grassland  perennials.  Appl.  Biochem.  Biotech.,  2009.  154(1):  59-­‐66.   139.   Zhang,   B.,   M.   von   Keitz,   and   K.   Valentas,   Thermochemical   liquefaction   of   high-­‐diversity   grassland  perennials.  J.  Appl.  Anal.  Pyrol.,  2009.  84(1):  18-­‐24.   140.   Florine,   S.E.,   et   al.,   Yield   and   composition   of   herbaceous   biomass   harvested   from   naturalized  grassland  in  southern  Iowa.  Biomass  Bioenerg.,  2006.  30(6):  522-­‐528.   141.   Tilman,  D.,  J.  Hill,  and  C.  Lehman,  Carbon-­‐negative  biofuels  from  low-­‐input  high-­‐diversity   grassland  biomass.  Science,  2006.  314(5805):  1598-­‐1600.   142.   Adler,   P.R.,   et   al.,   Plant   species   composition   and   biofuel   yields   of   conservation   grasslands.   Ecol.  Appl.,  2009.  19(8):  2202-­‐2209.   143.   Sheehan,  J.,  et  al.,  Energy  and  environmental  aspects  of  using  corn  stover  for  fuel  ethanol.   J.  Ind.  Ecol.,  2003.  7(3-­‐4):  117-­‐146.   210       144.   Barnes,   R.F.,   C.J.   Nelson,   and   G.W.   Fick,   Terminology   and   classification   of   forage   plants,   in   Forages,  volume  II:  The  science  of  grassland  agriculture,  R.F.  Barnes,  et  al.,  Editors.  2007,   Wiley-­‐Blackwell  Publishing:  Ames,  Iowa,  USA.  p.  3-­‐14.   145.   Silanikove,  N.  and  D.  Levanon,  Inter-­‐relationships  between  acidic  and  alkali  treatments  of   cotton  straw  and  wheat  straw  and  their  fibre  chemical  properties.  J.  Sci.  Food  Agr.,  1987.   38(2):  117-­‐124.   146.   Gross,   K.L.   and   S.M.   Emery,   Succession   and   Restoration   in   Michigan   Old   Field   Communities,   in   Old   Fields:   Dynamics   and   Restoration   of   Abandoned   Farmland,   V.A.   Cramer  and  R.J.  Hobbs,  Editors.  2007,  Island  Press:  Washington  D.C.  p.  162-­‐179.   147.   Sluiter,   J.B.,   et   al.,   Compositional   analysis   of   lignocellulosic   feedstocks.   1.   Review   and   description  of  methods.  J.  Agr.  Food  Chem.,  2010.  58(16):  9043-­‐9053.   148.   Rhee,   K.C.,   Determination   of   Total   Nitrogen,   in   Current   Protocols   in   Food   Analytical   Chemistry.  2001,  John  Wiley  &  Sons,  Inc.   149.   Hames,   B.,   C.   Scarlata,   and   A.   Sluiter,   Determination   of   protein   content   in   biomass,   in   Laboratory   Analytical   Procedures   (LAPs).   2008,   National   Renewable   Energy   Laboratory:   Golden,  CO.   150.   Bals,  B.,  et  al.,  Evaluation  of  ammonia  fibre  expansion  (AFEX)  pretreatment  for  enzymatic   hydrolysis   of   switchgrass   harvested   in   different   seasons   and   locations.   Biotechnol.   Biofuels,  2010.  3(1):  1.   151.   Jiang,   L.,   L.   He,   and   M.   Fountoulakis,   Comparison   of   protein   precipitation   methods   for   sample  preparation  prior  to  proteomic  analysis.  J.  Chromatogr.  A,  2004.  1023(2):  317-­‐320.   152.   Sluiter,   A.,   et   al.,   Determination   of   sugars,   byproducts,   and   degradation   products   in   liquid   fraction   process   samples,   in   Laboratory   Analytical   Procedures   (LAPs).   2008,   National   Renewable  Energy  Laboratory:  Golden,  CO.   153.   Tilley,   J.M.A.   and   R.A.   Terry,   A   two-­‐stage   technique   for   the   in   vitro   digestion   of   forage   crops.  Grass  Forage  Sci.,  1963.  18(2):  104-­‐111.   154.   Jung,  H.-­‐J.G.,  Analysis  of  forage  fiber  and  cell  walls  in  ruminant  nutrition.  The  Journal  of   Nutrition,  1997.  127(5):  810S-­‐813S.   155.   Blackshaw,  R.E.,  et  al.,  Differential  response  of  weed  species  to  added  nitrogen.  Weed  Sci.,   2003.  51(4):  532-­‐539.   156.   Carpita,  N.C.  and  M.C.  McCann,  The  functions  of  cell  wall  polysaccharides  in  composition   and  architecture  revealed  through  mutations.  Plant  Soil,  2002.  247(1):  71-­‐80.   211       157.   Henry,   R.J.   and   P.J.   Harris,   Molecular   distinction   between   monocotyledons   and   dicotyledons:  More  than  a  simple  dichotomy.  Plant  Mol.  Biol.  Rep.,  1997.  15(3):  216-­‐218.   158.   Åman,   P.,   Composition   and   structure   of   cell   wall   polysaccharides,   in   Forage   Cell   Wall   Structure  and  Digestibility,  H.G.  Jung,  et  al.,  Editors.  1993,  ASA-­‐CSSA-­‐SSSA:  Madison,  WI.  p.   183-­‐199.   159.   Mosier,   N.,   et   al.,   Features   of   promising   technologies   for   pretreatment   of   lignocellulosic   biomass.  Bioresource  Technol.,  2005.  96(6):  673-­‐686.   160.   Engels,   F.M.   and   H.G.   Jung,   Alfalfa   stem   tissues:   Cell-­‐wall   development   and   lignification.   Ann.  Bot.-­‐London,  1998.  82(5):  561-­‐568.   161.   Wilson,   J.R.,   Organization   of   forage   plant   tissues,   in   Forage   Cell   Wall   Structure   and   Digestibility,  H.G.  Jung,  et  al.,  Editors.  1993,  ASA-­‐CSSA-­‐SSSA:  Madison,  WI.  p.  1-­‐32.   162.   Chen,   L.,   et   al.,   Lignin   deposition   and   associated   changes   in   anatomy,   enzyme   activity,   gene   expression,   and   ruminal   degradability   in   stems   of   tall   fescue   at   different   developmental  stages.  J.  Agr.  Food  Chem.,  2002.  50(20):  5558-­‐5565.   163.   Jung,   H.G.   and   D.A.   Deetz,   Cell   wall   lignification   and   degradability,   in   Forage   Cell   Wall   Structure  and  Digestibility,  H.G.  Jung,  et  al.,  Editors.  1993,  ASA-­‐CSSA-­‐SSSA:  Madison,  WI.  p.   315-­‐346.   164.   Shen,   H.,   et   al.,   Developmental   control   of   lignification   in   stems   of   lowland   switchgrass   variety   alamo   and   the   effects   on   saccharification   efficiency.   Bioenergy   Res.,   2009.   2(4):   233-­‐245.   165.   Dien,  B.,  et  al.,  Improved  sugar  conversion  and  ethanol  yield  for  forage  sorghum  (Sorghum   bicolor  L.  Moench)  lines  with  reduced  lignin  contents.  Bioenergy  Res.,  2009.  2(3):  153-­‐164.   166.   Zhou,   Y.P.,   et   al.,   The   use   of   natural   abundance   stable   isotopic   ratios   to   indicate   the   presence  of  oxygen-­‐containing  chemical  linkages  between  cellulose  and  lignin  in  plant  cell   walls.  Phytochemistry,  2010.  71(8-­‐9):  982-­‐993.   167.   Kruse,   S.,   et   al.,   Evaluation   of   genotype   and   environmental   variation   in   fibre   content   of   silage  maize  using  a  model-­‐assisted  approach.  Eur.  J.  Agron.,  2008.  28(3):  210-­‐223.   168.   Parrish,   D.J.   and   J.H.   Fike,   The  biology  and   agronomy   of   switchgrass   for   biofuels.  Crit.   Rev.   Plant  Sci.,  2005.  24(5):  423  -­‐  459.   169.   Lewandowski,   I.,   et   al.,   The   development   and   current   status   of   perennial   rhizomatous   grasses  as  energy  crops  in  the  US  and  Europe.  Biomass  Bioenerg.,  2003.  25(4):  335-­‐361.   170.   McLaughlin,  S.B.  and  L.  Adams  Kszos,  Development  of  switchgrass  (Panicum  virgatum)  as   a  bioenergy  feedstock  in  the  United  States.  Biomass  Bioenerg.,  2005.  28(6):  515-­‐535.   212       171.   Hultquist,  S.J.,  et  al.,  Chloroplast  DNA  and  nuclear  DNA  content  variations  among  cultivars   of  switchgrass,  Panicum  virgatum  L.  Crop  Sci.,  1996.  36(4):  1049-­‐1052.   172.   Casler,   M.D.,   et   al.,   Latitudinal   adaptation   of   switchgrass   populations.   Crop   Sci.,   2004.   44(1):  293-­‐303.   173.   Fike,   J.H.,   et   al.,   Switchgrass   production   for   the   upper   southeastern   USA:   Influence   of   cultivar  and  cutting  frequency  on  biomass  yields.  Biomass  Bioenerg.,  2006.  30(3):  207-­‐213.   174.   Cassida,  K.A.,  et  al.,  Biofuel  component  concentrations  and  yields  of  switchgrass  in  south   central  U.S.  environments.  Crop  Sci.,  2005.  45(2):  682-­‐692.   175.   Kim,   Y.,   et   al.,   Comparative   study   on   enzymatic   digestibility   of   switchgrass   varieties   and   harvests   processed   by   leading   pretreatment   technologies.   Bioresource   Technol.,   2011.   102(24):  11089-­‐11096.   176.   Yang,   Y.,   et   al.,   Dilute   acid   pretreatment   of   oven-­‐dried   switchgrass   germplasms   for   bioethanol  production.  Energ.  Fuel.,  2009.  23(7):  3759-­‐3766.   177.   Adler,   P.R.,   et   al.,   Biomass   yield   and   biofuel   quality   of   switchgrass   harvested   in   fall   or   spring.  Agron.  J.,  2006.  98(6):  1518-­‐1525.   178.   Montgomery,  D.C.,  Design  and  Analysis  of  Experiments.  7th  ed.  2009,  New  York:  Wiley.   179.   Ma,   H.,   et   al.,   Enhanced   enzymatic   saccharification   of   rice   straw   by   microwave   pretreatment.  Bioresource  Technol.,  2009.  100(3):  1279-­‐1284.   180.   Pan,   X.J.,   et   al.,   Bioconversion   of   hybrid   poplar   to   ethanol   and   co-­‐products   using   an   organosolv   fractionation   process:   Optimization   of   process   yields.   Biotechnol.   Bioeng.,   2006.  94(5):  851-­‐861.   181.   Berlin,   A.,   et   al.,   Optimization   of   enzyme   complexes   for   lignocellulose   hydrolysis.   Biotechnol.  Bioeng.,  2007.  97(2):  287-­‐296.   182.   Gao,   D.H.,   et   al.,   Mixture   optimization   of   six   core   glycosyl   hydrolases   for   maximizing   saccharification   of   ammonia   fiber   expansion   (AFEX)   pretreated   corn   stover.   Bioresource   Technol.,  2010.  101(8):  2770-­‐2781.   183.   Hari   Krishna,   S.   and   G.V.   Chowdary,   Optimization   of   simultaneous   saccharification   and   fermentation   for   the   production   of   ethanol   from   lignocellulosic   biomass.   J.   Agr.   Food   Chem.,  2000.  48(5):  1971-­‐1976.   184.   Balusu,   R.,   et   al.,   Optimization   of   critical   medium   components   using   response   surface   methodology  for  ethanol  production  from  cellulosic  biomass  by  Clostridium  thermocellum   SS19.  Process  Biochem.,  2005.  40(9):  3025-­‐3030.   213       185.   Irick,   T.,   et   al.,   Comparison   of   colorimetric   and   HPLC   techniques   for   quantitating   the   carbohydrate  components  of  steam-­‐treated  wood.  Appl.  Biochem.  Biotech.,  1988.  17(1):   137-­‐149.   186.   Biomass   feedstock   composition   and   property   database.   Last   Updated:   25   January   2006   2006.   Date   Accessed:   12   March   2010.   Available   from:   http://www1.eere.energy.gov/biomass/feedstock_databases.html.   187.   Alizadeh,   H.,   et   al.,   Pretreatment   of   switchgrass   by   ammonia   fiber   explosion   (AFEX).   Appl.   Biochem.  Biotech.,  2005.  124(1):  1133-­‐1141.   188.   Donohoe,   B.S.,   et   al.,   Surface   and   ultrastructural   characterization   of   raw   and   pretreated   switchgrass.  Bioresource  Technol.,  2011.  102(24):  11097-­‐11104.   189.   Bals,   B.,   et   al.,   Evaluating   the   impact   of   ammonia   fiber   expansion   (AFEX)   pretreatment   conditions   on   the   cost   of   ethanol   production.   Bioresource   Technol.,   2011.   102(2):   1277-­‐ 1283.   190.   Chundawat,  S.P.S.,  B.  Venkatesh,  and  B.E.  Dale,  Effect  of  particle  size  based  separation  of   milled  corn  stover  on  AFEX  pretreatment  and  enzymatic  digestibility.  Biotechnol.  Bioeng.,   2007.  96(2):  219-­‐231.   191.   Kumar,  R.,  et  al.,  Physical  and  chemical  characterizations  of  corn  stover  and  poplar  solids   resulting   from   leading   pretreatment   technologies.   Bioresource   Technol.,   2009.   100(17):   3948-­‐3962.   192.   Chundawat,   S.P.S.,   et   al.,   Multi-­‐scale   visualization   and   characterization   of   lignocellulosic   plant  cell  wall  deconstruction  during  thermochemical  pretreatment.  Energy  Environ.  Sci.,   2011.  4(3):  973-­‐984.   193.   Igarashi,  K.,  M.  Wada,  and  M.  Samejima,  Activation  of  crystalline  cellulose  to  cellulose  IIII   results  in  efficient  hydrolysis  by  cellobiohydrolase.  FEBS  J.,  2007.  274(7):  1785-­‐1792.   194.   Dien,   B.S.,   et   al.,   Enzyme   characterization   for   hydrolysis   of   AFEX   and   liquid   hot-­‐water   pretreated   distillers'   grains   and   their   conversion   to   ethanol.   Bioresource   Technol.,   2008.   99(12):  5216-­‐5225.   195.   Gao,   D.,   et   al.,   Hemicellulases   and   auxiliary   enzymes   for   improved   conversion   of   lignocellulosic  biomass  to  monosaccharides.  Biotechnol.  Biofuels,  2011.  4(1):  5.   196.   Humbird,   D.,   et   al.,   Process   design   and   economics   for   biochemical   conversion   of   lignocellulosic  biomass  to  ethanol.  2011,  National  Renewable  Energy  Laboratory:  Golden,   CO.   197.   Sokhansanj,   S.,   et   al.,   Engineering   aspects  of   collecting  corn   stover   for   bioenergy.   Biomass   Bioenerg.,  2002.  23(5):  347-­‐355.   214       198.   Atchinson,   J.E.   and   J.R.   Hettenhaus,   Innovative   methods   for   corn   stover   collecting,   handling,  storing  and  transporting  2004,  National  Renewable  Energy  Laboratory:  Golden,   Colorado.  p.  1-­‐52.   199.   Kim,   S.   and   B.E.   Dale,   Life   cycle   assessment   of   various   cropping   systems   utilized   for   producing  biofuels:  Bioethanol  and  biodiesel.  Biomass  Bioenerg.,  2005.  29(6):  426-­‐439.   200.   Mann,  L.,  V.  Tolbert,  and  J.  Cushman,  Potential  environmental  effects  of  corn  (Zea  mays  L.)   stover  removal  with  emphasis  on  soil  organic  matter  and  erosion.  Agr.  Ecosyst.  Environ.,   2002.  89(3):  149-­‐166.   201.   Wilhelm,   W.W.,   et   al.,   Corn   stover   to   sustain   soil   organic   carbon   further   constrains   biomass  supply.  Agron.  J.,  2007.  99(6):  1665-­‐1667.   202.   Lau,   M.W.,   B.E.   Dale,   and   V.   Balan,   Ethanolic   fermentation   of   hydrolysates   from   ammonia   fiber  expansion  (AFEX)  treated  corn  stover  and  distillers  grain  without  detoxification  and   external  nutrient  supplementation.  Biotechnol.  Bioeng.,  2008.  99(3):  529-­‐539.   203.   Teymouri,   F.,   et   al.,   Ammonia   fiber   explosion   treatment   of   corn   stover.   Appl.   Biochem.   Biotech.,  2004.  115(1):  951-­‐963.   204.   Teymouri,   F.,   et   al.,   Optimization   of   the   ammonia   fiber   explosion   (AFEX)   treatment   parameters   for   enzymatic   hydrolysis   of   corn   stover.   Bioresource   Technol.,   2005.   96(18):   2014-­‐2018.   205.   Wyman,   C.E.,   et   al.,   Comparative   sugar   recovery   data   from   laboratory   scale   application   of   leading   pretreatment   technologies   to   corn   stover.   Bioresource   Technol.,   2005.   96(18):   2026-­‐2032.   206.   Nevins,   D.J.,   Analysis   of   forage   cell   wall   polysaccharides,   in   Forage   Cell   Wall   Structure   and   Digestibility,  H.G.  Jung,  et  al.,  Editors.  1993,  ASA-­‐CSSA-­‐SSSA:  Madison,  WI.  p.  105-­‐129.   207.   Pordesimo,  L.O.,  et  al.,  Variation  in  corn  stover  composition  and  energy  content  with  crop   maturity.  Biomass  Bioenerg.,  2005.  28(4):  366-­‐374.   208.   Shinners,   K.J.   and   B.N.   Binversie,   Fractional   yield   and   moisture   of   corn   stover   biomass   produced  in  the  Northern  US  Corn  Belt.  Biomass  Bioenerg.,  2007.  31(8):  576-­‐584.   209.   Crofcheck,  C.L.  and  M.D.  Montross,  Effect  of  stover  fraction  on  glucose  production  using   enzymatic  hydrolysis.  T.  ASAE,  2004.  47(3):  841-­‐844.   210.   Hoskinson,   R.L.,   et   al.,   Engineering,   nutrient   removal,   and   feedstock   conversion   evaluations  of  four  corn  stover  harvest  scenarios.  Biomass  Bioenerg.,  2007.  31(2-­‐3):  126-­‐ 136.   215       211.   Montross,   M.D.   and   C.L.   Crofcheck,   Effect   of   stover   fraction   and   storage   method   on   glucose   production   during   enzymatic   hydrolysis.   Bioresource   Technol.,   2004.   92(3):   269-­‐ 274.   212.   Duguid,   K.B.,   et   al.,   Screening   for   sugar   and   ethanol   processing   characteristics   from   anatomical  fractions  of  wheat  stover.  Biomass  Bioenerg.,  2007.  31(8):  585-­‐592.   213.   Akin,  D.,  et  al.,  Corn  stover  fractions  and  bioenergy.  Appl.  Biochem.  Biotech.,  2006.  129(1):   104-­‐116.   214.   Sluiter,   A.,   et   al.,   Determination   of   ash   in   biomass,   in   Laboratory   Analytical   Procedures   (LAPs).  2008,  National  Renewable  Energy  Laboratory:  Golden,  CO.   215.   Sluiter,   A.,   et   al.,   Determination   of   structural   carbohydrates   and   lignin   in   biomass,   in   Laboratory   Analytical   Procedures   (LAPs).   2008,   National   Renewable   Energy   Laboratory:   Golden,  CO.   216.   Sluiter,   A.,   et   al.,   Determination   of   extractives   in   biomass,   in   Laboratory   Analytical   Procedures  (LAPs).  2008,  National  Renewable  Energy  Laboratory:  Golden,  CO.   217.   Aden,  A.,  et  al.,  Lignocellulosic  biomass  to  ethanol  process  design  and  economics  utilizing   co-­‐current   dilute   acid   prehydrolysis   and   enzymatic   hydrolysis   for   corn   stover.   2002,   National  Renewable  Energy  Laboratory:  Golden,  CO.   218.   Tolera,  A.  and  F.  Sundstøl,  Morphological  fractions  of  maize  stover  harvested  at  different   stages  of  grain  maturity  and  nutritive  value  of  different  fractions  of  the  stover.  Anim.  Feed   Sci.  Tech.,  1999.  81(1-­‐2):  1-­‐16.   219.   Tolera,  A.,  F.  Sundstøl,  and  A.N.  Said,  The  effect  of  stage  of  maturity  on  yield  and  quality   of  maize  grain  and  stover.  Anim.  Feed  Sci.  Tech.,  1998.  75(2):  157-­‐168.   220.   Russell,  J.R.,  Influence  of  harvest  date  on  the  nutritive  value  and  ensiling  characteristics  of   maize  stover.  Anim.  Feed  Sci.  Tech.,  1986.  14(1-­‐2):  11-­‐27.   221.   Sewalt,  V.J.H.,  et  al.,  Fiber  composition  and  in  vitro  digestibility  of  corn  stover  fractions  in   response  to  ammonia  treatment.  J.  Agr.  Food  Chem.,  1996.  44(10):  3136-­‐3142.   222.   Rencoret,   J.,   et   al.,   Lignin   composition   and   structure   in   young   versus   adult   Eucalyptus   globulus  plants.  Plant  Physiol.,  2011.  155(2):  667-­‐682.   223.   Tarkow,  H.  and  W.C.  Feist,  A  mechanism  for  improving  the  digestibility  of  lignocellulosic   materials  with  dilute  alkali  and  liquid  ammonia,  in  Cellulases  and  Their  Applications,  G.J.   Hajny  and  E.T.  Reese,  Editors.  1969,  American  Chemical  Society:  Washington,  D.  C.  p.  197-­‐ 218.   216       224.   Myers,   D.K.   and   J.F.   Underwood.   Harvesting   Corn   Residue.   Agronomy   Facts.   Last   2007.   Date  Accessed:  3  July.  Available  from:  http://ohioline.osu.edu/agf-­‐fact/0003.html.   225.   Pordesimo,   L.O.,   S.   Sokhansanj,   and   W.C.   Edens,   Moisture   and   yield   of   corn   stover   fractions  before  and  after  grain  maturity.  T.  ASAE,  2004.  47(5):  1597-­‐1603.   226.   Sokhansanj,   S.   and   A.F.   Turhollow,   Baseline   cost   for   corn   stover   collection.   Appl.   Eng.   Agric.,  2002.  18(5):  525-­‐530.   227.   Zaldivar,   J.,   J.   Nielsen,   and   L.   Olsson,   Fuel   ethanol   production   from   lignocellulose:   a   challenge   for   metabolic   engineering   and   process   integration.   Appl.   Microbiol.   Biotech.,   2001.  56(1):  17-­‐34.   228.   Reijnders,   L.,   Ethanol   production   from   crop   residues   and   soil   organic   carbon.   Resour.   Conserv.  Recy.,  2008.  52(4):  653-­‐658.   229.   Shinners,  K.J.,  et  al.,  Single-­‐pass,  split-­‐stream  harvest  of  corn  grain  and  stover.  T.  ASABE,   2007.  50(2):  355-­‐363.   230.   Shinners,   K.J.,   et   al.,   Comparison   of   wet   and   dry   corn   stover   harvest   and   storage.   Biomass   Bioenerg.,  2007.  31(4):  211-­‐221.   231.   Silverstein,   R.A.,   et   al.,   A   comparison   of   chemical   pretreatment   methods   for   improving   saccharification  of  cotton  stalks.  Bioresource  Technol.,  2007.  98(16):  3000-­‐3011.   232.   Varga,  E.,  Z.  Szengyel,  and  K.  Réczey,  Chemical  pretreatments  of  corn  stover  for  enhancing   enzymatic  digestibility.  Appl.  Biochem.  Biotech.,  2002.  98-­‐100(1):  73-­‐87.   233.   Jørgensen,   H.,   J.B.   Kristensen,   and   C.   Felby,   Enzymatic   conversion   of   lignocellulose   into   fermentable  sugars:  challenges  and  opportunities.  Biofuel.  Bioprod.  Bior.,  2007.  1(2):  119-­‐ 134.   234.   Chen,  F.  and  R.A.  Dixon,  Lignin  modification  improves  fermentable  sugar  yields  for  biofuel   production.  Nat.  Biotechnol.,  2007.  25(7):  759-­‐761.   235.   Yang,  B.  and  C.E.  Wyman,  Pretreatment:  the  key  to  unlocking  low-­‐cost  cellulosic  ethanol.   Biofuel.  Bioprod.  Bior.,  2008.  2(1):  26-­‐40.   236.   Hisano,  H.,  R.  Nandakumar,  and  Z.-­‐Y.  Wang,  Genetic  modification  of  lignin  biosynthesis  for   improved  biofuel  production.  In  Vitro  Cell.  Dev.-­‐Pl.,  2009.  45(3):  306-­‐313.   237.   Li,   X.,   J.-­‐K.   Weng,   and   C.   Chapple,   Improvement   of   biomass   through   lignin   modification.   Plant  J.,  2008.  54(4):  569-­‐581.   217       238.   Jackson,   L.,   et   al.,   Improving   saccharification   efficiency   of   alfalfa   stems   through   modification  of  the  terminal  stages  of  monolignol  biosynthesis.  Bioenergy  Res.,  2008.  1(3):   180-­‐192.   239.   Fu,  C.,  et  al.,  Downregulation  of  cinnamyl  alcohol  dehydrogenase  (CAD)  leads  to  improved   saccharification  efficiency  in  switchgrass.  Bioenergy  Res.,  2011.  4(3):  153-­‐164.   240.   Dien,  B.S.,  et  al.,  Enhancing  alfalfa  conversion  efficiencies  for  sugar  recovery  and  ethanol   production  by  altering  lignin  composition.  Bioresource  Technol.,  2011.  102(11):  6479-­‐6486.   241.   Studer,  M.H.,  et  al.,  Lignin  content  in  natural  Populus  variants  affects  sugar  release.  P.  Natl.   Acad.  Sci.  USA,  2011.  108(15):  6300-­‐6305.   242.   Grabber,   J.H.,   et   al.,   p-­‐Hydroxyphenyl,   guaiacyl,   and   syringyl   lignins   have   similar   inhibitory   effects  on  wall  degradability.  J.  Agr.  Food  Chem.,  1997.  45(7):  2530-­‐2532.   243.   Stewart,   J.J.,   et   al.,   The   effects   on   lignin   structure   of   overexpression   of   ferulate   5-­‐ hydroxylase  in  hybrid  poplar.  Plant  Physiol.,  2009.  150(2):  621-­‐635.   244.   York,   W.S.,   et   al.,   Isolation   and   characterization   of   plant   cell   walls   and   cell   wall   components,  in  Methods  in  Enzymology,  H.W.  Arthur  Weissbach,  Editor.  1986,  Academic   Press.  p.  3-­‐40.   245.   Fukushima,   R.S.,   B.A.   Dehority,   and   S.C.   Loerch,   Modification   of   a   colorimetric   analysis   for   lignin  and  its  use  in  studying  the  inhibitory  effects  of  lignin  on  forage  digestion  by  ruminal   microorganisms.  J.  Anim.  Sci.,  1991.  69(1):  295-­‐304.   246.   Rolando,   C.,   B.   Monties,   and   C.   Lapierre,   Thioacidolysis,   in   Methods   in   Lignin   Chemistry,   S.Y.  Lin  and  C.W.  Dence,  Editors.  1992,  Springer-­‐Verlag:  Berlin.  p.  334-­‐349.   247.   Albersheim,  P.,  et  al.,  A  method  for  the  analysis  of  sugars  in  plant  cell  wall  polysaccharides   by  gas  liquid  chromatography.  Carbohyd.  Res.,  1967.  5:  340-­‐345.   248.   Updegraff,   D.M.,   Semimicro   determination   of   cellulose   in   biological   materials.   Anal.   Biohem.,  1969.  32(3):  420-­‐424.   249.   Santoro,  N.,  et  al.,  A  high-­‐throughput  platform  for  screening  milligram  quantities  of  plant   biomass  for  lignocellulose  digestibility.  Bioenergy  Res.,  2010.  3(1):  93-­‐102.   250.   Bals,   B.,   et   al.,   Low   temperature   and   long   residence   time   AFEX   pretreatment   of   corn   stover.  Bioenergy  Res.:  1-­‐8.   251.   Chundawat,   S.P.S.,   V.   Balan,   and   B.E.   Dale,   High-­‐throughput   microplate   technique   for   enzymatic   hydrolysis   of   lignocellulosic   biomass.   Biotechnol.   Bioeng.,   2008.   99(6):   1281-­‐ 1294.   218       252.   Kumar,   R.   and   C.E.   Wyman,   Effects   of   cellulase   and   xylanase   enzymes   on   the   deconstruction  of  solids  from  pretreatment  of  poplar  by  leading  technologies.  Biotechnol.   Progr.,  2009.  25(2):  302-­‐314.   253.   Li,   L.G.,   et   al.,   Combinatorial   modification   of   multiple   lignin   traits   in   trees   through   multigene  cotransformation.  P.  Natl.  Acad.  Sci.  USA,  2003.  100(8):  4939-­‐4944.   254.   Hu,   W.-­‐J.,   et   al.,   Repression   of   lignin   biosynthesis   promotes   cellulose   accumulation   and   growth  in  transgenic  trees.  Nat.  Biotechnol.,  1999.  17(8):  808-­‐812.   255.   Kajita,   S.,   et   al.,   Improvement   in   pulping   and   bleaching   properties   of   xylem   from   transgenic  tobacco  plants.  J.  Sci.  Food  Agr.,  2002.  82(10):  1216-­‐1223.   256.   Wagner,  A.,  et  al.,  Suppression  of  4-­‐coumarate-­‐CoA  ligase  in  the  coniferous  gymnosperm   Pinus  radiata.  Plant  Physiol.,  2009.  149(1):  370-­‐383.   257.   Voelker,  S.L.,  et  al.,  Antisense  down-­‐regulation  of  4CL  expression  alters  lignification,  tree   growth,  and  saccharification  potential  of  field-­‐grown  poplar.  Plant  Physiol.,  2010.  154(2):   874-­‐886.   258.   Whitney,   S.E.C.,   et   al.,   Structural   aspects   of   the   interaction   of   mannan-­‐based   polysaccharides  with  bacterial  cellulose.  Carbohyd.  Res.,  1998.  307(3-­‐4):  299-­‐309.   259.   Ragauskas,  A.J.,  et  al.,  From  wood  to  fuels:  Integrating  biofuels  and  pulp  production.  Ind.   Biotechnol.,  2006.  2(1):  55-­‐65.   260.   Willför,   S.,   et   al.,   Polysaccharides   in   some   industrially   important   hardwood   species.   Wood   Sci.  Tech.,  2005.  39(8):  601-­‐617.   261.   Choi,   J.,   D.-­‐H.   Choi,   and   O.   Faix,   Characterization   of   lignin-­‐carbohydrate   linkages   in   the   residual   lignins   isolated   from   chemical   pulps   of   spruce   (Picea   abies)   and   beech   wood   (Fagus  sylvatica).  J.  Wood  Sci.,  2007.  53(4):  309-­‐313.   262.   Wadenbäck,   J.,   et   al.,   Lignin   biosynthesis   in   transgenic   Norway   spruce   plants   harboring   an   antisense  construct  for  cinnamoyl  CoA  reductase  (CCR).  Transgenic  Res.,  2008.  17(3):  379-­‐ 392.   263.   O'Connell,   A.,   et   al.,   Improved   paper   pulp   from   plants   with   suppressed   cinnamoyl-­‐CoA   reductase  or  cinnamyl  alcohol  dehydrogenase.  Transgenic  Res.,  2002.  11(5):  495-­‐503.   264.   Piquemal,   J.,   et   al.,   Down-­‐regulation   of   cinnamoyl-­‐CoA   reductase   induces   significant   changes  of  lignin  profiles  in  transgenic  tobacco  plants.  Plant  J.,  1998.  13(1):  71-­‐83.   265.   Chabannes,   M.,   et   al.,   Strong   decrease   in   lignin   content   without   significant   alteration   of   plant   development   is   induced   by   simultaneous   down-­‐regulation   of   cinnamoyl   CoA   219       reductase   (CCR)   and   cinnamyl   alcohol   dehydrogenase   (CAD)   in   tobacco   plants.   Plant   J.,   2001.  28(3):  257-­‐270.   266.   Leplé,   J.-­‐C.,   et   al.,   Downregulation   of   cinnamoyl-­‐Coenzyme   A   reductase   in   poplar:   Multiple-­‐level  phenotyping  reveals  effects  on  cell  wall  polymer  metabolism  and  structure.   Plant  Cell,  2007.  19(11):  3669-­‐3691.   267.   Marita,  J.M.,  et  al.,  NMR  characterization  of  lignins  in  Arabidopsis  altered  in  the  activity  of   ferulate  5-­‐hydroxylase.  P.  Natl.  Acad.  Sci.  USA,  1999.  96(22):  12328-­‐12332.   268.   Franke,   R.,   et   al.,   Modified   lignin   in   tobacco   and   poplar   plants   over-­‐expressing   the   Arabidopsis  gene  encoding  ferulate  5-­‐hydroxylase.  Plant  J.,  2000.  22(3):  223-­‐234.   269.   Pedersen,  M.,  A.  Viksø-­‐Nielsen,  and  A.S.  Meyer,  Monosaccharide  yields  and  lignin  removal   from  wheat  straw  in  response  to  catalyst  type  and  pH  during  mild  thermal  pretreatment.   Process  Biochem.,  2010.  45(7):  1181-­‐1186.   270.   Li,   W.,   et   al.,   Rapid   dissolution   of   lignocellulosic   biomass   in   ionic   liquids   using   temperatures  above  the  glass  transition  of  lignin.  Green  Chem.,  2011.  13(8):  2038-­‐2047.   271.   Yan,   M.M.   and   C.B.   Purves,   Extraction   of   a   lignin   fraction   from   maple   wood   by   liquid   ammonia.  Can.  J.  Chem.,  1956.  34(12):  1747-­‐1755.   272.   Chundawat,   S.P.S.,   et   al.,   Restructuring   the   crystalline   cellulose   hydrogen   bond   network   enhances  Its  depolymerization  rate.  J.  Am.  Chem.  Soc.,  2011.  133(29):  11163-­‐11174.   273.   Hall,   M.,   et   al.,   Cellulose   crystallinity  –  a   key   predictor   of   the   enzymatic   hydrolysis   rate.   FEBS  J.,  2010.  277(6):  1571-­‐1582.   274.   Laureano-­‐Perez,   L.,   et   al.,   Understanding   factors   that   limit   enzymatic   hydrolysis   of   biomass.  Appl.  Biochem.  Biotech.,  2005.  124(1):  1081-­‐1099.   275.   Zhu,   L.,   et   al.,   Structural   features   affecting   biomass   enzymatic   digestibility.   Bioresource   Technol.,  2008.  99(9):  3817-­‐3828.   276.   Li,  C.,  et  al.,  Influence  of  physico-­‐chemical  changes  on  enzymatic  digestibility  of  ionic  liquid   and  AFEX  pretreated  corn  stover.  Bioresource  Technol.,  2011.  102(13):  6928-­‐6936.   277.   Eronen,   P.,   et   al.,   Interactions   of   structurally   different   hemicelluloses   with   nanofibrillar   cellulose.  Carbohyd.  Polym.,  2011.  86(3):  1281-­‐1290.       220