WWW!llll‘lll‘l’i‘lll’l‘llflfil'lfllfl’illll‘lfil

3 1293 00790 8431

 

_\

F LIBRARY
Michigan State
University J

‘Lr‘,:_fj)l5

 

 

M

This is to certify that the

dissertation entitled

APPLICATIONS OF TRIPLE QUADRUPOLE MASS SPECTROMETRY
TO THE DIRECT DETERMINATION OF MICROBIAL BIOMARKERS
AND THE RAPID DETECTION AND IDENl'IFICATION OF
MICROORGANISMS

presented by

Mark James Cole

has been accepted towards fulfillment
of the requirements for

PhD. degree in Analytical Chemistry

flflfl/

Major professor

Date December 5th. 1990

 

\)
C" ‘)
04;)
\‘u

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

—_-V__..___ *_ r77 7

Kw.
'N.

u 3
ES
Ox)

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

DATE DUE DATE DUE DATE DUE I
W
L________

!

 

   
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

fl

MSU II An Affirmative Action/Equal Opportunity Institution
cwmmd

 

 

 

 

APPLICATIONS OF TRIPLE QUADRUPOLE MASS
SPECTROMETRY TO THE DIRECT DETERMINATION OF
MICROBIAL BIOMARKERS AND THE RAPID DETECTION

AND IDENTIFICATION OF MICROORGANISMS

By

Mark James Cole

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Chemistry
1990

A

[245- 60%

ABSTRACT

APPLICATIONS OF TRIPLE QUADRUPOLE MASS SPECTROMETRY TO
THE DIRECT DETERMINATION OF MICROBIAL BIOMARKERS AND
THE RAPID DETECTION AND IDENTIFICATION OF
MICROORGANISMS

By

Mark James Cole

Methodolog' has been developed that takes full advantage of the
triple quadrupole mass spectrometer for detecting and identifying
microbial biomarkers. Complete tandem mass spectrometry (MS/ MS)
data were acquired on the pyrolysis products obtained from bacteria.
These data showed that very little additional information could be
obtained from pyrolysis using tandem mass spectrometry over that
using conventional mass spectrometry. A rapid. sensitive, and
automated method was developed for the direct determination of
phospholipid structures, from crude lipid extracts. using fast atom
bombardment/MS/MS. This method takes advantage of neutral loss.
parent ion, and daughter ion scan modes to provide independent
diagnostic mass spectra for each of many specific phospholipid
classes, and the compositions and positions of the fatty acid
constituents on each phospholipid. A rapid extraction procedure
combined with the automated instrument control program produces

this information in a few minutes from a single sample. Examples of

similar data were also obtained, using this same methodology, from
glycolipids. This technique has been applied, in this work, to the
detection and identification of microorganisms - including bacteria.
fungi, and amoebae. Applications in urinary tract infections and food-
borne contaminants are presented The need for, and general
structure of, a searchable database of the detected biomarkers in a

large variety of microorganisms is discussed.

The ability of the triple quadrupole mass spectrometer to
perform ion-molecule reactions was exploited for differentiating
phospholipid classes through reactions with ethyl vinyl ether in the
second quadrupole. The protonated molecules and reaction product
ions observed permitted the differentiation of various phospholipid
classes. Particular neutral gain scans are specific for each
phospholipid class. Ion dissociation products can be observed in the
same scan as the ion reaction products to provide data on the fatty
acid composition and position on the glycerophosphate core, along

with the phospholipid class.

A preliminary investigation shows great promise for the
application of tandem mass spectrometry in the detection and
identification of sterols in crude lipid extracts. Particular parent scans
are shown to be specific for particular classes of sterols. The
identification of sterols may be especially valuable for fungal

identification.

Cepyright © by
MARK JAMES COLE
1990

“At least I knew who I was
when I got up this morning.
but I think I must have been changed
several times since then.”

Alice, in Alice's Adventures in Wonderland.

“What kind of data did you expect to get
after using a hand-grenade for sampling?”

Dr. David C. White, upon hearing of our pyrolysis results.

ACKNOWLEDGEMENTS

Foremost, I would like to thank my advisor. Chris Enke.
Through his intellect, insight, sensitivity, and creativity, he has
provided me an outstanding role-model on which to base my scientific
career. I am forever grateful to him for allowing me the freedom to
pursue my research interests. Joining his research group was one of
the best decisions I ever made. I would also like to thank the
members of my graduate committee, Dr. Jack Watson, Dr. George

Leroi, and Dr. James Tiedje for their advise and assistance.

I would like to thank Eric Hemenway and Kate Noon for their
help in co-organizing the lst Annual Conference for Users of the Enke
Group's TSQ-7OB in Microbial Analysis (UEG’I‘MA). This small, but
fanatical, users-group meets on Lake Huron once a year (or more.
depending on the wind) to discuss important stuff. I would also like to
thank Kevin Hart. Jon Wahl, Ron Lopshire, and the rest of the Enke
group for their help and friendship.

Three important friends deserve special mention. Tom Wiese
has been the greatest friend a person could have. He is intelligent.
compassionate, knows how to sail, and can sneak me in skiing for free.
I owe much of my sanity, along with much of my insanity, to him.
Secondly, I would like to thank Mary Puzycki. She has always made
my world a little more surrealistic, and I’m probably the better for it.
Our “discussions” always motivated me to spend long hours working

alone in the lab. I am grateful for her love and friendship. Of all the

friends I have made during my stay at MSU, Kris “Mel-Mauler” Kurtz
is one who truly knows how to let loose. She alone is responsible for
the rest of my insanity. I am grateful to her for making the last five
years anything but boring.

Dr. Dave White has played an important role in my career. His
friendship and to-the-point advice has made an impact on me. If I
ever get rich, he deserves the credit. Also deserving mention are Dr.
James Tiedje, Dr. Larry Fomey, and Dr. Mike King. These three have

been a great help to me and my research.

My parents have provided me with love and support throughout
my life. I am also grateful to my grandparents for their moral and

financial support through difllcult‘times.

My boat, Fetish, has provided me with more fun than any
graduate student deserves. She has never failed to be there when I
needed her: she is always gentle, and never complains. I feel very
fortunate to have such an escape from frustration. I have overcome

many problems in my research while aboard her.

Last, I would like to thank those who financially supported my
research: the National Science Foundation’s Center for Microbial
Ecoloy at Michigan State University, the United States Department of
Agriculture, the State of Michigan’s Research Excellence Fund, the
MSU Biomedical Research Support Grant, and the Yates Research
Fellowship.

TABLE OF CONTENTS

Iistof’l‘ables ........................................................................................................... xi
UstofFigmee ........................................................................................................ xii
IistofStmctures ................................................................................................ .xv

CHAPTER 1 Introduction

Introduction .................................................................................................... l
Resesrch Goals ............................................................................................... 4
Backgamd ...................................................................................................... 5
Traditional and Non-Instrumental
Methods of Analysis .......................................................................... 6
Instrumental Analysis of Whole Cells ...................................... 10
Instrumental Analysis. of Specific
Microbial Biomarkers .................................................................... 12
Thesis Organimtion .................................................................................... 14
References ..................................................................................................... 18

I

Introduction .................................................................................................. 20
Backgound .................................................................................................... 21

Pyrolysis .............................................................................................. 21

Pyrolysis/Mass Spectrometry ..................................................... 25

Pyrolysis [Tandem Mass Spectrometry ................................... 27
Experimental ................................................................................................ 28
Results and Discussion .............................................................................. 29

Pyrolysis Inlet System ................................................................... 29

Collection of Complete

Pyrolysis/MS/MS Data Fields ..................................................... 33
Conclusions .................................................................................................. .35
References ..................................................................................................... 41

viii

CHAPTERS Direct Determination of Phospholipid
Structures in Bacteria by Fast Atom
Bombardment / Triple Quadrupole Mass

Spectrometry
Introduction ................................................................................................. 43
Background ................................................................................................... 44
Eiqaerlmental ............................................................................................... 51
Results and Discussion ............................................................................. 52
General Profiling Scheme ........................................................... 53
Sensitivity .......................................................................................... 61
Precision ............................................................................................ 61
Extraction Process ......................................................................... 64
Conclusions ................................................................................................... 66
References .................................................................................................... 69

CHAP’I‘ER4 Fast Atom Bombardment / Ion-Molecule
Reactions for the Difl'erentiation of Phospholipid

Classes
Introduction and Background ............................................................... 7 1
Experimental ............................................................................................... 71
Results and Discussion ............................................................................. 73
Unique Reaction Products .......................................................... 73
Where the Resciions Occur ........................................................ 77
Collisionally Induced
Dissociation Data ............................................................................. 77
Neutral Gain Scans ......................................................................... 80
References ..................................................................................................... 85

CHAPTER 5 Applications of Phospholipid Analysis to the
Detection and Identification of Microorganisms

Introduction and Background ................................................................ 86
Experimental ................................................................................................ 88
Results and Discussion .............................................................................. 90

Comparing the Data ........................................................................ 9O

ix

Microorganisms Analyzed ............................................................ 92

Detection and Identification

of Urinary 'Itact Infections ......................................................... 95
Difi'erentiation of Food -Bome Contaminants ..................... 104
Database/ Library of Microbial Profiles .................................. 109
References .................................................................................................. 112

CHAPTER 6 Applications of Other Biomarker Analyses to the
Detection and Identification of Microorganisms

Introduction and Background ............................................................. l 13
Experimental ............................................................................................. 119
Results and Discussion ........................................................................... l2 1
Glycolipids ...................................................................................... 121
Sterols .............................................................................................. 123
Other Potential Lipid Biomarkers ......................................... 128
References .................................................................................................. 131

LIST OF TABLES

1.1 Three of the biochemical tests used for identification

of bacterial genera. ............................................................................................ 8
2.1 Curie-point temperature as a function of composition .................... 24
3.1 Characteristic positive ion neutral losses for various head

groups and fatty acid anions in the daughter ion spectra ............... 50
3.2 Phospholipid recovery for various values of log number

of E. mli cells extracted ............................................................................... 65
3.3 Phospholipid recovery from E. call for various

extraction times ............................................................................................... 67
4.1 Composition of addition species ............................................................... 76
5.1 listing of microorganisms characterized by this technique .......... 93
5. 1 cont. ...................................................................................................................... 94

xi

1.1

1.2

2.1
2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3.1

3.2

3.3

LIST OF FIGURES

Schematic of a triple quadrupole mass spectrometer

along with the difi'erent scan modes possible ...................................... 3
Chart of a few of the morphological and biochemical

tests used for the identification of bacterial genera .......................... 7
Schematic diagram of a Curie-point pyrolyzer .................................. 23
Pyrolysis/ chemical ionization mass spectrum of

Corynebaaenum nephruii ...................................................................... 26
Schematic of the slow-leak, pyrolyzate containment,

inlet system ..................................................................................................... 3 l

Pyrolyzate profiles from different versions of the
pyrolyzate inlet system: a) leak hole too large,

b) leak hole too small. c) leak hole correct size ............................... 32
Pyrolysis/ chemical ionization mass spectra for

three different bacteria .............................................................................. 34
Complete pyrolysis / MS/ MS data space obtained

franE (nit ..................................................................................................... 36
Daughter spectra of m/z 196 obtained from the

pyrolyzate of three difi'erent bacteria .................................................... 37
Daughter spectra of m/z 209 obtained from the

pyrolyzate of three different bacteria .................................................... 38
Daughter spectra of m/z 225 obtained from the

pyrolyzate of three difi'erent bacteria .................................................... 39
Examples of the head groups from five

phospholipid classes .................................................................................... 47
Low-energy CID fragmentation of phospholipids

for positive and negative ions ................................................................... 49

The FAB/MS/MS data space obtained by this technique.

The masses of the phospholipids in each class are

obtained by neutral loss or parent ion scans specific for

each class. The daughter spectrum of each phospholipid

species provides the masses and positions of the fatty

acids .................................................................................................................... 54

 

xii

0) Conventional mass spectrum of a Proteus vulgar-ls
extract. b) Neutral loss spectrum of 141 11; specific for
PE. c) Neutral loss spectrum of 172 11: specific for PG ................. 56

Negative ion daughter spectrum of a phosphatidylglycerol
species at m/z 733 showing the two fatty acid ions and

their relative intensities. Cl6:0 is a 16-carbon fatty acid

with no unsaturation: C17 :cyc is a l7-carbon propyl fatty

acid ...................................................................................................................... 59

An examme of the use of neutral loss and daughter

spectra for obtaining a partial profile of E. coli. The

neutral loss scans obtain the (class) mass profiles while

the daughter scans obtain the (class)(fatty acid) formula

data. As indicated, seven other neutral loss scans are

normally performed for a "complete" profile. Daughter

spectra are obtained for the major peaks in each (class)
nusssesn. ......................................................................................................... 60

a] PE mass profile of l picogram PE extracted from

E. colt: equivalent to 10 cells. b) PE mass profile of
l microgram PE extracted from E. coli: equivalent to

107mlls ............................................................................................................ 62

Percent of total PE as a function of the log number of
cells for the major species present in E. coli. ................................... 63

Mass spectra of the reactions of four different

phospholipid classes: a) PE Cl6:0,Cl7:cyc b) PG

Cl6:0,C16:0 c) PC C18:0,C18:2 d) PI C15:0,C12:0.

Cl6:0 is a l6-carbon fatty acid with no unsaturation,

Cl7:cyc is a l7-carbon propyl fatty acid, Cl8:2 is an

18-carbon fatty acid with 2 points of unsaturation, etc.

The order in which the fatty acids are listed correspond

to their positions on the glycerol core ................................................. 74

omt. .................................................................................................................... 75

Mass spectra of two phosphatidylethanolamine analogues.
These lipids are of the same class but differ in their fatty
acid content: a.) PE C16:O,Cl7:cyc b) PE 018:0,Cl7:cyc ............. 78

Full daughter scan of a selected phosphatidylethanolamine
parent showing CID fragments as well as the characteristic
resction products .......................................................................................... 79

0) Conventional mass spectrum of a sample containing

phosphatidylglycerol. b) Neutral gain scan for the addition
of 26 u. This addition is specific for phosphatidylglycerol ........... 82

xiii

4.5

5.1

5.2

5.2
5.3

5.4

5.5

5.6

6.1

6.2
6.3

a) Neutral gain of 57 u: specific for PI.

b) Neutral loss of 260 11: also specific for PI ....................................... 83
Phosphatidylethanolamine mass profiles from three

difi'erent isolates of Pseudomonas aeruginosa. ................................... 99
Phosphatidylethanolamine mass profiles from six

bacterial species known to cause urinary tract infections ......... 101
cont. .................................................................................................................. 102

Phosphatidylethanolamine mass profile from an
infected urine sample ............................................................................... 103

Phosphatidylethanolamine mass profiles from three
bacteria, grown in tetrathionate broth, important to
the food industry ......................................................................................... 106

Phosphatidylethanolamine mass profiles from an
unknown contaminant and Salmonella abaetetuba. ....................... 107

Phosphatidylethanolamine mass profiles from three
bacteria, grown on trypticase soy agar. important to

the food industry ......................... . ............................................................... 108
Neutral loss of 252 11, specific for monogalactosyl-

diacylglycerol, from Legionella pneumophila knox ....................... 122
Parent scan for m/z 271 from fungal sample 138-1 ..................... 126
Parent scan for m/z 273 from fungal sample 138-1 ..................... 127

xiv

LIST OF STRUCTURES

3. 1 Basic glycerophospholipid structure ...................................................... 46
6. 1 General glyeosyldiacyiglycerol structure ............................................ 1 14
6.2 Monogalactosyldiacylglycerol .................................................................. l 16
6.3 General sterol structure ............................................................................ 1 17
6.4 Ergosterol. ...................................................................................................... 118

CHAPTER 1

Introduction

Introduction

 

icrobiologists have long held the desire to possess a

simple method for the rapid. reliable, and sensitive

detection and identification of microorganisms without
prior cultivation. Traditional microbial methodology was developed
many years ago and has remained relatively unchanged since then:
much of its roots are imbedded in the very beginnings of microbiolog.
While these methods are currently used to differentiate and identify
thousands of microbial species and strains, they are time-consuming,
prone to errors. and often rely on numerous subjective decisions
during the course of an identification. All of these methods require
prior isolation and cultivation of the organism - precluding
identification of organisms that will not grow outside their natural
environment. More current methods employ highly developed
routines that use sophisticated morphological, serological and
biochemical procedures. Some of them are so specific that they have
no general applicability and can only be used when the identification is
nearly complete. Thus, an unfulfilled need exists in microbiology for
new methodology which can achieve most of the requirements of

cultivation. speed, objectivity, and applicability.

Because it meets these requirements, mass spectrometry is
finding increased applications in microbiology. Tandem mass
spectrometry becomes an even more powerful tool for the detection of
microbial biomarkers by providing the high sensitivity characteristic of
mass spectrometry along with a great improvement in selectivity.
This selectivity allows for rapid and complete characterization of
difi'erent components of a microbe without their prior separation. In
addition. tandem mass spectrometry has the ability to detect trace

quantities of these components in complex mixtures and matrices.

The triple quadrupole mass spectrometer (TQMS) provides
many advantages for detecting and profiling biomarkers. The TQMS
provides high selectivity due to unit resolution for both mass analyzers,
high sensitivity due to the high efficiency of the collisional dissociation
process, a full array of ionization methods, and the ability to be highly
automated - allowing for rapid scan rates and easy implementation of
the MS/MS operating modes. A schematic of a TQMS, along with the
different scan modes possible. are shown in Figure 1.1. The daughter
scan produces a mass spectrum of all the fragment ions formed from a
selected parent ion. This is accomplished by setting the first
quadrupole to pass ion current at a particular parent m/z value and
scanning the third quadrupole. Daughter spectra are useful for
structure determination of selected ions. The parent scan produces a
mass spectrum of all the parent ion masses that dissociate to form a
particular daughter ion. This is obtained by scanning the first
quadrupole, while the third quadrupole is set for the desired daughter

m/z value. Parent spectra provide compound class

2

 

conversion
dynode (RUN)

AA

02 I
___.._/~' _, 0 b
3% I A 333 I elec‘tron
“tailor
l 03

 

 

 

 

 

 

 

DAUGHTER SCAN

 

 

 

 

 

 

 

 

 

 

 

 

 

 

F1 0F=fi 6:;
E r —g. :7.
«IF air-:5 Fm
SELECT m/z 500 SCAN m/s 50-510

PARENT SCAN

a F 171:,

#fi F1 #fi
SCAN m/s 140-500 SELECT m/z 150

NEUTRAL LOSS / GAIN SCAN

a G=H E

 

 

 

 

 

 

—_. _' 'é I. —_.—.
qfi==1 “Ffi r3=fi
SCAN m/z 100-500 SCAN m/s 50-450

SCAN m/z 150-550

 

Figure 1.1 Schematic of a triple quadrupole mass spectrometer along with
. the different scan modes possible.

determination. By choosing a daughter m/z value that corresponds to
a fragment characteristic of a particular compound class, a mass
spectrum is obtained containing peaks representing only those ions
formed in the source that belong to that class. The neutral loss and
neutral gain scans also provide compound class determination. A
neutral loss spectrum contains peaks representing all of the parent
ion masses that dissociate to lose a particular neutral mass, while a
neutral gain spectrum contains peaks representing all of the parent
ion masses that react with a reactant gas to form product ions of a
particular mass gain. This is implemented by scanning the first and
third quadrupoles concurrently, but with a mass offset equal to the
selected neutral loss or neutral gain. By choosing a neutral loss
characteristic of a particular compound class, a mass spectrum is
obtained containing peaks representing only those ions that belong to
that class. Neutral gain scans provide even greater selectivity of
compound class due to their reliance on a particular chemistry of the
chosen class of compounds. Only those ions that have the correct
chemistry to react in a selected manner will be detected. By choosing
a reaction unique to the compound class of interest, a mass spectrum
is obtained containing peaks representing only ions that can

chemically undergo the particular reaction.

Research Goals

 

This research is concerned with taking full advantage of the triple
quadrupole mass spectrometer in detecting and identifying microbial

biomarkers. Microorganisms contain a multitude of different

biomolecules that are needed to perform the functions required for
each microbial species to fill, and survive in, its niche in nature. The
natural selective forces imposed on different microorganisms have
required microbes to adapt and optimize specific sets of biomolecules
for survival in specific environments. Examples of these biomolecules
include the lipids needed for membrane function and integrity, the
respiratory quinones needed for metabolism, the polysaccharides
needed for protection from the environment, and other molecules
required for survival. The collection of the molecules needed by each
microbial species is unique to that species. In fact. individual subsets
of this complete collection of biomolecules are often unique to a
particular species. Identification of these unique subsets can provide
biomarkers for identification of microorganisms. This research is
centered around developing tandem mass spectrometry methodoloy
for identifying these biomarkers and applying it to the rapid detection
and identification of microorganisms. Special attention is given to
speed and sensitivity, as these are areas where current techniques fall

far short of being truly useful.

Bac und

Microbial identification techniques have undergone an evolution that
has taken them from early visual observations, through highly
sophisticated molecular biolog’, and into very sensitive instrumental
techniques. Putting this evolution in perspective is important in
recognizing where this research fits into the scheme of things. Three

areas in particular are important in the development of different

technologies: 1) traditional and non-instrumental methods of analysis,
2) instrumental analysis of whole cells, and 3) instrumental analysis of

specific biomarkers.

Traditional and Non-Instrumental Methods of Analysis

Numerous methods exist for monitoring the presence of microbes in
biological samples. These methods involve the direct detection of the
organism, or some part of the organism, or the indirect detection of a
characteristic metabolite of the organism. The most widely used
method is the selective agar pure culture assay. The microorganism in
question is grown on a selective medium in which the suspect
microbe is known to react in a consistently predictable manner.
Examples of typical reactions include the change of agar or solution
color, evidence of motility, the evblution of gas, and the production of
characteristic metabolites. A chart showing a few of these
morphological and biochemical tests for the identification of bacterial
genera is shown in Figure 1.2. In this example, bacteria are grouped
according to their responses to these tests. This chart is only for
identification of bacterial genera, species determination involves even
more tests and can be much more difficult. While widely used, these
techniques are time-consuming; many require 2 to 4 days to obtain
presumptive results and 5 to 7 days to obtain complete results.
Furthermore, confidence in these tests may be compromised by
reliance on subjective judgments such as a change of color or
production of gas. Three of these tests along with their negative
results, ideal positive results, and real-life positive results are listed in

Table 1.1. The positive results most often obtained in real working

6

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I 7 I I
"I?“ ”r film I ' coca
l rmfi fi" :5, t""' m... {mm
m as... man... w... W m
w am. W W
| I
"I" ”rm
I I I j
01'1“! (t) . H m M all“! H

 

l
M m H was) would?

IBM IIIH I ' I
III-cum slim-nan m m

 

 

 

Figure 1.2 Chart of a few of the morphological and biochemical tests used for the
identification of bacterial genera.

 

Table 1.1 Three of the biochemical tests used for bacterial

genera determination

REAL-LIFE
IESI W NEGAIliLEBESilLI ITIV R LT

CARBOHYDRATE YELLOW, WITH GAS RED, NO GAS REDDISH-YELLOW,
FERMENTATION BUBBLES FORMED POSSIBLE GAS FORMED
METHYL RED RED" YELLOW REDDISH-YELLOW
MIXED ACID
FERMENTATION

CITRATE UTILIZATION BLUE GREEN BLUE-GREEN

* Requires a 5-day incubation

 

conditions require a subjective judgment in interpreting the results; a
mistake can easily be made on a color test. and many bacteria do not

produce gas readily enough for a positive identification.

Another widely used set of techniques is the irnmunoassays; the
three most common being enzyme immunoassay (EIA),
radioirnmunoassay (RIA). and irnmunofluorescence assay (IFA). Each
method consists of using a labeled antibody specific for an antigen
associated with the suspect organism. Enzyme immunoassay uses an
enzyme-labeled antibody, RIA uses an isotopically-labeled antibody. and
[FA uses a fluorescently-labeled antibody. False-positive results. due to
non-specific binding, cross-reactions. and interferences of other
components in the sample are common [1.2]. While more rapid than
the pure culture assay. these techniques still require up to 27 hours
and numerous steps [2,3]. although newer developments have
facilitated less cumbersome assays which can be completed in an 8

hour working day [4].

More recently. genetic methods of detection and identification
by DNA colony hybridization have been studied [5]. Detection is
accomplished by the hybridization of isotopically—labeled DNA probe
segments, that are specific for the selected organism, with the
complementary strands of the DNA from the selected organism. While
this method has a 95 percent accuracy [5], it is expensive and requires
many elaborate and technically difficult procedures. Also. the DNA

probe used must be unique and specific to some function of the

microbe. requiring that the microbe be well characterized. This initial

characterization can be very time and labor intensive.

Indirect methods of detection depend on the ability to detect
metabolites produced by the microorganism. Chromatographic and
colorimetric techniques are often used. These methods must be
specific for the particular microbe of interest and some require
elaborate sample workup. Chromatographic identification is based on
retention values that can lead to greatly decreased confidence in the

results when an interfering matrix is introduced.

Instrumental Analysis of Whole Cells

The first efforts toward instrumental analysis of bacteria were
performed using analytical pyrolysis with gas chromatography. and
later mass spectrometry, on whole microbial cells. The use of whole
cells presented a problem to gas chromatographs and mass
spectrometers in that they were not volatile. Therefore pyrolysis was
used to break the whole cells up into volatile components which are
more amenable to the instrumental analysis. Pyrolysis is defined as a
chemical transformation through the application of energr in the form
of heat alone. resulting in a series of more volatile fragment molecules
(pyrolyzate) [6]. More simply, a non-volatile sample is rapidly heated
in an inert environment. No net oxidation can occur. so molecules in
the sample cleave at their weakest bonds resulting in the formation of
more volatile components from the sample. Since the molecules

fragment at specific points. the final mixture of low molecular weight

10

products is expected to contain information characteristic of the

original sample [7].

The first application of analytical pyrolysis to the detection of
microorganisms was by Wilson and Oyama in conjunction with the
Viking mission to Mars [8]. Gas chromatography was used as the
means of detection. and this work triggered an extensive series of
studies by other researchers [9.10.11]. Recent work in Py/GC has
been reviewed by Bayer [12] and Fox [13]. Despite the rapid
development of the technique. its use for microbe characterization
remains limited due to poor standardization and the lack of inter-
laboratory reproducibility [14]. This is due to several practical gas
chromatographic problems. such as insufficient column resolution.
gradual deterioration of column performance. or sudden differences
caused by column replacement. These problems pose constraints on
the establishment of a library of standard fingerprints. thereby
restricting this technique to a single laboratory for successful

completion of a given project.

The advances occurring in mass spectrometry instrumentation
and computing soon had researchers turning toward pyrolysis-mass
spectrometry for the study of microbes. In an attempt to overcome
those problems inherent in pyrolysis-gas chromatography. Meuzelaar
and coworkers developed techniques for obtaining reproducible mass
spectra of bacteria [15]. They were able to differentiate the few
species studied at the genus. species. and subspecies level without gas

chromatography. Recent work on the application of PY/MS in

11

microbiologr has been reviewed by Wieten. et al. [16]. So far. analyses
by pyrolysis/mass spectrometry of over 35 bacterial strains have been
reported [17]. However. spectral reproducibility was found to vary
between different instruments and laboratories [18].

Instrumental Analysis of Specific Microbial Biomarkers

The complexity of the data obtained with pyrolysis techniques
presented problems with interpretation. Involved mathematical and
computer tools were developed to aid in this interpretation. However.
it became apparent to some researchers that most of the data
contained in pyrolysis is not very useful for microbial identification
because microbes shared many common biomolecules. These
researchers started looking for specific sets of biomarkers that would

vary among microbial species and 'thus be useful for identification.

By far the most successful technique in this area is the profiling
of cellular fatty acids by gas chromatography [19] and gas
chromatography] mass spectrometry [20]. In this method. fatty acids
are extracted from whole microbial cells and derivatized to their
methyl esters. The fatty acid methyl esters are analyzed by G0 or
GC/MS. and a quantitative profile is obtained for the microbe studied.
The cellular fatty acids are found to be good biomarkers for microbial
identification. as their content and abundance remain unique and
predictable under strict growth conditions. The problems with this
technique lie in the need for a fairly large number of cells. the large
amount of sample handling and preparation in the extraction and

derivatization. and the general reproducibility problems associated

12

with GC. The use of GC / MS has greatly improved the sensitivity of this
technique [21]. but the other problems still remain.

A recent twist on this method is the use of pyrolysis to obtain
fatty acid information. One approach uses a methanolic solution
applied directly to the whole cells contained on the pyrolysis filament
[22.23]. Pyrolysis produces fatty acid methyl esters which are
analyzed by GC. Fatty acid profiles are obtained which are nearly
identical to the previously described method. The other approach
uses pyrolysis to directly produce fatty acid fragments from whole
cells. with subsequent analysis by mass spectrometry [24]. However.
since only the nominal masses of the fatty acids are obtained. this
method suffers from an inability to identify fatty acid isomers,
branched chain fatty acids. and hydroxy fatty acids. Therefore, much

useful information is lost.

Research has recently concentrated on the use of microbial
phospholipids for identification. The phospholipids contain all of the
fatty acid information as well as the information regarding the
presence of their classes and species within each microorganism.
Much mass spectrometry work has been performed toward elucidating
phospholipid structure. but only a few researchers are attempting to
use phospholipids for microbial identification. Some of the first work
involved fast atom bombardment. plasma desorption. and laser
desorption of phospholipids directly from lysed cells [25.26]. This
work demonstrated the usefulness of cation and anion spectra for

obtaining phospholipid information from bacteria. The sensitivity was

13

good. and little sample preparation was required. In further work. the
anion spectra were used with a linear regression analysis for library-
matching of unknown spectra [27]. Pramanik et al. applied this
method to other bacterial species [28].

This is the current extent of the evolution of analytical microbial
identification techniques. Research must now focus on increased
sensitivity. speed of analysis. new biomarkers. and improvements in
analysis methodology. The research presented here directly
addresses these issues and is a logical next step in the continuing

development of analytical techniques for microbial identification.

Thesis tion

Chapter two contains the details of an attempt to develop a method for
microbial identification based on pyrolysis and tandem mass
spectrometry. Previous work with pyrolysis and conventional mass
spectrometry showed promising results for identifying specific
portions of the data obtained as unique to specific bacterial species.
Because pyrolysis is such a harsh fragmentation technique. the
assumption made at the time was that the spectra obtained from
pyrolysis would contain peaks representing many isomass ions. and
that unique components could be found by resolving these spectra.
This work set out to resolve and identify the differences between
spectra by taking complete tandem mass spectrometric “maps” of the
entire pyrolysis data space. The design. construction. and operation of
a slow-leak pyrolyzate containment inlet system is described. This

14

system was necessary to keep the pyrolyzate around long enough to
collect complete MS/MS maps of the sample. The results obtained
and reasons for abandonment of this system for other methods are

discussed.

A rapid. sensitive. and automated method for the direct
determination of phospholipid structures in bacteria is presented in
Chapter three. The pyrolysis work showed the need for identifying
biomarkers which could be used for microbial identification. Previous
work with phospholipids showed their usefulness for this purpose.
Full advantage was taken of the triple quadrupole mass spectrometer
operating modes to develop a general scheme that characterizes the
phospholipid content of crude lipid extracts using fast atom
bombardment ionization. Neutral loss and parent scans provide
independent diagnostic mass spectra for each of many specific
phospholipid classes. while daughter scans provide the composition
and positions of the fatty acid constituents on each phospholipid. An
automated MS/MS instrument performs an extensive phospholipid
screening on a single sample. A useful mass profile of the
phosphatidylethanolamine species present in a one-picogram sample
of mixed phospholipids (equivalent to ten E. coli cells) has been
obtained. The spectra are reproducible and proportional to
concentration over at least the 5-log range of cell concentration
studied. A rapid extraction procedure combined with the automated
instrument control program produces profiles of the phospholipid
classes along with fatty acid composition and position information on

selected phospholipid species in a few minutes from a single sample.

15

Although applied here to bacterial phospholipids. this method is a
general one for any sample containing phospholipids.

In Chapter four. the phospholipid work is extended through the
use of some interesting chemistry. The ability of the triple quadrupole
mass spectrometer to perform ion-molecule reactions was exploited
for differentiating phospholipid classes through reactions with ethyl
vinyl ether in the second quadrupole. The protonated molecules and
reaction product ions observed permit the differentiation of various
phospholipid classes. The pattern of addition reaction products
formed is shown to depend solely on the functionality of the lipid
polar head group and not on the fatty acyl constituents. Neutral gain
scans are performed which are specific for each phospholipid class.
Ion dissociation products are observed in the same scan as the ion
reaction products to provide data on the fatty acid composition and
position on the glycerophosphate core along with the phospholipid
class. Although this method is less sensitive than the method
described in Chapter 3 for most phospholipid classes. it can identify
phospholipids that do not readily lose their head group. and detect
phospholipids in mixtures containing species that give interfering

neutral losses.

Chapter five presents the application of the phospholipid
analysis technique described in Chapter 3 to the detection and
identification of microorganisms. This technique has been applied to
bacteria. fungi. and amoebae: a complete listing of the number and

types of samples analyzed is tabulated. Of particular interest is the

16

application in the detection and identification of urinary tract
infections and food-borne contaminants. Actual clinical and industrial
samples were analyzed with promising preliminary results. The need

for. and general structure of. a searchable database is discussed.

The thesis is concluded in Chapter six with the application of
other lipid biomarker analyses to the detection and identification of
microorganisms. Lipid biomarkers. other than phospholipids. that
have been investigated in this research include glycolipids and sterols.
Other potential lipid biomarkers are discussed along with their

amenability to mass spectrometry and possible applications.

17

References

10.

ll.
12.

13.

14.
15.

Davis. B. D.; Dulbecco. R.; Eisen. H. N.; Ginsberg. H. S.
Microbiology, 3rd ed.; Harper and Row: Hagerstown. 1980;
Chapter 15.

Aleixa. J. A. G.; Swaminathan. B.: Minnich. S. A. J. of Microbial
Methods. 1984. 2. 135-145.

Minnich. S. A.; Hartman. P. A.; Heimsch. R. C. Appl. Environ.
Microbiol.. 1982. 43. 877-883.

Anderson. J. M.; Hartman, P. A. Appl. Environ. Microbiol.. 1985.
49. 1124-1127.

Hill. W. E.; Payne. W. L. J. Assoc. 01}: Anal. Chem. 1984. 67. 801-
807.

Levy. R. L. Chrom. Rev.. 1966. 8. 48-89.

Odham. G.; Larsson, L.; Mardh. P. A. Gas Chromatography/Mass
Spectrometry Applications in Microbiology; Plenum Press: New
York. 1984; Chapter 10.

Oyama. V. 1.. Nature. 1963. 200. 1058- 1059.

Reiner. E., Nature. 1965. 206. 1272- 1274.

Reiner. E.; Hicks. J. J.: Ball, M. M.; Martin, W. J. Anal. Chem,
1972. 44. 1058-1061.

Irwin. W. J .; Slack, J. A. Analyst. 1978. 103. 673-704.

Bayer. F. L.: Morgan. S. L. The Analysis of Biopolymers by
Analytical Pyrolysis Gas Chromatography. Levy E. J .; Liebman. S..
Eds.: Dekker: New York. 1985; Chapter 6.

Fox. A.; Morgan, S. 1. Rapid Detection and Identification of
Microorganisms. Nelson. W. H.. Ed.; VCH: Deerfield Beach. 1985;
Chapter 5.

Gutteridge. C. S.; Norris. J. R. J. AppL Bacteriol. 1979. 47. 5-43.

Meuzelaar. H. L. C.; Kistemaker. P. G. Anal. Chem. 1973. 45. 587-
590.

16.

17.

18.

19.

20.

21.

22..

23.

24.

25.

26.

27.

28.

Wieten. G.; Meuzelaar. H. L. C.: Haverkamp. J. Gas
Chromatography / Mass Spectrometry Applications in
Microbiology. Odham. G.; Larsson. L.; Mardh. P. A., Eds.: Plenum
Press: New York. 1984; Chapter 10.

Meuzelaar. H. L. C.: Haverkamp. J .: Hileman. F. D. Pyrolysis Mass
Spectrometry of Recent and Fossil Biamaterials: Compendium and
Atlas: Elsevier Scientific: New York. 1982: Chapter 7.

Whitehouse. M. J.; Boon. J. J.; Bracewell. J. M.; Gutteridge. c. 3.;
Pidduck. A. J .: Puckey. D. J. J. Anal. Appl. Pyrolysis. 1985. 8. 515-
532.

Moss. C. W. Analytical Microbiology Methods. Fox, A.: Morgan. S.
L.; Larsson. L.; Odham. G.. Eds; Plenum Press: New York, 1990:
Chapter 4.

Tunlid. A.: Ringelberg. D.: Phelps. T. J.; Low, C.: White. D. C. J.
Microbial. Methods. 1989, 10. 139-153.

Odham. G.; Tunlid. A.: Westerdahl, G.; Larsson. L.; Guckert. J. B.;
White. D. C. J. Microbial. Methods. 1985. 3. 331-344.

Holzer. G.; Bourne. T. F.: Bertsch. W. J. Chromatography. 1989.
4689 181-190.

Dworzanski. J. P.: Berwald. L.; Meuzelaar. H. L. C. AppL Environ.
Microbiol.. 1990. 56. 1717-1724.

DeLuca. S.; Sarver. E. W.; Harrington. P. B.; Voorhees. K J. Anal.
Chem. 1990. 62. 1465-1472.

Heller. D. N.; Fenselau. C.: Cotter. R. J .; Demirev, P.: Olthoff. J. K.;
Honovich. J.; Uy. O. M.; Tanaka. T.: Kishimoto. Y. Biochem.
Biophys. Res. Comm. 1987. 142, 194-199. ' '*

Heller, D. N.; Cotter. R. J.; Fenselau. C.: Uy. O. M. Anal. Chem.
1987. 59. 2806-2809.

Platt. J. A.: Uy. O. M.; Heller. D. N.; Cotter. R. J.; Fenselau. C. Anal.
Chem. 1988. 60. 1415-1419.

Pramanik. B. N.; Zechman. J. M.; Das. P. R.: Bartner. P. L. Blamed.
Environ. Mass Spectrom. 1990. 19. 164-170.

CHAPTER 2

Use of Pyrolysis / Triple Quadrupole Mass
Spectrometry for Microbial Identification

Introduction

 

very species of microorganism contains unique chemical

components and/or combinations of components which

could, in principle. be distinguished from those of all other
species. The ability to measure the quantities required to make this
distinction is another story. The differences in chemical composition
that make microbes unique among each other are related to very
important and specialized functions required for the survival of each
particular microorganism. Because of this. these distinguishing
components are often so highly incorporated into the cell’s
biochemical machinery that their extraction from the cell is difficult.
or they present physical or chemical difficulties. such as high
molecular weights or complex structures that differ in only minor
details. Analytical instrumentation has only recently been able to
overcome some of these difficulties. Although recent advances in mass
spectrometry instrumentation have made increasingly less volatile
substances amenable to analysis. whole microbial cells. and many of
their components. still remain out of reach. Methods must be used
that either break the cells down into smaller fragments or extract the

components that are more amenable to mass spectrometric analysis.

Z)

The former method was the first method chosen to achieve the overall

goal of microbial detection and identification.

The key to producing useful fragments from whole cells is in
generating fragments that still retain some of the uniqueness and
characteristics of the original distinguishing molecules. Important to
a rapid and easy analysis is the production of these fragments with a
minimum of sample preparation and workup. Pyrolysis satisfies these
needs by producing fragmentation from whole cells within seconds.
and requiring no sample preparation other than loading the sample
onto the pyrolysis wire. As mentioned in Chapter 1. the pyrolyzate
obtained is expected to contain information characteristic of the
original sample. Working on the hypothesis that the pyrolyzate is
composed of many components that overlap in the conventional
spectrum. tandem mass spectrometry was employed in an attempt to
resolve these overlays. Identification could then be made by wading
through the large tandem mass spectrometry data space obtained and
identifying the unique components. The thrust of the research
presented in this chapter is an attempt to find those portions of the
MS/MS data space that could be used for microbial detection and
identification.

“Mum!

Pyrolysis
Three common methods exist for pyrolysis: laser pyrolysis. heated

filament or “pyroprobe”. and Curie-point pyrolysis. Of these three

21

techniques. Curie-point pyrolysis has been shown to be the most
reliable for reproducibility of results [1.2]. and is the method that was

chosen for this work.

Curie-point pyrolysis. first described by Giacobbo and Simon [3].
utilizes a high frequency coil to inductively heat a ferromagnetic wire
to its Curie-point: the temperature at which the wire becomes
paramagnetic and ceases to absorb rf energy. A schematic diagram of a
Curie-point pyrolyzer is shown in Figure 2.1. A small amount of
sample is placed directly on the ferromagnetic wire. and the wire is
placed inside a quartz reaction tube. This whole assembly is inserted
into a high-frequency rf coil and the wire is heated inductively. As the
temperature of the wire reaches its Curie-point. the wire changes
from ferromagnetic to paramagnetic and stops absorbing energy.
Below the Curie-point temperature. the wire becomes ferromagnetic
and again starts to heat. As long as the high-voltage rf field is applied.
the wire stays at its Curie-point temperature. The final temperature is
therefore determined by the wire's Curie-point temperature which, in
turn. is a function of the metal alloy composition of the ferromagnetic
wire. Various mixtures of iron. nickel. chromium. and colbalt are
generally used. A range of temperatures between 358° and 980° C are
available from different alloys. as shown in Table 2.1. These
temperatures are extremely reproducible. with temperature rise times
of 20-30 milliseconds from room temperature to the Curie-point.
Curie-point pyrolysis offers the steadiest pyrolysis conditions of all the
pyrolysis techniques. This has been found to be the most important
factor for obtaining good spectral reproducibility [4].

22

 

HIGH FREQUENCY COIL

   

QUARTZ REACTION TUBE

 

 

 

 

 

 

 

 

 

 

 

 

 

FERROMAGNETIC WIRE SAMPLE

 

Figure 2.1 Schematic diagram of a Curie-point pyrolyzer

 

 

Table 2.1 Curle-Polnt Temperature as a Function of Composlilon

WM
100 of Ni
48 of Fe:51 of Nl:1 of Cr
49 of Fe:51 of Ni
40 of Fez60 of Ni
30 of Fe:70 of NI
67 of Ni:33 of Co
100 of Fe
40 of Ni:60 of Co
50 of Fe:50 of Co

W
358
440
51 0
590
61 O
660
770
900
980

 

 

Pyrolysis/Mass Spectrometry

The chemical complexity of bacteria leads to pyrograms that have a
peak at almost every m/z value up to around m/z 500. A
pyrolysis/chemical ionization mass spectrum of Carynebacterium
nephridii is shown in Figure 2.2. Because of the large amount of data.
the original method of trying to visually identify pyrolysis mass spectra
of whole bacteria has been mostly abandoned. Present means of
interpretation now center around two approaches: the identification of
chemical components specific to a species [5]. and the mathematical
deconvolution of whole spectra to identify those areas where they
differ [6]. Unfortunately. both of these approaches suffer from serious
disadvantages.

In order to detect a component which is specific for a particular
organism. that component must be found. identified. and its pyrolysis
pattern characterized. While this procedure has the potential for
detection of microbes in mixtures. the process of finding and
identifying the pyrolysis characteristics of components specific to
each particular species is time consuming. Also. much sample clean-
up is required for the detection of these components. In the other
approach. identification is based on finding minor spectral differences
between species by the mathematical deconvolution of whole pyrolysis
mass spectra. Other investigators have put much effort into
developing programs toward this purpose: some of the specialized
tools developed include applications of factor and discriminant
analysis, linear regression. graphical rotation. variance diagrams. and

various mapping techniques. among others [7.8.9]. Although there is

%

 

81

 

 

 

 

 

 

 

 

100-
Ion
210
so“
8 61 225
257
8 98
D so-
c:
:3
.o
<
m 281
.5 40-
g 237 391
a)
(I
20-
OJ

 

 

100 150 200 250 ' 300 350 400 450 500
Mass/Charge

Figure 2.2 Pyrolysis/chemical ionization mass spectrum of Carynebacterium
nephridii.

very little. sample clean-up needed. these approaches depend on high-
ordered mathematics that do not work well for minor components of a

mixture or data sets that exceed 20 samples [10].

Pyrolysis/ Tandem Mass Spectrometry

Tandem mass spectrometry achieves a dramatic increase in available
information over that from conventional mass spectrometry by
providing not just the mass. but an entire mass spectrum for the ions
of each m/z value in the conventional mass spectrum. This added
dimension of data allows for the determination of trace components
and the resolution of overlaying contributors to a particular ion peak in
the primary spectrum. There have been isolated reports on the use of
tandem mass spectrometry to examine the structures of the ions from
conventional pyrolysis mass spectra [11.12.13]. but they have shown

limited success.

Only two reports exist on the use of the full capabilities of
pyrolysis/tandem mass. spectrometry for bacterial differentiation
[14.15]. These experiments were exploratory in nature. and made
limited use of daughter. parent. and neutral loss scan modes for the
differentiation of three bacterial species. An interesting scan mode
labeled “mixed daughter scans” was employed, in conjunction with
factor analysis and graphical rotation. in an effort to determine which
daughter ions would be useful ones for which to collect parent ion
scans [14]. Mixed daughter scans are total daughter ion scans in
which the first quadrupole is operated in an rf-only mode. collision gas
is added to the second quadrupole, and the third quadrupole is

27

scanned [15]. The resulting spectrum contains all the daughters of all
the parents in the pyrolyzate. The use of parent scans for 3 different
daughter ions showed limited success. By far the biggest achievement
of these experiments was the demonstration of the usefulness of
tandem mass spectrometry for reduction of chemical noise and
identification of small portions of characteristic data from large

uncharacteristic data sets.

These experiments raise the possibility of finding unique parent-
daughter ion-pairs and neutral losses. in the complete MS/MS
pyrolyzate data space. that would be useful for microbial identification.
To find these unique transitions. methodology for collecting and
analyzing complete MS/MS maps needed to be developed. This was
the goal of the research presented in this chapter.

Experimental

All experiments were performed using a Finnigan (San Jose. CA) T59-
70 triple stage quadrupole instrument equipped with a specially -
designed pyrolysis inlet system. The pyrolysis inlet is described in
detail later in this chapter. Pyrolysis was accomplished using a Varian
Aerograph/ Fischer (Walnut Creek. CA) model 9425 Curie-point
pyrolyzer power supply (1.5 kW. 1.1 MHz) and a Curie-point pyrolysis
unit [16] modified to fit the pyrolysis inlet system. The pyrolysis unit
was obtained from Dr. Kent Voorhees at the Colorado School of Mines.
Curie-point wires rated at 510° C were used for analysis. Spectra were

acquired and processed using the Finnigan TSQ-70 data system and

software.

All bacteria were obtained from Dr. Jim 'I‘iedje at Michigan State
University. Cultures were grown on trypticase soy agar plates at 37° C
for twenty four hours prior to analysis. For analysis. the bacteria were

scraped off the plates and coated onto the Curie-point wires.

The samples were pyrolyzed at 510° C for 4 seconds. The
pyrolysis inlet system was maintained at 200° C. Positive ion chemical
ionization spectra were collected using methane as the reagent gas.
Daughter spectra were obtained using argon as a collision gas. The
collision gas pressure was 0.5 mtorr and the collision energy was

between 30 and 40 eV.

Results and Discussion

 

Pyrolysis Inlet System

In order to find those combinations of the few parent. daughter. and
neutral loss values that are characteristic of an organism. a daughter
spectrum of every ion in the conventional spectrum needed to be
collected. This required the pyrolyzate to be presented to the ion
source long enough for the instrument to perform all the necessary
daughter scans. Because of this time requirement. a slow-leak.
pyrolyzate-containment inlet system was designed and interfaced with
the TSQ-70 to provide a constant source of pyrolyzate to the mass

spectrometer.

A basic schematic of the final design is shown in Figure 2.3.
This design consists of the pyrolysis unit connected to a quartz
expansion chamber via a viton O-ring and clamp (not shown). The
expansion chamber fits through the GC inlet Cajun fitting on the TSQ-
70 and mates with the ion source via a vespel ferrule. The expansion
chamber is double-walled: the inner wall (actual chamber) is wrapped
with NiChrome wire. and the outer wall provides the seal with the
Cajun fitting. The NiChrome wire allows the chamber to be heated to
a controlled temperature by connection to a Variac. The end of the
chamber connecting to the mass spectrometer source has a small leak

hole that allows pyrolyzate to slowly leak into the source.

The TSQ-70 needs about ‘100 seconds to collect a complete
MS/ MS map of the ions obtained from the pyrolysis of whole bacteria.
Therefore, the leak hole needed to be adjusted to optimize the flow of
pyrolyzate out of the chamber. If the pyrolyzate is allowed to stay
inside the chamber at higher pressures and longer times. secondary
reactions will occur and the spectra will change as a function of the
storage time. Also. secondary reactions will cause the spectra to be
irreproducible between samples. A compromise was sought: keep the
pyrolyzate around only as long as needed for the instrument to collect
data. The pyrolyzate profiles from the different versions of the
chamber are shown in Figure 2.4. Reconstructed ion chromatograms
were obtained for the total ion current produced after pyrolysis. The
profile in Figure 2.4a shows a leak-hole that is too large - the
pyrolyzate does not stay around long enough to permit all needed data

30

.889? no.5 .EoEESCOO Baum—aha .xmflspai B: CO canEonam RN 9:6:-

 

 

 

ooSom co_
.aLmEoctmam mmmz

 

 

 

83 95>
N .23: 869592
.38ch

 

cemcmaxm Ego

 

 

 

 

 

 

 

 

_I 823m 6.5 _ _ ac: m_m>_9>a II—

 

31

 

b)

Relative Abundance

Ioo-I

 

 

80

 

' ""b""150'"'IE'"'260“"80""£D""350"”460"'T4§o""sbo

10.?
I.‘
6"
40‘

 

 

 

Figure 2.4

Pyrolyzate profiles from different versions of the
pyrolyzate inlet system: a) leak hole too large, b) leak
hole too small. c) leak hole correct size.

32

to be collected. After changing the hole size. the hole became too
small. as shown by the results in Figure 2.4b. In this configuration.
secondary reactions were occurring. and a full minute was needed
before spectra acquisition could even be started. After a number of
tries. the correct configuration was obtained. This is shown in Figure
2.4c. Fifty percent of the pyrolyzate has been evacuated from the
chamber by the time the instrument has finished collecting data. and
almost none of the pyrolyzate remains after 4 minutes.

Collection of Complete Pyralysis/MS/MS Data Fields

Most of the pyrolysis/MS work done so far has used electron impact
ionization. Only one use of chemical ionization has been reported [15].
Chemical ionization was chosen in this study for two reasons: 1) to
retain as much of the high-mass information as possible. and 2) to
increase the sensitivity of the technique. Pyrolysis / CI mass spectra for
three different bacteria are shown in Figure 2.5. These three bacteria
provide a good test group: Escherichia call and Pseudamanas
aeruginasa are similar organisms in that they are both gram-negative
rods. while Carynebacterium nephridii is a gram-positive club-shaped
bacilli which. taxonomically. is significantly different from the other
two. These spectra are quite similar to each other; they contain a
peak at almost every m/z value. and differ only in the relative
intensities of these peaks. As expected. the spectra from E. call and
P. aeruginasa were similar. while the C. nephridii spectrum showed

the most contrasts among the three spectra.

 

Escherichia coli
l 1
l 00‘

 

 

 

100‘

Relative Abundance

 

 

 

IN 160 2W 250 300 350 400 450 500

Carynebacterium nephridii
‘ ”I U
W

 

 

 

1” 150 2W 250 no can an ‘50 5M

Mass/Charge

 

Figure 2.5 Pyrolysis/ chemical ionization mass spectra for
three different bacteria.

34

Next. daughter spectra were taken of all the ions present in
these. conventional spectra. In contrast to results obtained by
Voorhees et al. [14]. good fragmentation was observed from most of
the parent ions. and rich daughter spectra were obtained. The
complete pyrolysis/ MS/ MS data space obtained from E. coil is shown
in Figure 2.6. The parent ions occur along the front axis and the
daughter ions occur along the trailing axis. Examples of typical
daughter spectra of m/z 196 obtained from the three bacteria are
shown in Figure 2.7. Taking the noise level into account. these three
spectra contain the same daughter ions. A visual search of all the
daughter spectra showed that very few differences existed between
these three bacteria. Some of the differences that were found in the
daughter spectra are shown in Figures 2.8 and 2.9. The ions that
formed peaks at m/z 209 and m'/z 225 appear to be different for E.
call and C. nephridii Daughter spectra of these ions contain unique
daughter ion masses. along with having some daughter ion masses in
common. This observation supports the original hypothesis that the
pyrolyzate contains isomass fragments that would be overlayed in the
conventional mass spectrum. Unfortunately. very few of these
differences were found. Those differences that were found existed

only in minor components of the pyrolyzate.

Conclusions

 

The original hypothesis that was made appears to be sound:
differences due to overlapping components do exist in the

conventional spectrum and can be resolved using tandem mass

35

 

100‘

 

 

 

 

 

 

Q)
o
C
CD
‘0
c
3
g 50-
a)
.2.
3
<1)
0:
o —
50
Figure 2.8 Complete pyrolysis / MS/ MS data space obtained from E. cali.

 

100'
80“
60"
40"

20"

 

100‘

80“

60‘

40“

20“

Relative Abundance

 

o-I

100‘
80“
60“
40‘

20“

 

 

 

0d

57
55
20 40 60
57
55 L
20 40 60
55 57
20 40 60

70

 

 

 

 

 

 

34 P. aerugmasa 196
81
9°” 122
, «14*-
80 100 120 140 160 180 200
7° 84 . ..
C. nephridii ‘96
81
x 99108 ‘22 152
I *4"
80 100 120 140 180 180 200
196
E. coli
70
84
“so 110
122
80 100 120 140 160 180 200
Mass/Charge

 

Figure 2.7 Daughter spectra of m/z 196 obtained from the pyrolyzate of
three difi'erent bacteria.

 

E. 00” 209

 

 

 

 

 

 

 

Ioo-
_ 7 84
° 131
- 55
0. “null i..”-- A A-“ - .n “A .1
a, , . . , . . .
2 60 so 100 120 140 160 180 zoo
(11
E Mass/Charge
<
9
“a
L5
m s as
a: .
100 _ C nephridii . 193
q x
209
0 " ML
so so 100 120 140 160 180 zoo
Mass/Charge

 

Figure 2.8 Daughter spectra of m/z 209 obtained from the pyrolyzate of three
different bacteria.

 

E. cali 225

 

 

 

 

 

 

 

 

 

Ioo-

_ 84

[as
‘ 180
197

- 56 71 140 J 207
8 l .JJJhlnl IILLLI I- 11111. .1 .LL L44 L J.I.Ill . I 1111
c 0 1 T I ‘ ' ' ' r 7 fl r r v v v
.5 so 100 150 zoo
13
g Mass/Charge
<
9
g 0 O.
can) 100. C. nephrIdII 20,

‘ 73 193 225

o i. AL 44%.. .14 n-4,; . “Ann 1.. -M-.nll A. I; Mill
T I I ' I r
so 100 150 zoo
Mass/Charge

 

Figure 2.9 Daughter spectra of m/z 225 obtained from the pyrolyzate of three
different bacteria.

spectrometry. However. only a few differences were found among the
limited number of bacteria used for this study. These differences will
probably not provide enough information to be useful for microbial
identification on a larger scale. The probability of absolute uniqueness
in the tandem mass spectrometry data space of every organism in a
large grouping seems low. Due to this reason. pyrolysis was abandoned
as an identification technique. Instead. methods for selective analysis
of known classes of biomarkers were investigated for their usefulness

in microbial detection and identification.

References

 

10.

ll.

12.

13.

14.

Meuzelaar. H. L. C.: Kistemaker, P. G.; Eshuis. W.; Engel. H. W. B.
Rapid Methods and Automation in Microbiology. Learned
Information: Oxford. 1976; pp 225-230.

Meuzelaar. H. L. C.: Haverkamp. J .: Hileman. F. D. Pyrolysis Mass
Spectrometry of Recent and Fossil Biamaterials. Elsevier
Scientific: Amsterdam. 1982: Chapter 4.

Simon. W.; Giacobbo. H. Chem. Ing. Techn.. 1965. 37. 709-714.

Whitehouse. M. J .: Boon. J. J .: Bracewell. J. M.; Gutteridge. C. S.;
Pidduck. A. J .: Puckey. D. J. J. Anal. Appl. Pyrolysis. 1985. 8. 515-
532.

Fox. A.: Morgan. S. L. Rapid Detection and Identification of
Microorganisms. Nelson. W. H.. Ed.: VCH Publishers: Deerfield
Beach. 1985: Chapter 5.

Windig. W.; Meuzelaar, H. L. C. AnaL Chem. 1984. 56. 2297-2303.
Meuzelaar. H. L. C.: Haverkamp. J .: I-Iileman, F. D. Pyrolysis Mass
Spectrometry of Recent and Fossil Biamaterials. Elsevier
Scientific: Amsterdam. 1982: Chapter 6.

Windig. W.; Haverkamp. J.; Kistemaker. P. G. Anal. Chem. 1983.
55. 81-88.

Windig. W.; McClennen, W. H.; Stolk. H.; Meuzelaar. H. L. C.
Optical Engineering. 1986. 25. 117-122.

Voorhees. K. J.; Durfee. S. L.; Updegraff, D. M. J. Microbial.
Methods. 1988. 8. 315-325.

Levsen. H.; Schulten. H. R. Blamed. Mass Spec.. 1976. 3. 137-
139.

Louter. G. J.; Stalmeier. P. F. M.; Boerboom. A. J. H.; Haverkamp.
J.; Kistemaker, J. Z. Naturfarsch. 1980. 35. 6-11.

Haverkamp. J.; Wieten. G.; Boerboom. A. J. H.; Dallinga. J. W.;
Nibbering. N. M. M. Analytical Pyrolysis, Voorhees. K.. Ed.:
Butterworths: London. 1984: Chapter 10.

Voorhees. K. J .: Durfee. S. L.; Holtzclaw. J. R.; Enke. C. G.; Bauer.
M. R. J. Anal. Appl. Pyrolysis. 1988. 14. 7-15.

41

15.

16.

Bauer. M. R. Ph.D. Dissertation. Michigan State University. East
Lansing. MI. 1986.

Leeuw. J. W. de: Maters. W. L.; Meent. D. v.d.: Boon. J. J. Anal.
Chem.. 1977. 49. 1881-1884.

CHAPTER 3

Direct Determination of Phospholipid
Structures in Bacteria by Fast Atom
Bombardment / Triple Quadrupole Mass
Spectrometry

Introduction

 

' yrolysis proved to be too harsh a technique for accessing
the unique components of microorganisms. Classes of
molecules useful as biomarkers were fragmented to such an

extent as to lose their identity and uniqueness. The results of the
pyrolysis experiments underlined the importance of identifying and
selectively obtaining specific information on whole biomarkers. The
need for selective analysis of known classes of biomarkers led to an

exploration of microbial lipids.

The diversity of lipids in microorganisms signifies a diversity of
functions. Lipids. either directly or indirectly. play a significant role
in the specialized functions each species maintains to survive in its
particular environmental habitat. Lipids carry out many
responsibilities including storage. membrane structure and function.
photosynthetic processes. and most energy-generating processes.
The types and distribution of lipids contained in microorganisms can

be very species specific [1-3].

Not all of the cellular lipids are easily accessed. Phospholipids
are responsible for the structure of the cell membrane, and thus.
ultimately its function. In addition. these membrane phospholipids
are easily accessed through simple extractions and provide a useful
and abundant biomarker for microbial detection and identification.
Our interest in phospholipids stems from these reasons. This chapter
describes a general method for rapidly characterizing the
phospholipid content and structures in crude lipid extracts. with the
ultimate goal toward application of this information to microbial

detection and identification.

Bee und

 

 

The resurgence of interest in the mass spectrometry of phospholipids
is due primarily to the development of fast atom bombardment (FAB)
ionization and tandem mass spectrometry (MS/MS). Fast atom
bombardment ionization allows the nonvolatile phospholipids to be
analyzed without prior derivatization. and MS/ MS allows phospholipid
mixtures to be fully characterized without prior separation.
Phospholipids are found in abundance in the cell membranes of living
organisms. They have been used as biomarkers for bacteria [4-6] and
algae [7]. for detection of the generation of platelet activating factor [8-
10]. and for the determination of general membrane function in living

matter.

The phospholipids in which this research is concerned are the
glycerophospholipids (Structure 3.1). Glycerophospholipids consist of
four primary functional groups: a glycerol-3-phosphate core on which
two fatty acids (R. R') have been esterified to the two free hydroxyl
groups in the sn-1 and sn-2 positions. and a second alcohol (Y) is
esterified to the phosphate group in the sn-3 position (except for
phosphatidic acid. which contains a protonated phosphate group).
This head group (Y) is the functional group which defines the specific
class to which the phospholipid belongs. while the fatty acids
distinguish the individual phospholipid molecular species within each
class. Examples of different phospholipid classes are shown in Figure

3.1.

For a method to be useful in obtaining a profile of bacterial
phospholipids in natural samples. it must have high sensitivity as well
as high selectivity. While most of the phospholipid information is
contained in the conventional FAB/MS spectrum [4.6]. the data are
difficult to interpret: signals from lipid species at low concentrations
are often buried in the high chemical background associated with
FAB/MS. and FAB/MS does not work well for mixtures of
phospholipids. Fast atom bombardment / tandem mass spectrometric
analysis of daughter ions produced by collisionally induced dissociation
(CID) of molecular ion precursors has proven useful for phospholipid
structure analysis [11-17]. While this method is useful for obtaining
information on the content. structures. and relative positions of the
fatty acids present on individual phospholipids. screening a mixture of

phospholipids for the classes present would be a complex procedure

45

 

 

Structure 3.1 Basic glycerophospholipid structure.

 

LIPID
Phosphatidylethanolamine
( E)

Ph h tid I I l
osp a (£89!er

Phosphatidylcholine
(PC)

Phosphatidylserine
- (PS)

Phosphatic: lll)nosliol

HEAD GROUP (Y)

—CH'2—CH;—NH2

—CH—CH —CH
2 | I 2
0H 0H

4.
—CH2—CH2—N (CH3 )3

—CH;-CH—COO H

 

Figure 3.1 Examples of the head groups from five phospholipid

classes.

47

requiring a mathematical analysis of the data set composed of the
daughter spectrum from the ion current at every mass. Constant
neutral loss scanning for polar head functional groups has been shown
to be very useful for detection and differentiation of phospholipid

classes in complex matrices and mixtures [13.18].

When phospholipid ions undergo low energy CID. only a few
different fragment ion masses are formed. However. the fragment
ions produced are the result of cleavages at specific points which are
common to all classes of phospholipids. These cleavages occur around
the phospholipid functional groups and. thus. provide significant
structural information about the parent ion. The low energy CID
fragmentation of phospholipids for both positive and negative ions is
shown in Figure 3.2. In the positive ion mode. the major reaction
occurring is cleavage of the phosphate/ glycerol bond resulting in the
loss of the polar head group as a neutral. while the rest of the ion
retains the charge. In the negative ion made. one or the other fatty
acid is cleaved and retains the charge while the rest of the ion is lost
as a neutral. Examples of some of the possible positive ion neutral
losses and fatty acid anions are given in Table 3.1. This research has
taken advantage of the fact that phospholipids always cleave around
their functional groups to develop a general scheme for the rapid
characterization of phospholipid classes present in bacteria. including
information on the fatty acids present on the species within each
class. This scheme is a sequence of positive ion neutral loss and

parent scans. and negative ion daughter scans.

 

v.
a 4
H
6 I4 ounuv cup—IO
_I I. o1.... . o ._.
w _ 2
I R IR 0 90 H2 H H
_ _ _ H CICIC
one one OHPIO
_ _ _ v
._. n_v ._. _
2 I 0
H2 H H 4" _ H
OHJV + n_v O
O 2 _ _2
O H H H
CICIC
a a
_.II V.
W I — I
( R R n_u v_. _ R R
_ _ I H _ _
H
O O O I OHPIO + O O
.2 _ _.. _ .2 _ 2
H H H m H H H
CICIC CICIC

 

Figure 3.2 Low-energy CID fragmentation of phospholipids for positive
and negative ions.

 

Table 3.1 Characteristic Positive Ion Neutral Losses
for Various Head Groups and Fatty Acid
Anions in the Daughter Ion Spectra

W main/sums.”

PE of 141 u C14:0 at m/z 28

PG of 172 u 014:1 at m/z 226

ps of 135 u 015:0 at m/z 241

p| of 260 u 015:1 at m/z 239

PC of 183 u 016:0 at m/z 255
PA of 98 u 016:1 at m/z 253
LPG of 300 u ' C17:0 at m/z 269
PMME of 155 u 018:0 at m/z 283
PDME of 169 u C18:1 at m/z 281

aPE=phosphatid lethanolamine. PG=phosphatidylg| cerol
PMME- - phosp atidyelmonomethylethanolamine. ME =
phosphatldyldimethye thanolamine PS phosphatidyl-
serine, Pl=phosphatidylinositol, PC= phosphatidylcholine,
PA=phosphatidic acid, LPG: -lysylphosphatidylglycerol.

b CI 5 :0 Is a 15-carbon fatty acid with no unsaturation, 015:1
is a 15 -carbon fatty acid with one point of unsaturation,
etc.

 

EXPERMNTAL

 

All experiments were performed using a Finnigan (San Jose. CA) TSQ-
70B triple stage quadrupole instrument equipped with a standard
Finnigan FAB source and a JEOL (Boston. MA) MS-009 charge-transfer
FAB gun and power supply that has been modified to fit and operate on
the TSQ-70B. Spectra were acquired and processed using the
Finnigan TSQ-70B data system and software.

All bacteria, except E. cali were obtained from researchers in
the National Science Foundation's Center for Microbial Ecologt at
Michigan State University. Cultures were grown on trypticase soy agar
plates at 37° C for twenty four hours prior to harvesting. E. call was
obtained lypholyzed from Sigma Chemical Co. (St. Louis. MO).

Suspensions of E. cali were made by weighing out a known
quantity of lypholyzed cells and suspending them in enough
distilled/deionized water (DDI) to make a 1x109 cells/m1

concentration using the conversion factor 5.9x1012

cells/gram dry
weight [19]. Dilutions of this suspension were made to obtain

solutions of different cell concentrations.

Crude lipid extracts were obtained from the bacteria by a
modified Bligh-Dyer procedure [20]. One ml of a cell suspension. or a
transfer loop full of bacteria. was placed in a clean. sterile 25-ml
sample vial to which 15 ml of a 2:1 methanol:chloroform mixture was
added. After the lipids were extracted. enough DDI water was added

51

to separate the chloroform from the MeOH/HZO. The chloroform

layer. containing the lipids. was drawn off for analysis. For those
experiments requiring a known cell concentration on the probe tip.
the chloroform was blown off. using a stream of nitrogen. until just
enough solution remained to be placed on the probe tip for analysis
(about 5 ill).

For the analysis. 3-5 (11 of the chloroform solution was placed on
the probe tip and dissolved in a drop of nitrobenzyl alcohol. Xenon
was used as the FAB gas and the FAB gun was operated with a filament

current of 10 mA and a xenon beam energy of 8 keV.

Neutral loss and parent ion spectra were obtained in the positive
ion mode with no CID gas added ‘to the center quadrupole other than
the residual gas remaining in the analyzer manifold. The
fragmentation observed may be predominantly the result of metastable
decomposition. Daughter spectra were obtained in the negative ion
mode using argon as a collision gas. The collision gas pressure was 0.5

mtorr and the collision energy was 30 eV.

RESULTS AND DISCUSSIONS

Phospholipid data are obtained by this technique on two levels. The
first level is referred to as the (class) mass profiles which are the
separate mass spectra for each phospholipid class. From these data.
information is obtained regarding the phospholipid classes present.
the masses of individual phospholipid species within each class. and

52

the relative intensities of the species within each class. The second
data level consists of the daughter spectra of individual phospholipid
species and is referred to as (class)(fatty acid) formula data. These
data can be interpreted to provide empirical formula information on
the fatty acids and structural information on each phospholipid
species. The data levels and their interrelation are shown as a data
space diagram in Figure 3.3. In this space. each phospholipid class
forms its own data plane with the mass profile of its species contained
along one dimension of the plane and the fatty acid masses for each of
its species along the other dimension. As discussed later. low-energy
CID does not yet allow the complete structure of the fatty acid groups
to be obtained. The method described here collects (class)(mass)
profiles and (class)(fatty acid) formula data. Research is underway

toward obtaining complete fatty acid structural information.

General Profiling Scheme

The first step in the general scheme is to obtain the positive ion
neutral loss scans for loss of the masses of the different phospholipid
polar head groups. The instrument is programmed to scan a series of
possible polar head group neutral losses. thereby obtaining (class)
mass profiles of all the phospholipid classes scanned. These profiles
approzdmate the relative abundances of all species within each
phoSpholipid class. The resulting spectra are free of any non-
phospholipid peaks. and many peaks of interest which might have
been hidden previously in the chemical background of the
conventional mass spectrum show up clearly. An example of the use of

neutral loss scans for collecting (class) mass profiles of two of the

53

 

 

 

 

 

 

 

 

 

 

 

 

PHOSPHOLIPID GLASSES

 

 

 

DEDEDI

 

Figure 3.3 The FAB/MS/MS data space obtained by this technique. The
masses of the phospholipids in each class are obtained by
neutral loss or parent ion scans specific for each class. The
daughter spectrum of each phospholipid species provides the
masses and positions of the fatty acids.

phospholipid classes contained in Proteus vulgaris is shown in Figure
3.4. Note that the peaks due to phosphatidylglycerol (neutral loss
172) are completely buried in the chemical background of the
conventional spectrum. This added detectability is a direct result of
the additional selectivity afforded through tandem mass analyzers.

The relative intensities of the phospholipid species obtained in
each (class) mass profile reflects the relative intensities of the species
in the conventional mass spectrum. Apparent differences in
intensities between the conventional spectrum and the neutral loss
spectra are seen when more than one phospholipid class contributes
to the abundance of a particular peak in the conventional mass
spectrum. These overlaying peaks are resolved by the neutral loss
spectra. The only exception that was found to using neutral loss scans
is in the characterization of phosphatidylcholine. Phosphatidylcholine
fragments so that the head group usually retains the charge. While the
neutral loss scan does work for phosphatidylcholine, greater
sensitivity was obtained through the use of a parent scan for the head
group ion. A parent scan for m/z 184 produces the (class) mass

profile for phosphatidylcholine.

The two remaining pieces of information needed for a complete
phospholipid profile are the composition. structure. and relative
positions of the fatty acids on the glycerophosphate core. Much of this
information can be directly obtained from the negative ion daughter
spectra of the individual phospholipid species detected in the series
of (class) mass profiles obtained from the neutral loss scans previously

55

 

' o .1 70

80-1

80-1

 

 

 

b)

Relative Abundance

 

 

 

1001

8 0"
7.

00"

 

 

 

 

Mass / Charge

 

Figure 3.4 (1) Conventional mass spectrum of a Proteus vulgaris
extract. 5) Neutral loss spectrum of 141 11: specific
for PE. c) Neutral loss spectrum of 172 u: specific for
PG.

performed. As shown in Figure 3.2 for [M-Hl' ion dissociation. the two
major fragments present in a negative ion daughter spectrum are the
peaks due to the two fatty acids contained on the phospholipid ion.
The mass/charge value of the fragment ion corresponds to the
molecular weight of the free fatty acid. less one hydrogen. as shown in
Table 3.1. In this manner. the number of carbons contained in the
fatty acid may be determined. along with the degree of unsaturation.
It is important to note that from these spectra. only the empirical
formulae of the fatty acids may be determined with certainty. At the
present time. low-energy daughter spectra cannot determine the
location of double bonds on a fatty acid, differentiate between a point
of unsaturation and a cyclic (cyclopropyl) structure. or differentiate
normal from branched fatty acids. Research using low-energy

methods to solve these problems has begun in this laboratory.

The negative ion daughter spectra also can provide information
on the relative positions of the two fatty acids on the phospholipid.
Zirrolli. et al. [9] have postulated preferential formation of the
carboxylate anion from the fatty acid at the sn-2 position (closest to
the head group) over that of the fatty acid at the sn-1 position. When
this is true, the daughter ion peak with the greater intensity is due to
the fatty acid from the sn-2 position. While this general rule
apparently does not apply to phospholipids containing highly
unsaturated fatty acids or when large differences in the chain lengths
of the two fatty acids occur [21]. it does hold for phospholipids
containing fatty acids with fewer than three points of unsaturation and
with chain lengths differing less than 10 carbons from each other.

57

Because fatty acids which fit the exceptions to this rule are extremely
rare in bacterial phospholipids. this rule is being used to obtain
positional information about the fatty acids present in our samples.
The exceptional fatty acids are easily diagnosed and the data treated
accordingly. Figure 3.5 is an example of the use of negative daughter
ions for Salmonella abaetetuba. The peak at m/z 733 represents the
[M-Hl' ion of a phosphatidylglycerol species. The two daughter ion
peaks obtained at m/z 255 and m/z 267 correspond to a C16:0 fatty
acid and a Cl7:cyc fatty acid. respectively. As shown by the difference
in peak intensities. the Cl7:cyc fatty acid occurs at the sn-1 position
on the phospholipid.

An automated instrument control procedure has been written
that performs the necessary neutral loss or parent scans. in the
positive ion mode. for all 9 phospholipid classes in which this
research is interested. switches the instrument into the negative
daughter ion mode. and collects the individual daughter spectra from
the 10 major peaks present in each of the phospholipid classes
previously detected. The resulting data file contains up to the 9
possible (class) mass profiles and the daughter spectra of up to 90
phospholipid species. The automated procedure can collect this
information from a sample in under ten minutes. A small portion of
the complete data set used to obtain a phospholipid profile for E. call
is shown in Figure 3.6.

 

[M-Hl'

 

 

 

100 733
80-1
g c
16:0
60-
2 255
2’
O
_>
g 40-
35
CI
017:cyc
20- 267
0"
200 300 400 500 600 700

Mass I Charge

 

Figure 3.8 Negative ion daughter spectrum of a phosphatidylglycerol species
at m/z 733 showing the two fatty acid ions and their relative

intensities.

C16:0 is a 16-carbon fatty acid with no unsaturation: Cl7:cyc is a
17-carbon propyl fatty acid.

 

 

 

 

 

 

13
ALL PI CLASS ALL PO CLASS
MOLECULAR NOLBCULAN

IONS IN SAMPLE IONS IN SM!

11L.

HIIIIII

 

 

 

o—th!

 

 

 

 

 

. W on
“SW”
616:1
c1010
cut.
so
A _l ,
" do '31» ' is ab do "is is T a so as no

 

Figure 3.8 An example of the use of neutral loss and daughter spectra for
' obtaining a partial profile of E. call. The neutral loss scans obtain
the (class) mass profiles while the daughter scans obtain the
(class)(fatty acid) formula data. As indicated. seven other neutral
loss scans are normally performed for a "complete" profile.

Daughter spectra are obtained for the major peaks in each (class)
mass scan.

Sensitivity

The increase in selectivity due to the use of tandem mass
spectrometry gives outstanding detection limits which are
unattainable for these samples in conventional mass spectrometry.
The top spectrum in Figure 3.7 is a neutral loss scan for
phosphatidylethanolamine (neutral loss of 141 u) of one picogram of a
mixture of phospholipids which are mostly phosphatidylethanolamine
species extracted from E. call. This is roughly the phospholipid
content of ten E. call cells. For comparison. the bottom spectrum
shows the same scan of 1 microgram of the same mixture: equivalent
to 107 cells. All of the phosphatidylethanolamine species present in
the l-microgram mixture have been detected in the one picogram
spectrum. This extremely low sample size requirement allows for the
analysis of phospholipids at bacterial levels found in natural samples

without prior cultivation of the organisms.

Precision

For this technique to be useful in bacterial phospholipid profiling. the
spectra must be reproducible though varying concentrations of
bacterial cells extracted. If the spectra changed with varying cell
concentrations. the information regarding the classes. individual
species, and fatty acid content of the phospholipids would be retained.
but all information pertaining to the relative abundances of the
individual species within each class would be dependent on the total
quantity of phospholipid present. The degree of spectral variance
obtained on our instrument for various cell concentrations is shown in

Figure 3.8. Using neutral loss scans. the percent of the total

61

 

704

 

 

 

 

 

 

 

 

 

a) 100‘
go.
50-.
40-1
690
732
20‘ 718
g 664
U W
c 0-
B 600 650 700 750 800
< .
a)
b) 1; 100‘ 70‘
m
76
a:
so.
so.
40-1 soot
2°“ 718
661
732
O-Wv—y—J‘vk-JILJL‘LMW,“*1 : r , 1
600 650 700 750 800
Mass/ Charge
Figure 3.7 a) PE mass profile of l picogram PE extracted from E. cali:

equivalent to 101 cells. b) PE mass frame of 1 microgram PE
extracted from E. call: equivalent to 10 cells.

62

 

 

 

 

 

 

 

33 5°“ LlElD_|.th§
2‘ 40. R 664
5 A 692
I- . 704
(‘5 30‘ A 718
°\° I 732
20-
1 25d: :3:
10- A
w W
o I j T . I I j
o 1 z 3 4 s 6
LOG NUMBER OF CELLS

 

Figure 3.8 Percent of total PE as a function of the log number of cells for the
major species present in E. cali.

phosphatidylethanolamine ion current was plotted as a function of the
log number of cells extracted for the major species present in E. cali.
High precision was obtained. with the relative intensities of the
masses within the spectra varying no more than 5 percent over at

least a 5-log range of bacterial cell concentration.

Extraction Process

Related to the problem of spectral precision as a function of sample
size is the problem of phospholipid recovery as a function of cells
extracted. The recovery of this method was checked by analyzing cell
extracts of E. call at various concentrations. The cells were extracted
for two hours before sampling. Neutral loss scans were performed for
phosphatidylethanolamine and phosphatidylglycerol. the two most
abundant phospholipids in E. cali: the total ion current obtained was
used as a measure of phospholipid abundance. Table 3.2 shows the log
of the total phospholipid ion intensity obtained for several values of the
log number of cells extracted. These data follow a linear fit with a
correlation coefficient of 0.997. showing that the amount of
phospholipid recovered in an extraction. relative to the number of

cells extracted. remains constant.

With an automated instrument control procedure that can
produce a phospholipid profile in under ten minutes. the time
required for extraction can become the limiting step with regards to
total analysis time. Typically. the samples are extracted for 24 hours
[22]. with reports of more rapid extractions of 2 hours [23]. In an

effort to decrease the total analysis time for a sample. the dependence

64

 

Table 3.2 Phospholipid Recovery for Various Values
of Log Number of E. cali Cells Extracted

Log Number of Cells Log Lipid Ion Current
5.0 5.58
6.0 6.66
7.0 7.70
8.0 ‘ 8.54

Correlation Coefficient = 0.997

 

~ ‘I

of phospholipid recovery on the total extraction time was determined.
Neutral loss scans were performed on a series of five E. cali samples
which had been extracted for 1 minute. 10 minutes, 1 hour. 2 hours,
and 24 hours. respectively. The total phospholipid ion intensities
were used as a measure of the phospholipid recovery. Table 3.3 shows
the phospholipid recovery. relative to the recovery obtained from the
24-hour extraction. for various extraction times. These data suggest
that little recovery advantage is gained after the first 10 minutes of
extraction. In fact. a 1-minute extracan yielded greater than 80%
phospholipid recovery. These rapid extractions significantly shorten
the total analysis time and open up the possibility of performing on-
line extractions coupled with a technique such as continuous-flow FAB.

CONCLUSIONS

 

The use of neutral loss. parent. and daughter scans in combination
with rapid extractions provides a quick and sensitive method for
profiling the phospholipid content and distribution in bacteria. Since
bacteria maintain a characteristic and rather constant membrane
phospholipid content. this method has a great potential for detecting
and identifying bacteria based on phospholipid profiling. Research in
this area is described in detail in Chapter 5.

This technique is not restricted to use in bacterial phospholipid
profiling: it can be used as a general method for any sample containing
phospholipids. We are using this same method to characterize the

phospholipid content of plants. fungi and amoebas. This research is

$

 

Table 3.3 Phospholipid Recovery from E. call
for Various Extraction Times

Average Relative
Extraction Time Number of Phos ho ipid Standard
(minutes) Samples IC* Deviation

1 3 3,867,690 23.4%

10 3 4,544,030 23.3%

60 ’ 3 4,487,700 28.2%

120 3 4,765,410 28.9%

1440 3 4,791,830 23.9%

* RIC = Reconstructed Ion Current

 

also described in Chapter 5. However. the caveat about fatty acid
position data must be taken into account when working with samples
which might contain fatty acids included in the exceptions mentioned
earlier in this chapter. Many plant species contain highly unsaturated
fatty acids and/or large differences in the chain lengths of the two
fatty acids present on the phospholipids.

Finally. the basic concepts used in this approach also can be
used to profile types of lipids other than glycerophospholipids. These
concepts are being applied to the analysis of the content and
distribution of glycolipids in bacteria. and sterols in fungi. This is
discussed in greater detail in Chapter 6. Applications to
phosphonolipids. ether lipids. and other types of lipids are certainly
possible. and also are discussed in' Chapter 6.

References

 

10.

ll.

12.

13.

14.

15.

. Ratledge. C.: Wilkinson. S. G. Microbial Lipids, Volume 1,

Academic Press: San Diego. 1988.

Kates, M. Advances in Lipid Research. Volume 2. Paoletti. R.. Ed..
Academic Press: New York, 1964: Chapter 1.

Lechevalier. M. P. CRC Crit. Rev. Microbiol.. 1977. 5, 109-210.

Heller. D. N.; Cotter. R.J.: Fenselau. C.: Uy. O. M. Anal. Chem.
1987. 59. 2806-2809.

Balkwill. D. N.; Leach. F. R.; Wilson. J. T.: McNabb. J. F.: White. D.
C. Microb. Ecol. 1988. 16. 73-84.

Pramanik. B. N.; Zechman. J. M.; Das. P. R.: Bartner. P. L. Biamed.
Environ. Mass Spec. 1990. 19(3). 164-170.

Ross. M. M.; Neihof. R. A.: Campana. J. E. Anal. Chem. Acta. 1986,
181. 149-157.

Plattner. R. D.: Stack. R. J.; Al-Hassan, J. M.; Summers. B.;
Criddle. R. S. Org. Mass Spec... 1988. 23. 834-840.

Zirrolli. J. A.: Clay. K. L.; Murphy, R. C. LIPIDS (In Press).

Weintraub. S. T.: Lear. C. S.: Pinckard, R. N. J. Lipid Res., 1990.
31. 719-725.

Jensen. N. J.; Tamer. K. B.: Gross. M. L. LIPIDS. 1986. 21. 580-
588.

Jensen. N. J.; Tamer, K. B,: Gross. M. L. LIPIDS, 1987. 22. 480-
489.

Cole. M. J.; Enke. C. G. Paper presented at the 37th Annual
Conference on Mass Spectrometry and Allied Topics. May 21-26.
1989. Miami Beach. FL.

Hayashi. A.: Matsubara. T.: Masanori. M.; Kinoshita. T.: Nakamura.
T. J. Biochem. 1989. 106. 264-269.

Sweetman. B. J .; Tamura. M.; Higashimori, K.: Inagami. T.: Blair, 1.
A. Paper presented at the 35th Annual Conference on Mass
Spectrometry and Allied Topics. May 24-29. 1987. Denver. CO.

16.

17.

18.

19.

20.

21.

22.

23.

Munster. H.; Budzikiewicz. H. Biol. Chem. Happe-Seyler. 1988.
369. 303-308.

Kayganich, K. A.: Murphy. R. C. Paper presented at the 38th
Annual Conference on Mass Spectrometry and Allied Topics. June
4-9, 1990. Tucson, AR.

Heller. D. N.; Murphy. C. M.; Cotter. R. J .: Fenselau. C.: Uy. O. M.
Anal. Chem.. 1988. 60. 2787-2791.

Stratford. B. C. An Atlas of Medical Microbiology: Common Human
Pathogens. Blackwell Sci. Publ.: Edinburgh. 1977. 97-107.

Bligh. E. G.; Dyer. W. J. Can. J. Biochem. Physiol.. 1959. 37. 911-
917.

Huang. Z. H.; Gage, D. A.: Sweeley. C. C. Journal American Society
for Mass Spectrometry. Submitted, 1990.

Odham. G.; Tunlid. A.: Westerdahl, G.; Larsson. L.; Guckert. J. B.;
White. D. C. Journal of Microbiological Methods. 1985. 3. 331-
344.

Tunlid. A.: Ringelberg. D.: Phelps. T. J.; Low. C.: White. D. C.
Journal of Microbiological Methods, 1989. 10. 139-153.

CHAPTER 4

Fast Atom Bombardment / Ian-Molecule
Reactions for the Differentiation a f
Phospholipid Classes

Introduction and Dec und

onstant neutral loss and parent scanning for polar head

functional groups is a very useful technique for

differentiation of phospholipid classes and analysis of
phospholipids in complex matrices and mixtures. However. in low-
energ CID. some classes of phospholipids do not readily lose their
polar head group while others fragment to give a non-unique neutral
loss [1,2]. Therefore. it is not always possible to differentiate
phospholipid classes using low-energy CID. Reactions occurring in the
second quadrupole between protonated phospholipids and ethyl vinyl
ether were found to be useful for differentiating phospholipid classes.
In this chapter, these ion-molecule reactions are explored as an
alternative to low-energy CID methods for the differentiation of

phospholipid classes.

Esperlmental

All experiments were performed using a Finnigan (San Jose. CA) T89-
70 triple stage quadrupole instrument equipped with a standard
Finnigan FAB source and an Ion Tech (Teddington, UK) saddle field

71

gun. Spectra were acquired and processed using the Finnigan TSQ-70
data system and software.

All phospholipids were obtained from Sigma Chemical Co. (St.
Louis. MO). Ethyl vinyl ether and glycerol were obtained from Aldrich
Chemicals (Milwaukee. WI).

The phospholipid samples were dissolved in the glycerol matrix
and 1-3 111 of this mixture was placed on the probe tip for analysis.
Xenon was used as the FAB gas and the FAB gun was operated with a
current flow of 2.4 mA and a beam energr of 7.5 keV. The individual
protonated lipids formed in the ion source were selected by the first
quadrupole to pass through to the second quadrupole (rf-only) for
reaction. Ethyl vinyl ether was introduced into the second quadrupole
to a pressure of 3.5 - 4.0 mtorr as indicated by the instrument's
convectron gauge. which is connected to the reaction chamber. The
reaction and dissociation products were analyzed by scanning the
third quadrupole over a wide mass range above and below the mass of

the selected parent ion.

The formation and abundance of the reaction products depends
on the instrumental parameters. The most important of these
parameters are the DC offset on Q2 and the voltages on the lenses
immediately before and immediately after 92. These potentials
control the kinetic energy of the reactant ion being passed by 91. The
instrumental parameters were optimized by adjusting them to

maximize the abundance of a product ion of m/z 101 produced in Q2

72

when protonated glycerol ions react with ethyl vinyl ether. This ion is
due to the loss of acetaldehyde from the complex of protonated and
neutral ethyl vinyl ether molecules [3]. The optimum DC offset on Q2
was always found to be between +2.5 V and +4.0 V; the lens
immediately preceding Q2 was found to optimize between +5.0 V and
+6.0 V: the optimum voltage for the lens immediately following Q2 was
found to be between -1.0 V and -5.0 V. These conditions are optimum
for the ethyl vinyl ether reactions used in this study.

Results and Discussion

 

Unique Reaction Products

Upon reaction with ethyl vinyl ether. each phospholipid class gave a
unique pattern of reaction products which permitted the
differentiation of the various phospholipid classes studied. Reaction
product spectra from representative members of four different classes
of phospholipids are shown in Figure 4.1. The selected reactant ion is
the lowest mass peak in the spectrum. The masses of the neutral
additions are given in parenthesis above each product ion peak in the
mass spectrum. Each phospholipid class except phosphatidylcholine
gave at least one unique reaction product with ethyl vinyl ether.
Phosphatidylcholine gave a non-unique reaction product. but it was the
only class to give a single reaction product. The mechanisms of the
reactions that are occurring are not yet known. However. the most
probable compositions of product species in the observed addition

reactions are listed in Table 4.1.

 

a)

Ihflafiuslnxuxkuxni

 

[MO-H] . Ir100 I
7114 WWW: 0169.017zcyc

 

 

 

 

 

 

 

 

100'
:01
00'1
401 (+31)
735
(+57) 10‘
20‘ 7“ (+72) (+000) (+117)
- A 1 L A
U I I f f ' I '
700 720 740 700 700 soc szo s40
IMsss/(Huugo
(+23)
100‘ 7‘ ' Phosphsidyldycsrol: C160. 016:0
90‘
(”+111 ’
.0-1 72 3
40‘ ‘
(+72)
79 5
20‘
44
(+ ) (+98) (+124)
707 '2‘ 047
_ .L ._ A. Ah; I.
o f I r ' r ' r ' I V r '
720 740 780 780 800 I 20 840
lasso/(Hung.

 

Figure 4.1

Mass spectra of the reactions of four phospholipid
classes: a.) PE C16:0. Cl7:cyc b) PG C16:0. 016:0 c)
PC C18:0. C18:2 d) PI C15:0. 012:0.

C16:0 is a 16-carbon fatty acid with no unsaturation.
Cl7:cyc is a l7-carbon propyl fatty acid. C18:2 is an
18-carbon fatty acid with 2 points of unsaturation.
etc. The order in which the fatty acids are listed
correspond to their positions on the glycerol core.

74

 

[MM] *

 

 

 

 

 

 

 

 

 

 

 

 

E 75 I
no . . .
C) mo. Phosphabdyldtolm. 0199,0192
1101
(+31)
g 00:1 1117
,2
g 40-
E
20‘
L
U I ' I I ' 1 m f v I . ra—
700 000 020 040 000 000 900 920 940
Mass/Charge
mm] + [11250 R
d) 100- 73’ Phosplliidylinosibl: 015:1.01221
829 (+92)
(1'57)
801 794
00-1
2 40‘
3 (+72)
(+31) 000
70s
20‘
(+44)
701
AAA ‘44 n I]; L l
U ' I ' TI I I I ' I f I
740 700 700 000 020 040 sec sso 900
Mass/Charge

 

Figure 4.1 cont.

 

Table 4.1 Compositions Of Addition Species

ethyl vinyl ether (Neu): CH2= CH-O-CH§CH3

MAS SPECIES MAS_S_ SPECIES
26 (32 H2 92 2Neu - C‘H 8
31 CHaO 98 Neu + CZH 2
44 CzH‘O . 104 Neu + CH 30H
57 caHso 117 Neu+CzH50

72 0‘ Ha o (Neu) 124 Neu + 04H)3

 

76

Where The Reactions Occur

For application in phospholipid class differentiation, only the reactions
of ethyl vinyl ether with the polar head group will be useful. If the
differences in reaction products were dependent on the differences in
fatty acyl groups, then different reaction products would be obtained
within the same class. The spectra of two phosphatidylethanolamine
analogues which differ only in their fatty acyl content are shown in
Figure 4.2. The same reaction products (relative to the reactant ion)
were obtained in each spectrum. These data demonstrate that the '
reactions are not occurring on the fatty acyl moieties of the
phospholipid, but are occurring on the polar head group. When
fragment ions which have lost their polar head group are reacted, all
classes of phospholipids form the same reaction products. This is
further evidence that the reactions are occurring on the polar head

group and are thus indicative of the phospholipid class.

Collisionally Induced Dissociation Data

A full daughter scan of m/z values above and below the mass value of
the protonated molecule. can provide both ion-molecule reaction data
and collisionally induced dissociation data. A full daughter scan of a
selected phosphatidylethanolamine parent ion that contains a 16-
carbon saturated fatty acid on the first position of the
glycerophosphate core. and a l7-carbon propyl fatty acid on the
second position is shown in Figure 4.3. Phospholipids dissociate to
form ions that contain information on the fatty acid composition.
These ions appear as the glycerol core with a loss of the head group

and one or the other of the fatty acids. The characteristic dissociation

77

 

 

[Mi-H] * [:100 l

 

 

 

 

 

 

 

    
    

 

 

 

 

a) m- 7‘” Phomandyimmia: 616:0, 017:cyc
80"
60“
40. (+31)
735
(+57)
201 75‘ (+72) (+312? (+117)
8 (+44) 775 821
C 748
(U - -v . f. -..L .Il. .l. v-.- 1., ll -
g D 700 720 740 760 no 000 - 020 340
Q
.2
H O
.59 [“le lxzoo |
b) g ‘00. 7.32 . We: 018:0.Ci7zcyc
80"
60‘
(+31)
763
40"
2°. 117
(+72) (+104) (+850)
004 .3.
. - - L. . . . v i -
750 800 850 900
Mass/Charge

 

Figure 4.2 Mass spectra of two phosphatidylethanolamine analogues.
These lipids are of the same class but differ in their fatty acid
content. a) PE C16:O,C 17zcyc b) PE C18:O.Cl7:cyc

78

 

 

 

CID I ‘— mm —|

 

 

 

100‘
80“
3
2
0
i 40"
c:
20‘
OJ

 

 

 

 

 

 

 

 

cmo’
W200 l 553 chm lxsoo I
cum’
313 FOR
[M+Hf
704
+
on
. . [man
325 F6174)” 735
[M+117]+
821
295
281
va'rtvvav'vv:['*1fr "']"'fr"'A'["
200 300 400 500 600 700 800

Mass/Charge

 

Figure 4.3 Full daughter scan of a selected phosphatidylethanolamine parent
showing CID fragments as well as the characteristic reaction

products.

fragments shown in Figure 4.3 appear along with the characteristic
reaction products. In the positive ion mode, the peak corresponding
to a loss of the fatty acid from the sn-2 position is more intense than
the peak corresponding to a loss of the fatty acid from the sn-1
position. This conclusion is drawn from observations of the
fragmentation patterns of different phospholipid classes containing
various fatty acids. The same pattern is also observed in the
conventional FAB mass spectrum of each phospholipid. These relative
intensities observed in the positive ion spectra are reversed from
those found by Zirrolli et al. in the negative ion spectra [4], where the
relative daughter ion peak intensity for the fatty acid in the sn-2
position is greater than the daughter ion peak intensity for the fatty
acid in the sn-1 position. Huang et al. have shown that this general
rule apparently does not apply .to phospholipids containing highly
unsaturated fatty acids or when large differences in the chain lengths
of the two fatty acids occur [5]. It is not yet known if these exceptions
also occur in the positive ion CID spectra: this should be kept in mind
when analyzing these types of data. However. for phospholipids which
do not fall under these exceptions, a single scan can structurally
characterize a phospholipid by giving the class of the phospholipid.
the fatty acids contained on the phospholipid, and the relative
positions of the fatty acids on the phospholipid.

Neutral Gain Scans
In addition to identifying classes of phospholipids through the unique
reaction products of selected individual peaks in the mass spectrum,

specific neutral gain scans can be performed that are selective for

a)

each particular class. In this way, all peaks in a mass spectrum that
belong to a particular class are identified in the same scan. An
example of a neutral gain scan is shown in Figure 4.4. The top
spectrum is the conventional FAB mass spectrum of a crude lipid
extract containing phosphatidylglycerol. The bottom spectrum is the
neutral gain scan for the addition of 26 u, an ethyne moiety. This
addition is specific for phosphatidylglycerol. The neutral gain scan
selected only that peak in the conventional mass spectrum resulting
from the protonated molecule of the one phosphatidylglycerol species
present. The neutral gain scan greatly increases specificity and signal-
to—background ratio, thereby improving the detection limit. This
technique is similar to neutral loss scanning with the advantage that
those ions that do not fragment easily to lose their head group as a

neutral loss. do readily react to add a neutral gain.

Figure 4.5 is a comparison of a neutral loss spectrum and a
neutral gain spectrum. The top spectrum is the neutral gain of 57 u.
which is specific for phosphatidylinositol, while the bottom spectrum
is the neutral loss of 260 u for the same sample. This neutral loss
value also is specific for phosphatidylinositol. The actual intensity axis
has been left on each spectrum for comparison. All of the
phosphatidylinositol species present in the sample were detected by
both scan modes. For phosphatidylinositol. the neutral loss spectrum
is an order of magnitude more intense than the neutral gain spectrum.
This fact is a function of the individual phospholipid class; the relative
sensitivities of the two methods depend upon the facility with which

each phospholipid class fragments to lose its polar head group as a

81

 

745

a) 1 00"

‘°‘ 7f1
son

40"

 

713
857
071

 

 

    

lf3

  

”7

      
  

         
 
   

     

 

 

 

20 710 'l’
H 795 upsrnJo
1? - .1 . , ‘ p 2.01
g "‘l‘nl‘i ails” 1“.-1i1'.|(M
g 700 800 900 1000
Q)
-§
-§
b) m 100‘ 7:3
00‘
00-
400
201
u :_r'A_:AI'A:A‘v_‘v~A—r—‘f rilvrffifi}v"_'vvr" vvvvvvvv r
7°° '00 000 1000
Mass/Charge

 

Figure 4.4 0) Conventional mass spectrum of a sample containing
phosphatidylglycerol. b) Neutral gain scan for the
addition of 26 u. This addition is specific for
phosphatidylglycerol.

 

738

 

 

 

a) 1001 .103
-o
I”
U
z
3
3 H
O
(
W
Z
:2 -4
.J
L“
C
721 - 2
. 70° '50 763 793 307 F
. no
I
Q . Lo
70° coo
MASS/CHARGE
. 730
b) ‘00 110‘
a: "6
2
§
3
< 721
s r‘
)2
(
d
LU
I
750 '1

 

 

 

 

 

 

MASS l CHARGE

 

Figure 4.5 a) Neutral gain of 57 u; specific for PI. b) Neutral loss of 260
also specific for PI.

neutral loss. These IMR neutral gain scans do not provide information
different from that obtained from neutral loss spectra. but they are
highly advantageous in situations where neutral loss data are

ambiguous or unavailable.

References

 

 

l. Munster, H.; Budzikiewicz. H. Biol. Chem. Hoppe-Seyler, 1988.
369. 303-308.

2. Cole, M. J .; Enke, C. G. Paper presented at the 16th Annual Meeting
of the Federation of Analytical Chemistry and Spectroscopic
Societies, October 1-6. 1989. Chicago, IL.

3. Pachuta, R. R.; Kenttamaa. H. 1.: Cooks. R. G.; Zennie, T. M.; Ping.
C.: Chang, C.: Cassady, J. M. Org. Mass Spec.. 1988, 23, 10-15.

4. Zirrolli. J. A.: Clay, K. L.; Murphy, R. C. IJPIDS (In Press).

5. Huang, Z. H.; Gage, D. A.: Sweeley. C. C. Submitted to Jow'nel of the
American Society for Mass Spectrometry. 1990.

CHAPTER 5

Applications of Phospholipid Analysis to the
Detection and Identification of Microorganisms

Introduction and Bac und

 

he sensitivity of mass spectrometry - from the standpoint

of the uniqueness of specific phospholipid patterns to

individual microorgansims - should lead to its widespread
use for microbial detection, characterization. and identification. In
fact. this has not been the case. While many researchers have used
mass spectrometry to analyze phospholipid species [1-4], only one
other use of phospholipids for the. general identification of bacteria has
been reported [5]. This method relies on linear regression of patterns
in conventional spectra of phospholipid extracts from bacteria.
Although accurate identifications were made from the limited library
used, this method is highly dependent on spectral reproducibility.
Also, difficulties will arise when the number of cells in the sample
becomes so low that the high chemical background associated with

fast atom bombardment becomes a major obstacle.

The low chemical background inherent in tandem mass
spectrometry makes the method described in Chapter 3 a powerful
means of identifying microorgansims. Almost all of the phospholipid

information can be obtained. including the classes and number of

86

species present, the empirical formulae of the fatty acids present on
each species. and the positions of the fatty acids present on each
species. This combined information forms a unique and detailed

fingerprint which can be used to identify microorganisms.

This chapter describes the application of the general
phospholipid analysis technique from Chapter 3 to the detection and
identification of microorganisms. Fingerprint profiles have been
collected from a number of bacteria. fungi, and amoebae: a listing is
given of microorganisms analyzed up to this point in the research.
Results from two of the countless potential clinical and industrial
applications are described. In the clinical application. the causative
agents of urinary tract infections are profiled. These profiles are then
used to identify bacteria present in infected urine samples. Sensitivity
and speed are exemplified. The industrial application described is in
the context of the food industry. Three bacteria important in food-
borne contamination are differentiated. Here, the focus is on the
general ability of the technique obtain the unique data. The three
bacteria studied are very difficult to differentiate using the common
known methodologies. This chapter concludes with a discussion on
the need for. and general structure of. a database/ library containing
microbial phospholipid profiles.

Experimental

All experiments were performed using a Finnigan (San Jose, CA) 1’89-
703 triple stage quadrupole mass spectrometer equipped with a
standard Finnigan FAB source and either an Ion-Tech (Teddington,
UK) saddle field FAB gun or a JEOL (Boston, MA) MS-OOQ charge-
transfer FAB gun and power supply that has been modified to fit and
operate on the TSQ-7OB. Spectra were acquired and processed using
the Finnigan TSQ-7OB data system and software.

Infected urine samples were obtained from Dr. John Dyke.
Sparrow Hospital, Lansing, Michigan. Borrelia bw‘gdorferi strains and
Leptospira interrogans icterohemorrhagiae were obtained from Dr.
Michael Kron, Department of Pathology and Medicine, Michigan State
University. The gram positive bacteria were obtained from Dr. Harold
Sadoff, Department of Microbiology, Michigan State University. The
bacteria responsible for food-home contamination were obtained from
Dr. Edward Richter. Silliker Laboratories of Ohio. Inc. Legionella
pneumophila knox, Naegleria 5049, and the 7 deep-probe subsurface
isolates were obtained from Dr. David C. White, Instituted for Applied
Microbiology, University of Tennessee. All other microorganisms were
obtained from researchers in the National Science Foundation's

Center for Microbial Ecologr at Michigan State University.

Crude lipid extracts were obtained from the microorganisms by

a modified Bligh-Dyer procedure [6]. One ml of a cell suspension, or a

transfer loop full of sample, was placed in a clean. sterile 25-ml
sample vial to which 15 ml of a 2:1 methanol:chloroform mixture was
added. After the lipids were extracted. enough distilled/deionized
water was added to separate the chloroform from the MeOH/H20.
The chloroform layer, containing the lipids. was drawn off, evaporated
to dryness. and the remaining lipids reconstituted in 1 ml of

chloroform.

For the analysis, 3-5 ill of the chloroform solution was placed on
the probe tip and dissolved in a drop of nitrobenzyl alcohol. Xenon
was used as the FAB gas and the FAB gun was operated with a filament

current of 10 mA and a xenon beam energy of 8 keV.

Neutral loss and parent ion .spectra were obtained in the positive
ion mode with no CID gas added to the center quadrupole other than
the residual gas remaining in the analyzer manifold. The
fragmentation observed may be predominantly the result of metastable
decomposition. Daughter spectra were obtained in the negative ion
mode using argon as a collision gas. The collision gas pressure was 0.5

mtorr and the collision energt was 30 eV.

All other experimental procedures. equipment, and conditions
are given in the specific sections in which they are discussed in this
chapter.

Results and Discussion

Comparing the Data

The differentiation of the spectra presented in this chapter is based
on an acquired familiarity with the similarities and differences that
occur among the patterns that characterize difierent microorganisms.
An appreciation of which differences constitute the normal variance
for a single organism. and which are valuable in distinguishing
organisms. has been obtained through the observation of many spectra.
Before this chapter proceeds. it is important to discuss the inferred
significance of the observed features and differences and the rationale

behind these conclusions.

Differences among the spectra obtained from replicates of the
same microbial species can be due to a number of factors. The
foremost causes of spectral variance are small differences in the age
and growing conditions of the cultures. Since microbes present a
dynamic living system. small variances in their chemical composition
can be expected between replicate cultures. Small spectral variances
also can occur in the day to day operation of the instrument. As long
as these spectral variances remain within certain predescribed limits.
they can be taken into account when matching spectra.

Now it becomes important to decide in what ways. and to what
extent, spectra may vary and still be considered identical.
Observations of the spectra obtained from the microorganisms used in

this research have led to some conclusions on how these spectra differ

90

among organisms. First, many organisms differ in the classes of
phospholipid they contain. This leads to an unambiguous
identification regardless of how much data they have in common.
Likewise. many organisms differ in their content of individual
phospholipid species within each class. This also leads to an
unambiguous identification. The gray area arises when two or more
microorganisms contain the same phospholipid classes and the same
phospholipid species within each class. Here. identification is based
solely on differences in intensities of the phospholipid species. In this
research. the base peak(s) is considered most important and. as such.
is considered invariant for a given species. The overall pattern of the
phospholipid peak intensities is the next important criterion. If the
pattern of relative intensities differs to a large extent. the spectra are
considered different. Next. individual peaks may vary within
predescribed “zones”. These zones are defined as major peaks (75%
to 90% of the base peak). intermediate peaks (20% to 75% of the base
peak). and minor peaks (less than 20% of the base peak).

At this time. pattern recognition is done crudely; all matching is
performed by eye on limited sets of data. However. this work clearly
shows that distinct differences in phospholipid content and
distribution do exist among the microorganisms studied so far.
Recognition of these differences. and identification of the similarities.
by the “human pattern recognition method” has not presented any
difficulties. or even near-ambiguities, for the limited data set available.
Given the large number of distinguishing parameters available in the

complete phospholipid profiles. it is reasonable to assume that

91

computer techniques can be developed to recognize these patterns on

a broader scale.

For an automated pattern recognition method to succeed. it will
have to determine those criteria that are essential for judging spectra.
In addition. such a method will need to identify and quantitate the
variations that can be tolerated for a correct match. The database
acquired so far is a limited one: this technique needs to be applied to
many more samples to determine which features and differences are
most important for taxonomic distinction. In this way. the pattern

recognition can be refined and optimized.

Microorganisms Analyzed

To date. about 50 separate microbial species have been characterized
using this technique. A general listing of these microorganisms is
given in Table 5.1. Where known. the strain(s) of the particular
species is also given. The cricket gut isolates. Savannah River isolates
and fungi isolates are presently unidentified species and form the basis
of three ongoing projects to compare different analysis techniques.
The Savannah River isolates were obtained from a deep-well drilling
probe at the Savannah River nuclear site and are part of a collaborative
study with Dr. David C. White at the Institute of Applied Microbiology.
University of Tennessee and Dr. David Stahl at the University of
Illinois. The cricket gut and fungi isolates are part of collaborative
studies with Dr. Mike King and Dr. Pete Stahl at the Kellogg Biological
Station. Michigan State University.

 

Table 5.1

Listing of Microorganisms Characterized by
this Technique

W
W
Pseudomonas fluorescens ARC 33512 Biotype ll
Pseudomonas fluorescens Aruna AK-15
Pseudomonas fluorescens Janet 88-1
Pseudomonas aeruginosa - many strains
Pseudomonas aureofaciens
Escherichia coli - H8101, and many other strains
Alcaligenes faecalis ATCC 15554
Alcaligenes eutrophus - many strains
Citrobacter freundii - many strains
Salmonella abaetetuba
Klebsiella pneumophila - many strains
Klebsiella oxytoca - many strains
Morganella morganii
Proteus vulgaris
Borrelia Burgdorleri - many strains
Leptospira interrogans ecterohemorrhagiae
Legionella pneumophila knox
2 Cytophagus species

W

Carynebacterium nephridii ATCC 11425
Bacillus subtilis

Bacillus megaterium

Bacillus cereus

Mycobacterium phlie

Leuconostoc mesenteroides
Streptococcus faecalis

 

83

 

Table 5.1 cont.

I A' 3 - . 0.- 0“ n‘ ...:....x‘i q'l
Ralf-4 Possible Shigella sp.
MIL-3 Possible Citrobacter sp.
AD-25 Unknown
Bac-4 Possible Citrobacter sp.
LEG-2 Possible Shigella sp.
3 l- 04: 'll"‘°‘i‘ '10-. ._ {II1I::
C484 C564
C553 0678
C570 C651
C528
EUNJ-J
5| Ill IEII E'I'ISII'
101-4B&D 101-3 101-6
112-1 B&D 113-2 113-2 B&D
138-1 106-1
AM BAE
Naegleria 5049

 

Detection and Identification of Urinary Tract Infections

The urinary tract is the most common site of infection during
hospitalization, and is considered infected when bacterial counts of
1x105 cells/ ml are obtained from properly collected urine specimens
[7]. Most infected urine contains only a single species of bacteria; at
most. only two or three species may be present. In the uninfected
state. urine is sterile and contains no microbes. For these reasons,
urinary tract infections presented the least complex samples and were
chosen as the first real-life clinical problem in which to apply this
technique.

The normal procedure in identifying a urinary tract infectant is
the use of the selective agar pure culture assay described in Chapter 1.
This assay requires 48 hours for isolation and incubation of the
infecting microbe. and is vulnerable to the inaccuracies and
uncertainties inherent to these tests. While urinary tract infections
may not of themselves be life threatening, in their advanced stages
they can be quite painful and serve as a source of invasion of the
bloodstream. This can lead to renal damage, infection of other tissues.
and endotoxic shock. A rapid and sensitive identification procedure
would help to avoid such problems by quickly detecting the presence.

and identifying the causative agent. of an infection in its earliest

stages.

The experimental procedure developed for extracting the urine
samples is rapid and simple. The urine samples are centrifuged at

1800 rpm for 3 minutes to spin the cells down to the bottom of the

%

sample vial used to collect the urine. The urine is then carefully
drawn off from the top until about 1 ml of urine remains in the vial.
This remaining 1 ml contains most of the cells from the 5 ml of urine
originally present in the sample vial. The sample is then extracted
using the procedure described in the experimental section of this
chapter.

The first attempts in analyzing urine samples met with failure.
In these experiments. American Type Culture Collection (ATCC)
cultures of bacteria were used as the standards with which to compare
the bacteria obtained from the urine samples. The first problem
encountered involved the urine samples themselves. _ Infected urine
can contain many more cells from sources other than just those from
the bacteria causing the infection. In addition to the bacteria. cells
from blood and. urethra sloughage may also be present. These
additional cells can comprise the majority of the cells present in the
urine and lead to very intense interferences in the analysis. As a

result. none of the urine spectra were similar to the standard spectra.

This problem was overcome through the use of 10 um Acrodisc
syringe filters obtained from Gelman Sciences (Ann Arbor. MI). Since
the bacteria responsible for urinary tract infections have cell
dimensions smaller than 10 um. and the eukaryotic cells associated
with the human host have cell dimensions greater than 10 um,
filtering the urine through 10-um filters prior to extraction removed
all of the non-infectant cell matter from the urine. The resulting
samples were then centrifuged and extracted as previously described.

%

Filtering the samples worked so well it uncovered a second
problem hidden in these experiments. A hospital environment
presents to bacteria many selective pressures that can cause the
occurrence of new strains that overcome these pressures.
Specifically. the widespread use of antibiotics in a hospital often
results in bacterial strains that are resistant to those antibiotics that
were previously effective. Therefore. the specific strain of bacteria
responsible for a urinary tract infection in a particular hospital is not
necessarily the same as the ATCC strain used as a standard. The
experiments described here clearly indicate the importance of the
standards upon which the identification is based.

The spectra obtained after filtering the urine were still not
identifiable as any of the ATCC species used as standards. Since this
problem was still thought to be a result of the sampling procedure.
bacteria, from urine samples previously identified by the hospital.
were cultured on trypticase soy agar (TSA) plates at 37° C for 24 hours
in an attempt to discover the differences due to the sampling problem.
Single colonies were removed from these plates. transferred to a
second series of TSA plates. and grown at 37° C for another 24 hours.
These bacteria were harvested. extracted. and analyzed. The spectra
obtained compared reasonably well with those spectra obtained from
the initial filtered urine extract. Examples of
phosphatidylethanolamine mass profiles obtained from an ATCC strain
of Pseudomonas aeruginosa, a filtered urine sample infected with

Pseudomonas aeruginasa. and a culture from the same urine sample

97

after two passes of growing on TSA are shown in Figure 5.1. The
noteworthy points in these spectra are the increased intensities of the
peaks at m/z 704 and m/z 744, in the bottom two spectra. over those
in the top spectrum. Also. a peak at m/z 732 occurs only in the
bottom two spectra. The large differences in the spectra between the
ATCC strain and the strain grown on two passes through TSA lends
strong evidence to the existence in the hospital of a different strain of
Pseudomonas aeruginasa. The similarity between the filtered urine
spectra and the spectra obtained from the bacteria grown on TSA is
corroboration that the sampling procedure is sound; the filters are
indeed filtering away interfering cell matter. The differences seen in
these two spectra could be the result of differences in the sample
histories. The growth phase and conditions of the bacteria collected
straight from the urine are different than the rich conditions offered
those isolated and cultured on TSA. Just how much these spectra will
be allowed to vary. and still be considered to belong to the same

species of organism, requires further research on many more samples.

The next step in developing these experiments was to use
bacteria obtained from actual infected urine samples - ones in which
the infecting bacteria had been previously identified by the hospital -
as the known standards with which to compare the unknowns. This
approach has the advantage of using the same strains of bacteria for
the standards as those occurring in the urinary tract infections.
Examples of phosphatidylethanolamine mass profiles for Escherichia
coli. Pseudomonas aeruginasa. Klebsiella pneumoniae. Klebsiella

oxytoca. Citrobacter freundii. and Morganella Morganii are shown in

%

 

P. aeruginasa
mq (ATCC strain) 7..

71st

 

 

 

P. aeruginosa
(From filtered urine) 71s

100-

716

 

 

Relative Abundance

 

 

Q0 ”0 7” 750 soo 850

P. aeruginasa
(Isolated from urine on TSA)

7‘8

 

 

 

 

 

100
744
730
704
o ‘1*“- .Lflr? r . H4
000 GO 700 750 sec 850
Mass/Charge

 

Figure 5.1 Phosphatidylethanolamine mass profiles from three different
isolates of Pseudomonas aeruginasa.

Figure 5.2. These small pieces of the complete profiles only represent
some of the differences among the bacteria; the complete information
for these six bacteria takes up considerable space and for this reason

has been excluded from this thesis.

A phosphatidylethanolamine mass profile from an infected urine
sample in which the causative bacteria was unknown is shown in
Figure 5.3. Comparing this part of the data space with the
corresponding part for the six bacteria in Figure 5.2. the unknown
spectrum most closely resembled that from the standard of P.
aeruginasa. The identification was based on the occurrence of the
peaks at m/z 716 and m/z 718 as the two most intense peaks in the
spectrum. coupled with the peak at m/z 744 as the next highest in
intensity. Either of the peaks at- m/z 716 or m/z 718 could be the
base peak. but the other should be within 75% of the base peak. In
addition, the lack of a peak at m/z 664 was unique to the spectrum
obtained from the P. aeruginosa standard. P. aeruginosa was the only
standard whose phosphatidylethanolamine (class)mass profile
contained this pattern of peaks. This was. in fact. the organism
responsible for the patient’s urinary tract infection. It must be
pointed out that the possibility exists that individual strains could
differ from one hospital to the next. In routine application. standards
would need to be run that were specific for each particular hospital.
In addition, these standards would need to be checked on a routine
basis to assure their integrity against the occurrence of new strains.
The data presented here are the results of a preliminary study. Many

replicates of each species need to be performed so that the

100

 

Escherichia coli

704

 

100-0
1
664 590 73 732 7“
676
°'1
6(0 650 700 750 ”0 $0

Pseudomonas aeruginosa

718
1&1

716

 

 

 

Relative Abundance

Klebsiella pneumophilia

744

 

 

100q
768
704 718 71.
ass 732
0-
can no 7“ 750 800 50

Mass/Charge

 

Figure 5.2 Phosphatidylethanolamine mass profiles from six bacterial species
known to cause urinary tract infections.

101

 

Klebsiella oxytoca

 

 

 

 

 

 

 

 

 

 

 

100- 7“
741
710 700
55“ 000
0

000 000 700 700
8 Citrobacter freundii
c 500
(u 100-
2 704 710
a
2 730
(D 664 J 744
.2 07s
a 0 ALI AL; IALUAf I‘ll Tflhfi l T1 #‘IIIA AU 4“"
B 000 050 700 750
a:

Morganella morganii
1001 59° 704 7“
710 700
700
664
_ I. L LL ..... hall _.m. .L. ..
O r I f T r fi r r
600 650 700 750
Mass/Charge
Figure 5.2 cont.

102

 

716

 

 

 

 

 

100. 71a
80-
§§ 744
:0
E
.3
2540-
a)
E
70‘ 73732
0 1
2°“ 690 l
' l
*
oJ ‘
600 650 700 750 800 350
Mass/Charge

 

Figure 5.3 Phosphatidylethanolamine mass profile from an infected urine
sandpkL

103

occurrence of different strains can be monitored. In addition,
mathematical data analysis techniques need to be developed for the
identification of co-infecting bacteria.

Differentiation of Food-Bome Contaminants

Salmonella species are very difficult to correctly differentiate from
Escherichia species using common microbial techniques. Escherichia
and Salmonella have up to 50 percent of their DNA sequences in
common [8]. In the food testing industry. this presents a formidable
problem. Since Escherichia are rarely pathogenic to human beings.
government regulations will allow food products to contain some
Escherichia. In contrast. Salmonella are usually pathogenic to human
beings: government regulations do not allow the presence of
Salmonella in foods. The methodology currently used to differentiate
Salmonella from Escherichia involves complex immunological assays
that take up to 7 days for results. This presents other problems: the
cost of storage and the spoilage incurred can be very expensive. Also,
the implications of time and expense in discovering a Salmonella
infection 7 days downstream of a production process could be
enormous. The food industry is very interested in shortening the time
period needed to obtain results. Research has begun in this laboratory

toward a solution to this problem.

The Salmonella abaetetuba. Escherichia coli. and Citrobacter
freundii used in this study were grown at 37° C for 48 hours in both
tetrathionate broth. containing potassium iodide and brilliant green.
and trypticase soy agar plates. Tetrathionate broth is partially

104

selective for Salmonella. and is the medium used for this type of
testing in industry. Other bacteria. such as the E. coli and C. freundii
used in this study. also can grow in this medium. but they do not grow
as easily or as fast as Salmonella. Citrobacter freundii is another
microbe of importance to the food industry. Like E. call. it also is
usually non-pathogenic to humans and can be difficult to differentiate

from Salmonella.

The phosphatidylethanolamine mass profiles for the three
bacteria grown in the tetrathionate broth are presented in Figure 5.4.
As shown by the patterns obtained. the three bacteria are easily
differentiated from each other. Again. for the sake of brevity only
enough information needed to make this point is presented. The total
data space provides the unique and characteristic information set for
each microorganism. A vial of previously sterile tetrathionate broth
was received that had become contaminated with one of the
organisms. This contaminant was readily identified as Salmonella
abaetetuba as shown by the phosphatidylethanolamine mass profiles in
Figure 5.5. Comparing the spectra. the spectrum from the unknown
most closely resembles that from S. abaetetuba, and differs greatly

from those of E. coli and C. fieundii

The results obtained from the organisms grown on TSA plates
were very interesting. The phosphatidylethanolamine mass profiles
are shown in Figure 5.6. While C. fieundii is easily difi'erentiated from
the other two bacteria. the phospholipid (classimass profiles of E. coli

and S. abaetetuba are nearly identical. These results illustrate the

105

 

Salm one/Ia abae te tuba
1001 7|)!

0

 

 

 

 

 

 

Citrobacter fre undii
m can

8 716

g w

'0

C

3 at

.0

<

CD

.2

E

a)

a:

 

 

Escherichia coli

1w

0

 

 

Mass/Charge

 

Figure 5.4 Phosphatidylethanolamine mass profiles from three
bacteria. grown in tetrathionate broth. important to

the food industry.
106

 

 

Unknown Contaminant

 

 

 

 

   

 

 

 

704
100-
60‘
§ 718
<6
3 so~ 7:12
I:
is 744
E; 40s
5!
Q)
a:
20'-1
664676 772
0«600 650 700 750 600 650
Mass/Charge
Salmonella abaetetuba
704
100-
60-
g; 716
'2 004
I:
2
CD
3 10-
% 732
a:
20-1
o-

 

 

600 650 700 750 600 650

Mass/Charge

 

Figure 5.6 Phosphatidylethanolamine mass profiles from an

unknown contaminant and Salmonella abaetetuba.

107

 

.... S. abaetetuba n

 

 

(L!
1-3

 

“aor"'do “'16? '710"T'0bffir‘?0

100- C. freundii

§

 

 

Relative Abundance

 

100- E. CO” 704

”.73

”-
mm 7“
IL- [ii 75le - A

do"'dé"'16o*‘;$' 1507's;

Mass/Charge

 

 

 

 

Figure5.6 Phosphatidylethanolamine mass profiles from
three bacteria. grown on trypticase soy agar.
important to the food industry.

108

importance of the role the growth medium can play. Tetrathionate
broth puts a selective pressure on Escherichia and Citrobacter, while
Salmonella grows freely. This may account for the differences in the
membrane phospholipid content. In contrast. all three organisms
grow well on TSA. Comparing Figure 5.6 to Figure 5.5. Salmonella
shows only small differences between the two growth media. These
differences remain small throughout the entire data space. On the
other hand. E. coli and C. freundii give quite different (classlmass
profiles between the two media. No new phospholipid classes occur
with the different media. rather the variations are in the form of
differences between the relative abundances of phospholipid species
within particular classes. Experiments of this type may be a potential
means of exploring how microorganisms respond to stress. Also. the
possibility is raised of subjecting very similar organisms to stressful
growth conditions in order to obtain more distinctive phospholipid
profiles.

Database / Library of Microbial Profiles

On a limited scale. comparison and matching of spectra among a small
number of microorganisms can be done using the human brain for
pattern recognition. For this technique to be useful on a large scale.
however. computerized methods must be developed for storing these
data. and for searching the resulting database. Figure 3.3 in Chapter 3
is a visual representation of how this database might look in its

simplest form.

109

The top level of the database would contain the phospholipid
class quantitative data. These data consist of the relative abundances
of the phospholipid classes contained in each microorganism. The
next level contains the phospholipid (class)mass profiles segregated by
class. The third level would contain the (class)(fatty acid) formula
data. Because the complete data set may not always be needed for an
identification. the search would be performed from the top level down
and continue until a positive match had been made. At the (class)mass
profile level. some type of intelligence will be needed to accommodate
small differences in the relative abundances of individual phospholipid
species between the unknown profile and those contained in the
database. As discussed earlier in this chapter. these differences can
be due to many factors. including instrumental parameters and small
differences in the age and growing conditions of the microbes.
Identification based on matching of the (class)(fatty acid) formula data
should be relatively straight forward. and not require any additional

intelligence.

This data base structure will need to be flexible and relational
among a variety of spectral and microbial characteristics. In some
cases. certain characteristics might correlate to a subset of the total
number of microorganisms whose profiles are contained in the
database. These characteristics could include the presence or absence
of a particular phospholipid class. the occurrence of a certain
combination of phospholipid classes. the presence of a particular

phospholipid species. or any combination of the above. This. by no

110

means. is a complete list of all possible correlating characteristics that

might be found in such a database.

Each of these characteristics would point to the subset of
microorganisms that contains them. Searching for these
characteristics first would limit the complete phospholipid profile
search to a subset of microorganisms that share correlations. These
subsets could be as broad as a whole phylum. or as narrow as a subset
that contains different strains of the same species. In this manner,
portions of the data space that might have had limited usefulness for
identification because they were shared by many microorganisms in
the total database. will become more useful through limiting the
search to a particular subset of microorganisms. As these correlations
are discovered. the database must be able to easily accommodate their
use. In addition, routines should be developed to search the database

for correlations that are not immediately obvious.

111

References

 

l. Heller. D. N.; Cotter. R.J.: Fenselau. C.: Uy. O. M. Anal. Chem.
1987. 59. 2806-2809.

2. Pramanik. B. N.; Zechman. J. M.; Das. P. R.: Bartner. P. L. Biomed.
Environ. Mass Spec. 1990. 19(3). 164-170.

3. Jensen. N. J.; Tomer, K. B.: Gross. M. L. LIPIDS. 1986, 21. 580-
588.

4. Jensen. N. J.; Tomer. K. B.: Gross. M. L. LIPIDS, 1987. 22, 480-
489.

5. Platt. J. A.: Uy. O. M.; Heller. D. N.; Cotter, R. J .; Fenselau. C. Anal.
Chem.. 1988. 60. 1415-1419.

6. Bligh. E. G.; Dyer. W. J. Can. J. Biochem. Physiol.. 1959, 37. 911-
917.

7. Boyd. R. F. General Microbiology; Times Mirror/Mosby College
Publishing: St. Louis. 1984: Chapter 33.

8. Brock. T. D.: Madigan. M. T.- Biology of Microorganisms. 5th ed.;
Prentice Hall: Englewood Cliffs. 1988: Chapter 19.

112

CHAPTER 6

Applications of Other Biomarker Analyses to
the Detection and Identification of
Microorganisms

Introduction and Dec und

n the previous chapters, the focus of the research presented
in this thesis has been on microbial phospholipids. However.
many other types of lipids exist that are useful biomarkers for
detection and identification of microorganisms. In combination with
the phospholipids. these lipids can provide an even greater level of
distinction on which to base a microbial identification. A preliminary
exploration of two other classes of lipids. glycolipids and sterols. is
presented here. Glycolipids contribute substantially to the total lipid
content of algae and Gram-positive bacteria [1], while sterols can be

used for fungal detection [1] and identification [2,3].

The glycolipids investigated in this research are the
glycosyldiacylglycerols (Structure 6. 1). The general term “glycolipid”
will be used in this chapter when referring to glycosyldiacylglycerols.
Like the phospholipids. glycosyldiacylglycerols consist of four primary
functional groups: a glycerol-3-phosphate core on which two fatty
acids (R. R’) have been esterified to the two free hydroxyl groups in

the sn-1 and sn-2 positions, and a sugar moiety (X) attached to the

113

Structure 8. 1 General glycosyldiacylglycerol structure.

114

sn-3 position. An example of the glycolipid
monogalactosyldiacylglycerol is shown in Structure 6.2. The sterols
studied in this research have the general structure shown in Structure
6.3. This structure consists of 4 rings (A. B. C. D). with an alkyl group
(R) attached to ring “D”. Many different configurations of sterols are
possible through unsaturation. hydroxylation. and methylation of the
ring system. and differences in the attached alkyl groups. An example
of the most common fungal sterol. ergosterol, is given in Structure

6.4.

For glycolipids. structural information is obtained using the same
methodology as for the phospholipids. This information can be
obtained with the same sample, and in the same analysis. as the
phospholipids. The sterol information is obtained by electron impact
ionization, off a heated direct exposure probe (DEP), using parent ion
scans. Sterols also are contained in the same extract as that for the
phospholipids. This chapter describes the tandem mass spectrometry
methodology developed. along with the type of information obtained.
for characterizing the content and structures of glycolipids and sterols
in crude microbial extracts. This information will, ultimately. be used
in combination with the phospholipid information toward the
detection and identification of microorganisms. In this chapter. the
thesis is concluded with a discussion on other potential lipid

biomarkers and their possible applications.

115

 

 

CHz—O—C —R
| 1?
HOCH2 CH—O—C —R’
0 l
“0 OCH2
OH
I
OH

 

Structure 6.2 Monogalactosyldiacylglycerol

116

 

 

Structure 8.3 General sterol structure.

117

 

HO

 

Structure 6.4 Ergosterol

118

 

Experimental

All experiments were performed using a Finnigan (San Jose. CA) TSQ-
7OB triple stage quadrupole instrument equipped with a standard
Finnigan El/ FAB source. a Finnigan direct exposure probe and a JEOL
(Boston. MA) MS-009 charge-transfer FAB gun and power supply that
has been modified to fit and operate on the TSQ-70B. Spectra were
acquired and processed using the Finnigan TSQ-70B data system and

software.

All bacteria and fungi. except Legionella pneumophila knox.
were obtained from researchers in the National Science Foundation's
Center for Microbial Ecology at Michigan State University. Legionella
pneumophila knox was obtained from Dr. David C. White at the
Institute for Applied Microbiolog. University of Tennessee. Cultures
were grown on trypticase soy agar plates at 37° C for twenty four hours
prior to harvesting.

Crude lipid extracts. containing both the glycolipids and sterols.
were obtained from the bacteria by a modified Bligh-Dyer procedure
[4). One. ml of a cell suspension. or a transfer loop full of bacteria. was
placed in a clean. sterile 25-ml sample vial to which 15 m1 of a 2:1
methanol:chloroform mixture was added. After the lipids were
extracted. enough DDI water was added to separate the chloroform
from the MeOH/Hzo. The chloroform layer. containing the lipids. was
drawn off and evaporated to dryness. The samples were then

reconstituted in 1 ml of chloroform for analysis.

119

For the FAB analysis of glycolipids. 3 to 5 pl of the chloroform
solution was placed on the probe tip and dissolved in a drop of
nitrobenzyl alcohol. Xenon was used as the FAB gas and the FAB gun
was operated with a filament current of 10 mA and a xenon beam
energy of 8 keV. For sterol analysis using the DEP. 3 to 5 pl of the
chloroform solution was placed on the probe filament. and the
chloroform was allowed to evaporate. For analysis. the DEP filament
was ballistically ramped to 250 mA. which corresponds to 250° C. and
held at that temperature for the length of the analysis. Electron
impact ionization was performed with a source filament current of 200

mA and an electron energy of 70 eV.

For the glycolipid analysis. neutral loss spectra were obtained in
the positive ion mode with no CID gas added to the center quadrupole
other than the residual gas remaining in theanalyzer manifold. The
fragmentation observed may be predominantly the result of metastable
decomposition. Daughter spectra were obtained in the negative ion
mode using argon as a collision gas. The collision gas pressure was 0.5
mtorr and the collision energy was 30 eV. For the sterol analysis.
parent spectra were obtained in the positive ion mode using argon as a
collision gas. Collision gas pressure was 0.5 mtorr and collision energy

was 10 eV.

Results and Discussion

Glycolipids

In addition to sharing structural similarities with the phospholipids.
glycolipids also undergo collisionally induced dissociation to fragment
in a manner similar to the phospholipids. During CID of the
protonated molecule. glycolipids fragment to lose their sugar moiety
as a neutral loss. while the rest of the ion retains the charge. In the
negative ion mode. one or the other fatty acid is cleaved from the
molecular anion and retains the charge. while the rest of the ion is

lost as a neutral.

Because these fragmentation patterns are identical to those for
the phospholipids. the same characterization scheme used for the
phospholipids can be applied to the glycolipids. Positive ion neutral
loss scans. specific for the loss of each possible sugar moiety. are first
performed. These scans identify the classes of glycolipids present in
the sample. An example of a neutral loss scan of 252 u. specific for
monogalactosyldiacylglycerol. from Legionella pneumophila knox is
shown in Figure 6. 1. To obtain the fatty acid information. the negative
ion daughter spectra can be collected for each of the glycolipid
species detected in the neutral loss scans. The daughter ions present
in these scans would represent the fatty acids contained on the
glycolipids. The neutral loss and daughter scans can be added. along
side those for the phospholipids. to the automated instrument control
procedure that collects the data. In this manner. both the glycolipid

121

 

747
1 00 T

80‘

60‘

731
401

 

Relative Abundance

761

2°- 701 717
799

‘1
.1111

 

’7 o 1891 : 0 i '09
Mass/Charge

§

 

FigureBJ Neutral loss of 252 11. specific for monogalactosyl-
(diacylglycerol, from Legionella pneumophila knox.

and phospholipid data may be collected from the same sample. and in
the same analysis.

Sterols

Sterols are major lipid components of fungi and can be used for
detection and identification of fungal species. Ergosterol is present in
almost all species of fungi. and is not present in other microorganisms
[1]. Therefore. the presence of ergosterol can be used as a biomarker
for the detection of fungi in a sample. The content and distribution of
the other sterols have been found to be unique to each particular
species [2.3]. and can be used in much the same way as phospholipids
for species identification. This provides much needed information for
fungi. since this research has found that the phospholipid and fatty
acid content does not vary as widely among fungal species as it does

among other species of microorganisms.

The volatility of sterols is relatively high compared to that of
phospholipids. and is well within the range of a heated inlet probe.
Since FAB produces a considerable amount of chemical background.
which can interfere with an analysis. the heated direct exposure probe
was chosen. along with electron impact ionization. This configuration
has the advantage of producing cleaner spectra. but it also has the
disadvantage of a more rapid loss of sample from the probe. Electron
impact ionization (E1) was initially chosen to study ergosterol.
Ergosterol contains conjugated double bonds in the B ring that cause

the M1" radical cation to predominate over the [M+H]"' protonated

molecule in the FAB spectrum [5]. Therefore, the assumption was
made that chemical ionization (CI) would not offer any advantages over
EI for this form of sterol. However. the usefulness of CI for other

forms of sterols has yet to be investigated.

Under collisionally induced dissociation. the sterols produced
fragmentation patterns that were identical for all of the common
sterols. The most characteristic fragmentation was the neutral loss of
the hydrocarbon chain attached to ring D (Structure 6.2). Taking
advantage of this fragmentation. parent scans are performed that
monitor the formation of the fragment ion corresponding to the loss of
the alkyl chain from the ring system. The resulting spectrum contains
peaks that represent the molecular weights of the sterols present in

the original sample.

Since only a few different ring systems are used for the majority
of the sterols. parent ion scans can be performed for those fragment
ions corresponding to the ring systems: these parent ion scans
identify most of the sterols present in the sample. In addition. the
sterols are all expected to have the same relative responses to
ionization. Therefore. the relative abundances of the peaks present in
a spectrum should correlate well with the relative abundances of the

sterols in the sample.

Isomers present a problem to this method. Some of the sterols
differ only in the position of the double bond within their ring system.

These sterols have identical molecular weights. and produce isomass

124

parent ions. Therefore. a peak in the parent ion spectrum might
represent several sterol isomers. The possibility exists for
differentiation of these isomers using daughter spectra. One of the
daughter ions formed from sterols is a result of the opening of ring A,
with subsequent cleavage of a portion of the ring. Since this
fragmentation might be controlled through the position of the double
bond(s) within ring B, different daughter spectra might be obtained
from different isomers. Research has begun in this laboratory toward

solving this problem.

A parent scan for m/z 271 from fungal sample 138-1. an
unknown fungal isolate from the Kellogg Biological Station. is shown in
Figure 6.2. This particular parent ion value (m/z 271) represents a
fragmentation from ergosterol (nominal molecular weight 396) in
which an ion is formed from the ring structure. This ion may not be
unique to ergosterol: any sterol containing a ring structure of this
mass would be detected. However. the limited information available
regarding fungal sterol content does not contain the presence of these
sterols in fungi [1]. As mentioned earlier in this chapter. ergosterol is
a sterol contained only in fungi. Therefore. this parent scan can be
used as a quick screening method for detecting the presence of fungi

in a sample.

A parent scan for m/z 273 from the same sample is shown in
Figure 6.3. This value represents a fragmentation from a multitude of
different sterols whose basic ring structure contains one less

unsaturation than that shown in Structure 6.4. Each of the peaks

125

 

396

 

 

 

 

100-
80‘
8 90-
§
<
2
'8 ‘°‘
0
m
20‘
0*‘1 *r‘r‘r*4*:fia‘—ét ,fi*fif
300 350 400 450

 

Figure 6.2 Parent scan for m/z 271 from fungal sample 138-1.

 

398
100 ‘

80"

60‘

302 316 346
40* 426

Relative Abundance

 

20- 330 959 999
369 414

 

 

 

 

300 . o o 450
Mass/Charge

 

Figure 6.3 Parent scan for m/z 273 from fungal sample 138-1.

present in this spectrum are due to one or more sterol species. Four
of the peaks have been tentatively identified: the peak at m/z 386 is
cholesterol: the peak at m/z 398 could be episterol. fecosterol.
brassicasterol. or a combination of these: the peak at m/z 414 is
sitosterol: and the peak at m/z 426 could be stigmasterol. fucosterol,
or a combination of the two. The remainder of the peaks have not yet
been identified. but their patterns in the mass spectrum can still be
used for identification. A series of different parent ion scans can be
used that cover a broad range of sterol species. Research is underway
in this laboratory in using these sterol profiles for detection and

identification of fungi.

Other Potential Lipid Biomarkers

The research presented in this-thesis has focused on three lipid
biomarkers: phospholipids. glycolipids. and sterols. These are the
lipids most common to microorganisms, and have the broadest
applications. However. many other types of lipids exist that could be
very useful for microbial detection and identification. These lipids
range from being useful for a relatively wide range of organisms. to
being specific for a particular group, or even a particular species. of
microorganisms. Three of these lipids. the isoprenoid quinones.
capnoids, and isopranyl ethers. are of particular interest to this
research due to their applicability in identifying microorganisms.
Developing the methodologies for their detection is the next goal in

this research.

lsoprenoid quinones have shown potential value as
chemotaxonomic markers for bacterial classification [6), and as such.
they are of interest to this research. The isoprenoid quinones are
involved with electron-transport and oxidative phosphorylation: thus.
they are associated with the plasma membranes of bacteria and the
photosynthetic membranes of phototrophic organisms. Because
quinones are widely distributed throughout all microorganisms. they
could be used for general identification. Both the classes of quinones.
and the different structural details contained within each class. vary
among microorganisms and can provide significant information for
microbial identification. In addition. certain types of quinones can be
specific for individual species of microorganisms. These could be used
as biomarkers for the detection of specific organisms in complex

communities containing many microbial species.

The capnoids, a sub-group of the sulphonolipids. occur only in
gliding bacteria. and can serve as biomarkers for the presence of these
bacteria [7]. As mentioned in Chapter 5. biomarkers specific to small
groups of organisms can be very useful when searching a database by
narrowing the focus of the search. In addition, these lipids have many
variations and can be used in the identification of the microorganisms
that contain them.

Isopranyl ether is a general term for any of the acyiglycerols.
such as phospholipids or glycolipids. in which the fatty acids are
attached to the glycerol core through ether-linkages. Like the
capnoids, they occur in a defined group of bacteria - in this case the

129

archaebacteria - and can be used for detection and identification of
this group of bacteria [1]. The archaebacteria occupy environments of
extreme conditions. such as high temperatures. high salt content, or
high acid content. Because their major lipids consist of isopranyl
ether lipids. these biomarkers will be very important when attempting
to identify this type of organism. The information regarding the
classes of isopranyl ethers present in a sample can be obtained in the
same way as that for the phospholipids and glycolipids. since they
share the same head groups and general structures. However,
obtaining the structural information regarding the ether-linked fatty
acids might present a problem to low-energy CID methods. This
requires further investigation.

The biomarkers mentioned- in this thesis are by no means a
complete list. Many other biomarkers exist that can be useful. at some
level, for the detection and identification of microorganisms. Tandem
mass spectrometry is emerging as a promising technique for accessing
this information. The outlook for this technique in the field of
microbial taxonomy is very bright indeed. Many newly discovered
microorganisms do not fit into any of the taxonomic catagories that
exist today. New data obtained on the molecular and genetic levels are
forcing many microorganism out of their old categories. New
taxonomic systems, based on chemical. molecular. and genetic
similarities and differences. are needed. Tandem mass spectrometry

promises to be at the forefront of this field.

13)

References

 

1. Ratledge. C.: Wilkinson, S. G. Microbial Lipids. Volume 1.
Academic Press: San Diego. 1981: Chapter 2.

2. Jones, M. G.; Mallet. A. I. Microbios Lett.. 1983. 24, 2841.

3. Howell, S. A.: Noble. W. C.: Mallet. A. I. Rapid Commun. Mass
Spectrom, 1989, 3, 230-232.

4. Bligh. E. G.; Dyer. W. J. Can. J. Biochem. Physiol.. 1959, 37, 911-
917.

5. Kinoshita. T.: Ohki. Y.: Nagaki. H.; Nakamura. T. Paper presented
at the 37th Annual Conference on Mass Spectrometry and Allied
Topics. May 21-26, 1989, Miami Beach. FL.

6. Collins. M. D.: Jones. D. Microbiol. Rev.. 1981. 45. 316-354.

7. Godchaux, W.; Leadbetter. E. R. Arch. Microbiol.. 1988. 150, 42-
47.

131

"‘lllllillllilillli“