- — .. 4 _ ‘.
‘ _. , 1.4 4 J - . 3-
4 x ‘ 9 l “ ‘ ’- I~ " ' r ; ’J’ . "31” what-1:
., . _ ‘ _ .h. . .4 ,. ~ ,4 .. , 4. - -_ , , . {in-n W: m.
. '4 » , l g — .4 Agar” , " l" ‘5‘" .v.
. cf _ .0, . :3
3
y
F
.
A 'V
1*" fl
“#15 Ltd-1’
:."“ Ln?“
', .1 5"
-‘ :‘J’
.u
-..

PR

M32

-.4: .
:7 “$5
«51'

mm
X
Rm

fl

:3?!
2:. '

1-HT: 9425?”

.v . cl“ LE!“
1,: - "Lt,“ .w -" 32.5.?
.1. A {I . i: J lav-g:

w:
‘3; ~‘veé‘i‘:;.; . £41246;
:3?”

4“-¢-.-a...-r.

’4

1;: Lgéf‘
$3.4. . ' 11;; x 34v:
"3&2 mgr». ‘5': r7. 4‘ :u:fi;§i; V . 7"}
‘54:: “"44“- 4 ‘

,. . A f
a - ». 1 W l- , 4.; 7w." b
I . . . , \ s , “fin. , mi 4 “E . 1-, PEI. .— «I “‘1’

un
"a
.

w.‘

1"
33.":

r
ézw

3r -
'h r. «.-

«(I

 

an:

3'
1
'

 

 

-‘y1‘-

n

 

 

(at “(K-(WHOM? Hut .I I. now

V ‘—
05,- _
, . ,_ - .

'3

 

. .. _.
"‘ "15‘ 2.3.!" v-01 .4:
[.7

_ . ::
. $4144“; .
. ‘13.,“ 1,.
"I

 

35;.“ '.

MICHIGAN STATE UN IES

l u 1mm / lI/uI/‘Illll .‘llllllll

3 1293 00898 2138

ll

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled

ELECTRON CAPTURE NEGATIVE IONIZATION MASS SPECTROMETRY OF
STEROIDS AND SIMILAR ELECTROPHILIC COMPOUNDS

presented by

Helen Kathe Mayer

has been accepted towards fulfillment
of the requirements for

 

 

Ph . D . degree in Chemistry

/ Major professor

  

 

MSU is an Aflirmutiw Action/Equal Opportunity Institution 0-12771

 

LIBRARY
Michlgan State
Unlverslty

 

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU lo An Affirmetlvo Action/Equal Opportunlly Inotnution
em

 

M31“

 

ELECI‘RON CAPTURE NEGATIVE IONIZATION MASS SPECTROMETRY OF
STEROIDS AND SIMILAR ELECTROPHILIC COMPOUNDS
By
Helen Kathe Mayer

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

DOCTOR OF PHILOSOPHY

Department of Chemistry
1991

 

ABSTRACT

ELECTRON CAPTURE NEGATIVE IONIZATION MASS SPECTROMETRY OF
STEROIDS AND SIMILAR ELECTROPHILIC COMPOUNDS

By
Helen Kathe Mayer

Electron capture negative ionization mass spectrometry (ECNI-MS) is a
sensitive and selective technique for highly halogenated compounds or those
containing tt-conjugation. Extended conjugation is particularly important for the high
ECNI responses of certain ketosteroids. Combinations of functional groups within the
steroid nucleus have been correlated to ECNI response. Because steroids Often have
other potential electron-stabilizing substituents, monocyclic, bicyclic, and quinone
compounds also have been studied to evaluate individual Structure-response effects.
Steroids that are chemically oxidizable to molecules with extended conjugation have
been successfully used assayed by ECNI-MS.

Combinations of double bonds, carbonyl groups, halogens, cyano groups, and
epoxides within the steroid nucleus have been studied. For enhanced ECNI response,
the following structural features are important: a cross-conjugated system with a
neighboring group, such as l,4—dien-3,11-dione; a linearly-conjugated system, such as
4-en-3,6-dione; a fluorine atom alpha to a carbonyl group in a highly-conjugated
steroid; or an epoxide alpha to a carbonyl group.

Sensitivity of the ECNI-MS technique is dependent on instrumental parameters
as well as the electron-capture properties of the molecule. The choice of reagent gas,
source temperature, and source pressure are important factors in ECNI response and
can differ for different types of compounds. To capture an electron, a compound must
have a positive electron affinity. Electron affinities for most compounds studied are

not known; estimates were obtained by measuring the reduction potential.

Some steroids, usually steroidal drugs, can be chemically oxidized to effective
electron-capturing species. Not only are chromatographic properties improved, but a
degree of selectivity is imparted because these oxidized products have a higher ECNI
response than the oxidized product of endogenous steroids. Pyridinium
chlorochromate has been the oxidant of choice, but alternative methods, including
electrochemical synthesis, were studied.

Several assays for steroids were developed using the chemical oxidation
technique. These include the determination of dexamethasone in human plasma and
the determination of prednisolone in mouse plasma. Anabolic steroids do not oxidize
to highly electron-capturing species, except for fluoxymesterone; preliminary studies
Show that conventional ECNI derivatives like heptafluorobutyrate may be the best

choice for activating this group of compounds.

 

ACKNOWLEDGMENTS

I would like to acknowledge the faculty of the Department of Chemistry,
especially Dr. J. T. Watson for support and advice, Dr. J. Gaudiello for assistance with

the electrochemical experiments, Dr. R. Hollingsworth for the use of the HPLC and

' UV-Vis, Dr. J. Funkhouser for the use of the polarograph, and Dr. W. Reusch for his

gift of samples and his vast knowledge of organic chemistry. A special thanks to the
following members of the staff of the MSU-NIH Mass Spectrometry Facility for the
advice and assistance they have given me over the years: Doug Gage for his JEOL
expertise and his patience with after-hours phone calls, Mel Micke for fixing my
computer, Mike Davenport for aligning the filament countless times, Bev Chamberlin
for helpful hints with derivatization, and Melinda Beming for help with word
processing. I would like to thank all the Watson group members, past and present, for
their help and encouragement. Thanks to Ellen Yurek for her help in the lab, to Jennifer
Johnson for cleaning sources, and especially to Kathleen Kayganich for teaching me the
principles of mass spectrometer operation as well as extraction and oxidation
techniques. Thanks to Kate Noon for her careful proofing of this manuscript. I would
like to thank the friends who made my journey through graduate school a lot easier,
especially Steve and Judy Mullen, Doug Motry, and Jessalin Faulkner. To my family,
my sisters Linda and Heidi and my parents Michael and Elisabeth, I owe immeasurable

thanks for their love, patience, and understanding.

 

TABLE OF CONTENTS

List of Tables ........................................... ix

List of Figures ......................................... xiii

List of Abbreviations ..................................... xvii

Chapter 1: Introduction .................................... 1

Formation of Negative Ions in Mass Spectrometry ..................... .3

Types of Reactions .................................... 3

How are Thermal Electrons Formed? ......................... .5

What Happens After Electron Capture? ........................ 6

Processes Directly Following Electron Capture ................ 6

Stabilization vs. Autodetachment vs. Dissociation .............. 6

Mechanisms Involving Negative Ions ...................... 8

The Formation of "Unusual" Ions ....................... 12

Molecular Parameters Affecting Ion Formation ................... 16

Electron Affinity ................................. 16

Ion Lifetimes ................................... 18

Experimental Parameters Influencing Ion Abundances ............... 19

Source Temperature ............................... 19

Source Pressure .................................. 20

Reagent Gas .................................... 21

Concentration ................................... 25

Electron Energy ................................. .25

Emission Current ................................. 25

Lens Potentials .................................. 25

Applications of ECNI-MS ................................... 26

Selectivity ......................................... 26

Sensivity ......................................... .27

Types of Compounds .................................. 28

Electron Capture Response Studies .............................. 28
Electron Capture Negative Ionization Mass Spectrometry Response

Studies ........................................... 28

GC-ECD Response Studies ............................... 30

The Analysis of Steroids Using ECNI-MS ......................... .34

Analytical Techniques for Steroid Analysis ..................... 34

The Analysis of Corticosteroids Using Mass Spectral Techniques ........ 38

Electron Capture Negative Ionization Mass Spectrometry ............. 39

Use of the Oxidation Methodology .......................... 41

Objectives of This Study .................................... 45

References ............................................ 48

 

Chapter 2: Electrochemical Studies of Steroids .................... .53

Electrochemical Synthesis of Ketosteroids ......................... 54
Electrosynthesis on Planar Electrodes ........................ .55
Experimental ................................... 56
Preliminary Cyclic Voltammetry ........................ 59
Electrosynthesis .................................. 62
Bulk Coulometry of Prednisolone ....................... 66
Electrochemistry of Steroids Other than PrednisolOne ........... 69
Why Is the Electrosynthesis of Steroids So Inefficient? ........... 7 1
Probable Methods to Overcome Fouling ................... 71
Desorption Off an Electrode-Direct Probe .................. 71
Electrochemical Deposition on a Surface--FAB ............... 76
Modified Nickel Electrodes: An Alternate Synthesis Method .......... 77
Experimental ................................... 78
Results ....................................... 80
Conclusions ........................................ 81
The Determination of Reduction Potentials ......................... 81
Cych Voltammetry . . . . . . ............................. 85
Experimental ................................... 85
Re stsul ....................................... 86
Differential Pulse Polarography ............................ 87
Experimental ................................... 88
Results ....................................... 88
Discussion ..................................... 90
Conclusions ........................................ 94
References ............................................ 95
Chapter 3: The Analysis of Anabolic Steroids by ECNI-MS ............. 97
Profile of Steroid Abuse .................................... 98
Why Take Steroids? ................................... 98
Why N91 Take Steroids: Possible Side Effects ................... 99
Who are the Users of Anabolic Steroids? ...................... 100
What Steroids are Most Commonly Abused? .................... 100
Commonly Used Detection Techniques ........................... 102
Radioimmunoassay ................................... 102
High Pressure Liquid Chromatography ....................... 102
Gas Chromatography/Electron Impact Mass Spectrometry ............ 103
Extraction and Hydrolysis of Urine ...................... 103
Derivatization .................................. 106
Electron Capture Negative Ionization Mass Spectrometry ................ 107
Commonly-Used Electron Capture Derivatization Agents ............ 108
Preliminary Results ................................... 1 l 1
TMS Derivatives of Anabolics ........................ 111
Selected Fluorinated Derivatives ....................... 113
Conclusions from Preliminary Studies .................... 114
Future Work ....................................... 116
Proposed Work with Conventional Derivatization ............. 116

applications of Chemical Oxidation to the Analysis of Anabolic

ids ......................................

Fast Chromatography .............................. 118
Validation of Analysis ............................. 119
References ............................................ 120

vi

 

Chapter 4: ECNI-MS Structure-Response Study ................... 124
Determination of Optimum Instrumental Conditions ................... 126
Experimental .......................................... 135
Materials ......................................... 135
Internal Standard .................................... 135

Gas Chromatography-Mass Spectrometry ..................... 137

Gas Chromatography .................................. 138
Sample Preparation ................................... 139

PCC of Selected Steroids ............................... 139
Preparation of Derivatives ............................... 139
Determination of Concentration Using Effective Carbon Number ....... 141
ECNI Response Studies of Various Classes of Compounds ............... 145
Preliminary ECD Study ................................ 145
ECNI Responses of Ketosteroids ........................... 146
ECNI Fragmentation of Ketosteroids ........................ 152
ECNI Responses of Epoxysteroids .......................... 153
ECNI Responses of Halogenated Steroids ..................... 155
ECNI Relative Responses of Derivan'zed Steroids ................. 157
ECNI Spectra of Bicyclics ............................... 161
ECNI Responses of Bicylic Compounds ...................... 166
Relative Responses of Cyclohexanones ....................... 168
ECNI Mass Spectra of Quinones ........................... 169
Relative Responses of Quinones ........................... 171
Relative Responses of Arene Epoxides ....................... 175
Relative Responses of Mycotoxins and Trichothecines .............. 176
Other ECNI Studies ...................................... 178
[M-I-Il' Formation: An Indicator of Low Response? ............... 178

Size Comparison Study ................................ 183
Comparison of Reduction Potentials to ECNI Responses ............. 183
Ketosteroids ................................... 183

Bicyclic Compounds .............................. 184
Comparison of UV Spectra to ECNI Response ................... 185
Reproducibility ..................................... 187
Reagent Gas Study ................................... 189
Comparison of Ionization Techniques ........................ 193
Conclusions ........................................... 195
References ............................................ 197

Chapter 5: Methodology for the Determination of Steroids in Biological
Fluids ...................................... 200

Alternative Oxidation Methods ............................... 201
Determination of Corticosteroids in Newts ........................ .203
The Determination of Dexamethasone in Human Plasma ................ 205
Experimental ...................................... 206
Changes to the Oxidation Methodology ....................... 209
The Contamination Problem ............................. 214
Practice Standard Curves ............................... 215
Dexamethasone Precision Study ........................... 217
The Interference of 6B—Hydroxydexamethasone .................. 222
Post-Dexamethasone Nighttime Study ....................... 223
Dexamethasone Precision and Patient Study .................... 224
Conclusions ....................................... 229

vii

 

no"

Determination of Prednisolone in Mouse Plasma ..................... 230

Experimental ...................................... 232
Preliminary Study ................................... 234
Mouse Plasma Sample Set #1 ............................ .235
Possible Interferences ................................. 235
Mouse Plasma Sample Set #2 ............................ .237
Conclusions ....................................... 240
References ........................................... 241
Chapter 6: Conclusions ................................... 243
References . . . .’ ....................................... 247

 

Table 1.1

Table 1.2

'Table 1.3

Table 1.4

Table 2.1

Table 2.2

Table 2.3

Table 2.4
Table 2.5
Table 2.6
Table 2.7
Table 2.8
Table 2.9

Table 2.10

Table 2.11

LIST OF TABLES

Lifetimes of Molecular Anions (adapted from reference 17) ...... 18

Electron Thermalization rates (Kf), number of collisions required
for vibrational quenching (Z), and ionization potentials (IP) for
various reagent gases (adapted from reference 42) ............ 21

Relative Response of the Electron-Capture Detector to some
Haloacyl Derivatives (adapted from reference 63) ............ 31

Electron Absorption Coefficients Of Steroids (adapted from
reference 64) ._ ................................. 33

Accessible Potential Range for Some Solvents (Adapted from
reference 10) .................................. 6O

Accessible Potential Ranges in Dimethylsulfoxide Containing
Different Supporting Electrolytes (0.1 M) (Adapted from reference
10) ........................................ 60

Oxidation Potentials of Various Steroids in Acetonitrile. N. B.:
All oxidation potentials were obtained using a quasi-reference
electrode. Because these data were obtained on different days and
no internal standard was used, the values cannot be directly

compared. .................................... 61
UV Data for Selected Steroids in Methanol ................ 65
Reverse Phase HPLC of Electrochemical Solution ............ 66
Suspected Identity of HPLC Fractions ................... 70
Literature Values for the Reduction Potentials of Ketosteroids ..... 83
Reduction Potentials of Steroids in Acetonitrile .............. 8 6

Comparison of Experimental and Literature Half-Wave
Potentials. ................................... 89

Reduction Potentials Using the DMF/I-Izo Solvent System. NB:
-2.35 V is the maximum reduction potential ................ 91

Polarography of Steroids Using Acetonitrile. NB: -2.97 V is the
maximum reduction potential ........................ 92

ix

Table 3.1
Table 3.2

Table 3.3
Table 3.4

Table 3.5

Table 3.6
Table 4.1
Table 4.2
Table 4.3
Table 4.4
Table 4.5
Table 4.6
Table 4.7
Table 4.8
Table 4.9
Table 4.10
Table 4.1 1

Table 4.12

Table 4.13
Table 4.14
Table 4.15
Table 4.16
Table 4.17
Table 4.18
Table 4.19

IOC Forbidden Anabolic Steroids (Adapted from reference 6) . . . . 101

Urinary Metabolites of Anabolic Steroids (references in
parenthesis) .................................. 104

Anabolic Steroids Excreted in Urine* (from reference 17) ....... 106
ECD Relative Responses of TMS Derivatives of Anabolic Steroids
1 12

ECD Relative Response of Selected Anabolic Steroid

Derivatives .................................. l 1
Relative ECNI Responses of Oxidized Anabolic Steroids ....... 118
Chromatography Under ECNI Conditions ................ 134
Steroids Prepared by Chemical Oxidation ................ 140
Contributions to the Effective Carbon Number ............. 142
Estimated Concentration Using the ECN Method ............ 143
Concentrations Estimated by ECN Method ............... 144
Relative ECD Responses of Steroids ................... 145
ECNI Relative Responses of Steroids ................... 147
ECNI Relative Responses of Epoxysteroids ............... 148
ECNI Responses of Fluorinated Steroids ................. 149
Relative ECNI Responses of Dexamethasone Derivatives ....... 158
Relative ECNI Responses of Fluorinated Dexamethasone
Derivatives .................................. 159
Relative ECNI Responses of 5a—Androstane-3,17-diol
Derivatives .................................. 160
ECNI Mass Spectra of Bicyclic Compounds ............... 164
ECNI Relative Responses of Bicyclic Compounds ........... 167
Relative ECNI Reponses of Cyclohexanones .............. 169
ECNI Spectra of Quinones ......................... 171
Relative Responses of ECNI-MS and GC-ECD of Quinones ..... 172
ECNI Relative Responses Arene Epoxides ................ 177
ECNI Relative Responses of Mycotoxins and Trichothecines ..... 177

X

Table 4.20
Table 4.21
Table 4.22
Table 4.23
Table 4.24
Table 4.25

Table 4.26
Table 4.27
Table 5.1
Table 5.2
Table 5.3
Table 5.4

Table 5.5
Table 5.6

Table 5.7

Table 5.8

Table 5.9

Table 5.10

Table 5.11

Table 5.12

Steroid Epoxides with Molecular Anions ................. 181
Steroid Epoxides with [M-Hl' Ions .................... 181
Size Comparison Study ........................... 183
UV Spectra of Steroids ........................... 186
Reagent Gas Study .................. I ............ 191

ECNI spectra of 1,4-Androstadien-9a-fluoro-16a-methyl-3,11,17-
trione (Oxidized Dexamethasone) Using Different Reagent

Gases ...................................... 193
Comparison of Various Ionization Techniques ............. 194
Comparison of ECNI Relative Responses to Literature Values . . . .196
Comparison of FCC and TPAP Oxidations ............... 202
Isotope Peaks of the Molecular Anion of Flumethasone ........ 213
Unknowns from Practice Standard Curve ................ 216
Unknowns from Practice Standard Curve with Less Internal

Standard (all concentration values are in ng/mL) ............ 216
Unknowns from Practice Standard Curve ................ 217

Dexamethasone Precision Study--RIA Values for Replicate
Analyses of the Indicated Sample Pools (all concentrations in
ng/mL) ..................................... 220

Comparison of Dexamethasone Concentrations Experimentally
Obtained by RIA and ECNI-M MS All results expressed in
ng/mL ...................................... 220

Dexamethasone Precision Study--ECNI-MS Values for Replicate
Analysis of the Indicated Sample Pools (all concentrations in
ng/mL) ..................................... 221

Post-Dexamethasone Nighttime Study. Experimentally
Determined Value for Dexamethasone in Plasma by Indicated
Methods at the Indicated Times ...................... 225

Precision Study for Dexamethasone. Values for the Replicate
Analysis of the Indicated Sample Pools Expressed in nymL ...... 226

Dexamethasone Patient Study. Comparison of RIA and ECNI-MS
Experimental Results ............................. 227

Estimated 6B—Hydroxydexamethasone Concentrations for Patient
Samples .................................... 228

 

Table 4.20
Table 4.21
Table 4.22
Table 4.23
Table 4.24
Table 4.25

Table 4.26
Table 4.27
Table 5.1
Table 5.2
Table 5.3
Table 5.4

Table 5.5
Table 5.6

Table 5.7

Table 5.8

Table 5.9

Table 5.10

Table 5.11

Table 5.12

Steroid Epoxides with Molecular Anions ................. 181

Steroid Epoxides with [M-H]' Ions .................... 181
Size Comparison Study ........................... 183
UV Spectra of Steroids ........................... 186
Reagent Gas Study .................. i ............ 191

ECNI spectra of 1,4-Androstadien-9a-fluoro—160t-methyl-3,11,17—
trione (Oxidized Dexamethasone) Using Different Reagent

Gases ...................................... 193
Comparison of Various Ionization Techniques ............. 194
Comparison of ECNI Relative Responses to Literature Values . . . . 196
Comparison of FCC and TPAP Oxidations ............... 202
Isotope Peaks of the Molecular Anion of Flumethasone ........ 213
Unknowns from Practice Standard Curve ................ 216
Unknowns from Practice Standard Curve with Less Internal

Standard (all concentration values are in ng/mL) ............ 216
Unknowns from Practice Standard Curve ................ 217

Dexamethasone Precision Study--RIA Values for Replicate
Analyses of the Indicated Sample Pools (all concentrations in
ng/mL) ..................................... 220

Comparison of Dexamethasone Concentrations Experimentally
Obtained by RIA and ECNI-MS. All results expressed in
ng/mL ...................................... 220

Dexamethasone Precision Study--ECNI-MS Values for Replicate
Analysis of the Indicated Sample Pools (all concentrations in

ng/mL) ..................................... 221
Post-Dexamethasone Nighttime Study. Experimentally
Determined Value for Dexamethasone in Plasma by Indicated
Methods at the Indicated Times ...................... 225

Precision Study for Dexamethasone. Values for the Replicate
Analysis of the Indicated Sample Pools Expressed in ng/mL ...... 226

Dexamethasone Patient Study. Comparison of RIA and ECNI-MS
Experimental Results ............................. 227

Estimated 6|3—Hydroxydexamethasone Concentrations for Patient
Samples .................................... 228

 

Table 5.13

Table 5.14

Determination of the Amount of Prednisolone in Mouse Plasma
Sample Set #1 ................................ 235

Determination of the Concentration of Prednisolone in Mouse
Plasma Sample Set #2 ........................... 239

 

LIST OF FIGURES

Figure 1.1 Potential energy curves for negative ion formation. (a) Ion-pair
formation. (b) Dissociative electron capture. (0) Electron
attachment, EA of MX < 0. (d) Electron attachment, EA of MX >
0. (from reference 17). ............................. 7

Figure 1.2 Mechanisms of formation of unusual negative ions (from reference
8) ......................................... 14

Figure 1.3 Single ion reconstructed chromatograms of the methane ECNI-MS
analysis of floranil. The peak at m/z 180 is the molecular anion;
m/z 162 corresponds to [M+H-F]'; m/z 142 corresponds to
[M-HFT (from reference 8) .......................... 15

Figure 1.4 Steroid numbering system. . . Q ...................... 35
Figure 1.5 ECNI mass spectrum of pentaf‘ L y‘ ime, TMS—ether

derivative of testosterone. The base peak at m/z 181 represents the
pentafluorobenzyl anion (from reference 93) ................ 40

 

Figure 1.6 The oxidized product of 6B-hydroxycortisol is 4-androsten-
3,6,11,17-tetraone. A sample of urine was spike with 615-
hydroxycortisol, extracted, and oxidized using PCC. (a) TIC from
GC-EI-MS analysis (about 56 ng of analyte on-column). (b) TIC
from GC-ECNI-MS analysis (about 44 ng of analyte on-column).
Note the improvement in selectivity and sensitivity (adapted from
reference 69) ................................... 4 3

Figure 1.7 Prednisolone and prednisone are oxidized to the same product, 1,4—
androstadien-3,l 1,17-trione ......................... 45

Figure 2.1 One-compartment electrochemical cell used for cyclic
voltammetry. All solvent transfer is done under vacuum; the
connection to vacuum is not shown. (a) Silver wire reference
electrode, (b) Platinum working electrode, (c) Platinum counter
electrode, (d) Sample reservior (adapted from reference 14) ....... 57

Figure 2.2 Electrochemical cell used for bulk electrolysis. All solvent
transfer is done under vacuum; the connection to vacuum is not
shown. (a) Platinum counter electrode, (b) Silver wire reference
electrode, (c) Platinum working electrode, ((1) Platinum gauze
electrode, (e) Sample reservoir (adapted from reference 14) ....... 57

Figure 2.3

Figure 2.4

Figure 2.5

Figure 2.6

Figure 2.7

Figure 2.8

Figure 2.9

Figure 3.1

Figure 3.2

Figure 4.1

Figure 4.2

Figure 4.3

Cyclic voltammagram showing that electrochemical oxidation of
prednisolone is irreversible. The oxidation potential of
prednisolone is 2.3 V vs. QRE (quasi—reference electrode) ........ 62

HPLC of electrochemically oxidized prednisolone (not extracted).
Suspected identity of peaks: (1) TBAF, (2) prednisolone, and (3)
oxidized product. (a) Electrochemical oxidation product. (b)
Electrochemical oxidation product spiked with 1,4-androstadien-
3,11,17-trione. Peak 4 is the spiked compound ............... 67

Fractions collected from the HPLC analysis of the electrooxidation
product of prednisolone ............................ 67

Cyclic voltammagrams of prednisolone. (a) The current stays
approximately the same if the potential is scanned from -1.5 V to
+2.5 V. (b) However, if the potential is scanned from 0 to -2.2 V
repeatably, the amount of current decreases, proving that material
is adhering on the surface ........................... 72

A platinum electrode that fits into the direct probe of either a
I-IPS985 or JEOL 505 mass specu'ometer .................. 73

A platinum electrode that fits into the FD-FAB probe of the JEOL
I-IX-l 10 mass spectrometer .......................... 76

Mechanism for the electrochemically regenerated nickel oxide
hydroxide electrode ............................... 7 8

Steroids used for pilot study on derivatization: (a) stanozolol, (b)
methyltestosterone, and (c) fluoxymesterone .............. 113

Similarity of oral contraceptive metabolites to anabolic steroid
metabolites. (a) Major metabolite of norethisterone (an oral
contraceptive). (b) Major metabolite of methenolone (an anabolic
steroid) ..................................... 117

ECNI response as a function of source temperature. The source
pressure was held constant at 2 x 10‘5 torr. (a) The response of
octafluoronaphthalene decreases with increasing source
temperature. (b) The response of 4-androsten-3,6,l7-trione is
maximized at a source temperature of 200 °C ............... 126

ECNI response as a function of source temperature and pressure.
(a) The optimum conditions for the analysis of a low-response
compound such as l—androsten-3,17-dione were at a source
temperature of 150 °C and a source pressure of 4 x 10'5 torr. (b)
The optimum conditions for the analysis of a high-response
compound such as 1,4-androstadien-3,1l,17-trione were at a
source temperature of 250 °C and a source pressure of 1 x 10'5
torr ........................................ 128

The ratio of the peak areas of the ion at m/z 298 (from 1,4»
androstadien—3,ll,17-trione) to ion at m/z 285 (from l-androsten-
3,17-dione) as a function of source temperature and pressure. . . . . 130

xiv

Figure 4.4

Figure 4.5
Figure 4.6

Figure 4.7

Figure 4.8

Figure 4.9

Figure 4.10

Figure 4.11

Figure 4.12

Figure 4.13

Figure 5.1
Figure 5.2

Figure 5.3

Figure 5.4

Reconstructed mass chromatograms of the major ions of the three
components of the internal standard. (a) Note the tailing on the
DB- 1 30-m x 0.32-mm x 0.-25 -p.rn column. (b) The tailing is
minimized on aDB-l 30-mx0.25-mmx1.0-ttmcolumn,butthe
retention time is doubled ........................... 136

Structure (b) is a vinylog of structure (a) .................. 151

Proposed mechanism for the formation of a phenolate ion in 1,4-
dien-3-one steroids .............................. 153

Proposed mechanism for the formation of [M-2]‘ and [M-16]'
anions in the ECNI spectrum of 4-androsten-3,11,17-trione ...... 154

Fragmentation of 4-androsten-9a—fluoro—16a-methyl- -,3 11,17-
trione ...................................... 156

Structures Of the bicyclic compounds analyzed in this dissertation.
162

Fragmentation of C6H80 fiom Compound 10 results' In an ion at
m/z 138 ..................................... 165

Proposed mechanism for the formation of a more stable anion by
the loss of a hydrogen alpha to a carbonyl carbon in saturated
ketosteroids ................................... 180

Proposed mechanisms for the capture of an electron in an or-
ketoepoxide. After the opening of the epoxide ring. to relieve
strain, two rearrangements are possible. If a hydrogen 15 available
on the fl-carbon to the epoxide, a [M- H] anion will be formed.
Otherwise, the M' anion will be formed after the breakage of the
ring ....................................... 182

Half-wave reduction potentials of ketosteroids as a function of
ECNI relative response. All responses are relative to androstan-3-
one ........................................ 184

CC Chromatograms of (a) newt plasma and (b) newt limb ....... 205

Sample preparation scheme for the determination of
dexamethasone or prednisolone in plasma ................ .207

On a DB- 1 lS-m x 0. 25- -mm x 1.0-um column, the minor isotope
peak of oxidized flumethasone (top mass chromatogram) overlaps
with the major isotope peak of oxidized dexamethasone (bottom
mass chromatogram) ............................. 211

ECNI mass spectra of (a) 1,4-androstadien-6a,9a-difluoro—l6a—
methyl-3,11,17-trione (oxidized flumethasone), (b) 1,4-
androstadien-90t-fluorO-16a- -methyl- -3, 11,17-trione (oxidized
dexamethasone), and (c) 1,4-androstadien-90t-fluorq-l6a-methyl-
3, 6, 11, 17- tetrone (oxidized 6B—..,.i.....,‘ ....... 212

XV

 

Figure 5.5

Figure 5.6

Figure 5.7

Figure 5.8

Standard curve for the determination of dexamethasone in human
plasma. The equation of the line was y = 0.0646 + 0.299x with a
correlation coefficient of 1.000. The LOD (YB + 3 83) was 0.096
ng/mL ...................................... 218

Standard curve for the determination of 6B-hydroxydexamethasone
in human plasma. The equation of the line was y = 0.0605 +
0.0199x with a correlation coefficient of 0.974. . , ............ 228

The biosynthesis of andrenocorticosteroids. Adapted from Orten,
J. M.; Neuhaus, O. W. Human Biochemistry, Mosby: St. Louis,
1982, p. 618 .................................. 238

Standard curve for the determination of prednisolone in mouse

plasma. The equation of the line is y = 0.0535 + 0.0604x with a
correlation coefficient of 0.998. The LOD was 0.3 ng ......... 240

xvi

 

2292;
>

DFI'PP
DIP
DMF
DPP
DST
E1/2
EA

EC
ECD
ECN
ECNI
EFSM
EI
ELISA

LIST OF ABBREVIATIONS

Abscisic acid

Octadecahydryl

Chemical ionization

Coefficient of variance

Cyclic voltammetry

Decaflnurnu 'K ‘ I ‘r ‘ .cphinp

Direct insertion probe
Dimethylformamide

Differential pulse polarography
Dexamethasone suppression test
Half-wave reduction potential
Electron affinity

Electrochemical detection

Electron capture detector

Effective carbon number

Electron capture negative ionization
Electric field switching selected ion monitoring
Electron impact

Enzyme-linked immunosorbent assay
Ethanol .

Fast atom bombardment

Field desorption

Flame ionization detector
o-Pentafluorobenzyloxime

Gas chromatography

Gas chromatography-mass spectrometry
High density lipoprotein
Heptafluorobutrate

Hewlett Packard

xvii

 

HPLC

LOD
LUMO

MO-TMS
MS
MS/MS
MSD
MSTFA
MSU

TMCS

High performance liquid chromatography
International Olympic Committee

Limit of detection

Lowest unoccupied molecular orbital
Magnetic field switching selected ion monitoring
Methoxyamine-trimethylsilyl

Mass spectrometry

Tandem mass spectrometry

Mass selective detector
N—methyl-N-Limet. , 1.;1,‘ ifl... - 'J
Michigan State University

Negative chemical ionization
Octafluoronaphthalene

Polyaromatic hydrocarbon

Pyridinium chlorochromate

Positive chemical ionization

 

 

Perfluorokerosene
Pentafluoropropionyl
Quasi-reference electrode
Radioimmunoassay
Signal-to-noise ratio

Standard calomel electrode
Standard hydrogen electrode
Selected ion monitoring

Selected reaction monitoring
Tetrabutylammonium fluoroborate
Trifluoroacetyl

Total ion chromatogram
Thin-layer bulk electrolysis
Thin-layer chromatography
Trimethylchlorosilane
Trimethylsilylimidazole
Trimethylsilyl
Tetra-n-propylammonium perruthenate
Ultraviolet

xviii

 

Chapter 1: Introduction

Detection of small amounts of analyte in complex matrices is an important
problem in analytical chemistry, especially with the current emphasis on
environmental and biological samples. In the biological and environmental sciences,
for example, the problem becomes exceedingly difficult because of the amount and
variety of interfering compounds which are often unknown to the investigator.
Sensitive and selective methodology for sample analysis needs to be further developed
to detect small amounts of analyte with a minimum of sample preparation.
Confirmation of the analyte of interest is another problem. Proving with certainty the
presence or absence of a component is critical in many analyses in both the
environmental and biomedical fields. In environmental samples, the absence of a
component may mean a costly and time-consuming clean-up of a ecosystem could be
avoided. In the biomedical field, the absence or presence of a substance may mean the
confirmation of a disease, the success of a treatment, or the detection of the use of
illicit drugs.

An example of an analysis problem in the biological sciences is the detection of
drugs in urine or plasma. High pressure liquid chromatography (HPLC), gas
chromatography (GC), and immunoassays have been commonly used. Although they
can confirm the absence or presence of a drug, specificity can suffer, especially for
structurally-similar compounds. Often, all interfering compounds must be removed in
time-consuming extraction steps. Also, a chromatographic peak provides only
retention time data as a parameter for comparison to standards. Gas chromatography-
mass spectroscopy (GC-MS) is a superior technique for trace analysis because it also
provides structurally-related information along the mass axis.

Electron-impact mass spectrometry (El-MS) is the most popular identification
technique. For many classes of compounds, the molecular ion is small or absent.

Major fragmentation peaks within a class of compounds are often similar in m/z values

1

2
and cannot be used for quantitation. Chemical ionization (CI) is often used to obtain
complementary fragmentation patterns to those obtained by EI. CI spectra can provide
additional structural information in the form of a molecular ion or an adduct ion due to
adduct formation with the reagent gas (1).

For qualitative analysis, fragmentation patterns provide a "fingerprint" for the
analyte of interest. For quantitative analysis, fragmentation is often more a burden
than a blessing. The most abundant ion in the spectrum may not be the one which
distinguishes the analyte from the matrix. Molecular ions of low abundance make
detection of small amounts of analyte difficult.

Electron-capture negative ionization mass spectrometry (ECNI-MS) has been
utilized in quantitative analysis because less fragmentation is obtained. Because
negative ions usually have low internal energies, they are unable to fragment (2). The
most abundant ion in the spectrum is usually the molecular anion. In the mass spectra
of steroids, the molecular ion is often the base peak in the spectrum, although losses of
radicals such as hydrogen and CH3 and small neutrals like HF have been Observed. In
selected ion monitoring (SIM), ion currents at a few m/z values are transmitted to the
detector. By using the SIM mode in ECNI and selecting only the molecular anion,
small amounts of analyte can be detected. The sensitivity of ECNI is also impressive.
In fact, the formation of a molecular anion under ECNI conditions is about 400 times
more rapid than the formation of [M+H]+ under positive chemical ionization (PCI)
conditions or the formatiOn of M" under negative chemical ionization (N CI)
conditions (3). The reaction rate for gas phase PCI and NCI reaction is about 10'9
cm3s". For electron capture, the rate constant is 4 x 10'7 cm3s'l, or about 400 times
greater.

Another problem plaguing the analysis of biological samples is the high
background from the matrix; for example, the presence of endogenous steroids in the

analysis of urine can cause interference in the detection of synthetic corticosteroids.

3
Corticosteroids are substituted pregnanes, some of which occur naturally in the adrenal
cortex. ECN I is a selective technique that responds best to highly electrophilic
species, i.e., highly halogenated or highly conjugated (mB—unsaturation) compounds.
Compounds which have negative electron affinities will not respond, eliminating much
of the background and often simplifying sample clean-up.

Many electrophilic compounds have been analyzed via ECNI, especially in the
biological and environmental sciences for quantitative analysis. Polychlorinated
compounds and pesticides (4), aromatics (5), derivatized amino acids (6), mycotoxins
in food (7), amines (8), prostaglandins (9), pharmaceuticals (10), and steroids (11)
have been analyzed using this technique.

Formation of Negative Ions in Mass Spectrometry
Types of Reactions
Negative ions can be formed in a mass spectrometer ion source by the
following mechanisms:
(1) AB + e' (~0.l eV) —-> [AB']* —) AB" Resonance capture
(2) AB + e' (0-15 eV) -—) [AB‘]* -—) A + B' Dissociative resonance capture
(3) AB + e' (> 10 eV) —) A' + B+ + e' Ion-pair formation
(4) AB + C' —9 AB‘ + C Negative ion-molecule reaction
Resonance electron capture is the most useful of all these pathways because the

molecular anion predominates. Resonance capture requires a thermal electron (an

electron of energy 0.1 eV or less). The anion is produced in the vibrationally excited

4
state, [AB']*, which is dependent upon molecular structure and electron energy. If the
excess vibrational energy is not removed or delocalized, dissociation will occur. The
molecular anion can be stabilized by delocalization of charge via a conjugated system,
collisional stabilization, or emission of infrared photons.

Dissociative electron capture occurs when an electron of 0 to 15 eV is captured.
Direct rapid cleavage or molecular rearrangement (12) can occur. This mode is not as
useful in quantitative analysis as the resonance capture mode. In resonance capture, all
of the ion current is concentrated in the molecular anion. The ion current in
dissociative resonance capture is shared between two or more fragments, resulting in a
decrease in detectability.

Ion-pair formation occurs when the electron has over 10 eV of energy. Both a
positive and a negative ion are formed, with the electron carrying away the excess
energy. In an ion source operated under EI conditions, the ion-pair formation
mechanism predominates. The majority of negative ions formed are of low mass and
are not indicative of the parent molecule (e. g., OH', CN', N02", etc.); consequently,
there is little molecular ion information unless the anion is unusually long-lived.
Negative ion formation under low pressure and high electron energy conditions have
been reviewed by Bowie (2) and Budzikiewicz (13).

Negative ion-molecule reactions truly deserve the designation of negative
chemical ionization (NCI) because a negative ion in the source (e. g., H', OH‘, CH3O')
reacts with the analyte. Many reactions can occur, including charge transfer, proton
transfer, hydride transfer, oxygen exchange with either halogens or hydrogen, and
anion-molecule adduct formation (14).

Of most interest to this work are those reactions which occur under ECNI

conditions, namely resonance and dissociative capture.

 

How are Thermal Electrons Famed?

In ECNI, a high~pressure chemical ionization (CI) source is used. A
nonreactive reagent gas (such as methane, ammonia, isobutane, etc.) is introduced into
the source at the pressure of 0.1 to l torr and bombarded by a beam of electrons,
generating positive ions and thermal electrons. Using methane as an example, the

following reactions can occur ( 15):

CH4 + e'pfim —> CH4+' + e'primary + {secondary
CH4+' + CH, —) CH5+ + CH3'

esecondary + CH4 ‘9 eRhea-ma] + CH4

The ionizing energy is about 200 eV to ensure good penetration of electrons
into the high pressure source. Ionization of methane removes about 30 eV from the
bombarding electron while more energy is removed from the electron by nonionizing
collisions with methane (14). One primary electron can potentially produce about six
to ten ion-electron pairs. The positive ions formed undergo other collisions, such as
those to form CH5+. The secondary electrons collide with methane, producing the
thermal electrons, which can be captured by high electron affinity analytes. The high
pressure gas also provides stabilization of the molecular anion.

A low-energy electron distribution is formed by the modulating gas. This
distribution varies from instrument to instrument and can cause differences in ECNI
spectra. In one particular experiment, it was shown that 48% of the low-energy
electrons had energies near thermal, 5% had energies less than 0.15 eV, and 12% had
energies ranging from 0.15 to 0.45 eV (16).

 

What Happens Afier Electron Capture?

Wynn. After the capture of a thermal
electron, an unstable molecular anion [AB']* is formed. The processes that this
unstable anion can undergo are the following: stabilization by collisions with the
modulating gas or radiative emission, dissociation into A' 4- B, or autodetachment of
the electron. After forming the negative anion, reactions which can neun’alize or
change the anion can occur, including associative detachment of the electron,
displacement of certain functional groups on the molecule, proton transfer, charge
exchange, and association (clustering) (17). In addition to these one-step reactions,
Sears and cO-workers (8) have proposed four mechanisms to explain the formation of
unusual ions (see Figure 1.2).

Wagon. Stabilization of the anion,
dissociation, and autodetachment of the electron are the three primary mechanisms by
which the unstable intermediate [AB']* forms stable products.

If excess internal and translational energy are present, the capture of an
electron can lead to the process depicted in Figure 1.1 (b). For dissociative capture
reactions, energy maxirna are determined by energy differences between unoccupied
orbitals, the potential energy of the molecule, and the Franck-Condon factors
describing the interaction between the molecule and the dissociating pair (18).

Electron capture can occur by two methods. If the electron affinity (EA) is
negative (Figure 1.1 (c)), then the extra electron on intermediate [AB']* may
autodetach or the intermediate may dissociate to A' + B' if the Franck-Condon
transition is above the dissociation limit. The autodetachment process is quite rapid;
therefore, these compounds rarely form detectable negative ions. Rapid
autodetachment leaves little possibility for collisional stabilization unless the source

presure is very high. Stabilization by radiative emission is also improbable.

MX*
MX , _ _
M+X MX MX
M+X WX
M+X'
MX‘ _
MX M+X MX W M“
M+X' M+Xn

(c) (d)

Figure 1.1: Potential energy curves for negative ion formation. (a) Ion-pair
formation. (b) Dissociative electron capture. (c) Electron attachment, EA of MX < 0.
(d) Electron attachment, EA of MX > 0. (from reference 17).

 

8

When the electron affinity is positive (Figure 1.1 (d)), two types of electron
capture can occur. If the potential energy curve for AB‘ intersects the vertical Franck-
Condon region, the reactions are similar to those discussed for Figure 1.1 (c). If the
potential energy curve does not intersect the Franck-Condon region and the excess
energy of the unstable intermediate [AB']* can be distributed into vibrational modes,
the lifetime of the molecular anion may be such that collisional stabilization can occur
and it can be observed. This is a resonance process because no electron is produced to
carry away the excess energy.

If the unstable molecular anion has a positive electron affinity and an
incompletely filled molecular orbital of sufficiently low energy, three mechanisms are
possible--stabilization, dissociation, or autodetachment. If the unstable molecular
anion is long-lived and has excess energy equal to the electron affinity of the molecule
itself, then either dissociation or stabilization may occur. Stabilization occurs both
collisionally with the reagent gas and by radiative emission. At gas pressures of about
one torr, the lifetime of the [AB']* intermediate must be above one microsecond for
effective collisional stabilization and for detection in a mass spectrometer.

W5. After a negative ion is formed by either
resonance or dissociative processes, several reactions can be responsible for changing
or neutralizing the ion. These mechanisms include associative detachment reactions,
displacement reactions, charge exchange reactions, association reactions, and proton
transfer reactions. Most of the extensive research in this field has beendone in the
context of atmospheric chemisn'y rather than with species of interest to ECNI studies.

1. WW

AB' +CD—iABCD+e‘
In this case, the analyte reacts with a neutral (designated by CD) in the ion source or is
lost on the ion source walls. At times, the rate of this reaction could almost equal the

rate of formation of AB‘, resulting in the loss of signal (17).

 

9
Another source of negative ion and electron loss is in recombination reactions
with positive ions (19). Because both negatively and positively charged species are
present, the gas in the chemical ionization volume can be considered a plasma. In a
plasma, ambipolar diffusion predominates, meaning that electrons and positive ions
diffuse at the same rate if their populations and energies are equal. Ambipolar
diffusion is responsible for much of the ion and electron loss on the walls of the
source.
2. Disnlacrrnentlseactinns
AB'+CD-—>ACD+D‘ or AB'+CD—)AC'+B+D
These reactions occur if the thermodynamics are favorable and proton transfer is
unfavorable. For example, the OH' ion will displace the C1‘ in CH3C1. However, in
the reaction of OH' with CH3CN, proton abstraction is favorable and will occur rather
than the displacement of CN'. These reactions are analogous to SN2 reactions in the
solution phase, and even result in inversion of configuration (17). Reaction
efficiencies can differ greatly depending on the nature of the nucleophile and the
leaving group.
3. W
AB' + CD —) AB + CD'
These reactions can occur if the electron affinity of CD is greater than the electron
affinity of AB. It is estimated that these reactions are 30 to 100% efficient (17).
4. i . . [Cl . I E .
AB’ + R —) ABR'
These reactions are analogous to the reactions with reagent gas observed in positive CI
spectra. Most of the kinetics for these reactions have been studied in the context of

atmospheric chemistry.

 

10
5. RtntnnltansfeLReactinns
A‘ + YI-I —) AH + Y'

These reactions occur when using CI gases that are strong bases such as OH', F',
NHz', or OCH3'. Anion-adduct formation may occur but, more commonly, [M-H]'
ions are formed. This process has been used advantageously in analytical processes;
for example, the OH’ ion has been used in the negative ion detection of substituted
cholestanes and cholesteryl esters (20). The OH“ ion was formed in the ion source
using a 2:1 mixture of CH4 and N20. In the spectra of all but two of the 35 steroids
studied, [M-H]' ions were present, often as the base peak. For about one-third of the
compounds, additional ions are found at [M+43]‘, [M+25]‘, and [M+15]‘, which are
attributed to reactions of [M-HJ' with the N20 reagent gas. The [M—3]' ion was found
in compounds containing hydroxy groups or double bonds. This ion is due to H2 loss
following proton abstraction. The [M-5]' ion present in some compounds probably is
formed in the same way after loss of two H2 molecules. Other, usually low abundance
ions, such as [M-19]', [M-21]', and [M-37]', are attributed to proton abstraction
followed by losses of water, water and H2, and two water molecules, respectively. Not
many fragment ions are noted because the fragmentation of Y‘ is usually limited due to
the exotherrnicity of this reaction appearing as vibrational energy of the AH bond (21).

Often an oxygen-containing anion is net added on purpose; however, proton
abstraction can occur as a by-product of low-level contamination of oxygen-containing
species such as O', 02', or OH' in the ion source. Some molecules undergo both
electron-capture and ion-molecule reactions; for example, in the presence of oxygen,
diazepam produces a [M-H]' ion as well as a M" ion (22). The H' is lost from the
carbon adjacent to the amide carbonyl after M" is formed, resulting in a more
conjugated anion in which the charge is delocalized. It is postulated that the same

bond is reduced in polarographic analysis. In a study that compared the ECNI spectra

 

l 1
obtained on different instruments, diazepam was used to determine the presence of
oxygen in different ion sources (15).

The rate of the electron capture reaction vs. the rate of the deprotonation
reaction will determine whether M" or [M-H]' is Observed. For example, it has been
observed that the change in selectivity of polyaromatic hydrocarbons (PAHs) with
instrumental parameters is due to the competitive formation processes of [M-H]' and
M" (23). PAHs with electron affinities greater than 0.5 eV produce M" while those
with electron affinities less than 0.5 eV produce [M-H]' as their base ion in methane
ECNI spectra. Another study with PAHs has confirmed that the [M-H]' ion is from a
reaction of the PAH molecule with OH' (24). In this study, water was added to the
nin‘ogen reagent gas to produce OH‘, (HZO)OH‘, and (HZO)20H' in the ion source,
and the ECNI mass spectra of fluorene, anthracene, and fluoranthrene were recorded.
The production of [M-H]' from fluorene was enhanced, as was the production of [M-
H]' from anthracene. Fluoranthrene, however, produced only M" ions because it
rapidly captures an electron.

The ECNI spectra of PAHs are not the only cases in which the electron affinity
of the compound contributes to the formation of [M-H]' ions. Crow and co-workers
(25) studied a series of bromobenzene and chlorobenzenes with one to six substituents.
Under methane ECNI conditions, only those with a significant amount of halogenation
(pentachlorobenzene, hexachlorobenzene, tetrabromobenzene, and hexabromo-
benzene) had M" ions; the rest produced [M-H]' as their highest mass anion.

The parent structure may also have some role in the production of [M-Hl‘ ions
over M" ions. In the methane ECNI study of polychloroanisoles with three, four, or
five chlorine substitutions (26), compounds with a hydrogen ortho to a methoxy group
produced [M-H]' ions, while the others, containing ortho chlorines, produced

molecular anions.

12

The mass spectra of many other compounds display [M-H]' as their highest
mass ion, rather than M". In these studies, the cleanliness of the source with respect to
oxygen content is not known. ECNI studies of brassinosteroids (27) and barbiturates
(28) show that the spectra of both have [M-H]' ions. The ECNI spectra of many
derivatives have [M-H]' ions, including the trimethylsilyl derivatives of cytokinins
(29) and the pentafluorobenzyl derivative of 3-ketovalproic acid (but none of the other
derivatized valproic acid metabolites studied) (30).

The presence of oxygen can result in unusual ions other than [M-H]'. Many of
these have been reviewed by Budzikiewicz (13) and Stemnrler and Buchanan (24).
Chlorinated compounds can give species such as [M+O-Cl]' and [M+O-Cl-H7j‘.
PAHs produce ions corresponding to [M+O-H]‘ and [M+02]" from ion-molecule
reactions with 02". Ions corresponding to [M+O-2H]" and [M+20-2H]" are

attributed to the wall-catalyzed reactions discussed in the next section.

WWW. Much of the fragmentation of negative ions
is simple; losses such as HF, CH3, and C1 are often observed. Many other fragment
and adduct ions are not so easily explained. Competitive processes in the source can
play a large role, and these processes can often involve multistep mechanisms if the
anion is sufficiently stable. Sears and co-workers have treated this subject in great
detail (8).

The basis of the Sears, et a1. model is that there are many competing reactions
in a chemical ionization source. What is actually observed in the mass spectrum

depends on the rates of these reactions as shown by the following scheme:
M _°., M—
process X1

Mi_.e_, Mi—

 

13
In this case, the analyte molecule, M, can either capture an electron to forrn M" or can
undergo process X to form Mi, which potentially can capture an electron. If Mi has a
similar or greater electron-capturing ability as M, Mi' may appear in the spectrum or
could even be represented by the most abundant peak. If Mi has a much greater
electron-capture capability than M, process X does not have to be efficient for Mi' to
dominate the spectrum.

Four possible mechanisms have been proposed for the production of unusual
negative ions (8) as illustrated in Figure 1.2.

Route A first involves electron capture to form M". After electron capture,
neutralization by positive ions is possible. Altemately, wall reactions can occur,
forming the neutral species Z, which can potentially capture an electron. M" must be
at least as intense as Z' in the ECNI spectrum for this pathway to occur. All pathways
A through D have the potential of ending in neutralization.

Route B involves a reaction between the analyte molecule M and a positive ion
to form M+' (this ion does not necessarily have the same structure as M). After a
reaction with an electron, a neutral, X, is formed which could potentially capture an
electron.

Route B explains the formations of [M-O]" ions observed in the mass spectrum
of the trifluoroacetyl derivative of aminophenanthrene (8). Prominent ions appear at
M", [M-HF]", [M-CH3OH]", [M-H20]", [M-O]", and [M+CI-I2]". The [M-O]" ion
is difficult to justify using conventional ideas about ECNI spectra. Route A is unlikely
because there is no precursor ion of sufficient intensity. Route C is also unlikely
because the [M-O]" ion is fairly prominent. Route D does not occur in the other
fluorinated derivatives of aminophenanthrene, so it is unlikely it would occur here;
therefore, route B is the only other alternative. In this molecule, oxygen is present
only in a carbonyl group. It is expected that the positive ion formation of this

derivatized aminophenanthrene by CH5+ and C2H5+ regent ions is at the usual fast

Mo
e
A.” M walls
wollso [9‘ (b)
- . R'

M M
P'lwolls Elwotls

Z X MR W

is 1: la la

Z ' X ' M R ' W -
P’lwalls P’lwotls Piwotls P’ wolls

neutrals " " "

@O©©

Figure 1.2: Mechanisms of formation of unusual negative ions (from reference 8).

rate of ion-molecule reactions. To test this hypothesis, reagent gas is doped with a
substance which blocks the initial positive ion—molecule reaction while not disturbing
the electron reactions. In this experiment, triethylamine was used as the dopant. In the
presence of triethylamine, a very low abundance of [M-O]" was observed; therefore,
the mechanism for the formation of [M-O]" must first involve a positive-ion reaction
with the reagent gas. At this time, the identities of the species M+ and X are not
known; not much is known about the products generated by the recombination of large
positive ions and electrons or by the neutralization of positive ions on a metallic
source. The expected large energy release from these processes may make
unconventional rearrangements possible.

Route C involves a reaction of the analyte with a radical from the modulating

gas. The resulting species, MR, can capture an electron to form MR". Compounds

15
which can act as free radical traps include tetracyanoethylene (31), naphthalene (23),
and PAH isomers (32). The base ion for tetracyanoethylene is at mlz 103 which
corresponds to [M+H-CN]". Other prominent ions include those at [M+CH3-CN]"
and [M+CZH5-CN]". All these ions are from reactions with radicals formed from the
methane reagent gas before an electron is captured.

Wall reactions (route D) are very important pathways that can form unusual
ions. These reactions may have some analytical utility, as demonstrated by studies
using chlorine pretreatment of the source to detect such compounds as trimethylsilyl
derivatives of organic acids, fatty acid methyl esters, and trifluoroacetyl derivatives of
alcohols (33). Many wall-assisted reactions need some other species in the source in
order to occur. For example, the methane ECNI spectrum of both fluoranil and
chloranil contains the [M+H-Cl]" ion, which occurs through a wall-assisted reaction.
This ion is not observed using C02 gas, suggesting an absence of gas-phase or surface-

bound free radicals under 002 conditions.

 

 

 

 

1.0;-
>.
r:
g 0.8-
o
...)
C
H 0.6-
C
O
H
U 0.4
o
.t‘
g 0.2-
L
O
z

"'l" v' 'WIvvrvrvv—V—V—r'fi'
220 250

Time (sec)

Figure 1.3: Single ion reconstructed chromatograms of the methane ECNI-MS
analysis of floranil. The peak at mlz 180 is the molecular anion; mlz 162 corresponds
to [M+H-F]"; mlz 142 corresponds to [M-HF]" (fiom reference 8).

16

Ions formed from wall reactions tend to have non-Gaussian profiles. Single ion
reconstructed chromatograms of the methane ECNI analysis of fluoranil are shown in
Figure 1.3. In this case, the non-Gaussian profiles of peaks at mlz 142 (corresponding
to [M-2F]") and 162 (correponding to [M+H-F]") are characteristic of wall reactions
while Gaussian profile of ion current at mlz 180 corresponds to M". Neutral species
that can form ions often stick to the source wall, and they move out of the ion source
more slowly, causing the non-Gaussian profile.

Other researchers have used the Sears models to postulate mechanisms for
unusual ions. For example, an unusual ion, MH’, is observed in the ECNI spectrum of
chloroprothixene (34). Three mechanisms are possible for the formation of this ion:
(1) electron capture followed by H' transfer from the methane gas (route A in the Sears
model), (2) H' transfer followed by electron capture (route C), or (3) reaction on the
source wall (route D). To determine the mechanism, several experiments were
performed. When a gold-plated source is used, the MH‘ ion remains unchanged;
therefore, this ion does not result from a wall reaction, disproving mechanism 3. There
is also no change in the [M-H]'/MH' ratio with a change in emission current,
suggesting that MI-I' is independent of radical concentration. Mechanism 2, which is
dependent on the concentration of the radicals from the reagent gas, is not the likely

mechanism in this case. Mechanism 1 is the most probable in this case.

Molecular Parameters Affecting Ion Formation

Ion formation is affected by the electron affinity of the analyte, the electron
attachment rate (also known as the electron-capture cross section), and the lifetime of
the ion.

W. In order to capture an electron, a compound must have. low-
energy vacant orbital which correlates to a positive electron affinity (EA) (14).

Systems with extensive n-conjugation result in lower-energy unoccupied orbitals. For

17
example, the electron affinity of benzene is negative and it will not form a stable
anion. On the other hand, pentacene has larger orbitals which can overlap more
effectively, resulting in low energy unoccupied orbitals. The electron affinity of
pentacene is positive (35).

Fragmentation also is dependent on electron affinity. To fragment AB into A',
the A-B bond dissociation energy and the total fragment energy must be less than the
EA of A. Thus, the formation of molecular anions is favored by high, positive electron
affinities, as shown for PAHs (23).

According to Field (35), molecular electron affinities generally are affected by
the same structural factors that increase the wavelength of light absorption. In light
absorption, the electron is usually promoted fiom an occupied to an unoccupied
orbital, the same lowest unoccupied orbital that captures an electron. Not all the
absolute energies are the same, but the trends should be similar for a series of
compounds.

Several methods have been used to experimentally determine electron affinity.
For example, gas-phase electron affinities may be derived from gas-phase electron
transfer equilibria with a pulsed electron high-pressure mass spectrometer. Using this
method, the electron affinity values for 21 quinones were determined (36).

Electron affinity is difficult to measure experimentally; therefore, estimation or
approximation methods have been developed. Ab initio calculations have been used to
calculate electron affinity directly, but with great difficulty. Another method of
approximating the trends in electron affinities is by calculating the lowest unoccupied
molecular orbital (LUMO) energy. Laramee and co-workers (18) calculated the
LUMO energies for a series of polychlorodibenzofurans and polychlorodibenZo—p-
dioxins using the "Complete Neglect of Differential Overlap" method. No molecular

anion was observed for molecules with high LUMO energies, i. e., above 1.6 eV.

 

18

Another electron affinity estimation method was proposed by Chen and
Wentworth (37). The half-wave reduction potential in aprotic solvents (Em) was
correlated to experimental values for a series of compounds. The relationship is
expressed by EA = E1 ,2 + 2.49 i 0.2 eV.

The rate of electron attachment (cross section) can vary by orders of
magnitude, while the cross section for electron impact varies very little (17). Cross
sections are influenced by structural differences between molecules; for example,
electron attachment rate constants increase as chlorine substitution increases. The rate
constant also increases as the halogen changes from F to C1 (38).

W. The lifetime of an anion before it undergoes autodissociation
can also vary greatly from 10‘15 to 10'4 s. In order to have significant collisional
stabilization under CI conditions, an anion must have a lifetime of at least 10'6 s. The
ion lifetime is dependent on molecular structure and electron energy. At very long ion
lifetimes (>10 Its), molecular anions can be observed under low-pressure conditions.

Molecular anion lifetimes are listed in Table 1.1.

Table 1.1: Lifetimes of Molecular Anions (adapted from reference 17)

M' Husec)
C6H5N02- ~40
Phthalic anhydride’ 313
o—ClC6H4N02' 17
o—C6H4(N02)2‘ 463
m-N02C6H4COCH3‘ 3 10
0-CH3OC6H4N02- 16
cgnsir <1 x 106
C6F6- 12
cgng‘ 1x 1045
C6H5CN' 5
C6F5CN' 17

02 2 X 10'6

 

l9
Long-lived ions have positive electron affinities, usually imparted by conjugation
and/or electron-withdrawing substituents like halogens, CN', or N021 These
molecules tend to be large, so that the excess energy can be stabilized through many
degrees of freedom; therefore, this can be called a "resonance" process. At ion
lifetimes above 10 us, molecular anions are observed under EI conditions of low
pressure and high electron energy. Internal energies of a molecule increase the
fragmentation, as shown for a series of polychlorodibenzo—p—dioxins (18). The
branching ratio, log ([M-Cl]‘)/(Cl‘), is linearly related to the energies of the low-lying
unoccupied orbitals (LUMO). The internal energy of [M-Cl]' increases with

increasing degree of chlorination.

Experimental Parameters Influencing Ion Abundances

ECNI ion abundances and overall sensitivity are affected greatly by the choice
of instrumental conditions. The following instrumental parameters will be discussed in
more detail: source temperature, source pressure, reagent gas, sample concennation,
electron energy, emission current, and lens potential.

W. In general, ECNI spectra are highly dependent on source
temperature. As the temperature increases, dissociative electron capture increases
because negative ion lifetimes and the forward rate of clustering reactions are
decreased. At low source temperatures (< 100 °C), the molecular anion formation is
enhanced for a variety of compounds including hexabromobiphenyl (39), diazepam
(22), and PAHs (24).

Sensitivity is also affected by source temperature; however, different
compounds behave differently to increases in temperature. For example, Low and co-
workers (23) studied the response of three PAHs, . anthracene, fluorene, and
fluoranthene, at temperatures varying from 90°C to 230°C. The response of

anthracene and fluorene increased slightly with increasing temperature, but the

 

20
response of fluoranthrene fell off dramatically. Thus, compounds which exhibit
primarily deprotonation, such as fluorene, have best sensitivity at high temperatures (>
150°C) while those favoring resonance capture, such as fluoranthrene, prefer low
temperatures (< 100°C) (23). Compounds that undergo primarily dissociative
resonance capture usually have a higher response at low source temperatures (39).

Temperature also affects the formation of ions resulting from reactions other
than electron capture (8). Positive ion-electron recombination reactions decrease with
increasing temperature, while positive ion-negative ion recombination reactions
increase with increasing temperature.

In a study of fluorinated derivatives of chlorophenols (40), the effects of source
temperature were dependent on the type of derivative. Derivatives used included
pentafluorobenzyl, pentafluorobenzoyl, and 3,5-bis(trifluoromethyl)benzoyl.
Temperature effects are minimal for compounds undergoing dissociative electron
capture, such as the 3,5-bis(trifluoromethyl)benzoyl derivative of chlorophenol, but are
important for those which can undergo resonance electron capture, such as the
pentafluorobenzyl and pentafluorobenzoyl derivatives.

W. Control of the ion source pressure is thought to be the most
critical parameter for the reproducibility of ion abundances (14). As with source
temperature, the optimum source pressure is dependent on the compound. In a study
of PAHs (23), fluorene, which exhibits mainly [M-I-I]' ions, has an optimum sensitivity
at low source pressure (< 0.065 kPa). For fluoranthrene which produces mainly M",
the source pressure was not as critical.

Increasing the pressure increases the number of ion-electron pairs formed until
the electron beam is stopped by gas molecules. Also, an increase in pressure will
increase the collisional frequency, which affects the rate of electron thermalization and
collisional stabilization of the [A3]". intermediate. Usually, increasing the pressure

increases the response until a maximum is reached. The initial increase in response is

 

21
probably due to collisional stabilization (41). After the maximum, the loss in response
is probably from ion transmission losses (15). The pressure maximum is different for
different reagent gases (18).

W. The reagent gas can have three functions in negative chemical
ionization. The first is to convert primary electrons into' thermal electrons. This
allows resonance and dissociative electron capture. The gas can also react with the
analyte. Charge transfer can occur if electron affinity of the reactant gas is less than
the electron affinity of the substance. True chemical ionization can also occur between
ions from the gas and analyte molecules.

In Table 1.2, the second-order electron thermalization rate constants and the
number of collisions (Z) required for vibrational quenching of photoexcited
bromobenzene positive ions are listed. In general, electron thermalization increases
with increasing complexity of the reagent gas, i. e., increasing molecular size, dipole

moment, and 1: character (42).

Table 1.2: Electron Thermalization rates (Kf), number of collisions required for
vibrational quenching (Z), and ionization potentials (IP) for various reagent gases
(adapted from reference 42).

Gas Kimm’s'li Z 11!

He 6.4 x 10-12 150 24.59
Ar 1.3 x 10-13 77 15.76
02 1.3 x 1010 30 12.07
N2 2.2 x 1011 70 15.58
CH4 8.6 x 1010 40 12.51
CF4 -- -- 16.25
€02 5.8 x 109 18 13.77
NH3 5.9 x 109 -- 10.18
cos -- 8 11.17

H20 -- 8 12.61

22

In order to form stable molecular anions, the second-order rate constant for
collisional stabilization is important. This rate constant should be fast with respect to
the autodetachment rate constant. The collisional stabilization rate has not been
measured; however, an assessment of this rate can be determined using the Z number,
the number of collisions necessary for vibrational quenching. The lower the Z value,
the more efficiently excess internal energy is removed from the excited ion (43).
Quenching efficiency for gases not containing unsanuation or heavy atoms can be
correlated by the relation NMrl/2 where N is the number of atoms and Mr is the
reduced mass. Gases containing heavy atoms and unsaturation have vastly improved
quenching efficiencies (42).

Many studies have been done comparing the effects of the type of reagent gas
on ion abundances. If the major ion is M", as in the case of anthraquinone and his
(N JV-diethyldithiocarbamato) nickel (II), only minor differences are noted in the mass
spectra with various reagent gases including He, Ne, N2, CH4, Ar, Kr, Xe, i-C4H10,
and 002 (44). If the most abundant ion is not the molecular ion, changing the reagent
gas can have a more significant effect on the relative abundances of the ions. For
example, polybrominated biphenyls (PBB) and bromobenzenes were studied under
methane and nitrogen ECNI conditions (39). When nitrogen was used as the reagent
gas, the relative abundance of Br" was enhanced as compared to M". The ratio of Br“
to M" was small under methane ECNI conditions.

In studies with polychlorodibenzopcdioxins (18, 41), six different reagent
gases were used: methane, hydrogen, helium, sulfur hexafluoride, xenon, and argon.
Pressures were varied between 0.1 and 1.0 torr and the relative abundances of the
major fragment ions were recorded. Only a weak signal for C1“ was observed when
using sulfur hexafluoride as the reagent gas during the analysis of l,2,3,4-
tetrachlorodibenzo—p-dioxin because the reagent gas removed all of the thermal
electrons. When helium is used, the relative abundances of M", [M-Cl]‘, and Cl‘

23

remain constant over a pressure range of 0.1 to 0.8 mmHg. This can be used
advantageously in GC-MS work because helium is the carrier gas of choice. Also,
competing ion-molecule reactions are not observed with helium as they are with
hydrogen. Using hydrogen, M" is the dominant ion at low pressures, but at high gas
pressures, Cl' predominates. Argon shows similar pressure-related behavior. Xenon
exhibits unique behavior with varying pressures. At low pressures, M" dominates. As
the pressure increases, the relative abundance of M" decreases at pressures from 0.02
to 0.29 mmHg at which point M" again becomes the dominant ion. 1

The variation of ECNI response with type of reagent gas has also been studied.
One of the primary studies was performed by Gregor and Guilhaus (44) on
anthraquinone and his (NJV-diethyldithiocarbamato) nickel (11) using nine reagent
gases. The relative responses were in the following order: He < Ne ~ N2 < CH4 ~ Ar
~ Kr < Xe < i-C4H10 < C02. Both analytes form mainly molecular anions. In order to
prove that the responses were from just the effect of the reagent gas, it was determined
that the sample did not in any way participate in the thermal electron formation or
collisional stabilization processes. The gas pressure for maximum sensitivity varies
with each reagent gas; in general, the gas pressure corresponding to the highest
sensitivity decreases with increasing atomic or molecular weight of the gas. When
monoatomic gases are used as the reagent gas, the sensitivity of the analyte increases
with increasing atomic number of the gas. When polyatomic gases are used, the
sensitivity of the analyte increases with the available degrees of freedom of the reagent
gas, with the exception of C02. C02 is thought to be a very effective electron
thermalizer because it has a large quadrupole moment, a low-lying vibrational
threshold, and can thermalize sub-excitation electrons by resonance scattering
processes.

Other studies have confirmed the observations made by Gregor and Guilhaus
about reagent gas type. Laramee, er 01. studied l,2,3,4-tetrachlorodibenzo-p-dioxin

24
using argon, xenon, sulfur hexafluoride, hydrogen, and helium (41). Heavier mass
gases, such as methane, argon, and xenon, have the best sensitivities at low source
pressure.

Sensitivities of fluoranil as well as nitrobenzene with fluorine, bromine,
chlorine, nitro, and CF3 substitutions were determined using six different reagent gases
(43). In general, the order of relative response was Xe < He < Ar ~ N2 < CH4 ~ C02.
The use of C02 and CH4 as reagent gases increased the sensitivities of the analytes by
one to three orders of magnitude more than the other gases.

The relative response of fluorocarbons with seven different reagent gases has
been studied by Huang and co-workers (42). In general, the order of relative response
for hydrogen- or oxygen-containing compounds, such as perfluoro-t-butanol and
perfluorobis(isopropyl) ketone was He < CF4 ~ Ar < CH4 < C02 ~ air ~ NH3. For M"
forming compounds that included perfluorohexane, perfluorobenzene, and perfluoro-
trans-decalin, the relative response was He < CF4 < Ar ~ air < 002 ~ NH3 < CH4.
The actual order was very compound dependent. Carbon dioxide or ammonia is best
for the detection of hydrogen- and oxygen-eontainin g fluorocarbons in which
dissociative resonance capture predominates. For compounds which exhibit resonance
capture, methane is the best reagent gas.

Certain reagent gases can also simplify ECNI mass spectra. For example, C02
has been suggested as a model reagent gas for several reasons (43). The spectra are
simpler because only resonance and dissociative electron capture processes are
observed for compounds that often undergo source reactions involving gas-phase and
surface-bond free radicals. The fluoranil and substituted nitrobenzene compounds
tested still had high sensitivity, comparable to that of methane ECNI. '

In a study of polychlorinated compounds, such as heptachloroepoxide, dieldrin,
and hexachlorobenzene, the response and spectra of several reagent gases, including

nitrogen, ethylene, and methane were compared (45). When using N2, no rhenium

25
ions were observed, thus reducing the background. However, a lower ECNI response
for the compounds tested was obtained using nitrogen rather than methane. C2H4 was
a poor reagent gas because there is more background, lowered response, and a chance
of ion-molecule adduct formation.

Concentration. Concentration of the analyte can affect the appearance of the
spectra as well as the response. For certain compounds, such as the amino acid
carboxy-n-butyl eSter N-pentafluoropropionate, the ECNT mass spectrum is biased
towards high mass fragments at high concentrations (23). This may be due to
intermolecular reactions. This effect was not noted for compounds such as PAHs.
Radical incorporation reactions are enhanced by low sample concentration (15). At
high concentrations, the response becomes nonlinear and the chromatographic peak
shape becomes non-Gaussian (5). Often, saturation of the detector with analyte results
in flat-top chromatographic peaks.

W. The electron energy will determine the penetration of
electrons into the reagent gas as well as the efficiency of ion pair production. The
electron energy has a very minor effect on the relative ion abundances of
decafluorotriphenylphosphine (DFI'PP) and chlordane (5).

W The emission current determines the number of electrons
entering the ion source. As the emission current is increased, so is sensitivity because
the ion population is increased. The effects of emission current were studied using
DFI'PP and chlordene (5). The relative abundance of ions from DFI'PP does not
change with increasing emission current. Many of the chlordene ions, however, are
dependent on emission current. The types of ions most likely to change with
increasing ionization current result from radical reaction prior to ionization.

W. The lenses controlling the extraction and transmission of ions

can affect the relative ion abundances. The effects of lens potential on ion abundances

26
is complex, often depending upon interactions between the lenses. The most complete
work on this t0pic was done on a HP 5985 quadrupole mass spectrometer (5).

The repeller potential has a minor effect on ion abundances, but can affect the
voltage profiles of the other lenses. The drawout lens accelerates ions to the ion focus
lenses. Higher mass ions are slightly affected by the drawout lens. The ion focus lens
collimates the ion beam and has the largest effect on relative ion abundances, which is
especially pronounced for low-mass fragment ions such as C1‘ or C61: 5'. This lens has
an important effect on the comparison of ECNI mass spectra measured on different
instruments. To overcome the quadrupole fringing fields, the entrance lens is used

whose patential affects the abundance of the high-mass ions.

Applications of ECNI-MS
Selectivity

ECNI is more selective than positive CI because not all molecules have low-
lying vacant orbitals or virtual vacant orbitals. In conjugated systems, the n:* orbital is
delocalized and electron-withdrawing groups (like F, Cl, CN) lower the energy of
these orbitals making them more accessible. Positive ions are generated by removing
valence-level electrons, as can be done for all compounds (14). In order to generate a
negative ion, a compound must have a positive electron affinity. Because many
organic compounds have negative electrons affinities, ECNI can be very selective for
certain classes of compounds.

Ionization efficiency is determined by the electron affinity and the electron-
capture cross section of the analyte, which can vary over a wide range such as a factor
of 10‘5 between the values for hexane and carbon tetrachloride (16). After colliding
with a thermal electron, a molecule can lose this electron by collisions. For small

molecules, the electron loss process can be very rapid (about 10'13 s), but for large

27
molecules, the excess energy can be redistributed and the rate constant for electron loss

is about 10'6 5. Thus, the negative anion can be more readily observed.

Sensitivity

As with selectivity, the sensitivity of compounds to electron capture
techniques varies greatly. Sensitivity is dependent on the electron-capture cross
section which can vary greatly. In contrast, the cross section only varies by one or two
orders of magnitude in the positive ion mode (46). Therefore, one can expect widely
varying and unpredictable responses under ECNI conditions. For compounds
exhibiting electron-capture prOperties, less than 10 ng of sample can give a discernible
spectrum. For highly electrophilic compounds, saturation of the detector signal can
occur with as little as 100 ng. To saturate the detector with positive ion current,
micrograms of sample are needed. The reaction rate for gas phase positive and
negative ion reactions (PCI and NCI) is about 10‘9 cm35'1. For electron capture, the
rate constant is 4 x 10'7 cm3S'1 or about 400 times greater (3). Thus, sensitivity should
be 400 times greater by ECNI than by PCI or N CI.

For selected compounds with high electron-capture cross sections, ECNI can
be 100 to 1000 times more sensitive than positive E1 (3). Detection limits for highly
fluorinated compounds like perfluoromethylcyclohexane and perfluoro- l ,3-
dimethylcyclohexane are 2 to 3 fg (47). This is a three to four times better detection
limit than that obtained by GC-BCD. In theory, the detection limit of the GC-ECD is
on the order of 330 attomoles (48). Because ECNI-MS measures ion abundances
rather than a decrease in standing current, Hunt and Crow have estimated that ECNI
could be 10 to 100 times more sensitive than ECD (49). It has been noted (47) that
noise rather than signal determines the detection limit in ECNI. This noise is thought

to be general chemical background.

28
Types of Compounds

ECNI-MS has found utility in the analysis of many electrophilic compounds,
mainly highly halogenated ones (5). Highly conjugated compounds also have been
effectively analyzed (10). For compounds without electrophilic groups, halogenated
derivatives have been prepared prior to analysis (50). For a review of halogenated
derivatives, please refer to Chapter 3.

Compounds that have been analyzed by ECNI-MS with no further
derivatization include s-triazines (51), trichothecenes (7), aflatoxins (52), polycyclic
aromatic hydrocarbons (53), dihydropyridine calcium-channel blockers (54),
lorazepam (55), and steryl fatty acid acyl esters (56).

Many biological compounds can be analyzed by ECNI-MS only after
derivatization. Derivatization is necessary because the molecule is thermally labile
and/or has poor electron capture response. Corticosteroids have been analyzed after
MO-TMS derivatization (57). Valproic acid (30) and histamine (58) have been
analyzed as their pentafluorobenzyl derivatives. A heterocyclic amine thought to be a
carcinogen, 2-amino-3,8-dimethylimidazo [4,5-f] quinoxaline, has been analyzed as its
3,5-bis-trifluoromethylbenzyl derivative (59).

Electron Capture Response Studies
Electron Capture Negative Ionization Mass Spectrometry Response Studies

The effect on halogen substitution on ECNI-MS response has been studied for
a series of brominated dibenzodioxins and dibenzofurans (60). The presence of at least
two bromines or a bromine with at least one other chlorine is required for high
response. In the ECNI mass spectra of these compounds, Br' ions usually
predominate; in fact, an assay was developed for brominated compounds based on

monitoring only the ion current corresponding to the Br‘.

29

Crow and co-workers (25) compared the methane positive CI, methane ECNI,
metltane/CH2C12 chloride attachment NCI, and methane/02 N CI source conditions
using chlorobenzenes and bromobenzenes with one to six substitutions. The
compounds all have fairly similar positive CI responses, except for tetrabromo- and
hexabromobenzene which had low responses. In the ECNI mode, bromobenzene,
chlorobenzene, and dichlorobenzene cannot be detected. The tri- and
tetrachlorobenzenes have fairly low responses. In these cases, the major ion is Cl';
minor ions include [M-H]" at about 10% relative abundance. No molecular anion is
observed. The di- and tribromobenzenes have responses in ECNI similar to that
obtained during positive CI. Their spectra consist of mainly Br', with [M-H]' as a very
minor ion (< 0.5%). The only compounds that have molecular anions at significantly
higher responses than under positive CI conditions were pentachlorobenzene,
hexachlorobenzene, and tetrabromobenzene. Hexabromobenzene had a molecular
anion and about the same response as in positive CI. Interestingly, hexabromobenzene
has a lower response than tetrabromobenzene.

The methane/CI-I2C12 chloride attachment NCI spectra generally are not very
abundant. The monochlorides and dichlorides were not detected; the trichlorides have
molecular anions, but very poor sensitivity, worse than positive C1 or ECNI. Only
pentachlorobenzene and hexachlorobenzene have sensitivities comparable to that
achieved with ECNI.

The methane/02 NCI technique is the best choice for many of the compounds.
For all bronrinated benzenes except monobromobenzene, it is the most sensitive
detection method. This detection method is also superior for the chlorinated benzenes
containing more than three compounds.

In a study of the responses of chlorinated phenols, biphenyls, and aliphatic
bicyclics, the trends in response for ECD and ECNI were compared (63). In the

phenol and biphenyl systems, molar response increases with increasing chlorine

30
number for both ECD and ECNI. In each case, the ECNI response falls slightly for the
fully chlorinated species. However, for the chlorinated bicyclics, the increase of molar
response with increasing chlorine number is observed only in ECNI; the ECD response
does not vary.

Monochlorinated and dichlorinated ethyl acetates have recently been studied by
ECNI-MS and ECD (62). The position of the chlorine substitution is vitally important;
the highest ECD response monochloride is CHZCICOOCHZCH3 while the highest
response dichloride is CHZCICOOCHZCHZCI. The presence of two chlorine
substituents on the same carbon increases the response slightly, but not significantly.
The standard deviations for this ECD work were quite high (ranging from 10 to 16%).
The standard deviations for the ECN I work were not even reported; this is probably
because typically ECNI-MS standard deviations are higher than ECD standard
deviations. The trends in ECD and ECNI response are the same.

In a study of fluorinated the pentafluorobenzyl, pentafluorobenzoyl, and 3,5-
bis(trifluoromethyl)benzoyl derivatives of phenol (40), it was noticed that the relative
ECNI molar response showed the same trend regardless of source temperature. Each
of these derivatives have spectra consisting of essentially one peak. The relative trends

are the same when compared to the uends in the ECD response.

GC-ECD Response Studies

ECD differs from ECN I in the method of detection. In ECNI, thermal
electrons formed from a reagent gas are captured by the analyte and the resulting
negative ions are detected. In contrast, a standing current in ECD is formed from the
63Ni foil in the ECD detector. The decrease in standing current when an analyte
captures an electron is measured.

Many more response studies for steroids have been done using ECD rather than

ECNI-MS. In studies with haloacyl derivatives of testosterone and diethylstilbesterol

31
(63), monochloroacetyl and chlorodifluoroacetyl derivatives capture electrons more
efficiently than trifluoroacetyl (see Table 1.3). Pentafluoropropionyl and
heptafluorobutyryl derivatives capture electrons more efficiently than trifluoroacetyl,
but not quite as well as chlorodifluoroacetyl. However, the features impart by the
heptafluorobutryl derivatives are considered the best compromise between sensitivity
and volatility.

Unsaturated steroids have been studied using ECD. The earliest study was
done by Lovelock and co-workers (64), the person who invented the ECD detector. As
shown in Table 1.4, the 4-ene-3.one system is thought to be the basis for enhanced
response. A steroid with three isolated ketones has the same response as a steroid with
only one ketone. A l-ene-3-one does not have an appreciable increase in response;
however, a 4-ene-3-one increased the response as in 4-androsten-3,17-dione. The 17-
position seems significant to the authors because 4-cholesten- 3—one and testosterone
(4-androsten-3-one-17-ol) have lower responses than compounds with a 17-one. The

4-cholesten-3-one response is lower than that of 4wandrosten-3-one by a factor of two.

Table 1.3: Relative Response of the Electron-Capture Detector to some Haloacyl
Derivatives (adapted from reference 63)

Compound

Derivative Testosterone Diethylstilbesterol
Acetyl 1.0 --
Monochloroacetyl 40 2.7
Chlorodifluoroacetyl 340 --
Dichloroacetyl -- 2.6
Trichloroacetyl -- 2.1
Trifluoroacetyl 4 1.7
Pentafluoropropionyl 50 --
Heptafluorobutyryl 190 --

Perfluorooctonyl 600 --

32

Since the authors admit that from a single test the reproducibility of 4—androsten-3,17-
dione was within 17%, a difference of a factor of two may not be very significant.
Comparing the response for testosterone to the response for 4-androsten-3,17-dione is
unfair because testosterone has a hydroxy group which is more likely to remain on the
injector or column. Then the authors stated that the location of the carbonyl group in
the D-ring is not important because progesterone, 4—androsten-3,l7-dione, and 4-
androsten-3,16-dione have the same response. Perhaps this is true because the 17-
position is not important for enhanced response.

The 1.4-diene-3—one system is interesting. 1,4-Androstadien-3,l7-dione has
one fifth the response of 4-androsten-3,17-dione. However, when an ll-ketone is
added to the 1,4-androstadien-3,17-dione, the response increases by a factor of 100.

The other ECD response study for steroids used substituted 17a-
acetoxyprogesterones (65). Methyl groups in the A or B ring have no effect, but
substitution at the C-16 position are important. A methyl group at C-16 increases the
response by a factor of two to three, while a methylene group has two to six times the
response of a methyl group. The 1,4-diene-3-one group has a very small increase in
response over that of a 4-ene-3-one group. Much more significant are the responses of
the linearly conjugated 4,6—diene-3-one system. The addition of a 6-methyl group to
this system increases the response by six. The most sensitive compound in this study
is 4,6—pregnadien-3-one-65-methyl-16-methylene-Nor-hydroxy acetate (melengestrol
acetate), which. combines all the important features for enhanced response-~linear

conjugation, a 6-methyl group, and a 16-methylene group.

33

Table 1.4: Electron Absorption Coefficients of Steroids
(adapted from reference 64)

Compound Absorption Coefficient
Androstane 0.01
Cholestane 0.01
Cholesterol 0.03
Androstane-3,17-dione 0.10
Allopregnane-3,1 1,17—trione 0.50
4-Cholestene—3-one 1 1.3
Testosterone 8.4
19-Nortestosterone 9.1
3,5-Cholestadiene-7-one 96.5
17B—hydroxy-l ,4,6-androstatrien-3,17-dione 85.0
l-Androsten-3,17-dione ‘ 0.21
1,4—Androstadiene-3,17-dione 4.5
4»Androsten-3,17-dione 23.5
4-Androsten-3,l6.dione 19.2
Progesterone 22.2
4—Androsten-1 1-ol-3,l7-dione 76.0
1,4,6-Androstatriene-3,17-dione 167
6-Ketoprogesterone 143
4-Androsten-3,l 1,17-trione 248
4-Pregnene-3,11,20-trione 250
l,4-Androstadien-3,1 1,17-trione 535
4,4»Dimethyl-5-cholestene—3-one 0.03
Estrone 0.03

All values are relative to the absoption coefficient of chlorobenzene, which is taken as
unity, and are for electrons with thermal energy.

34
The Analysis of Steroids Using ECNI-MS
ECNI-MS has been shown to be a sensitive and selective technique for the
analysis of compounds that are either highly halogenated or highly conjugated.
Steroids are typically neither highly halogenated nor highly conjugated nor do they
survive the GC intact; therefore, steroids usually are derivatized before ECNI analysis.
Previous work in this laboratory (10, 66-70) has shown the feasibility of using
chemical oxidation before ECNI analysis for the analysis of selected corticosteroids,
namely dexamethasone in plasma and 6B—hydroxycortisol in urine.
A steroid is a fused four-ring system composed of three cyclohexane rings and
one cyclopentane ring. In the IUPAC system, the parent name of a steroid results from
the number of carbons in its skeleton; for example, a nineteen carbon steroid is an

androstane. The numbering of the carbon atoms is shown in Figure 1.4.

Analytical Techniques for Steroid Analysis

Radioimmunoassay (RIA) has been widely used in the analysis of steroids. In
general, radioimmunoassays provide a precise, low-cost, easy, and sensitive analysis
with minimal or no sample extraction. Very small amounts of sample are needed; this
makes RIA a valuable technique in cases were very little sample is available, as in the
analysis of dexamethasone in newborns (71). A problem with the RIA analysis is
precision; coefficients of variance are usually not better than 5%. Conuibutions to this
error include specific activity of the label, incorrect optimization of assay conditions,
non-specific binding effects, sample contamination, incomplete phase separation, and
errors in counting of the radioactivity (72). The most significant disadvantage of the
RIA assay is the cross-reactivity with structurally-similar steroids. Also, RIA ean
evaluate only one steroid at a time; another assay must be performed to quantitate

another steroid.

35

 

 

 

 

l 2
l I
19
1 C
9
10 3
A B
7
5
4 6
Key for Abbreviations

A = Androstane C19 structure
P = Pregnane 021 structure
C = Cholestane 02—, structure

a or [3 before a letter refers to the configuration at C-5 of steroids; e. g. OLA for 5a-
androstane

Superscripted number after a letter denotes a souble bond; for example, A4 = 4-
androstene

Figure 1.4: Steroid numbering system.

36

RIA analyses have been developed for all classes of steroids. For
dexamethasone, RIA analysis (73-77) is the technique of choice for clinical analysis.
RIA is fairly sensitive with a limit of detection (LOD) of 20 pg with a coefficient of
variation on the order of 8%. Known interferences include cortisol (2% as active as
dexamethasone), 9a-fluoroprednisolone (5%), paramethasone acetate (19%), and
flumethasone (11%). For the analysis of 6B-hydroxycortisol, the cross-reactivity with
cortisol is so high (17%) that an extraction step is necessary (78-79). The limit of
detection is 25 pg with a coefficient of variance of 5% using 1 [AL of urine or 50 [IL of
plasma (80).

Immunoassays based on enzymes have the same advantages of sensitivity and
selectivity as their RIA counterparts, but do not have the problems associated with the
handling and disposing of radioactive materials. An enzyme-linked immunosorbent
assay (ELISA) for urinary 6B-hydroxycortisol (86) has been developed. Like
radioimmunoassays, ELISA methods are quite sensitive, with a detection limit of 10
pg and quite good precision. The major problem with ELISA is the same as with RIA;
that is, cross-reactivity with other steroids. In this case, there was a 1.2% cross-
reactivity with cortisol and 6% with 6B-hydroxycortisone.

Thin-layer chromatography (TLC) is used in the quick screening of drug
formulations (81). To confirm the results of the assay, CI-MS using a direct-probe
inlet was used. Detection limits of this hour-long TLC assay were on the order of 25
ng. TLC is not a useful technique for determining steroids in biological fluids where
the concentration is much lower. On the other hand, a complementary high
performance liquid chromatography (HPLC) method had a detection limit of 2 ng.

HPLC techniques have been used in the analysis of corticosteroids (82-85),
often analyzing more than one steroid at a time. The major advantage of the HPLC
method is there is no need for derivatization; however, HPLC methods suffer from

problems with sensitivity and selectivity. The most common detector used is

37
ultraviolet (UV). Almost all steroids absorb at the same UV frequencies which makes
the extraction steps before analysis critical. In fact, some researchers will knowingly
sacrifice analyte to be sure that the interferences are removed. The best limit of
detection is 0.5 ng/mL for a HPLC method for the analysis of dexamethasone in
plasma. More typical detection limits range from 410 ng/mL for a variety of steroids,
including the simultaneous detection of prednisolone and prednisone.

HPLC-MS has been used to make liquid chromatography a more specific
detector for corticosteroids (88-90). Although the HPLC inlet offers the advantage of
direct analysis without derivatization, the sensitivity is not as good as comparable GC-
EI-MS methods. In one study comparing LC-MS with GC-MS (90) for the analysis of
serum cortisol, the LC-MS technique was half as sensitive as the GC-EI-MS technique
employing methoxyamine-trimethylsilyl (MO-TMS) derivatives.

Gas chromatographic methods have long been used in steroid analysis. The
hydroxy group on steroids accounts for their low volatility because of hydrogen
bonding and thermal lability. In order to improve thermal stability and
chromatographic properties, derivatives have been used (50). In the past, various
detectors, such as flame ionization (PD) and electron capture (ECD) detectors have
been used. At the present time, GC-MS is the detector of choice because of the
specificity imparted by the mass axis.

Derivatization reagents should have the following desirable characteristics
(87): i

(1) Only one derivative must be made from the parent compound. More than
one derivative complicates the GC chromatOgram and also makes quantization
difficult. '

(2) The reaction must be complete. Often an internal standard is reaction

simultaneously in order to compensate for incomplete reactions.

38

(3) The by-products should not interfere with the separation or mask peaks
which represent the analyte. Ideally, one or two peaks are obtained, one for the
reaction product and the other for the unreacted moiety. For example, a by-product
such as the leftover derivatization agent should be removed.

(4) The derivative should be stable so that it does nor degrade before GC
analysis.

(5) The derivatives must not interfere with the separation of the parent
compounds; i. e., the derivatives of two different compounds should be different.

(6) Other important advantages derivatization may have is the improvement of
chromatography and the incorporation of a clean-up step.

(7) For GC-MS, the majors ions should be dependent on the parent molecule,

not the derivatizin g moiety.

The Analysis of Corticosteroids Using Mass Spectral Techniques

The synthetic corticosteroid dexamethasone has been successfully detected
using GC-EI-MS techniques. One of these techniques, based on the trimethylsilyl-
enol-trimethylsilyl-ether (tetra-TMS) derivative of dexamethasone was used in the
determination of dexamethasone in human plasma (66). The detection limit was
approximately 0.15 ng/mL. Replicate injections of sample were not possible; the
extract was dissolved in 5 [LL of hexane and 3-4 uL of this extract was injected into the
GC-MS. The injection of such a large fraction of the sample resulted in the
accumulation of contaminants on the GC column, requiring frequent removal of short
sections of the column. The other method (91) used the tetra-TMS derivative to
confirm the presence of dexamethasone in bovine tissues after analysis by nomial-
phase HPLC. The sample extraction was complex: after a three-phase liquid-liquid
extraction, the sample was injected onto a coupled-column HPLC system. The

detection limit of this method was 6 ppb. Problems with the El analyses include

39
choice of ions to monitor. The molecular ion has a relative abundance of less than
10%. The ions monitored, [M-335]+ and [M-375]"', are not particularly diagnostic. In
order to overcome this difficulty, high resolution selected ion monitoring (HR-SIM)
was used.

Chemical ionization methods have found limited use in the analysis of
corticosteroids. Both MO-TMS and tetra-TMS derivatives were used for a GC—CI-MS
study of dexamethasone (92). The [M+1]+ ion of the tetra-TMS derivative had a
higher relative abundance than for MO-TMS (80% vs. 50%), so the tetra-TMS was
used for the development of an assay for the determination of dexamethasone in
human plasma. The detection limit of pure sample was about 100 pg; however, the
limit of detection in plasma was not determined. An estimate of the detection limit can

be obtained by noting that the lowest value on the calibration curve was 500 pg.

Electron Capture Negative Ionization Mass Spectrometry

ECNI-MS has been shown to be a very sensitive technique for compounds with
electrophilic substitutions. Not many steroids are inherently electron capturing;
therefore, a derivatization step is needed to improve sensitivity. Many conventional
electron-capturing derivatives have been used for this purpose. For example, the
pentafluorobenzyloxime derivative of testosterone has been analyzed by isobutane
ECNI (93), but its base peak was [C5F6CI-12]' and another prominent ion was [M-I-IFT'
(see Figure 1.5). The former ion could be found in any such derivative of a steroids
and thus lacks selectivity. Selected ion monitoring (SIM) of the [M-I-IFJ" ion gave a
detection limit of about 20 pg.

Having a large proportion of the ion current carried by the reagent-specific ion
causes many difficulties in the Specificity of the analysis. In order to overcome this
problem, researchers have developed MO-TMS methods for steroids with substitution
in the form of double bonds, carbonyl, and hydroxyl groups (57, 88, 94).

4O

 

 

 

 

400' 181
OSilCHgl3
q, CBFSCHzio-N
o :
g m/21814-i
‘U
S
_Q 50‘
m .
(D
.2
E
93
167 197 ~ -
' 100 200 300 460 500' ' 600
m/z

Figure 1.5: ECNI mass spectrum of pentafluorobenzyloxime, TMS-ether derivative
of testosterone. The base peak at mlz 181 represents the pentafluorobenzyl anion

(from reference 93).

41

Corticosteroids such as prednisone, prednisolone, dexamethasone, and betamethasone
fall into this category. The MO-TMS derivative does not confer any electron-
capturing properties to the steroid; however, the derivatization process enables the
compound to survive the GC intact. In the case of the MO-TMS derivative of
prednisolone acetate, the M" ion was formed and fragmentation was minimal. The
limit of detection was 80 pg. The spectra are very different for prednisolone,
dexamethasone, and betamethasone. M" or [M-H]' is very small, less than 10%
abundance. The base peak often is [M—2(NOMe)-TMS]', which does correspond to
the parent molecule. Fragmentation is much more prevalent. Even so, quantitation of
0.1 to 10 ng of corticosteroids in human aqueous humor are possible by this method.
Along with problems associated with fragmentation, other problems with the MO-
TMS derivative include the possibility of forming chromatographically- separable
isomers, which makes quantitation difficult. Also, the derivatization procedure occurs
in two steps; this may be less efficient than a one-step procedure.

A recent method for the analysis of dexamethasone in plasma and synovial
fluid is based on the use of a tri-TMS derivative followed by GC-ECNI—MS analysis
(95). The ion monitored corresponds to the [M-TMSOH-TMS]'; very little other
fragmentation is noted. The limit of detection is 0.1 ng/mL with a coefficient of

variance of 6%.

Use of the Oxidation Methodolog

When using conventional derivatization procedures to improve the vapor-phase
characteristics and the electron-capture properties of the analyte for further analysis by
ECNI-MS, there are two significant disadvantages. One is that any compound with the
derivatizable functional group (such as hydroxy or amino group) will react with the
derivatizing agent, thus forming species which can be detected by ECNI-MS and

which may interfere with the detection of the analyte. Another disadvantage often

42
observed is that the major ions are indicative of the derivatization moiety rather than
the parent molecule.

In order to overcome these disadvantages, an approach to derivatization using
chemical oxidation has been used in this laboratory (10, 66-70). Using a mild
oxidizing agent such as pyridinium chlorochromate (PCC), hydroxysteroids are
converted to ketosteroids. Under ECN I conditions, these ketosteroids either form a
molecular anion or undergo simple dissociation such as losing hydrogen, a methyl
group, or HF. With little or no fragmentation, sensitivity is improved.

The real advantage of the oxidation methodology is the selectivity against
background. Steroids amenable to this technique are oxidized to highly electrophilic
species; other endogenous steroids form species which do not efficiently capture
electrons. The decrease in background is best illustrated in Figure 1.6. In this
experiment, a urine sample was spiked with 6B-hydroxycortisol. The oxidized product
of 6B-hydroxycortisol, 4-androsten-3,6,l 1,17-tetrone, has an enhanced ECNI response
due to the linearly conjugated 4-ene-3,6-dione system. The EI-MS reconstructed total
ion chromatogram (Figure 1.6(a)) illustrates the response from a non-selective
detector. All of the steroids in urine elicit a response. The spiked 6B-hydroxycortisol
has a small response compared to the other steroids. If the selective ECNI-MS
detector is used instead, the background virtually drops off (Figure 1.6(b)), and the
signal due to the oxidized 6B-hydroxycortisol is significantly enhanced. Most of the
trouble in developing sensitive ECNI methods is not in the response of the compound
of interest but in the presence of background. The described oxidation methodology
can help overcome this background problem.

Assays developed for steroids using the oxidation methodology are relatively
sensitive. Steroids which can be successfully assayed include dexamethasone,
betamethasone, prednisolone, prednisone, and 6B-hydroxycortisol. The limit of

detection for the analysis of dexamethasone in plasma is 0.25 ng/mL (66). This is

43

by PCC

   

 

 

 

6,6-Hydroxycortisbl Oxidized product
a GC/El-MS
80
3 t
t:
8. 601
§
§ 40«
.33
0
'0
2 20"
§ 1 ' ' t}, 123
a: 0 rea 0

4567891011121314151617

 

b GC/EC-NI/MS
80
0
E’
8. 50'
§
8..
«3 40‘ Area=16800
3 1 x
3
o 20‘
.2.
fa"
a: o-

 

4567891011121314151617

Figure 1.6: The oxidized product of 6B-hydroxycortisol is 4-androsten-3,6,1 1,17-
tetraone. A sample of urine was spike with 6B-hydroxycortisol, extracted, and
oxidized using PCC. (a) TIC from GC-EI-MS analysis (about 56 ng of analyte on-
column). (b) TIC from GC-ECNI-MS analysis (about 44 ng of analyte on-column).
Note the improvement in selectivity and sensitivity (adapted from reference 69).

44
much higher than the detection limit of the instrument which is 160 fg pure standard.
One of the reasons for this disparity is the high value of the blank because the internal
standard used, 13C5,2I-I3-dexamethasone, had a significant amount of unlabeled
dexamethasone present.

In comparison to the ideal derivative described previously, the oxidation
methodology fares well. Only one derivative is made from each compound using the
PCC reaction. Reaction efficiencies have been estimated at 50-95% (96) depending on
the analyte. An isotopically-labelled or structurally-similar internal standard can
compensate for the inefficiency of the reaction or losses during the clean-up
procedures. By-products from the reaction itself are rare, but if the do exist, they are
usually not as electron-capturing as the analyte. Because the derivatives are extremely
stable, they can be stored for months in the freezer. After oxidation, chromatography
of the steroids is improved over parent compounds. Interferences are minimized
because the analyte oxidizes to a higher-response compound than the matrix; therefore,
stringent clean-up procedures before GC—MS analysis are not needed. The major ion
in the ECNI mass spectrum of oxidized steroids is always dependent on the parent
molecule as either the molecular anion or an ion based on a simple loss from the parent
molecule is formed.

The major problem with the oxidation methodology is that the derivatives of
two different compounds are not necessarily different. For example, the steroidal
drugs prednisone or prednisolone differ only by the substituent at C-1 1, which is either
a hydroxy or carbonyl group. Upon oxidation with PCC, both prednisone and
prednisolone form the identical product, 1,4-androstadien-3,1 1,17-t1ione (see Figure
1.7). One method of analyzing prednisone in the presence of prednisolone is by
careful choice of the oxidizing agent. If sodium bismuthate is used, the ll-hydroxy
group will not react; thus, prednisone can be analyzed without interferences.

45

oxrdatton

 

 

oxidation

   

Prednisolone

Figure 1.7: Prednisolone and prednisone are oxidized to the same product, 1,4-
androstadien-3,1 l,l7-trione.

In an attempt to improve the selectivity and sensitivity of the oxidation
methodology, ECNI-MS-MS was tried (68). Oxidized dexamethasone was analyzed
using selected reaction monitoring (SRM) of the ion currents corresponding to the
reactions rn/z 330 an") -) mlz 31o ([M-HF]"). When the GC inlet was used, there
was no noticeable difference between the samples analyzed by SRM and those
analyzed by conventional SIM. The major advantage to SRM is that the direct inlet
probe can be used, which significantly simplifies and shortens the analysis.

Objectives of This Study

The analysis of steroids in biological matrices, such as urine or plasma, is a
complex problem. One possible way to solve this problem involves the use of
chemical oxidation followed by GC-ECNI-MS analysis. The chemical oxidation step
converts the hydroxysteroids into ketosteroids, thus improving their chromatographic
properties. Also, this step imparts a degree of selectivity because the oxidized product
of the steroid of interest, usually a steroidal drug or 6B-hydroxycortisol, will have a

higher ECNI response than endogenous steroids.

46

In this study, three aspects of the chemical oxidation methodology were further
developed: ( l) the oxidation of steroids by alternative methods, (2) the study of the
relationship of structure to ECNI response of steroids and other cyclic compounds, and
(3) the development of specific assays for dexamethasone, prednisolone, and anabolic
steroids.

Pyridinium chlorochromate (PCC) has been the oxidizing agent of choice for
the conversion of hydroxysteroids into ketosteroids. Virtually no selectivity is
imparted by this technique; all hydroxy groups within the steroid nucleus are converted
into carbonyl groups. The efficiency of conversion ranged from 50 to 95% depending
on the parent compound. Optimum reaction times are in the six to eight hour range. If
this method is to be used in a clinical setting, the oxidation and clean-up procedure
must be on the order of eight hours, a typical work day. In an attempt to improve
selectivity, efficiency, and reaction time, other oxidation methods were investigated.
Electrochemical oxidation was primarily studied because of its potential to solve
problems inherent in the PCC methodology.

The success of the chemical oxidation methodology is dependent on the
suppression of the interfering compounds. This occur because these compounds do not
oxidize to highly electron-capturing species. In the investigation of the effects of
structure on the ECNI response of steroids, different combinations of double bonds,
carbonyl groups, halogens, and epoxides were studied. Smaller cyclic systems, such as
cyclohexanes, quinones, and terpenes, were studied to compare the effect of molecule
size on the ECNI response. In order to further understand the structure-response
correlation, the ECNI response will be compared to the electrochemical reduction
potential which can be considered the ability to capture an elecrron in the solution
phase. The results from the structure-response study will be used as a guide for
choosing steroids and other biological compounds amenable to ECNI assay

development.

47

Finally, this project also involved the use of chemical oxidation for assay
development. One particularly exciting area is in the development of assays for
anabolic steroids. Because many anabolics have low responses under ECNI
conditions, other electron-capturing derivatives such as trifluoroacetyl and
pentafluorobenzyl, were investigated.

Two collaborative projects involving the analysis of steroids were done using
the chemical oxidation methodology. The analysis of dexamethasone in human
plasma was done in collaboration with Dr. Clinton Kilts of Duke University. The
other, involving the determination of prednisolone in mouse plasma, was done in
collaboration with Dr. Pamela Fraker of Michigan State University.

The results of these studies will further the understanding of the process behind
the ECNI response of ketosteroids as well as provide further examples of the utility of
the chemical oxidation methodology.

References

1.

SQS‘P?!”

10.
11.
12.
13.
14.
15.
16.

17.

18.
19.

20.
21.
22.

Lin, Y. Y. Lipids 1980, 15, 756.

Bowie, J. H. Mass Spectrom. Reviews 1984, 3, 161.

Hunt, D. F.; Crow, F. W. Anal. Chem. 1978, 50, 1781.

Stemmler, E. A.; Hires, R. A. Anal. Chem. 1988, 60, 787.

Stemmler, E. A. Ph.D. Dissertation, Indiana University, 1987.

Low, G. K.-C.; Duffield, A. M. Blamed. Mass Spectrom. 1984, I I, 233.

Brumley, W. C.; Sphon, J. A. In Applications of Mass Spectrometry in Food
Science; J. Gilbert, Ed.; Elsevier: UK, 1987, p. 141.

Sears, L. J.; Campbell, I. A.; Grimsrud, E. P. Blamed. Mass Spectrom. 1987, I4,
401.

Doherty, L. C. Ph.D. Dissertation, Michigan State University, 1987.

Her, G. R.; Watson, J. T. Blamed Mass Spectrom. 1986, I3, 57.

Her, G. R. Ph.D. Thesis, Michigan State University, 1985.

Dougherty, R. C.; Roberts, J. D.; Bucs, F. J. Anal. Chem. 1975, 47, 54.
Budzikiewicz, H. Angew. Chem. Int. Ed. Eng. 1981, 20, 624.

Dougherty, R. C. Anal. Chem. 1981, 53, 625A.

Stemmler, E. A.; Hites, R. A.; et 01. Anal. Chem. 1988, 60, 781.

Hunt, D. F.; Crow, F. W. In Proceedings of the 9‘h Materials Research
Symposium, April 1978, NBS Special Publication 519; Washington, D. C.:
National Bureau of Standards, 1979, pp. 601-607.

Harrison, A. G. Chemical Ionization Mass Spectrometry; CRC Press: Boca
Raton, FL, 1984.

Laramee, J. A.; Arbogast, B. C.; Deinzer, M. L. Anal. Chem. 1988, 60, 1937.

Hass, J. R. In Mass Spectrometry, Part B.; C. Merritt, Jr. and C. N. McEwen,
Eds; Marcel Dekker: New York, 1979, p. 288.

Roy, T. A.; Field, F. H.; Lin, Y. Y.; Smith, L. L. Anal. Chem. 1979, 51, 272. '
Jennings, K. R. Phil. Trans. R. Soc. Land. A 1979, 293, 125.
Garland, W. A.; Miwa, B. J. Blamed. Mass Spectrom. 1983, 10, 126.

48

23.

24.
25.

26.
27.

28.

29.

31.

32.

33.

34.

35.

36.

37.
38.
39.

41.
42.

43.

45.

49

Low, G. K-C.; Batley, G. E.; Lidgard, R. O.; Duffield, A. M. Blamed. Mass
Spectrom. 1986, I3, 95.

Stemmler, E. A.; Buchanan, M. V. Org. Mass Spectrom. 1989, 24, 94.

Crow, F. W.; Bjorseth, A.; Knapp, K. T.; Bennett, R. Anal. Chem. 1981, 53,
619.

Busch, K. L.; Hass, J. R.; Bursey, M. M. Org. Mass Spectrom. 1978, I3, 604.

Schmidt, J.; Vorbrodt, H.-M.; Adam, G. Blamed. Mass Spectrom. 1986, I3,
663.

Jones, L. V.; Whitehouse, M. J. Blamed. Mass Spectrom. 1981, 8, 231.

Hocart, C. H.; Schlunegger, U. P. Rapid Cammun. Mass Spectrom. 1989, 3, 36.
Kassahun, K.; Burton, R.; Abbott, F. S. Blamed. Mass Spectrom. 1989, I8, 918.
McEwen, C. N.; Rudat, M. A. J. Amer. Chem. Soc. 1981, 103, 4343.

Dougherty, R. C.; Howard, S. V.; Wonder, J. D. Symp. Carcinogen PAH Mar.
Environ, EPA-600/9-82-013; 1982, pp. 14-25.

Kassel, D. B.; Kayganich, K. A.; Watson, J. T.; Allison, J. Anal. Chem. 1988,
60, 911.

Callery, P. 8.; Garland, W. A.; Fukuda, E. K. Org. Mass Spectrom. 1989, 24.
385.

Field, F. H. "Chemical Ionization: Negative Aspects," 28th ASMS Conference
on Mass Spectrometry and Allied Topics, May 1980, p. 2.

Heinis, T.; Chowdhury, 8.; Scott, S. L.; Kebarle, P. J. Am. Chem. Soc. 1988,
110, 400.

Chen, E. C. M.; Wentworth, W. E. J. Chromatogr. 1981, 217, 151.
Christophorou, L. G. Environ. Health Perspect. 1980, 36, 3.

Greaves, J.; Bekesi, J. G.; Roboz, J. Blamed. Mass Spectrom. 1982, 9, 406.
Trainer, T. M.; Vouros, P. Anal. Chem. 1987, 59, 601.

Laramee, J. A.; Arbogast, B. C.; Deinzer, M. L. Anal. Chem. 1986, 58, 2907.

Huang, S. K.; Despot, K. A.; Sarkahain, A.; Bierl, T. W. Blamed. Mass
Spectrom. 1990, I9, 202.

Sears, L. J.; Grimsrud, E. P. Anal. Chem. 1989, 61, 2523.
Gregor, I. K.; Guilhaus, M. Int. J. Mass Spectrom. Ian Process. 1984, 56, 167.
Oehme, M.; Stoeckl, D.; Knoeppel, H. Anal. Chem. 1986, 58, 554.

47.
48.
49.

50.

51.
52.
53.

54.
55.
56.

57.

58.

59.

61.

62.

63.

65.

68.

50
Otvos, J. W.; Stevenson, D. P. J. Am. Chem. Soc. 1956, 78, 546.
Begley, P.; Foulger, B.; Simmonds, P. J. Chromatogr. 1988, 445, 119.
Pellizzari, E. D. J. Chromatogr. 1974, 98, 324.

Hunt, D. F.; Stafford, G. C. Jr.; Crow, F. W.; Russell, J. W. Anal. Chem. 1976,
48. 2098.

Knapp, D. R. Handbook of Analytical Derivatization Reactions; Wiley: New
York, 1979 and references contained therein.

Huang, L. Q.; Mattina, M. J. I. Biomed. Mass Spectrom. 1989, I8, 828.
Park, D. L.; et al. J. Assoc. 017. Anal. Chem. 1985, 68, 636.

Bayona, J. M.; Barcelo, D.; Albaiges, J. Biomed. Mass Spectrom. 1988, I6,
461.

Ehrhardt, J. D.; Ziegler, J. M. Blamed. Mass Spectrom. 1988, 15 , 525.
Koves, E. M.; Yen, B. J. Anal. Toxicol. 1989, I3, 69.

Evershed, R. P.; Prescon, M. C.; Spooner, N.; Goad, L. J. Steroids 1989, 53,
285.

Midgley, J. M.; Watson, D. G.; Healey, T.; Noble, M. Blamed. Mass Spectrom.
1988, 15 , 479.

Payne, N. A.; Zirrolli, J. A.; Gerber, J. G. Anal. Biochem. 1989, I78, 414.

Murray, 8.; Gooderham, N. J.; Boobis, A. R.; Davies, D. S. Carcinogenesis
1989, 10, 763.

Buser, H.-R. Anal. Chem. 1986, 58, 2913.

Erhardt-Zabik, S.; Watson, J. T.; Zabik, M. J. Blamed. Mass Spectrom. 1990,
I9, 101.

Jalonen, J. Blamed. Mass Spectrom. 1990, I9, 253.
Zlatkis, A.; Poole, C. F. Anal. Chem. 1980, 52, 1002A.

Lovelock, J. E.; Simmonds, P. G.; Vandenheuvel, W. J. A. Nature (London)
1963, I97, 249.

Koshy, K. J. Chromatogr. 1976, 126, 641.

Kayganich, K.; Watson, J. T.; Kilts, C.; Ritchie, J. Biomed. Mass Spectrom.
1990, I9, 341.

Kayganich, K. A.; Heath, T. G.; Watson, J. T. J. Am. Soc. Mass Spectrom.
1990, l, 341.

69.
70.
71.
72.

73.

74.
75.
76.

77.
78.

79.
80.
81.

82.
83.
84.
85.
86.
87.

88.

89.

51
Watson, J. T.; Kayganich, K. Biochem. Soc. Trans. 1988, I7, 254.
Her, G. R.; Watson, J. T. Anal. Biochem. 1985, 151, 292.
Hichens, M.; Hogans, A. F. Clin. Chem. 1974, 20, 266.
Teale, J. D. In Screnttf c Foundations of in Clinical Biochemistry; D. L.
yizllgigam, R. F. N mm and V. Marks, Eds; Yearbook Medical: London, 1978,

Dugrlasaoia, M. C.; Chapman, D. I.; Moss, M. S.; O'Connor, C. Biochem. J. 1973,
I3 , 1.

Carroll, B. J.; et al. Arch. Gen. Psychiatry 1981, 38, 15.
Saenger, P.; Forster, E.; Kream, J. J. Clin. Endocrinol. Metab. 1981, 52, 381.

Arana, G. W.; Workman, R. J.; Baldessarini, R. J. Am. J. Psychiatry 1984, 141,
1619.

Carroll, B. J.; et al. J. Clin. Endocrinol. Metab. 1980, 51, 433.

Nahoul, K.; Adeline, J .; Paysant, F.; Scholler, R. J. steroid Biochem. 1982, 17,
343.

Park, B. K. J. steroid Biochem. 1978, 9, 963.
Kishida, S.; Fukushima, D. K. Steroids 1977, 30, 741.

Gaignage, R.; Lognay, G.; Marlier, M.; Severin, M.; Dreze, P.
Chromatographia 1989, 28, 623.

Rose, J. Q.; Jusko, W. J. J. Chromatogr. 1979, I62, 273.

Rocci, M. L.; Jusko, W. J. J. Chromatogr. 1981, 224, 221.

Bunch, E. et al. J. Assoc. Ofl'. Anal. Chem. 1987, 70, 967.

Lamiable, D.; et al. J. Chromatogr. 1986, 378, 486.

Zhiri, A.; Wellman-Bednawska, M.; Siest, G. Clin. Chem. Acta 1986, I5 7, 267.

Touchstone, J. C. CRC Handbook of Chromatography: Steroids; CRC Press:
Boca Raton, FL, 1986.

Houghton, B.: Teale, P. Dumasia, M. C.; Welby, J. K. Blamed. Mass
Spectrom. 1982, 8, 558.

Cairns, T.; Siegmund, E. G.; Stamp, J. J.; Shelly, J. P. Blamed. Mass Spectrom.
1983, 10, 203.

Gaskell, S. J.; Rollins, K.; Smith, R. W.; Parker, C. E. Blamed. Mass Spectrom.
1987, 14, 717.

91.
92.
93.
94.

95.
96.

52
McLaughlin, L. G.; Henion, J. D. J. Chromatogr. 1990, 529, 1.
Kasuya, Y.; et al. J. Pharm. Sci. 1984, 73, 446.
Gaskell, S. J.; Brooks, P. W. Intern. J. Mass Spectrom. Ian Phys. 1983, 48, 241.

Watson, D. G.; Midgley, J. M.; McGhee, C. N. J. Rapid Commun. Mass
Spectrom. 1989, 3, 8.

Girault, J .; Istin, B.; Fourtillan, J. B. Blamed. Mass Spectrom. 1990, I9, 295.
Kayganich, K. A. Ph. D. Dissertation, Michigan State University, 1989.

Chapter 2: Electrochemical Studies of Steroids

The oxidation of hydroxysteroids to ketosteroids is the first step in the chemical
oxidation assay for steroids in urine or plasma. In previous work in this laboratory (1),
pyridinium chlorochromate (PCC) was chosen as the best reagent for the oxidation
step. Several problems are encountered with the use of PCC. Depending on the
compound, the yields range from 50 to 100%; therefore, complete oxidation is not
available for all compounds. In order to achieve high yields, long oxidation times (5-
10 hours) are necessary (1). Because this assay is designed to be done in a typical
eight-hour work day in a clinical setting, a sufficient shortening of the PCC oxidation
step could mean that the extraction from plasma and the oxidation steps could be
accomplished in the same day rather than the two days now needed. Under PCC
conditions, all hydroxysteroids are oxidized to ketosteroids; often a greater degree of
selectivity is needed. For example, using PCC, both prednisone and prednisolone are
oxidized to the same compound, l,4—androstadien-3,11,l7-trione. A methodology that
could distinguish the two steroidal drugs would be useful.

Electrochemical oxidation has been proposed to overcome some of the
problems with PCC oxidations. If the oxidation potentials of two steroids are
different, electrochemistry could selectively oxidize the one with the less positive
potential. Electrochemical oxidation also has the potential for more efficient
conversions than by chemical means in a shorter period of time.

Steroid reduction potentials have been determined in order to compare the
capture of an electron in the condensed phase to the capture of an electron in the gas
phase. The electrochemical reduction potential in aprotic solution can be correlated

with the electron affinity of the compound as done by Chen and Wentworth (2).

53

54
Electrochemical Synthesis of Ketosteroids

Very few examples of direct oxidative conversion of alcohols to ketones exist
in the literature; this is due to the high oxidative potentials required (3). However,
many examples of the use of oxidation for the detection of steroids exists, especially
with HPLC-EC (electrochemical detection).

Several researchers have used electrooxidation for organic synthesis. Shono
and co-workers (4) used the anodic oxidation of the dienol acetates of steroids to
mimic liver microsomal y-hydroxylation of 4-en-3-ketosteroids. A constant potential
of 2.0 V (vs. SCE) with a platinum electrode was used. Electrochemical oxidation also
was used to synthesize 9-substituted products of 1,3,5(10)-estratrien-3-methoxy-17-
one (5). The oxidation was performed in an undivided cell with two platinum
electrodes at a constant current of 800 mA for 40 min. Substitution at the 9-position
depends on the solvent system used; for example, the use of methanol resulted in 9-
methoxy products. When methanolic sodium cyanide is used as the solvent, three
major products were formed with the cyano group at either the 2-, 3-, or 108-positions
(6).

Some electrosynthetic procedures used modified electrodes. The oxidation of
cholesteryl acetate dibromide (7) using a lead dioxide electrode resulted in an 85 to
93% yield of products. After debromination, the major products were 15-
hydroxycholesteryl acetate, 24-dehydrocholesteryl acetate, and 25-dehydrocholesteryl
acetate. The selective oxidation of 3-hydroxysteroids has been accomplished using a
nickel oxide hydroxide electrode (8). The yields for this selective oxidation are
generally low, ranging from 22 to 78%.

Electrochemical detection of steroids after HPLC separation has been
accomplished by oxidation. Ethynylestradiol (l,3,5(10)-estladien-l7a-ethynyl-
3B,]?l3-diol) can be oxidized with a mercury electrode (9). When the potential is held

for some time at the switching potential of the cyclic voltammagram, sharp reduction

55

waves are noted. These data suggest that the oxidation process produces an insoluble
mercury compound. Ethynylestradiol is not oxidized when either gold, platinum, or
glassy carbon electrodes are used. Phenolic steroids can be easily oxidized
electrochemically, but the oxidation is more difficult for saturated steroids. A HPLC
method (10) with electrochemical detection was developed using acetonitrile
containing sodium perchlorate as the mobile phase. The oxidation potentials (vs. SCE)
ranged from 1.08 V for ergosterol to 1.78 V for campesterol. The detection limits
varied between 10 and 100 ng. In some cases, this was an improvement in sensitivity
over that of UV detection by a factor of ten.

Some ketosteroids are more effectively detected if they are derivatized before
HPLC-EC. Because reduction of a ketone group occurs at fairly negative potentials, it
is imperative that oxygen be removed before analysis, which is difficult to do when
using HPLC. In order to efficiently oxidize ketosteroids, derivatization is necessary
(11,12), usually with phenylhydrazine. Detection limits are on the order of 5 ng.

Because the electrochemical oxidation of alcohols occurs at fairly high
oxidation potentials, some researchers have designed ingenious methods of
overcoming this problem. One such method involves the use of iodonium ion as the
catalytic electron carrier (13). The low anode potentials used (0.6-0.8 V vs. SCE) are
sufficient for the oxidation of the iodine, which serves as both a catalyst and an
electron carrier. Secondary aliphatic linear and cyclic alcohols have been converted to

their corresponding ketones in yields typically ranging from 72 to 100%.

Electrosynthesis an Planar Electrodes
In these experiments, steroids were directly oxidized using platinum electrodes.
After problems with extracting the products from the supporting electrolyte were

solved, bulk electrolysis was used for synthesis. Because the oxidization product

56
adhered to the electrode, direct desorption off an electrode in a mass spectrometer
source was an interesting possibility.
Experimental

Chemicals. All solvents used for electrochemistry were HPLC grade and
purified before use. All solvents were transferred under vacuum to ensure they were
pure and oxygen-free. The supporting electrolyte, tetrabutylammonium fluoroborate
(TBAF), (n-C4H9)4NBF4, was purchased from Southwestern Chemicals (Austin, TX)
or Aldrich (Milwaukee, WI). Steroids were purchased from Sigma (St. Louis, MO) or
Steraloids (Wilton, NH).

mm. For cyclic voltammetry studies, a specially designed
one-compartrnent cell was used (see Figure 2.1). The cell is designed for use with
three electrodes: a platinum wire auxiliary electrode, a silver wire reference electrode,
and either a platinum, glassy carbon, or gold disk working electrode (BAS, w.
Lafayette, IN). The design of the cell allows solvent transfer under vacuum. A
"dumpster" side arm of the cell holds the compound of interest; this allows a blank
analysis before the sample is added.

The silver wire reference electrode is really a quasi-reference electrode (QRE).
If the silver wire is placed in nitric acid, it undergoes the following reaction:

AgNO3 + e' 2 Ag° + NO3‘.
Because nitric acid will not be used in these analyses, no equilibrium will be
established. Also, the silver wire electrode potential will change with time. To
determine a reference point, an internal standard that has a well-known reduction
potential should be added to the cell. Ferrocene has been used for this purpose. All
the other peaks in the cyclic voltammagram (CV) then can be corrrpared to the
reference peak.

Electrochemical synthesis was performed in the three-compartment cell shown

in Figure 2.2. These three compartments are separated by frits, but all three

57

 

 

 

L %
a
/‘

 

 

 

 

 

 

 

 

 

Figure 2.1: One-compartment electrochemical cell used for cyclic voltammetry. All
solvent transfer is done under vacuum; the connection to vacuum is not shown. (a)
Silver wire reference electrode, (b) Platinum working electrode, (c) Platinum counter
electrode, (d) Sample reservior (adapted from reference 14).

 

____fi

(j

fir
ll—dn

 

 

 

 

L131, %
E t...

1 2 3

Figure 2.2: Electrochemical cell used for bulk electrolysis. All solvent transfer is
done under vacuum; the connection to vacuum is not shown. (a) Platinum counter
electrode, (b) Silver wire reference electrode, (c) Platinum working electrode, (d)
Platinum gauze electrode, (e) Sample reservoir (adapted from reference 14).

 

 

 

 

 

 

 

 

 

 

 

 

58

compartments contain supporting electrolyte and solvent. The actual conversion of
analyte using a platinum gauze electrode occurs in compartment 3. The silver wire
reference electrode, which is encased in glass with a small frit on the bottom, is also in
this chamber. The platinum auxiliary electrode is located in compartment 1 to prevent
any undesired side reactions. The auxiliary electrode in compartment 3 is used only
for the preliminary CV before beginning the electrolysis. All solvent transfer is done
under vacuum.

W. In CV, the potential of a stationary working
. electrode is increased linearly to a point and then decreased back to its starting point.
After scanning to a potential where one or more electrode reactions take place, the
direction of the scan is reversed and the reactions of the intermediates and products
formed during the forward scan can sometimes be detected. CV is not a routine
quantitative analysis technique, but has been used to study rates and mechanisms of
oxidation-reduction processes (15, 16). In a reversible reaction, the equilibrium
conditions between the oxidized and reduced material are maintained even with a
rapidly changing potential. To be a reversible reaction, the difference in peak
potentials between the anodic and cathodic peaks must be AEp = Ep(anode) -
Ep(cathode) = 57/n mV where n equals the number of electrons involved in the
electrochemical process. Reversibility is dependent on the scan rate used. The
forward and reverse rate constants for the expression

0 + e' :: R

are finite; therefore at high scan rates the equilibrium will not be maintained as the
potential changes. Quasi-reversible refers to reaction where the forward and reverse
rate constants are of the same magnitude. Irreversible refers to reactions where kb>>kf
for the anodic peak. In this case, AEp > 57/n mV. Often only the anodic peak is seen

because the material is not rereduced at a sufficient rate.

59

In this work, cyclic voltammetry was done using a one-compartrnent cell with
about 2-5 mg of steroid, 0.2 M TBAF, and 3-4 mL solvent. The potential was scanned
from 0 to 2.5 V at 100 mV/s and then reversed. All analyses were run on a BAS 100-
A Electrochemical Analyzer (W. Lafayette, IN).

W- Coulometry is a constant potential technique; the current
changes to keep the potential constant relative to a reference electrode. Bulk
coulometry was used for the oxidation of steroids on a synthetic scale. About 20 mg of
steroid were placed in compartment 3 of a three-compartment electrolysis cell
(illustrated in Figure 2.2). The solvent and 0.2 M TBAF were placed in all three
compartments. An EG&G Princeton Applied Research Potentiostat/Galvanostat
Model 273 (Princeton, NJ) was used in the potentiostat mode. Synthetic oxidation was
accomplished at a constant potential 200 mV above the anodic peak in the CV because
at the peak potential only 50% of the sample can be converted.

W. Thin-layer bulk electrolysis is an option
on the BAS-100A Electrochemical Analyzer system. It differs from bulk coulometry
in the way data are gathered and displayed. In bulk electrolysis, at nrinute-long
intervals, the ratio of the average current of that minute to that of the first rrrinute is
computed. The charge ratio is also computed at one-minute intervals. In TLBE, the
current and charge ratios are computed every second; therefore, a decision about the
fate of the electrolysis can be made every second, rather than every minute. TLBE is
useful for electrolysis of short duration; that is, only a few minutes.

Preliminary Cyclic Voltammetry

W. The solvent for oxidation must be capable
of dissolving the analyte as well as having the appropriate potential range. The
accessible potential ranges for several solvents are listed in Table 2.1. Note that these

figures can only be taken as a guide because the actual potential limit is dependent on

60

Table 2.1: Accessible Potential Range for Some Solvents
(Adapted from reference 10)

Solvent Cathodic Limit Anodic Limit
V vs. SCE V vs. SCE
Water -2.7 1.5
Methanol -2.2 ' 1.8
Acetonitrile -3.5 2.4
Dimethylformamide -3.5 1.5
Tetrahydrofuran -3.6 1.8
Methylene chloride -1.7 1.8

Table 2.2: Accessible Potential Ranges in Dimethylsulfoxide Containing Different
Supporting Electrolytes (0.1 M) (Adapted from reference 10)

Electrolyte Cathodic Limit Anodic Limit
V vs. SCE V vs. SCE
LiClO4 -2.68 2.10
KCIO4 -2.33 2.10
NaClO4 -2.08 2.10
KNO3 -2.33 2.10
KBF4 -2.33 2.10
K2S203 -2.33 2.10
LiCl -2.68 1.52
Me4NCl -2.40 1.52
Et4BClO4 -2.30 2.10

Bu4NBr -240 1.45

61
solvent purity, electrode material, and electrolyte. The potential range of some of the
common electrolytes is listed in Table 2.2.

In this work, methylene chloride, acetonitrile and tetrahydrofuran were used as
solvents for cyclic voltammetry because steroids are soluble in these solvents. TBAF
was the supporting electrolyte because tetrabutylammonium salts have at high anodic
limit. Prednisolone, prednisone, and 4-androsten-17B-ol-3-one were used as the test
compounds.

WW. Of all the commonly used
electrochemical solvents, steroids are generally more soluble in methylene chloride.
Using methylene chloride, both prednisolone and 4-androsten-17B-ol-3-one were
oxidized, and the oxidation was not reversible. However, the CV was not reproducible
on a day-to-day scale; the next time the experiment was attempted, no oxidative peak
was seen for either compound.

Table 2.3: Oxidation Potentials of Various Steroids in Acetonitrile
N. B.: All oxidation potentials were obtained using a quasi-reference electrode.

Because these data were obtained on different days and no internal standard was used,
the values cannot be directly compared.

Compound Oxidation Potential
(V vs. QRE)

Prednisolone 2.4
Prednisone Did not work
Cortisol - 2.1
Cholesterol 2.1
Dexamethasone 2.3
6B-Hydroxycortisone 2.4
4-Androsten-6B-ol-3,17—dione 2.1

Acetonitrile has a larger potential window than methylene chloride, which

allows more steroids to be oxidized. As shown in Table 2.3, many steroids were

62
oxidized using acetonitrile with TBAF as the supporting electrolyte. Prednisolone has
prominent oxidative peaks (see Figure 2.3), and the oxidation reaction was not
reversible so that the products will not be reconvened into starting material.

Prednisolone was chosen as the model compound for the bulk coulometry studies.

'2 §

umn run 1.! u 1.u.l.l.u.l.l.u.l.1 uul

  

CURRENT (uA)
J.
8

1.
P
8

-16.W

uu Int 1 LLMJLI

s
e

POTENTIAL

Figure 2.3: Cyclic voltammagram showing that electrochemical oxidation of
prednisolone is irreversible. The oxidation potential of prednisolone is 2.3 V vs. QRE
(quasi-reference electrode).

Steroids with 6B-hydroxy groups have large oxidative peaks. 6B-
Hydroxycortisone had a very large oxidative peak, but because it was very expensive
(about $8000 per gram), all further work was done using the much less expensive 4-
androsten-6B—ol-3,l7-dione.

Electrosynthesis

W. In order to determine the efficienCy

and the products of the oxidation using GC-MS, first the supporting electrolyte must
be removed. About 99% (w/w) of the solute is supporting electrolyte. Many methods

were developed to remove the TBAF.

63

The steroids are soluble in benzene, but not in water. TBAF is insoluble in
benzene and slightly soluble in water. A trial solution containing prednisolone, 1,4-
androstadien-3,l l,l7-tlione, TBAF, and acetonitrile was evaporated under nitrogen,
then redissolved in benzene. A cloudy solution resulted. Then, the salt was extracted
with water, because the salt is more easily dissolved in water than the steroid. After
five to six extractions, the benzene is evaporated and the residue is dissolved in ethyl
acetate. Using ECD—GC, the extraction efficiency was calculated to be 2.1% for
prednisolone and 14% for the expected oxidized product, 1,4-androstadien-3,11,l7-
trione. Another extraction that works just as well as benzene/water is ethyl
acetate/water. This extraction follows the same procedure as benzene/water with about
the same yield. Many other methods of separation of salt and steroid were tried, most
unsuccessfully. One reasOn for the limited success is that TBAF and the steroids used
have approximately the same solubility in a variety of organic compounds.

C18 solid-phase extraction columns have been used for desalting matrices
because salts pass through unretained (17). A C18 column was conditioned with
methanol and then water. Then 1 mL of trial solution containing TBAF, 1,4-
androstadien-3,l 1,17-trione, and acetonitrile was placed on column. After washing.
with water and then hexane, ethanol was used as the eluent. Less than 0.5% of the
steroid was recovered, and a great deal of noise was present in the GC chromatogram.
This desalting procedure is probably most effective with salts that are more water-
soluble then TBAF.

Because the methods of salt separation have very poor efficiency, "benzene
recrystallization" was used to separate out the salt. The electrochemical solution,
dissolved in acetonitrile, is evaporated under nitrogen. A small amount of benzene is
added, and the solution is cooled in an ice bath. After several hours of cooling,
crystals of salt form and are filtered. The filu'ate, containing the steroid, is evaporated

and reconstituted in ethyl acetate for further analysis. A trial electrochemistry solution

64
containing the supporting electrolyte tetrabutylammonium tetrafluoroborate (TBAF)
and l,4-androstadien-3,11,l7-trione went through this procedure. TBAF was
recovered with an 85% yield and the steroid was recovered with a 90% yield.

The benzene recrystallization method was also tried with a real sample, a
prednisolone electrochemical oxidation. The approximate recovery of TBAF was
93%. The recovery of the oxidized product (1,4-androstadien-3,11,17-trione) was less
than 1%. The recovery of starting material, prednisolone, was 25%. This suggests that
the extraction procedure works, and the low yield of product is due to something other
than poor extraction efficiency.

W. Ultraviolet spectroscopy (UV) studies were
undertaken for two reasons. One was to determine the maximum wavelength of
selected steroids for use with HPLC-UV detection. The other was to see if the
differences between the unextracted and the salt-extracted electrooxidation products of
prednisolone were observable by UV spectroscopy.

All studies were done using a Varian DMS 200 UV-Vis Spectrophotometer.
The maximum wavelengths for a series of steroids in methanol are shown in Table 2.4.
Most steroids have a maximum wavelength between 235-250 nm, which justifies the
use of 240 nm for the HPLC detector. Cholesterol, with a wavelength below 220 nm,
is one exception. The supporting electrolyte, TBAF, also has a maximum wavelength
under 220 nm.

The actual electrochemical reaction products, all dissolved in acetonitrile, were
also studied. The maximum wavelengths for the standards in acetonitrile were less
than those in methanol. Prednisolone had a maximum wavelength of 233 run, while
l,4-androstadien-3,1 1,17-uione had one of 240 nm. The unextracted electrooxidation
of prednisolone had two maxima, one at 236 nm and the other at less than 200 nm.
The maximum wavelength of the benzene recrystallized product is 236 nm, just the

same as one of the unextracted peaks. The extra peak in the unextracted sample is

65
probably not due to TBAF. An almost saturated solution of TBAF in acetonitrile was
analyzed and no response was noted in the 190-300 nm range.

Table 2.4: UV Data for Selected Steroids in Methanol

Maximum ' Molar absorptivity
Sample Wavelength (nm) (L mol'lcm'l)
l,4-Androstadien-3,l l, 17-trione 237.8 14500
Prednisolone 242.4 9770
Cortisol 241.0 13400
4-Androsten-3,l 1,17-trione 236.3 13800
Cholesterol < 220 ----
4-Cholesten-3,6-dione 250.1 7870
TBAF < 220 ----

Wally. In order to identify and quantitate

both the starting material and the oxidized products from the electrochemical
oxidation, a reverse phase HPLC method was developed. A Waters 600 multisolvent
delivery system coupled with a Waters 712 WISP autosampler, a Waters 740 data
module, and a Kratos Spectroflow 783 absorbance detector set at a wavelength of 240
nm with a range of 0.2 AUFS. Either 20 11L of the steroid standard or 10 11L of the
real reaction solution was injected into a 150 x 4 mm Biorad ODS-55 C18 column.
The flow rate was 1 mL/min.

The major objective of the HPLC method development was to separate TBAF
fiom prednisolone and l,4-androstadien-3,1 1,17-trione. Isocratic systems consisting
of varying amounts of water/methanol and water/methanol/acetic acid were not
successful. A gradient system which varied the methanol concentration from 100% to

80% in 20 minutes also did not work. An acetonitrile/water solvent system seemed

66
most promising. The results from this study are shown in Table 2.5. The solvent
system used for the analysis of the real electrochemical samples was 70% water/30%

acetonitrile.

Table 2.5: Reverse Phase HPLC of Electrochemical Solution

% Water % Acetonitrile Results
40 60 TBAF and prednisolone were not separated
60 40 TBAF and prednisolone were not separated
70 30 TBAF and prednisolone separated
80 20 Did not work

Bulk Coulometry of Prednisolone

W. About 35 mg of prednisolone was placed in the third
compartment of a tluee-compartrnent cell (see Figure 2.2) with acetonitrile as the
solvent and TBAF as the supporting electrolyte. Controlled potential coulometry was
run at 2.1 V. The decay of current was monitored. After two and a half hours, it did
not quite reach zero, but this may be due to impurities. During the experiment, 7.06
coulombs were passed.

Results. The oxidized solution was analyzed using HPLC with the 70%
water/30% acetonitrile solvent system. The chromatogram is shown in Figure 2.4(a).
From a comparison with istandards, the apparent formation of l,4-androstadien-
3,11,17-trione is 0.9%, and the recovery of the starting material (prednisolone) is 56%.
However, if the electrochemical sample is spiked with an authentic sample of 1,4-

androstadien-3,11,l7-trione, the peaks are separated as shown in Figure 2.4(b).

67

 

 

 

 

 

 

2
1
3 q 4
0 9 17 26
Time (minutes)

Figure 2.4: HPLC of electrochemically oxidized prednisolone (not extracted).
Suspected identity of peaks: (1) TBAF, (2) prednisolone, and (3) oxidized product.
(a) Electrochemical oxidation product. (b) Electrochemical oxidation product spiked
with 1,4-androstadien-3,1 1,17-trione. Peak 4 is the spiked compound.

 

 

 

 

 

 

 

La

0 9 1 7
Time (minutes)

Figure 2.5: Fractions collected from the HPLC analysis of the electrooxidation
product of prednisolone.

68
Table 2.6: Suspected Identity of HPLC Fractions

HPLC Fraction GC Peak Identity
1 1 nothing (perhaps solvent front)
2 1 plasticizer
2 unknown
3 1 unknown
2 plasticizer
4 1 unknown
2 plasticizer
3 m/z 282 (corresponds to prednisolone -
substituents at C-l7 — water)
4 l,4-androstadien-3,1 l, l7-trione
5 1 nothing
6 1 l,4-androstadien-3,l 1,17-trione
' 2 prednisolone

In order to determine the identity of the compounds represented by these peaks,
fractions were collected (see Figure 2.5). Six injections of 50 11L each were made.
The fractions were lyopholized, then reconstituted for further analysis by GC-MS. The
GC-EI-MS results are summarized in Table 2.6. Note that the supporting electrolyte
produced no GC peaks other than baseline noise because it breaks down in the injector.
Some fractions may have contained substantial amounts of supporting electrolyte.

The peak previously assigned to the supporting electrolyte was incorrect
because a sample of this compound at the concentration in the electrooxidation product
did not give a response. Further UV analysis shows that TBAF does not have a
response at 240 nm, the wavelength used for all HPLC analyses. This peak, at 4.3
minutes, disappears after the sample has had the supporting electrolyte removed, so
somehow whatever is eluting must have some relation to the supporting electrolyte or

is just extracted along with the supporting electrolyte.

69

The other problem is the peak correponding to the suspected oxidation product.
This retention time is shifted for all the electrooxidations, whether or not they have had
the salt extracted. If prednisolone is oxidized using PCC oxidation, the peak retention
times for oxidized product and 1,4-androstadien-3,l 1,17-trione coincide.

The results from the quantitation of the peaks again show the inefficiency of
the electrooxidation. By PCC oxidation, there was a 50% recovery of the oxidized
product. When the electrochemical product was run unextracted, 14% of the
prednisolone and 11% of what is thought to be the oxidized product were recovered.
The prednisolone peak is overlapped by several peaks; therefore, it is hard to
quantitate. After the electrochemical oxidation product has had the salt removed by
benzene recrystallization, 6% of the prednisolone and 5% of what is thought to be the
oxidized product is recovered. This shows that some steroid is lost in the extraction
procedure.

Electrochemistry of Steroids Other than Prednisolone

Prednisolone was the first steroid to exhibit any type of electrochemical
response. In the search for more steroids that might exhibit an electrochemical
response, prednisone was thought to be promising because it has an ll-carbonyl rather
than a ll-hydroxy group, thus having one less functionality to oxidize than
prednisolone. Prednisone, however, did not give an electrochemical response. This
may be because its oxidation potential is greater than that of the solvent (acetonitrile).

Cortisol (4-pregnen-1lB,17a,21-tliol-3,20-dione) seemed to be a logical next
choice because it is structurally similar to prednisolone, lacking only one double bond.
Cortisol did display an oxidative response (potential at 2.1 V vs. QRE). For the bulk
electrolysis, about 23 coulombs should have been passed if all the steroid had been
oxidized; only about 11 coulombs were passed. The resulting solution was yellow, but
the color may have been due to an impurity. Analysis by GC-MS showed that the
product has two components. One is 4-androsten-3,1 1,17-trione, the expected

70
oxidation product. The other was not identified, but could be the breakdown of
cortisol in the gas chromatograph. The starting concentration of cortisol was not
available because more was added during the course of the reaction; therefore, the
percentage oxidized is not known. Pyridinium chlorochromate oxidation (PCC)
oxidation of cortisol gave a 30% yield. '

Cholesterol was chosen as the next steroid to oxidize because it is relatively
inexpensive and has been used for oxidative electrochemical detection in HPLC.
Cholesterol has many oxidation products, even when oxidized chemically. For
example, chemical oxidization using PCC resulted in the formation of two products.
The major product is possibly 4-cholesten-3,6-dione. The minor product did not result
in an interpretable spectra and was not identified.

Cholesterol was electrolyzed at a voltage of 2.2 V. After an hour, five
coulombs had been passed. About seven coulombs should have been passed if all the
substrate had been oxidized. The changes in color during the experiment were
interesting. The solution first turned a pink, then a dusty rose, and ended as a forest
green solution. Not all of the cholesterol had dissolved, so the resulting solution was
filtered to get rid of the remaining crystals of cholesterol. Six products were formed,
three of which have been identified as 4-cholesten-3,6—dione, 4-cholesten-3—one, and
5-cholesten-3-one.

6B-Hydroxycortisone had an oxidation potential of 2.3 V vs. QRE, but also had
a larger current response (8.4 x 10'5 A) than even prednisolone (5.6 x 10‘5 A). This
larger response may be due to the ease of oxidation of the 6-hydroxy group. Bulk
coulometry studies were not performed on 6B-hydroxycortisone because it is a very
expensive steroid. Instead, 4-androsten-6l3-ol-3, 17 -dione was used, which gave the
same magnitude of response and was much less expensive. Almost no charge was

passed (0.8 coulombs) and, not surprisingly, no product was converted.

71
Why is the Electrosynthesis of S teralds So Inefficient?

As has been stated before, the recovery of oxidized steroids has been very low,
ranging from less than 1 to 10% depending on the analysis method used. One reason
for the inefficiency may be that the oxidized material is adsorbed to the electrode after
it is formed, thus rendering the electrode useless for the oxidation of the rest of the
material in solution. This conclusion was reached because the current decreased on
each successive cyclic voltammetric cycle when the voltage is scanned from 0 to 2.5 V
(vs. QRE). If the voltage is scanned to -1.5 V, the current stays almost constant (see
Figure 2.6). If, after oxidation, the voltage is scanned to -2.0 V, the electrode "cleans"
itself; i. e., this procedure reduces the oxidized product back to prednisolone and then
the cyclic voltammagram appears normal. This shows that the product of the oxidation
appears to remain on the surface of the electrode (see Figure 2.6).

Probable Methods to Overcome Fouling

Fouling of the electrode may depend on the nature of the analyte, the
characteristics of the solvent, and the material used for the working electrode. The
solvent and working elecuode were varied next, using prednisolone as the analyte.
When acetonitrile was used as the solvent, prednisolone stuck to the glassy carbon
electrode. Benzonitrile has the same potential window as acetonitrile, but different
solvent properties. When dissolved in benzonitrile, prednisolone adhered to both the
platinum and glassy carbon electrodes. No oxidative peak was present when using the
gold electrode. -

Desorption Of an Electrode-Direct Probe

After discovering that the electrooxidized products from steroids adhered to the
electrode, procedures were developed to advantageously use this phenomenon. If the
oxidized product absorbed on a metal surface is placed in a mass spectrometer, the
surface compounds may be desorbed and analyzed.

72

151 t.

.SE 1 e

.T -i.5 -2
Potential (V)

OE 1

Current (ILA)

-1E 1

 

 

“1.5E 1 _

SE 0

48 0

Current (11A)

1E O

 

DE 0

 

 

r- o’ . . “1.6 ‘1.8 '2

15 o - Potential (V)

-ZE 0 _

 

Figure 2.6: Cyclic voltammagrams of prednisolone. (a) The current stays
approximately the same if the potential is scanned from -l.5 V to + 2.5 V. (b)
However, if the potential is scanned from 0 to -2.2 V repeatably, the amount of current
decreases, proving that material is adhering on the surface.

73

An electrode that can fit into either the probe assembly of a HP5985 or a JEOL
AX505 mass spectrometer was designed (see Figure 2.7). A thin sheet of platinum foil
was cut into a 2 x 11 mm rectangle and attached to a platinum wire. The wire was
attached to the gold nest in the case of the HP5985 assembly with a series of electrical
connectors. For the JEOL assembly, the electrical connectors were placed into the
direct probe and held by an allen screw. It is estimated that the electrode, with a
surface area of 44 m2, can possibly hold up to 454 ng of oxidized steroid, which

should be more than enough to detect.

 

 

 

1:. 1/4"

 

 

platinum foil

electrical connector (fits inside conventional El probe)

Figure 2.7: A platinum electrode that fits into the direct probe of either a HP5985 or
JEOL 505 mass Spectrometer.

In order to deposit steroids on the platinum surface, 3-4 mg of steroid were
dissolved in acetonitrile with 0.2 M TBAF in a one-compartment cell. The TLBE
program on the BAS Electrochemical Analyzer was used to deposit the thin films.

' If a steroid solution is deposited directly on a platinum electrode using a
syringe, the low mass end of the spectrum is the same as if the steroid had been
introduced via direct probe (DIP); however, the high mass end is of lower abundance.
A quadrupole instrument such as the HP5985 forms steroid molecular ions in low
relative intensities; therefore, the JEOL AX505, a magnetic sector instrument, was
used for all subsequent experiments.

Several studies were undertaken to determine the optimum probe position in

the JEOL. Not surprisingly, it was better to have the sample face the filament than be

74
opposite the filament. The best reconstructed total ion chromatogram (TIC) and
spectrum was obtained when the electrode was perpendicular to the filament. In this
case, the TIC and spectrum looked just like the results obtained using a DIP.

From the area under the curve of a cyclic voltammagram, it was determined
that approximately 300 ng of material, possibly 1,4-androstadien-3,11,17-trione, the
oxidized product of prednisolone, had been deposited on the electrode. From the
analysis of the area under the oxidation curve, the oxidation of testosterone on an
electrode deposited about 100 ng of material. Only about 50 ng are necessary to detect
the oxidized compound by EI-MS if the steroid is placed on the platinum probe
directly (no electrochemistry involved). Therefore, enough compound should have
been electrochemically oxidized on the probe for detection by mass specuometry. The
supporting electrolyte can also be detected by mass spectrometry. The electrode was
usually dipped in acetonitrile before analysis, although sometimes this step was
omitted. When this step was omitted, there were a few more ions from the supporting
electrolyte, which desorbs at a later time than the steriod, but no other differences were
noted.

Two very careful studies were undertaken to determine the feasibility of
placing the electrode directly in the source. One experiment used EI as the ionization
technique; the Other, ECNI.

The contributions to the direct probe EI profile of the elecuode from
background ions in the source, from the clean electrode, from the supporting
electrolyte and from unoxidized material were each determined in separate
experiments. No contribution to the background was observed from either the
connectors used to hold the electrode in the direct probe apparatus or from the
platinum itself. Placed directly on the probe, the spectrum of both 1,4-androstadien-
3,11,17-trione (oxidized prednisolone) and 4-androsten-3,17-dione (oxidized
testosterone), looked like the spectrum obtained from a normal direct probe experiment

75

except for a diminished molecular ion. After prednisolone and testosterone were
oxidized with the TBLE program, each electrode was dipped in acetonitrile and then
placed inside the JEOL 505 ion source. In the case of both the "oxidized testosterone"
and the "oxidized prednisolone," no ions indicative of these oxidized products were
seen. In fact, the spectra looked surprisingly similar to spectra taken when the
electrode is merely dipped in the electrolysis solution (that is, the solution containing
just testosterone or prednisolone and supporting electrolyte). Perhaps, the amount of
oxidized material on the electrode is below the detection limit of electron impact
ionization.

The oxidized material electrochemically adsorbed on a platinum electrode was
also analyzed by ECNI. ECNI has less fragmentation, so if the steroid were present,
the molecular anion would be detected, thus simplifying the specu'um. Also, if the
material on the electrode was an intermediate, such measurements would provide
molecular weight information about this intermediate. Samples larger than 50 ng of
material can be detected.

In previous studies, an unusual ion isotope pattern was noted. Distinct patterns
were repeated at 305/307/309, 339/341/343/345, and 358/360/362/364. These are also
observed for clean electrodes when ammonia is used as the reagent gas. What these
ions represent is a mystery.

When prednisolone oxidized on a platinum surface is analyzed, only
background ions are present. The most intense peaks at mlz 233/235 represent
rhenium oxides and are present in all backgrounds. Again, if there is material oxidized
on the electrode, the amounts are too small to detect. In conclusion, neither E1 nor
ECNI is a viable technique for detection of electrochemically oxidized steroids an a

platinum surface.

76
Electrochemical Deposition an a Surface--FAB
Although the EI/ECNI desorption studies were not successful, it was thought
that fast atom bombardment (FAB) could possibly desorb a steroid that had been
electrochemically oxidized on a platinum surface. A FD-FAB probe (see Figure 2.8)
was modified, replacing the metal originally used with platinum. The only problem
was that the repeller contact was not notched into the ceramic used; therefore, the

repeller has to be adjusted in a previous analysis.

2 x 6 mm platinum toil

/

r—J—.

 

 

 

 

 

U U 4——— platinurn wire

Figure 2.8: A platinum electrode that fits into the FD-FAB probe of the JEOL HX-
110 mass spectrometer.

 

 

 

In preliminary studies using a regular FAB probe and the JEOL HX110 double
focussing mass spectrometer, l,4-androstadien-3,11,l7-trione was used to determine
the best matrix for steroid analysis. Glycerol, thioglycerol, nitrobenzyl alcohol, and
5:1 dithiothreitol:dithioerythritol did not work. With glycerol/HCI (0.1 M), a
substantial [M-l-H]+ peak was observed at m/z 299, but mlz 299 also corresponds to
[3G+Na]"'. The test compound then was changed to 4-androsten-3,17-dione, which
does not have any interferences with glycerol ion. In the positive ion FAB spectrum
with glycerol/HCI as the matrix, a substantial [M+H]"‘ peak was observed at mlz 287.
The addition of 20% TBAF does not diminish the analyte peak. The detection limit is
150 ng for 4-androsten-3,17-dione. In a study of ketosteroids, DiDonato and Busch

had to use derivatization to obtain FAB spectra of 1 ug of steroid (18).

77

Negative FAB matrices such as glycerol, hexamethylphosphoramide, and
triethanolamine were used for some more electrophilic steroids such as l,4-
androstadien-3,ll,l7-trione and 1,4-androstadien-9ct-fluorine-16a-methyl-3,1l,l7-
trione. No response was obtained for l,4-androstadien-3,ll,17-trione, but 1,4-
androstadien-9a-fluorine-160t-methyl-3,11,17-trione had a spectrum with major
contributions from peaks at mlz 329 (30%) corresponding to [M-H]" and m/z 310
corresponding to [M-I-IFJ'. ‘

To test the FAB probe under real electrochemical desorption conditions,
testosterone and dexamethasone were each separately oxidized onto the electrode.
Neither gave any signal from the analyte, oxidized or unoxidized, but the supporting
electrolyte was observed. The amount of dexamethasone oxidized on the electrode
was obtained using the electrochemical data. From analyzing the area under the curve,
there were 230 ng of material on the electrode. Because this is a two-sided electrode
and only one side faces the fast atom beam in the source, effectively only 115 ng of
material are available for analysis. Because 115 ng is less than the detection limit of
steroids by FAB, no signal could be observed. Therefore, desorption by FAB is not a

viable technique.

Modified Nickel Electrodes: An Alternative Synthesis Method

A modified nickel oxide hydroxide electrode has been reported to convert
primary alcohols, mar-diols, and secondary alcohols into their respective carboxylic
acids, dicarboxylic acids, or ketones (8). Because the oxidation rate decreases with
increasing steric hindrance, 3-hydroxysteroids can be selectively oxidized. For
example, 4-androsten-3B,l7B—diol was selectively oxidized to 4-androsten-3-one-l7fl-
01 in a 50% yield. Other products included a 38% yield of 4-androsten-3,17-dione and
a 7% yield of unconsumed starting material. Long reaction times (about 24 hours)

resulted in a 80% formation of 4-androsten-3,17-dione.

78

(1) Ni(OH)2 + OH- zNiOOH + H20 + e-

(2) NiOOH + RzCHOHabs —>- Ni((J)H)2 + RQCOH

 

(3) RQCOH —> R20=O

overall R20HOH + 20H: -—> ch-O + 2H20 + 2e-

Figure 2.9: Mechanism for the electrochemically regenerated nickel oxide hydroxide
' electrode.

In this method of oxidation, nickel hydroxide is coated on a nickel foil
electrode. The nickel hydroxide reacts with hydroxide ion from the supporting
electrolyte KOH in order to form nickel oxide hydroxide on the surface of the
electrode (see Figure 2.9 for the appropriate equations). The nickel oxide hydroxide
reacts with the absorbed alcohol in the rate-determining step with the transfer of the
proton to the surface, thus regenerating the nickel hydroxide on the surface of the
electrode. The steroid radical intermediate is easily oxidized to a ketone.

Experimental

We. The nickel (II) hydroxide is deposited according
to published methods (19). A 2.3 cm x 3.0 cm nickel foil was placed in a 0.1 N
NiSO4°6HzO, 0.1 N sodium acetate, and 0.005 N sodium hydroxide solution. A silver
wire reference and a platinum gauze auxiliary electrode were connected to the nickel
foil with a PAR Potentiostat/Galvanostat Model 273 used in the galvanostat mode. A

current density of 0.5 mA/cm2 was applied in the following manner:

79

a.) Step to -9 mA

b.) Hold for 60 s

c.) Step to +9 mA

d.) Hold for 60 s

e.) Repeat
This cycle was repeated for 8 to 10 minutes to ensure an even, black coating of the
electrode. To clean the elecuode after use, it was immersed in a 20% HQ solution for
a few seconds.

Electrolysis. About 10-20 mg of steroid were placed in a 30 mL TLC bottle
with 20-25 mL 50% t-butanol/50% 0.1 M KOH. The steroid is sparingly soluble in
this mixture, so rapid mixing during the entire elecuolysis procedure is necessary. A
silver wire reference electrode and platinum gauze auxiliary electrode are used. The
PAR Model 273 is used in the galvanostat mode; that is, the current is held constant
while the potential changes in order to keep a constant current. Therefore, everything
in the solution that can be oxidized will be in order to keep the current constant. A
current density of 1.2 mA/cm2 was maintained. The reaction time was on the order of
24-28 hours.

WW5. The authors (8) recommended the following isolation
procedure. About 0.2 g of NaCl is added. Ether extraction (3 x 25 mL) was followed
by drying the ether fraction. The ether was removed, and a solution of the solute
reconstituted in ethyl acetate. A practice extraction of 1 mg of pure 4-androsten-3,l7-
dione proved that the extraction efficiency was only 20%.

A second attempt at isolation was based on the assumption that KOH did not
dissolve in ethyl acetate. The 50% t-butanol/50% 0.1 M KOH solution was evaporated
to dryness and then dissolved in ethyl acetate. The KOH crystals did not precipitate as

expected, even after hours of cooling in an ice bath. Even so, the ethyl acetate solution

80

was analyzed by GC. The extraction efficiency was 60%, and no problems due to the
KOH were noted.
Results

Prednisolone (1,4-pregnadien-1lB,l7a,21-triol-3,20-dione) was oxidized using
the nickel hydroxide electrode for twelve hours. The solution turned yellow green
during the electrolysis in which 7.928 coulombs were passed. About one-fourth of the
electrode surface had degraded. Only prednisolone itself was recovered after
electrolysis. After all, 3-hydroxy groups are especially amenable to this procedure and
prednisolone does not have any 3-hydroxy groups. Also, the substitution at the 17-
position may be difficult to oxidize.

In the literature study, 4-androsten-3B,l7l3-diol was easily oxidized; therefore,
it was chosen as a test compound. The current was kept constant at -21 mA. After 24
home, 1387 Q of charge had been passed. The working electrode had lost all of its
coating, and some black material was floating in the green solution. The auxiliary
electrode was coated with a green layer, much like the color of the nickel sulfate

solution. The yields were:

4-androsten-3B,17B-diol (starting material) 1.3%
4-androsten-3-one-17B-ol 32%
4-androsten-3,17-dione 10%

The extraction procedure has a 60% extraction efficiency for 4-androsten-3,l7-dione.
AnOther attempt at oxidation was conducted with a constant current of ~18 mA.
After 26 hours, 1881 coulombs were passed. The nickel oxide surface on one side of

the electrode degraded, but the other side remained intact. The yields were:

4—androsten-3B,17l3-diol (starting material) 85%
4-androsten-3-one-17B-ol 13%
4-androsten-3,17-dione 3%

All the starting material is accounted for this time.

81

The nickel oxide hydroxide electrode is not a good method to prepare oxidized
steroids. The reaction times are long, there is no selectivity (everything in solution is
oxidized, including supporting electrolyte and solvent, in order to keep the current
constant), and too many side products are formed. The diketone (4-androsten-3, 17-
dione) is very difficult to obtain; it is perhaps obtainable if a high current density is
maintained, the electrode surface is redeposited many times during the reaction, and
the reaction is run for several days.
Conclusions

Electrochemical oxidation had been proposed as a more selective and more
efficient method of preparing ketosteroids. Synthetic oxidation on a planar electrode
had a low yield as the electrode was fouled by the reaction. If the fouled electrode was
placed directly inside a mass spectrometer source, the material on the electrode could
not be detected by either E1, ECNI, or FAB. A modified electrode, the nickel oxide
hydroxide electrode, was used to oxidize steroids. This electrode has found use in
preparing 3-keto steroids but, even under harsh conditions, synthesizing fully oxidized
steroids was very inefficient. Pyridinium chlorochromate oxidation, even with its
problems of low yield and lack of selectivity, is still the best choice for preparing

ketosteroids.

The Determination of Reduction Potentials

Reduction potentials . were determined for the comparison with ECNI studies
described in Chapter 4. In this section, the electrochemical aspects of the reduction
potential study will be undertaken.

Reduction potentials of selected ketosteroids have been obtained by varidus
researchers using polarographic techniques. Reduction potentials obtained in different
labs cannot be directly compared, even if the solvent and the supporting electrolyte are

the same. Small differences in the purity of solvents used, the amount of moisture in a

82
system, or the difference in the liquid junction potentials can result in large differences
in reduction potential. As a basis for comparison between two sets of data, at least one
(and preferably more) of the compounds must be repeated in each set. This criterion
was not observed in previous work on ketosteroids in the literature; therefore,
reduction potentials had to be obtained in this laboratory. Table 2.7 summarizes the
findings of the literature survey.

The effect of functional groups on the reduction potentials of steroids has been
extensively studied (25). It had been generally accepted that substitution in a position
remote from the electrophilic group has little effect on the half-wave potential.
However, in the case of steroid compounds, substituent effects can be transmitted from
other rings and from remote positions. It was assumed that in most cases a potential
difference of 0.01 V is significant, but in all cases, reduction potentials measured by
only one research group were compared.

Substitutents at 017 on the steroid nucleus affect a carbonyl group at C-3. For
example, in the comparison of the last three compounds in Table 2.7, 5a-androstan-
3,17-dione is much more difficult to reduce than 5a-cholestan-3-one. The influence of
the 17-substituent depends on the solvent system; in 50% ethanol at pH 10.5, the effect
of substitution at the 17-position on 4-en-3-one steroids is minimal.

Substitution at C-ll had much less effect on the carbonyl group at 03;
however, the author was comparing the half-wave porentials of two steroids that
differed by the presence or absence of an ll-hydroxy group. However, if the
substitution at C-ll is a carbonyl group, the steroid is more easily reduced than if a
hydroxyl group at C-ll is present. A carbonyl substitution at C-ll is more easily
reduced than a double bond at the same position, but the reduction potentials differ by
only 0.1 V.

83

mm cod-

8.7
~N.7

no.7
cm n07

wNN-

3am-

. 5N-
NN wm.-

2.4-

3.7 Jud-

34.7

8.7 :07

5.7

vm.7

8.7

no.7

mN 3.4-
NN.N- $5.7

mN.N- dfiN-

vm mud- .cfiN-

coerce: Gem .9 Z Sm

_U<E E nod .EOHm some

as no .mofi son

3 me doom see

Seam-r E fio .IOLm «moo

Us 3 fio £20 seen

HE E fie .335, E fio MED

neeeeeee

595%-oeoe-:.m-eoaeee<4
eaceeeoeoeéaeoapeie
ooord~b=97§835
Bozo-ew-oeo-m-esmEQG
Beau-en-oeo-m-esmseau
oEED-em-oene-m-esmo—onu

3508 3.52 -o:o_o-ON.m-=o§=w2m-w.v
3808 3-62 -o:oێN.m-:o_oaeon-v._

oeomo-oNfi-caeon-v
ace-m-eamo_o:U-v

ace-7:53.30
o:O€-¢~.m-§:mo.&
enema-S ._ fim-Szmeae:<
aeo€-:.m-§mo€=<

cacao-S ._ _.m-eo.m8ee<-e
ocean-26.753524
oeo$h_.weo_eamoee<-n.m
once-S.m-=oEsaeee<e.§
oeofitdeoesmeeéee
Season-8638634;
cede-26-538238?
2256:8533?
oeo%-h~.m-:2m8o:<-v

Bozo eroeartaeoaeeeza-e
895 Tacomo-Sfi-eameaofix-v
access_-oeoe-:.m-eoaeee<-e

egg—ECU

83.383. .e geese.— ee_eeeoz 2.. .8 82.3 9.383: 8." uses

84

mm

mud-
36-
mod-

oEoom Ezmcofifiaafioahoh 2mg.
coax: 835882532309 75,—.
cote—no Banana—beefing. Exam—H
35.—um 39m

323.55ng min

2e53>8en<

o8o5-fib:e-75=w£¢-v

he In 63.28 352 08893-8975:on
833m sea .mgm $8 each-ew-oeomo-:.m-:Bm8ee<

3.2—es EN 035,—.

85

The effect of a substituted methyl group in the 6-position of 3-keto-a,B-
unsaturated steroids is again very minimal. Only occasionally do methyl substituents
have a significant effect on half-wave potentials.

A linearly-conjugated system is much easier to reduce than a cross-conjugated
system. 1,4-Pregnadien-3,20—dione-l7a-ol acetate has a half-wave potential of -1.22
V (vs. SCE) in 50% ethanol at pH 6. 4,6-Pregnadien-3,20-dione-17or-ol acetate in the
same solvent has a potential of -1.06 V. The position of the linearly conjugated system
within the steroid is important. In 90% ethanol at pH 8.5, 4,6-cholestadien-3-one has a
half-wave potential 0.06 V more positive than that of 3,5-cholestadien-7-one.

The dependence of reduction potentials on the number of carbonyl groups is
slight. Both androstan-3,l7-dione and androstan-3,11,17-trione have the same
reduction potential.

Halogen subsitutions have been studied within the steroid nucleus. The
reduction potentials become more negative in the sequence I < Br < Cl < F. In rigid
systems, such as the B and C rings of steroids, it is possible to distinguish the epimers

of a-halo ketones from polarographic data.

Cyclic Voltammetry
Experimental

A one-compartment cell (illustrated. in Figure 2.1) was used with a platinum
wire auxiliary electrode, a silver wire reference elecuode, and a platinum disk working
electrode. The purified solvent was uansferred under vacuum. TBAF (0.2 M) was
used as the supporting electrolyte. Reduction potentials were obtained on a BAS
Electrochemical Analyzer. Typical conditions were the following:

3-4 mg of steroid dissolved in about 7 mL of solvent

Potential range = 0.0 to -2.0 V

Scan rate = 200 mV/s.

86
Results

Acetonitrile should be a good solvent for the reduction of steroids because it
has a large potential window. The CV obtained has no forward or reverse peaks. The
cell was open to atmosphere, and more steroid was added because it was thought that
the concentration was too low. Now, both forward and reverse waves were noted for
the test compounds, 1,4—androstadien-3,l l,l7-trione and 5a-androstane-3-one.

Silver wire is a QRE. In order to obtain reproducible values for oxidation-
reduction potentials, a compound with a known reduction potential must be added and
the CV reanalyzed. A good choice for the internal standard is ferrocene, which is
soluble in a variety of organic solvents and has well-characterized electrochemical

properties. Ferrocene has the following equilibrium (20):

Fe(C5H5)2 + e' <—-> Fe(C5H5)2 E1 ,2 = +01% :l: 0.007 V vs. SHE in acetonitrile

After adding more sample, the data obtained for reduction potentials are in Table 2.8.

Table 2.8: Reduction Potentials of Steroids in Acetonitrile

Sample E Sample E Ferrocene E 2 Sample
vs. QRE vvs. QRE {1 vs. SHE

1.4-11-3.1 1,17-tn'one" -1.35 +0.75 -1.91

Androstane-30m -1.48 +0.72 -2.01

*A = androstadien '

In a second attempt, about 10-15 mg of steroid was used. Only after the cell
was opened and more steroid was added did a reductive peak appear.

Methylene chloride was also tried as a solvent. With 1,4-androstadien-3,11,‘l7-
uione, a small peak appeared at -1.82 V (vs. QRE). After the cell was opened and
more steroid was added, it took about four reductive-oxidative cycles to get a

reproducible peak that by then had shifted to -1.60 V (vs. QRE).

87
1,4-Androstadien-3,l7-dione in methylene chloride was also analyzed. A very
small reductive peak that shifted its potential with successive cycles was observed, but
this could have been a contaminant. After the cell was opened and more steroid was
added, the reduction potential was at -1.6 V.
After the electrochemical cell was opened to air, the peak potential always
changed and grew larger. This is due to the oxygen wave. None of the steroids could

be analyzed by cyclic voltammetry using a platinum electrode.

Differential Pulse Polarograph y

As shown in the previous section, cyclic voltammetry does not work for the
determination of reduction potentials for steroids. Almost all of the reduction
potentials in the literature were determined using polarographic techniques.

Classical polarography involves measuring current as a function of potential at
a dropping mercury electrode. The electrode continually regenerates, with each drop
corresponding to a new experiment. Therefore, there are no problems with
contamination or with electrode fouling.

Because classical polarography involves recording the current continuously
over the lifetime of a drop, current fluctuations due to the growth and fall of the drop
are also recorded, generating a great deal of noise. In order to decrease the noise level,
sampled-dc polarography has been developed. In sampled-dc polarography, the
current is measured for the last 5-20 ms of the drop. The resulting polarogram is a
smooth curve with a minimum of noise. The detection limit and resolution are the
same as for classical polarography.

In differential pulse polarography (DPP), as in classical polarography, a dc
potential is increased linearly with time. However, a dc pulse of an additional 20 to
100 mV is applied for 60 ms before the detachment of the mercury drop. The current

measured near the end of the pulse is subtracted from the current measured just prior to

88

the pulse. The, resolution and sensitivity of this technique are better than that for
classical polarography.
Experimental

All polarograms were generated using a EG&G PAR Model 303A static
mercury dropping electrode connected to an EG&G PAR Model 273
Galvanostat/Potentiostat operated by EG&G PAR Model 270 Electrochemical
Analysis System software on an IBM computer. The cell had a capacity of about 20
mL of solution. At least 10 mg of steroid were used for each analysis. A platinum
wire auxiliary electrode and a Ag/AgCl reference electrode were used. All samples
were deaerated with nitrogen for four minutes before analysis. The conditions used for
sampled-dc polarography are the following:

drop time = 1 5

drop size = small

scan rate = 0.01 V/s

scan increment = 0.01 V

current range = 10 11A

potential range = -0.2 to -3.1 V (dependent on solvent)
For DPP, all of the above conditions were used as well as:

pulse height = 0.025 V

pulse width = 0.05 s
The reference electrode, Ag/AgCl, was supplied by the manufacturer. An SCE is not
available for this model dropping mercury elecuode (DME). According to the DME
manual (21), the Ag/AgCl electrode has a potential difference of 50 mA from a SCE.
Results 7

Several solvent systems in the literature provide a basis for comparison of

reduction potentials of steroids. Two of these were attempted. It is difficult to

89
compare the experimental results with the literature values because experimental
conditions and reference electrodes make a difference.

The first experiment compared two solvents systems and two methods of data
collection. Androstan-3,l7-dione was used as the test compound. Solvent system I
(22) consisted of 1.5 mM steroid made up in a 25 mL volumetric flask. After the
steroid was dissolved in a small amount of 95% ethanol, 1.25 mL of 1 M
tetrabutylammonium chloride (Southwestern Chemical, Austin, TX), 1 drop of Triton
X-100 (Aldrich), and 1 mL water were added. The solution in the volumetric flask
was diluted to mark with 95% ethanol. Solvent system 11 consisted of about 1.5 mM of
steroid in a 50% DMF (Mallinkrodt HPLC grade)/50% water with 0.1 M lithium
chloride (Mallinkrodt) as the supporting electrolyte (23). One drop of Triton X-100
was added before analysis.

In this experiment, sampled-dc polarography was compared to DPP. Using
solvent system I, the reduction potential of androstan-3,17-dione was 2.0 V (vs.
Ag/AgCl) under DPP conditions. No reductive wave was seen using sampled-dc. In
solvent system 11, the reduction potential under DPP conditions was -1.6 V and -l.5 V
for sampled-dc polarography. The reductive waves were sharper and larger using
solvent system 11; therefore, the use of solvent system I was discontinued.

In the next experiment, the reproducibility of dc-sampled and DPP were
compared. Both l,4-androstadien-3,17-dione and 4-androsten-3,17-dione had been
analyzed in the literature study. Each compound was analyzed at least twice under

both dc-sampled and DPP conditions. DPP values are more reproducible.

Table 2.9: Comparison of Experimental and Literature Half-Wave Potentials

Compound Literature Em Experimental Em AEI, (V)
V vs. SCE V vs. Ag/AgCl L't -2Exp.
1,4-Androstadien-3,l7—dione -l.54 -l.63 :1: 0.01 0.09

4-Androsten-3,l7-dione - l .74 -1.80 :l: 0.015 0.06

90
Table 2.9 summarizes the findings from this study. The reproducibility of the
experimental values is quite good. The difference in E1 ,2 values (column 4 of Table
2.9) should be the same for both compounds if they really track the literature study.
As shown by the data, there is a slight difference between them. '

A number of steroids and terpenes dissolved in DMF/water were analyzed by
DPP. Problems were encountered because not all steroids are soluble in this solvent
system. Table 2.10 summarizes the results of this study.

Acetontrile was used as a solvent for the determination of reduction potentials
because not all steroids are soluble in DMF/HZO. As shown in Table 2.1, acetontrile
has a larger potential window. Steroids (7-16 mg) were dissolved in 25 mL HPLC-
grade acetonitrile (Mallinkrodt) with 0.1 M TBAF.

Discussion

The 50% DMF solution was used for the determination of half-wave potentials
for terpenes because they all dissolved in this system. To compare these values to
those obtained in acetonitrile, it is necessary to subtract 0.44 V from the 50% DMF
reduction potentials.

Most terpenes have half-waves potentials greater than that of the solvent (-2.35
V); therefore, they have very little ability to capture an electron in solution. There are
some noteable exceptions. In the bicyclononanes (compounds l-S), compound 3
which has an (LB-unsaturated site is reducible while compound 1 is not. Compound 5
has an epoxide group which makes it fairly easy to reduce. In fact, compound 5 and
compound 12 have the same reduction potential, which shows that a dienone and a en-
epoxide have the same electron-capturing ability. Among compounds having a propyl
chain (compounds 16-20), only compound 19 has an observable half-wave potential.
The presence of a double bond in both the propyl chain and in the alpha position to the
carbonyl in the ring is important for reducibility. .

91

Table 2.10: Reduction Potentials Using the DMF/H20 Solvent System
NB: -2.35 V is the maximum reduction potential

Compound’ El/z (V)
Androstane-3,17-dione norhing
4-Androsten-3,l7-dione -1.84
1-Androsten-3,l7-dione -1.73
4-Androsten-3,11,l7-trione -l.8l
4-Androsten-3,6,17-tlione -0.93, - l .55
1,4-Androstadien-3, l7-dione -1.67
1,4-Androstadien-3,l 1,17-trione -1.62
1,4,6-Androstatrien-3,17-dione -1.44
4-Pregnen-3,20—dione -1.82
4-Pregnen-3,1 1,20-trione -1.82
4»Pregnen-3,6,20—trione -0.93, -2.08
4-Pregnen-16a,17a-epoxy-3,20-dione -1.84
Pregnane-4a,5a-epoxy—3,20-dione -1.89, -2.17
1 nothing
3 - 1.80

5 -0.94

7 nothing
10 - 1.83

12 -0.93

13 -1.55

14 -1.82
16 nothing
l7 nothing
18 nothing
19 -1.72
20 nothing
21 ' nothing
22 nothing
23 nothing
24 -2. 13
25 - 1.72
26 nothing

*The structures of compounds 1-27 are presented in Figure 4.9.

92

Table 2.11: Polarography of Steroids Using Acetonitrile

NB: -2.97 V is the maximum reduction potential

E1/2 (V)

Compound

Androstane-3-one

2—Androsten- 17-one
Androstane-3-one-4B,SB-epoxy-l7-ol Acetate
l-Androsten-3—one-4B,5B-epoxy-17-ol Acetate
l6-Androsten- 17-cyano
Androstane-2a-bromo—3J7-dione
Androstane-2a-iodo-3,17-dione
Androstane-3,17-dione
4-Androsten-3,17-dione
1-Androsten-3,17-dione

4-Androsten-3,l 1, l7-trione
4-Androsten-3,6, l7-trione
1,4-Androstadien-3, 17 -dione
1,4-Androstadien-3,l 1,17-tlione
1,4,6-Androstatrien-3, l7-dione
4-Pregnen-3,20-dione

4-Pregnen-3,1 1,20-trione
4-Pregnen-3,6,20—trione

Pregnane- 16a, 17a-epoxy-3,20-dione
4-Pregnen-16a,l7a-epoxy-3,20—dione
Pregnane-4a,5a-epoxy-3,20-dione
4,6-Cholestadiene
4,6-Cholestadien-3-chlorine
4,6-Cholestadien-10t,2a-epoxy-3-one
4,6-Cholestadien-1a,20t-epoxy-3-ol Acetate
5-Cholesten-3-one

Cholestane-3,6-dione
4-Cholesten-3,6—dione
Cholestane-5a,6a—epoxy-4-one
Cholestane-40,5 B-epoxy- 3-one

nothing
norhing
-2.41

- 1.04, -2.05
-2.59

-0.69

-0.72
nothing
-2.27

- 1.03, -2. 12
-2.21

- 1.26, -2.26
-2.03

- 1.95

- 1.75

-2.23

-2.22

- 1.28

-2.41 '

-2.25

-2.41
nothing
nothing
-1.50, -l.82
- 1.80

-2.31
nothing
-1.32, -2.30
-2.68

-2.41

ECNI Response.

1.0
1.1
26
200
1.2
0.35
0.35
1.4
1.2
0.98
3.2
300
2.1
360
300
1.4
. 9.1
280
24
29
21
0.47
0.35
58

0.17
1.6
280
6.1
9.2

*5a-Androstan-3-one has been assigned a relative response of 1.0.

93

Steroids were chosen for reduction potential studies based on several factors.
A representative selection of various structural features was included. Very expensive
steroids and those obtained from the Steroid Reference Collection (Queen Mary
College, London) could not be used because not enough was available to make 20 mL
of an 1.5 mM solution. Simple steroids without many structural features, such as
androstan-3-one, 2-androsten-17-one, and androstan-3,l7-dione did not have a
response; 1. e., the half-wave potential of these compounds is greater than -2.97 V vs.
Ag/AgCl. In the literature, it has been observed that the group at the 17-position
sometimes affects the reduction potential. In this study, the effect of changes at the
17-position is small. 4-Androsten-3,17-dione has a 151/2 of -2.27 V as compared to 4-
pregnen-3,20-dione with a El ,2 of -2.23 V. The addition of a 6-carbonyl group to a 4-
en-3-one compound makes it much easier to reduce. 4-Androsten-3,11,17-trione has
half-wave potentials at -l.26 and -2.26 V. In this case, the substitution at the 17-
position makes a difference. An alkyl chain at the l7-position makes the steroid more
difficult to reduce, as in the case of 4-cholesten-3,6-dione with E1 ,2 at -l.32 and -2.30
V. The compound with the pregnane substitution at the l7-position, 4-pregnen-3,6,20-
trione, has only one half-wave potential at -1.28 V.

Depending on the other substituents in the steroid, the presence of an 11-
carbonyl may or may not affect the reduction potential. The reduction potential of 4-
androsten-3,l 1,17-uione is 0.06 V lower than the reduction potential of 4-androsten-
3,17-dione. Similarly, the reduction potential of l,4-androstadien-3,l 1,17-trione is
0.08 V lower than that of 1,4-androstadien-3,17-dione. However, the difference in
reduction potentials between 4-pregnen-3,20-dione and 4-pregnen-3,l 1,17-tlione is
only 0.01 V. A ll-carbonyl group in conjunction with the 17-carbonyl group often
results in a more easily reduced compound.

The presence of an a-halogen results in a higher reduction potential. For

example, androstan-3,17-dione has no response; however, if a a-halogen such as

94
bromine or iodine is present alpha to the C-3 carbonyl, the half-wave potential is on
the order of -0.7 V. The iodine has a more negative pctential than the bromine; the
opposite trend is expected.

The presence of an epoxide group has a small effect on the reduction potential.
4-Pregnen-3,20-dione has almost the same reduction potential as 4-pregnen-l6a,l70t-
epoxy-3,20-dione. The position of the epoxide does not seem to be important; both
pregnan-4a,5a—epoxy-3,20-dione and pregnan-l6or,17a-epoxy-3,20-dione have an
El/Z of -2.41. This is the sarrre reduction potential as cholestan-4,5-epoxy-3-one and
androstan-3-one-4B,SB-epoxy-17-ol acetate. The substitution at the 17-position is not
very important.

The reduction potentials will be compared to electron capture negative

ionization mass spectrometric responses in Chapter 4.

Conclusions

Cyclic voltammetry did not allow the determination of reduction potentials for
steroids because the electrodes used do not allow the more negative potentials needed.
Polarographic analysis on a mercury electrode was a much better technique. Although
the 50% DMF/50% water solvent system allowed the determination of reduction
potentials of bicylic compounds, many steroids were not soluble in this system and
thus could not be analyzed. These steroids were much more soluble in acetonitrile
with TBAF as the supporting electrolyte. Certain structural features are important for
ease of reduction. Carbonyl groups in the 6-position, a ll-carbonyl group in
combination with a l7-carbonyl group, a halogen alpha to a carbonyl group, and an

epoxide group all result in less negative reduction potentials.

References

S”

95”.”???

11.
12.
13.

14.
15.

l6.

17.

18.
19.

20.

21.

Kayganich, K. A. Ph. D. Dissertation, Michigan State University, 1989.
Chen, E. C. M.; Wentworth, W. E. J. Chromatogr. 1981, 217, 151.

Bard, A. J .; Lund, H. Encyclopedia of Electrochemistry of the Elements:
Organic Section Volume XII; Marcel Dekker: New York, 1978.

Shono, T.; Toda, T.; Oshino, N. Tetrahedron Lett. 1984, 25, 91.

Ponsold, K.; Kasch, H. Tetrahedron Lett. 1979, 46, 4463.

Ponsold, K.; Kasch, H. Tetrahedron Lett. 1979, 46, 4465.

Cooper, A.; Mantell, C. L. 1&EC Prac. Design & Development 1966, 5, 238.
Kaulen, J .; Schaefer, H.-J. Tetrahedron 1982, 38, 3299.

Bond, A. M.; Heritage, 1. D.; Briggs, M. H. Anal. Chim. Acta 1982, 138, 35.
Samuel, A. J .; Webber, T. J. N. In Electrochemical Detectors: Fundamental
Aspects and Analytical Applications; Ed., T. H. Ryan; Plenum Press: New
York, 1984.

Shimada, K.; Tanaka, M.; Nambara, T. Anal. Lett. 1980, 13, 1129.

Bond, A. M.; et al. Anal. Chem. 1988, 60, 1023.

Shono, T.; Matsumura, Y.; Hayashi, J .; Mizoguchi, M. Tetrahedron Lett. 1979,
46, 165.

Smith, W. H.; Bard, A. J. J. Am. Chem. Soc. 1975, 97, 5203.

Skoog, D. A. Principles of Instrumental Analysis; Saunders: Philadelphia,
1985.

Evans, D. H.; O'Connell, K. M.; Petersen, R. A.; Kelly, M. J. J. Chem. Ed.
1983, 60, 290.

Van Horne, K. C. Sarbent Extraction Technology; Analytichem International:
Harbor City, CA, 1985.

DiDonato, G. C.; Busch, K. L. Blamed. Mass Spectrom. 1985, 12, 364.

Briggs, G. W. D.; Jones, E.; Wynne-Jones, W. F. K. Trans. Faraday Soc. 1955,
51 , 883.

Km”. H. M; Wendt, H.; Strehlow, H. Z. Elektrachem. Ber. Bensenges.
Physik. Chem. 1960,64, 483.

Model 303 Static Mercury Dropping Electrode Operating and Service Manual;
EG&G Princeton Applied Research: Princeton, N. J ., 1978.

95

22.
23.
24.
25.

26.
27.
28.

96
Kabasakalian, P.; McGlotten, J. Anal. Chem. 1959, 31, 1091.
Shinrike, N.; Nambara, T. Yakugaku Zasshi. 1971, 91, 6.
Cohen, A. 1. Anal. Chem. 1963, 35, 128.

Zuman, P. Substituent Efiects in Organic Polarography; Plenum Press: New
York, 1967; Chapter 9.

Robertson, D. M. Biochem. J. 1955, 61, 681.
Kabasakalian, P.; McGlotten, J. Anal. Chem. 1962, 34, 1440.
Delaroff, V.; Bolla, M.; Legrand, M. Bull. Soc. Chim. France 1961, 1912.

Chapter 3: The Analysis of Anabolic Steroids by ECNI-MS

In today's society, with the emphasis on physical fitness and "win-at-all-costs"
philosophy, it is no surprise that "miracle drugs" such as anabolic steroids are a
widespread problem. A glance at the headlines over the past few years shows a range
of abuse of steroids, from that of superstars such as Ben Johnson who was stripped of a
gold metal in the 1988 Summer Olympic Games, to studies showing that an estimated
8% of the high school students in Michigan used steroids in the past year (1). Even
lawmakers have not ignored this problem. In Michigan, the sale of steroids is now a
felony, rather than a misdemeanor. At the federal level, steroids are Schedule III
controlled substances (section 202C of Controlled Substance Act, Congressional
Record, November 2, 1990).

Even with this recent interest in steroids, scientific debate over the
effectiveness and the dangers of anabolics remains unsettled, even though it is
generally agreed that the disadvantages outweigh any probable advantages. The main
thrusts of the campaign against steroids have been education and enforcement by
random drug testing during sporting events. Drug testing today is hampered by several
problems. There is a need for better sensitivity, in order to detect drugs taken long
periods of time before an athletic event, and for long-term research projects. There
also is a need for better selectivity to avoid false positive results with other
compounds such as oral contraceptives and endogenous steroids.

Some analytical methodologies for anabolic steroids rely on conventional gas
chromatography/electron impact mass spectrometry (GC/EI-MS). Electron capture
negative ionization mass spectrometry (ECNI—MS) has been proven to have better
sensitivity than EI for compounds that are highly electrophilic (24). Because most
biological compounds are neither highly halogenated nor highly conjugated,

fluorinated derivatives that have been developed for GC/electron capture detection

97

98
(ECD) also have been used for ECNI-MS. This chapter will concentrate on the
development of suitable MS techniques for the quantitative detection of anabolic
steroids and metabolites in urine. Selective and sensitive methodology was sought via
two avenues involving ECNI-MS: a) one approach involved the use of conventional
ECD derivatization prior to analysis by ECNI-MS, and b) the other was patterned after

the novel oxidative approach developed in this laboratory for dexamethasone.

Profile of Steroid Abuse
Why Take Steroids?

Anabolic steroids are derivatives of testosterone which have been structurally
altered to retain the anabolic effect while minimizing the androgenic effect. This
alteration is usually done by adding a Net-methyl group. Conventional medical uses
of anabolics include correcting deficient testosterone levels, anemia due to bone
marrow failure, and hereditary angioneurotic edema (5). Medical uses of anabolics
also exist in sports; for example, anabolics can help improve recuperation of muscles
after strain and may help injuries heal more quickly (6).

There is conflicting scientific evidence for improvement of athletic
performance following administration of anabolic steroids. In a survey by Haupt and
Rovere (7), reports in the literature only agreed on one point: there is no improvement
in the aerobic performance of athletes treated with anabolic steroids. When comparing
the effect of anabolic steroids on strength, the literature was not conclusive; fourteen of
twenty-four studies reported significant strength increases with anabolic use. In all
cases, strength increase depended on a combination of having previous weight-training
and continuing this training during the period of steroid use. A high protein diet also
helps because anabolic steroids increase protein biosynthesis in muscles, which results

in an increase in muscle mass and strength (8). Interestingly, studies which showed

99
increases in body size and weight while using anabolics were the only ones which

showed increases in strength.

Why flat Take Steroids: Possible Side Efl'ects

The side effects of anabolic steroids also have been hotly debated. They range
from ”part of the price you pay" discomfort to those which are possibly life
threatening. Unfortunately, long-term effects have not been studied. One of the most
common side effects is mood shifts, usually in the form of aggressiveness. In fact,
double-blind studies are being discontinued (7) because athletes can discern whether
they are on anabolic steroids based on mood shifts alone. About 30% of the athletes
using anabolic steroids report subjective side effects (7). These side effects can
include aggressiveness, changes in libido, muscle spasms, irritability, headache,
dizziness, and nausea (9). These effects are reversible upon discontinuation of
steroids.

Endocrine functions are also impaired (9). These include decreased
spermatogenesis (which may result in temporary infertility), testicular atrophy, acne,
gynecomastia, and decreased amounts of luteinizing hormone, follicle-stimulating
hormone, and testosterone. Again, all these effects are reversible upon discontinuing
steroid use. Hepatic and cardiovascular side effects are not as widely observed, but
they may be linked to possible long-term effects (5). For example, HDL-cholesterol
sometimes decreases in plasma and low-density lipoproteins increase with sustained
steroid use, a factor which could increase the probability of a subsequent heart attack.
Elevated liver-function test values have occurred, but these are reversible, and long-
term liver damage appears to be rare (7 ). Both peliosis hepatitis and liver tumors have
occurred in fewer than 60 cases, all in patients treated with anabolic steroids for an

extended period (usually over two years) for medical purposes.

100
Of greater concern are the irreversible side effects observed in early adolescent
boys and in women of all ages (8). Young boys may experience deepening of the
voice, male-pattern baldness, and acne. Full adult height may not be obtained due to
premature epiphyseal closure. Women may experience the following irreversible
masculinizing side effects: deepening of the voice, hirsutism, growth of facial hair,
increased body hair, enlarged clitoris, and alopecia. Some side effects such as

irregular menstruation and decreased breast size can be reversed.

Who are the Users of A nabolic Steroids?

The actual extent of steroid abuse is difficult to determine, given the reluctance
of athletes to participate honestly in such studies. The number of Olympic athletes
who used steroids in training for the 1988 Summer Games has been estimated from
10% to 90% (10). In 1988, six percent of the National Football League‘s players tested
positive for steroid use (10). A survey of athletes in the 1987 National Championship
of US Powerlifters, 33% of the respondents admitted to steroid use (11). Even among
younger people, reported steroid use is high. A survey of high school juniors, both
athletes and nonathletes, found 11% of the boys and 0.5% of the girls admitted to
steroid use (12). Anabolics are not just a problem in competitive sports; some use is
directed toward improving personal appearance. In a 1988 survey of 1010 men at
three US colleges, 2% admitted steroid use, a gross underestimate by the authors'
standards (13). Of those who admitted steroid use, 24% were not competitive athletes.

What Steroids are Most Commonly Abused?

The International Olympic Committee (IOC) has forbidden the use of drugs in
five categories: stimulants, narcotic analgesics, beta blockers, diuretics, and anabolic
steroids (6). The list of forbidden anabolic steroids is found in Table 3.1. Many

studies agree that the most commonly self-reported abused steroid is

101
methandrostanolone (tradename Dianabol) (11, 13, 15-16). Next in popularity are the
injectable testosterone esters. Even though these steroids retain an androgenic effect,
they are popular because detection is more difficult. The use of testosterone is
suspected if the testosterone-to-epitestosterone ratio in urine is above 6:1 (14). Other
popular steroids among powerlifters include oxandrolone, nandrolone, stanozolol, and
oxymetholone (11, 15).

In a study of weight trainers (16), common patterns of anabolic steroid use
were obtained. Eighty percent of the athletes "stack ", or took two or more steroids
at a time. Typically, this consisted of an injectable testosterone ester and a fast-acting
oral formulation. After a cycle of six to ten weeks taking steroids, the minimum time
off was four weeks. Often, dosages of steroids were varied during the cycle, and
athletes usually took four to eight times the recommended medical dose of steroids.
Since steroids taken orally or as water-based injections can be detected for about two
weeks after discontinuation and oil-based injections can be detected for about two
weeks after administration (8), athletes carefully time cycles to maximize performance

and minimize their chances of being caught.

Table 3.1: IOC Forbidden Anabolic Steroids

(Adapted from reference 6)
Bolasterone Methyltestosterone
Boldenone Nandrolone
Clostebol N orethandrolone
Dehydrochloromethyltestosterone Oxandrolone
Fluoxymesterone Oxymesterone
Mesterolone Oxymetholone
Methandienone Stanozolol
Methenolone Testosterone

«And other chemically and pharrnacologically related compounds.

102

Commonly Used Detection Techniques
Radioimmunoassay

Radioimmunoassay (RIA) techniques for detecting anabolic steroids have been
in use for a number of years (17). Three groups of steroids can be detected by RIA
techniques: (a) nandrolone analogs, (b) norethandrolone analogs, and (0) 17a-
substituted analogs. Before RIA analysis can be performed, an extraction step is
necessary. The major advantage of this technique is its sensitivity, but RIA is not
routinely used in screening because of its many disadvantages. Cross-reactivity is
observed, especially with normal urinary steroids and oral contraceptives such as
norethindrone and norgestrel. The testosterone-to-epitestosterone ratio cannot be
measured without using another RIA (14). One anabolic steroid, methenolone, cannot
even be detected. During the 1976 Summer Olympic Games, RIA was used for
screening of urine samples with confirmation by GC/MS (18). By the time of the 1984

Summer Games, GC/MS was used for both screening and confirmation.

High Pressure Liquid Chromatography

High pressure liquid chromatography (HPLC) has been used mainly for the
analysis of anabolics in the veterinary, pharmaceutical, and forensic fields, rather than
human urine screening (17). Recent studies using HPLC include the analysis of
nandrolone in biological samples using on-line immunoaffinity sample pretreatment
(19) and the analysis of l7-hydroxy steroid esters (20).

One major problem with HPLC is in establishing absolute confirmation of the
analyte. Suggested means for identification include: ( 1) same retention time as that
of a standard, (2) the nearest peak in the chromatogram is separated by at least onefull
peak width at half maximum height, and (3) reanalysis with a standard, proving co-
elution (21). More promising techniques based on HPLC are LC/MS and LC/MS/MS
(22,23). The mass spectrometer provides confirmation of the analyte, while the LC

 

103
can separate even underivatized steroids. Also, the lengthy (often 3 to 48 hours)
hydrolysis step for conjugated steroids can be eliminated (24). The digestive juice
most often used for the hydrolysis step, Helix pomatia, has been observed to convert
free 3B~hydroxy-5-en-pregnanes into their 4-en-3-oxo analogs as well as to convert
androstenedione, androstenediol, and dehydroepiandosterone into their 4-en-3-oxo

analogs, which can lead to false positives.

Gas Chromatography/Electron Impact Mass Spectrometry

By far the most common and the most trusted analytical technique for detecting
anabolic steroids is GC/MS. Because good sensitivity (< 10 ng/mL of urine)(17) is
needed for the screening of anabolic steroids or their metabolites, selected ion
monitoring (SIM) is used. The few characteristic ions of a steroid or metabolite are
monitored during a specific retention time window. In one study (14), screening is
accomplished by monitoring two characteristic ions of each steroid. Confirmation of
positive screens is achieved by reanalyzing the sample while monitoring twenty
characteristic ions or, if possible, obtaining a full scan. Some anabolic steroids have
more than one major metabolite (see Table 3.2), the identification of which can help in

the establishing whether a particular steroid was used.

Extraction and Hydrolysis of Urine

To prepare urine for analysis by GC/MS, about 5 mL of urine is extracted using
a XAD-2 column or, more commonly, a C18 solid-phase extraction column. If
desired, the steroids can be separated into free and conjugated fractions at this point
i (see Table 3.3). Before analyzing the conjugated steroids, they are hydrolyzed with

Helix pomatia or E. coli B—glucuronidase for 3 to 24 hours.

 

104

 

was. 8058; so: _o_?at.8-.§oE.E rocaueucmém

.2...Funtgécmcooaemeocé
965:? one ocmcuoaémaocéF.>x9o>c_o.nn..-en

SQouoficaaeocép
o:0.oc£o§o..§8-mw
8228.288?

_o_o~8£m>xeu§u$.v
588993.253;
_o_o~ocm.m>x2u>z.w F
_o_o~oc£m>x9u>z-v
_o_o~oc2m>x20>z-.oioc_2
_o_o~ocm.mio_m2

85:53 .8 226
ocoK 7.995.865. 7.655.
ococenoaioaz
a r. . 1 .4 9%
83-8.8.2“-736-52.313.38:
5:89 one oc9n-o:£meocm-_o_o-nn F.no-_>£mE-et
299288.??? :m P .4. 252.5% 2: 38-9
.23: F.8-ocmaoeucmamxssoeanp
w:Pm-mcm~moficm-nm-_>£oE-enp
.258 use .23: 3228.29 7%-.EEEACLSE
8898:2685»
.1 l: . L h 1 ‘4 n0
ococo_uocm568>x2o>c-no-eo_ms.

 

«8:388. >355

Ava. 2228.55

E: 89282852

88 23.2952

$3 .2285

$8 wagon—om

:9 0:828:25:
2935 2.235 E28

3.85.5.3 5 30:222.
«2205 0:323 .0 85883. B055 nu.» 03a...

105

85380:. 88. 322-28.:
88288222512823 ......
88288822251222.23....o

88288522252123-8103:

28 222......
o:o.o.o:mx2.xeo>=-n$ use 8.
30.89.30.285 7.8.2
.828. 8888:...2m2

822828282... 5... . .... 7.8.82.2: 52. 28-8
:1: 7.8.22.3 P89322282Ba2v. P

.8253 we... o:o.o:o£oE§8oE.o
8288.882:

20.688 235 30359536....

552225818...
822.§>xo>xeoE-or.o_az

828608203292
38329533-?

gflggogfifiv
322823—5920...

.8. 380655928220...

30. 9.0.26:me

.8. 8288.22

.8. 828.855

.8. 8832228.”.

.8. 82888225

.988. a.» snap

106

Table 3.3: Anabolic Steroids Excreted in Urine"I
(from reference 17)

Conjugated Fraction Free Fraction
Bolasterone Dehydrochloromethyltestosterone
Boldenone Fluoxymesterone
Chlorotestosterone Methandienone
Dromostanolone Oxandrolone
Bthylestrenol Stanozolol
Mesterolone

Methandriol

Methenolone

Methyltestosterone

N androlone

Norethandrolone

Oxymesterone

Oxymetholone

Stanozolol

Testerosterone/Epitestosterone

*" ' ' 0' ° 0 o .
Anabolic sterord means either the parent steroid or one of its metabolites.

Derivatization

In order to improve chromatographic properties and possibly sensitivity,
derivatives must be made of anabolic steroids before analysis by GC. Many different
derivatives have been used for analysis by EI/MS; the most popular of these will be
described.

WW8) derivatives were among me first to be
used in the analysis of anabolic steroids. The methoxime moiety protects the carbonyl
functionality while the trimethylsilyl reacts with the hydroxyl groups. Reports of
detection limits for individual steroids or metabolites range from 3 ng/mL to 20 ng/mL
of urine (25-27). The ketone and hydroxyll groups are derivatized differently, a

process that can help in structure identification, by providing unique ions and retention

107

times for certain steroids. Problems associated with the MO-TMS derivative include
long reaction times, formation of multiple produCts in the form of syn/anti isomers,
acid lability. and the need for a liquid chromatography step after derivatization.

W8.) derivatives are now more widely accepted than MO-
TMS derivatives (17, 24, 27-33). Selective derivatization can be accomplished by
selecting the silanizing reagent and catalyst. For example, a 100:2 mixture of N-
methyl—N-u'imethylsilyluifluoroacetamide:trimethylchlorosilane (MSTFA:TMCS) will
form primarily TMS-ethers. If the derivatizing reagent is N-methyl-N-
u'imetltylsilylu'ifluoroacetamide:u'imethylsilylimidazole, MS'I'FA:TMIS, (1000:2 with
2 mg/mL dithioerythritol), TMS-ether/I‘MS-enol derivatives are formed. MSTFA is
usually chosen for derivatization because it promOtes derivative stability by preventing

loss of TMS groups from the steroid.

Electron Capture Negative Ionization Mass Spectrometry

In order to detect trace levels of anabolic steroids long after administration,
analytical techniques with better sensitivity and selectivity are needed. Electron
capture negative ionization (ECNI) mass spectrometry has the capacity to meet some
of these analytical needs. ECNI has the attractive features of high sensitivity and high
selectivity for molecules that are sufficiently electrophilic to capture thermal electrons.
Despite its attractive atu'ibutes, ECNI has not been as widely used as had been
anticipated; this is partially due to problems associated with finding proper derivatives
of the analyte which are amenable to the ionization technique. Successful applications
of the technique in the biomedical area have been limited principally to those reported
for selected drugs (34-36), biogenic amines (37, 38), metabolites of arachidonic acid
(39-42), and steroids (43).

Electron capture techniques have seen limited use in anabolic steroid analysis.

Heptafluorobutrate (HFB) derivatives of testosterone and nandrolone have been

108
analyzed using GC/electron capture detector (GC/ECD) (44). More recently, Ginkel,
et al. analyzed nandrolone in meat using a di-HFB derivative and GC/ECNI/MS (45).
The I-IFB derivative was chosen because the ECNI response to MO-TMS derivatives
was inadequate.

Electron-capturing derivatives also have been useful in EI/MS analyses. In
fact, one of the most common methods of analyzing Stanozolol involves the use of the
I-IFB-TMS derivative (14, 24, 46) because complete TMS derivatization leads to
unstable products. The HFB reagent derivatizes the nitrogen in the pyrazole ring,
while the TMS reagent reacts with the 17B-hydroxyl group. The EI/MS detection limit
of 1 ng/mL may be improved by using ECNI-MS. One significant problem in the
analysis of stanozolol, as well as in the analysis of many other anabolics, is that only a
small proportion of the steroid and/or metabolite is excreted in the urine. For

stanozolol, about 16% is excreted in urine, while 40-60% is excreted in feces.

Commonly-Used Electron Capture Derivatization Agents

Many electron capture derivatives of steroids have been developed (47 ), most
originally for use with GC/ECD.

1. W ethers have been used in methods that rely on
detection by electron capture techniques. The TMS—ether moiety in itself does n0t
affect electron capture (48), but after forming TMS ethers, thermally-labile steroids
survive transit through the GC and are more likely to reach the detector intact, thus
increasing sensitivity. The benefit of ECNI in this case is realized as a result of
discrimination against background as the TMS derivative has no significant elecuon-
capture activity. A comparison of the TMS derivative of dexamethasone to Other
derivatives is shown in Table 4.10.

2. WWW derivatives have been used for

ECNI/MS analysis of many corticosteroids with much success (49, 50). A common

109

problem with the conventional approach to derivatization for ECNI analysis is that
many of the derivatives produce abundant ions that relate only to the derivatizing
agent, and not to the parent structure of the analyte. For MO-TMS derivatives, major
high-mass ions are based on the parent molecule, not the derivatizing agent, which
allows for monitoring of specific ions with good sensitivity. Different reaction times
and conditions are used to derivatize hindered ketone and hydroxyl groups, which can
be used to help elucidate the structure of the original molecule. By using MO-TMS
derivatives, however, one is not taking advantage of any added electron capturing
moiety that possibly could increase the detector response.

MO-TMS derivatives also were studied as part of the comparison study of
dexamethasone derivatives under ECNI conditions. MO—TMS and TMS derivatives
have the same response as dexamethasone (when it is analyzed using the direct probe
inlet). Again, no additional electron capture ability is imparted.

3a. Wm provide an improvement in
sensitivity over that available with conventional TMS techniques. Among halogens,
electron capture response increases with increasing atomic weight, but volatility
decreases (51). If these derivatives have the same volatility and ease of preparation as
conventional TMS derivatives, these derivatives can be very useful in ECNI.

3b- WWW (known as flophemsyl
derivatives) are a unique variation on a TMS derivative that also has an electron-
capturing functionality (52-54). The pentafluorophenyl group is more sensitive to
GC/ECD than a chloromethyl group, and is more thermally stable than a HFB
derivative. The flophemesyl derivatives can be made to selectively react with
hydroxyl and/or ketone groups. For example, flophemesylamine reacts with certain
hydroxyl groups in the presence of ketones, but not with 173- or llB—hydroxyls. For
anabolic steroids, the most useful reagent seems to be flophemesylamine:flophemesyl

chloride (10:1) which will react with unhindered secondary hydroxyl groups and the

l 10
tertiary 17 B-hydroxyl groups. This reagent will not react with the 113-hydroxyl group;
the only anabolic steroid that exhibits this functionality is fluoxymesterone.

4. WW3, including pentafluoroctanoate (55), 9-H-
hexadecafluorononoate (56), chloroacetate (57), benzoate (58), and
pentafluorobenzoate (59), have all been used to derivatize steroids for analysis by GC
and detection by ECD. Their success with anabolic-type steroids or ECNI/MS analysis
is unknown.

5. W result from promising new reactions for steroids
(60). The perfluorotolyl derivative in particular is volatile, can be prepared in good
yields, and results only from reaction with hydroxyl groups. The major disadvantage,
from the anabolic steroid point of view, is that this derivative will not react with 1113-
hydroxyl or tertiary 17-hydroxyl groups. For urinary metabolites without these groups,
this could be a good derivative.

6. Wm is used to derivatize ketones (47). In this case,
the disadvantages far outweigh the advantages. The derivative itself is not very
volatile, and is acid-labile and unstable when exposed to air and direct light. Better
sensitivity can be achieved using another reagent specific for ketones, o-
pentafluorobenzyloxime.

7. Wm: (sold as Florox by Pierce) reacts with ketone
groups for complete derivatization. It has been used successfully for both ECD (61)
and ECNI/MS (62) analyses. of ketones. Sometimes, any remaining hydroxyl groups
must be protected (usually as TMS-ethers) to improve chromatographic properties. In
the ECNI spectrum, the base peak corresponds to the derivatizing agent, but there is a
prominent ion representing the parent molecule.

8. WWW derivatives react exclusively with hydroxyl
groups (63, 64). Under ECNI conditions, the molecular ion is advantageously
represented by the base peak, and the detection limit is 1 pg. The major disadvantage

1 11
is the possible formation of multiple products. Up to five different products can be
observed for a typical diol. For certain steroids, this problem can be compensated by
reaction time. Hindered hydroxyl groups, such as at C-1 1, are difficult to derivatize; a
tertiary l7B—hydroxyl also may not be derivatized.

9. Mam—51m include trifluoroacetyl (TFA) (65),
pentafluoropropionyl (PFP) (66), and heptafluorobutryl (HFB) (45, 62, 67, 68) groups.
In a study comparing the relative ECD responses of various haloacyl derivatives of
testosterone (66), the HFB derivative had the best response, about fifty times that of
the TFA derivative. Not surprisingly, the most common haloacyl derivative used for
ECNI is I-IFB (45, 62). The base peak in any HFB derivative ECNI spectrum is
representative of the derivative; any peaks representative of the parent molecule are
very small. For example, the peaks representing the parent molecule of derivatized
nandrolone are less than 5% (45). Other problems with I-IFB derivatives include the
fact that hindered hydroxyl groups may not react quantitatively and that some products
are chemically unstable.

10. Wm has been used in this laboratory for the determination
of dexamethasone in plasma (69) and 6B-hydroxycortisol in urine (70). This
methodology involves chemical oxidation of the analyte to a modified product with
extraordinarily high electrophilic properties. The synergism in selectivity available
from the combination of a selective chemical reaction, selective ionization conditions,
and selective mass-to-charge monitoring techniques by mass spectrometry offer the

potential for a uniquely simple and sensitive quantitative assay for particular steroids.

Preliminary Results
TMS Derivatives of A nabolics
TMS derivatives do not add any electrophilic character to a molecule, but they

do allow thermally labile steroids to pass through the GC intact; thus, steroids that

112

have other electron-capturing features can be detected. The only anabolics which
possibly could benefit from TMS derivatization are stanozolol and fluoxymesterone.
The chemically oxidized product of fluoxymesterone (1,4-androstadien- Nor-methyl-
17B—ol-9a-fluorine-3,11—dione) was also used for this study.

The steroids were derivatized with MSTFAzTMCS (100:2) at 60°C for 5
minutes. Approximate concentrations were determined by using GC-FID and the
effective carbon number concept (72). The ECD and ECNI responses for these

compounds are summarized in Table 3.4.

Table 3.4: ECD Relative Responses of TMS Derivatives of Anabolic Steroids

Compound Relative Rosponse‘I
l,4-Androstadien-3,l 1,17-trione 1.0
4-Androsten-3,17-dione (ox. testosterone) 0.003
Fluoxymesterone 0.004
Oxidized fluoxymesterone 0.63
Fluoxymesterone-TMS 0.027
Oxidized fluoxymesterone-TMS 1.2
Stanozolol-TMS 0.0002

I"l,4—Androstadien-3,1l,l7-trione has been assigned a relative response of 1.0.

In Table 3.4, l,4-androstadien-3,l l,l7-trione is a representative high-response
compound; that is, any compound with a response at this level will be suitable for
ECNI assay development. 4-Androsten-3,l7-dione is the oxidized product of the
endogenous steroid testosterone; any compound with a response at this level will not
be easily detectable by ECNI. Even as its TMS derivative, stanozolol is a very poor
electron capturer, much worse than oxidized testosterone. Underivatized

fluoxymesterone is a poor-response compound because it is thermally labile. The

113
response of the TMS derivative of fluoxymesterone is slightly better. To obtain a
reasonable response, fluoxymesterone must be oxidized. The presence of the 11-
carbonyl in conjunction with the 9-fluorine is significant for response. The TMS
derivative of the hindered 17-hydroxyl group in oxidized fluoxymesterone does help
the response slightly.

Selected Fluorinated Derivatives

Three steroids, stanozolol, methyltestosterone, and fluoxymesterone (see
Figure 3.1), were chosen for a preliminary evaluation of derivative formation and
detector response because they have a variety of functional groups. These steroids
were derivatized using published (but not optimized) methods. Derivatives tested
include those prepared by using reagents such as TFA, HFB, flophemesyl, Florox, bis-
trifluoromethylbenzoyl, and pyridinium chlorochromate (for chemical oxidation). If
the reaction proceeded in sufficient yield, the approximate concentration of the desired
product was calculated using the GC/flame ionization detector (GC./FID) and the
effective carbon number concept for assessing the detector response (72). Relative

responses of the derivatives were determined using GC/ECD.

 

Figure 3.1: Steroids used for pilot study on derivatization: (a) stanozolol, (b)
methyltestosterone, and (c) fluoxymesterone.

In the procedure attempted (73), TFA derivatization reactions proceeded with a

very low yield. Only fluoxymesterone reacted in any significant amount, and even

114
then only one of two hydroxyl groups was pr0tected. Problems with this procedure
may be solved by the choice of catalyst and clean-up procedure.

Many procedures for the formation of HFB derivatives have been published.
The two procedures attempted here used heptafluorobutyric acid anhydride (68) and
heptafluorobutrylimidazole (73). Both methods suffered from the formation of
multiple reaction products. Stanozolol was derivatized only at the nitrogen in the
pyrazole ring as expected (14). Methyltestosterone formed only the HFB derivative at
the ketone, leaving a free hydroxyl group. A future possibility is to form the di-HFB
derivative (45), which should be easier to detect than a mono-I-IFB derivative.

As stated previously, flophemesyl derivatives can selectively react with a
variety of functional groups depending on the choice of reagent and reaction
conditions. For the anabolic steroids used in this experiment, a 10:1 mixture of
flophemesyldiethyamine:flophemesylchloride did not work very well. Stanozolol did
not react at all. For fluoxymesterone, one hydroxyl group was derivatized, as would
be expected because the 115-hydroxyl group does not react under the conditions used.
Mono- and di-flophemesyl derivatives were formed for methyltestosterone.

The bis-trifluoromethylbenzoyl chloride derivatized testosterone which has
only one unhindered hydroxyl group in one hour at 60°C. After a five-hour reaction,
the hindered hydroxyl group on methyltestosterone barely reacted. A study of reaction
conditions and times should be undertaken to Optimize this yield.

Florox is a ketone-specific reagent. Not unexpectedly, stanozolol did not react.
Both methyltestosterone and fluoxymesterone were derivatized as expected. The

reaction conditions for the formation of Florox derivatives were 60 minutes at 50°C.

Conclusions from Preliminary Studies
Table 3.5 shows the relative ECD responses of anabolic steroid derivatives

relative to that of 1,4-androstadien—3,11,17-u'ione, a very good elecuon capturing

1 15
compound. Any compound with a response higher than that of 1.4-androstadien-
3,11,17-trione will have an advantageous detection limit by electron capture detector.
The ECD detector was used to estimate the electron capture response of a given
derivative. Generally, the trends in responses are the same for either ECD or ECNI-
MS; however, the absolute limits of detection may differ. Under ECNI conditions, the

derivatives can fragment, which leads to higher detection limits.

Table 3.5: ECD Relative Response of Selected Anabolic Steroid Derivatives

Derivative Relative Response"
Florox of Methyltestosterone 10

Florox of Fluoxymesterone 4

HFB of Methyltestosterone 0.5

I-IFB of Fluoxymesterone 0.2

HFB of Stanozolol 2

Oxidized Fluoxymesterone 100

*l,4-Androstadien-3,l 1,17-trione has been assigned a relative response of 1.0.

As shown by the results in Table 3.5, both mono-HFB and Florox derivatives
have a very good electron capture response. The addition of a TMS group to the 173-
hydroxyl group of either fluoxymesterone or methyltestosterone Florox derivatives has
no effect on the ECD response.

The most promising procedure for detection of fluoxymesterone involves
chemical oxidation because it gives a lOOJold greater response than 1,4-androstadien-
3,11,17-trione. This enhancement is achieved by oxidative transformation of the
analyte, a process which discriminates against producing chemical noise from the

sample matrix.

116
Along with improving the formation of the above derivatives, other derivatives
which are promising include the pentafluorobenzoate and silyl derivatives with
halogen substitution. 4-O-Perfluorotolyl ethers do not react with tertiary 17-hydroxyl
groups; however, they could be used for steroids and metabolites such as nandrolone

or boldenone that do not contain this group.

Future Work
Proposed work with conventional derivatization

Many qualitative screens exist for anabolic steroids; most are based on
GC/EI/MS of TMS-ether and/or TMS-ether-TMS-enol derivatives. More selective and
sensitive methodology is needed for the confirmation of a positive screen. Also, many
physiological and pharmacological studies of long-term steroid use require
methodology that can detect trace levels of steroids or metabolites.

The proper derivative for a steroid or a set of steroids must meet several
conditions. After appropriate reaction conditions are determined in order to optimize
the formation of one product, the derivatives will be evaluated for their sensitivity
under GC/ECNI/MS. Derivatives that generate major ions indicative of the parent
molecule rather than the derivatizing moiety are most desirable. Tradeoffs between
one-step reactions, complete reactions, short reaction times, and short GC analysis
times should be evaluated.

Derivatives also should be evaluated for their selectivity. One of ,the greatest
problems in anabolic steroid analysis today derives from the similarity of oral
contraceptive metabolite structures to those of certain anabolics. For example, the oral
contraceptive norethisterone is partially metabolized to l9-norandosterone, the major
metabolite of nandrolone (75). Another problem is the similarity of the major
metabolite of norethisterone (see Figure 3.2) to the major metabolite of methenolone

(75). During analysis by GC-MS, the TMS-ether-TMS-enol derivatives of these two

117
metabolites produce the same ions at the same retention time. If one uses a derivative
specific for ketones (such as Florox), only the methenolone metabolite will appear.
The norethisterone metabolite will not be derivatized; thus, it will not appear in the

chromatogram.

   

(a) (b)

Figure 3.2: Similarity of oral contraceptive metabolites to anabolic steroid
metabolites. (a) Major metabolite of norethisterone (an oral contraceptive). (b) Major
metabolite of methenolone (an anabolic steroid).

Applications of Chemical Oxidation to the Analysis of A nabolic Steroids

The proposed methodology is extremely promising for the selective detection
of fluoxymesterone. From comparison of preliminary data, the steroid most amenable
to the oxidation technique is fluoxymesterone which can be chemically oxidized to
1,4-androstadien—l7or-methyl-l7B-hydroxy-9or-fluorine-3,1l-dione. Not only does the
oxidized product have the 1,4-en—3,11-one structure that enhances the electron-capture
response (74), it also has a 9-fluorine group. In fact, the response of this compound
under ECNI conditions is 20 times that of 1,4-androstadien-3,1 1,17-trione, which is a
high-response compound (see Table 3.6). The presence of the 17B-hydroxy does not
seem to adversely affect the chromatography.

Many anabolic steroids have a tertiary 17-hydroxy group which cannot be
oxidized to the ketone in a simple, one-step reaction. Those anabolics that have
secondary 17-hydroxy groups, namely testosterone, nandrolone, and boldenone, are

easily oxidized to ketosteroids, but their response is poor in comparison to that of 1,4-

l 18
androstadien-3,11,17-trione (see Table 3.6). These results were expected from the
preliminary studies because these oxidized anabolic steroids have only one conjugated
site. The detection limit of oxidized boldenone is 2 ng/mL, far short of the 0.5 ng/mL
minimum needed for effective analysis of urine for anabolics. Also, the advantage of

selectivity over the endogenous steroids is lost.

Table 3.6: Relative ECNI Responses of Oxidized Anabolic Steroids

Steroid Relative Response.
Oxidized testosterone 0.0075
Oxidized nandrolone 0.007 5
Oxidized boldenone 0.013
1,4-Androstadien-3,l l, l7-trione 1.0

Oxidized fluoxymesterone l9

*1,4-Androstadien-3,ll,17-trione has been assigned a relative response of 1.0.

Fast Chromatography

The rate-limiting step in anabolic steroid analysis is usually the work-up and
derivatization of the urine samples, but long GC analysis times also can be frustrating.
Tandem mass spectromeu'y (MS/MS) methods can be used to shorten analysis times
by using shorter columns without substantial loss of chromatography (76). A twenty-
minute analysis on a 30-m column can be shrunk to seven minutes by monitoring
daughter ion scans on the effluent from a 4-m column (77). As well as shortening
analysis times, selectively also can be improved because one can monitor specific
daughter ions, even if the parent molecular ions are the same. Negative ion MS/MS
can be used to analyze the most promising optimized derivatives for even more

selectivity and perhaps better sensitivity (78).

1 19
Another method Of reducing analysis time is by using narrow bore columns.
For example, Other GC analysis Of urinary steroids usually takes about an hour using
conventional 0.25-mm inner diameter columns. If a 50-mm inner diameter column

with the same plate number is used, the analysis is complete in only 12 minutes (79).

Validation of Analysis

After a suitable derivative for a certain anabolic steroid or metabolite is
identified based on the ECNI sensitivity and reaction conditions, it should be tested in
a biological matrix for possible interferences. Extraction procedures from urine should
be Optimized. Possible interferences from urine Of females using oral contraceptives
also should be investigated. The use of short column chromatography in conjunction
with MS/MS should be evaluated, if indicated; should this approach be necessary, a
protocol involving a Finnigan TSQ 70 as previously published by this laboratory (78)
should be used. Sensitivity and selectivity should be compared to those of published
methods. After a derivatization method passes these tests, it should be evaluated using
real samples with parallel analyses by conventional, published methods to establish
validity.

References

5‘5”!”

10.

ll.
12.
l3.
14.
15.
16.
17.

18.
19.
20.
21.
22.
23.

Girard, F. "Steroids: The Terror Within." The Detroit News, July 8, 1990.
Dougherty, R. C. Anal. Chem. 1981, 53, 625A.
Begley, P.; Foulger, B.: Simmonds, P. J. Chromatogr. 1988, 445, 119.

Erhardt-Zabik, 8.; Watson, J. T.; Zabik, M. J. Biomed. Mass Spectrom. 1990,
19, 101.

Strauss, R. H. "Anabolic Steroids." In Drugs & Ped'ormance in Sports, R. H.
Strauss, Ed.; W. B. Saunders: Philadelphia, 1987, 59.

De3Jong, E. G.; Maes, R. A. A.; van Rossum, J. M. Trends Anal. Chem. 1988,
7, 75.

Haupt, H. A.; Rovere, G. D. Am. J. Sports Med. 1984, 12, 469.
Siegel, A. J. Your Patient & Fitness 1989, 3, 6.
Kibble, M. W.; Ross, M. B. Clin. Pharm. 1987, 6, 686.

Janofsky, M.; Alfano, P. "System Accused of Failing Test Posed by Drugs."
New York Times, November 17, 1988.

Yesalis, C. E.; Herrick, R. T.; et al. Phys. Sportsmed. 1988, I6, 91.
Duda, M. Phys. Sportsmed. 1988, I6, 16.

Pope, H. G.; Katz, D. L.; Champoux, R. Phys. Sportsmed. 1988, I7.
Hatton, C. K.; Catlin, D. H. Clin. Lab. Med. 1987, 7, 6.

Strauss, R. H; Wright, J. B.: et al. Phys. Sportsmed. 1983, II, 87.
Burkett, L. N.; Falduto, M. T. Phys. Sportsmed. 1984, 12, .

Sample, R. H. B.: Baenziger, J. C. In Analytical Aspects of Drug Testing; D.
Deutch, ed.; Wiley: New York, 1988, pp. 247-272.

Dugal, R.; Dupruis, C.; Bertrand, M. J. Br. J. Sports Med. 1977, II, 16.
Haasnoot, W.; Schilt, R.; et al. J. Chromatogr. 1989, I89, 1.

Noggle, F. T.; Clark, C. R.; DeRuiter, J. J. Chromatogr. Sci. 1990, 28, 2.
DeRuig, W. G.; Stephany, R. W.; Dijkstra, G. J. Chromatogr. 1989, 489, 89.
Edlund, P. O.; Bowers, L.; Henion, J. J. Chromatogr. 1989, 487, 3.

Metzler, M. J. Chromatogr. 1989, 489, 11.

120

25.
26.
27.
28.

29.

30.

31.
32.
33.

34.

35.
36.
37.
38.
39.

41.

42.

43.

45.

47.

121
Bjorkhem, 1.; Ek, H. J. steroid Biochem. 1983, 18, 4.
Bjorkhem, 1.; Lantoo, O.; Lof, A. J. steroid Biochem. 1980, I3, 1.
Bjorkhem, 1.; Ek, H. J. steroid Biochem. 1982, I 7, 4.

Diirbeck, H. W.; Biiker, I.; Scheulen, B.; Telin, B. J. Chromatogr. Sci. 1983,
21, 4.

Masse, R.; Laliberte, C.; Tremblay, L.; Dugal, R. Biomed. Mass Spectrom.
1985, 12, l.

Diirbeck, H. W.; Biiker, I.; Scheulen, B.: Telin, B. J. Chromatogr. 1978, 167,
1. '

Diirbeck, H W.; Biiker, I. Biomed. Mass Spectrom. 1980, 7, 4.
Masse, R.; Bi, H.; Ayotte, C.; Dugal, R. Biomed. Mass Spectrom. 1989, 18, 4.

Cartoni, G. P.; Ciardi, M.; Giarrusso, A.; Rosati, F. J. Chromatogr. 1983, 279,
5.

Hunt, D. F.; Stafford, G. C. Jr.; Crow, F. W.; Russell, J. W. Anal. Chem. 1976,
48, 2098.

Hunt, D. F.; Crow, F. W. Anal. Chem. 1978, 50, 1781.
Lirn, H. K.; et al. Anal. Chem. 1988, 60, 1420.

Lewy, A. J.; Markey, S. P. Science, 1978, 201, 7.
Ibrahim, K. B.: et al. Anal. Chem. 1984, 56, 1695.

Roberts, L. J .; et al. Adv. Prostaglandin, T hromboxane, Leukotriene Res. 1982,
10, 2.

Strife, R.; and Murphy, R. J. Chromatogr. 1984, 305,.
Waddell, K. A.; Blair, I. A.; Welby, J. Biomed. Mass Spectrom. 1983, 10, 83.
Balazy, M.; Murphy, R. C. Anal. Chem. 1986, 58, 109.

Gaskell, S. J. "Analysis of Steroids." In Methods of Biochemical Analysis,
Vol. 29; D. Glick, Ed; Wiley: New York, 1983, pp. 385-434.

Bodrogi, L.; Hollosi; I. J. Pucsok; et al. In Advances in Steroid Analysis '84, S.
Gorog, Ed.; Elsevier: Amsterdam, 1985, pp. 287-293. '

Van Ginkel; L. A.; Stephany, R. W.; et al. J. Chromatogr. 1989, 489, 1.
Choo, H.-Y. P.; Kwon, O.-S.; Park, J. J. Anal. Toxicol. 1990, I4, 1.

Knapp, D. R. Handbook of Analytical Derivatization Reactions, Wiley: New
York, 1979.

48.

49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.

61.

62.

63.

65.

67.
68.

70.
71.
72.
73.

122

Mayer, H. K. Department of Chemistry, Michigan State University,
unpublished results.

Midgley, J. M.; Watson, D. G.; et al. Biomed. Mass Spectrom. 198, 15 , 4.
Houghton, B.: Teale, P.; et al. Biomed. Mass Spectrom. 1982, 9, 4.
Zlatkis, A.; Poole, C. F. Anal. Chem. 1980,52, 1002A.

Morgan, E. D.; Poole, C. F. J. Chromatogr. 1974, 89, 2.

Poole, C. F.; Zlatkis, A.; et al. Lipids, 1980, 15 , 7.

Morgan, E. D.; Poole, C. F. J. Chromatogr. 1975, 104, 3.
Nakagawa, K.; et al. Steroids, 1966, 7, 3.

Kirschner, M. A.; Taylor, J. A. Anal. Biochem. 1969, 30, 3.
Landowne, R. A.; Lipsky, S. R. Anal. Chem. 1963, 35 , 5.
Fitzpatrick, F. A.; Siggia. 8. Anal. Chem. 1973, 45, 2310.
Zmigrod, A.; et al. Steroids, 1970, I5, 6.

Jarrnan, M. J. Fluor. Chem. 1989, 42, 3.

Koshy, K. T.; Kaiser, D. G.; VamDerSlik, A. L. J. Chromatogr. Sci. 1975, I3,
97.

Gaskell, S. J.; Brooks, P. W. Int. J. Mass Spectom. Ion Processes 1983, 48, 2.
Murray, 8.; Watson, D. J .steroid Biochem. 1986, 25, 2.

Watson, D.; Murray, S. et al. Biochem. Soc. Trans. 1988, I6, 7.
Vandenheuval; W. J. A. et al. Biochim. Biophys. Acta 1961, 48, 59.
Lehennin, L.; Reiffstock, A.; Scholler, R. J. Chromatogr. Sci. 1972, 10, 2.
Sarda, I. R.; et al. Steroids, 1968, 12, 6.

Wilson, A.; Mason, P. A.; Fraser, R. J. steroid Biochem. 1976, 7, 6.
Watson, J. T.; Kayganich, K. A. Biochem. Soc. Trans. 1989, l 7, 2.
Kayganich, K. A. Ph. D. Dissertation, Michigan State University, 1989.
Her, G.-R.; Watson, J. T. Biomed. Mass Spectrom. 1986, 13, 56.
Jorgensen, A. D.; Picel, K. C.; Stamoudis, V. C. Anal. Chem. 1990, 62, 6.
Pierce Catalog, Pierce Chemical Company, Rockford, IL, p. 11.

74.

75.
76.
77.

78.

79.

123

Mayer, H. K.; Reusch, W.; Watson, J. T. The Effect of Structure on the
Electron Capture Negative Ionizatigln Mass Spectrometric Response of Steroids
and Other Cyclic Compounds. 38 ASMS Conference on Mass Spectrometry
and Allied Topics, June 1990, Tucson, AZ,.

De Boer, D.; de Jong, B.: et al. Biomed. Mass Spectrom. 1988, I7, 1.

Kinter, M.; Shipe, J. R.; Savory, J. Clin. Chem. 1988, 34, 217.

Rossi, S.-A.; Johnson, J. V.; Yost, R. A. Rapid Chromatographic Separation
for the Determin tion Of Underivatized Steroids with Tandem Mass
Spectrometry. 38 ASMS Conference on Mass Spectrometry and Allied
Topics, June 1990, Tucson, AZ.

Kggganich, K. A.; Heath, T. G.; Watson, J. T. J. Am. Soc. Mass Spectrom.
l , I, 3.

Cramers, C.A.; Leclerq, P.A. Crit. Rev. Anal. Chem. 1988, 20, 1.

Chapter 4: ECNI-MS Structure-Response Study

The oxidation methodology developed in this laboratory for the analysis of
corticosteroids is advantageous, in part, because of its selectivity. The analyte
chemically oxidizes to a highly electrophilic species, whearas endogenous steroids do
not, thus decreasing the chemical background. In order to predict the relative
responses of ketosteroids and similar compounds under ECNI conditions, a study of
the effect of structure on response was undertaken as decribed in this chapter.
Interactions between double bonds, carbonyl groups, halogens, and epoxides were
examined in steroids and Other cyclic compounds such as cyclohexanones, terpenes,
quinones, and trichothecenes.

Previous work correlating ECNI responses with steroid structure has been
limited to only a few compounds (1). More extensive studies of steroid electron-
capture responses (2, 3) have been conducted with GC-ECD. These studies have
examined combinations of double bonds and carbonyl groups almost exclusively.

The work summarized in this chapter has concentrated primarily on
ascertaining the ECNI responses of cyclic compounds. After the determination of
optimum instrumental conditions, many sets of compounds were analyzed. The
electron-capture properties Of ketosteroids were investigated, with special emphasis on
epoxidized and halogenated compounds. The responses of conventional electron-
capturing derivatives, such as trifluoroacetyl derivatives, were compared to those of
compounds Obtained by PCC oxidations. Cyclohexanones and bicyclic compounds
were analyzed to provide comparisons Of compound size to response in additiOn to
providing functional orientation not found within a steroid nucleus. The ECNI
response of quinones was compared to published electron affinity values. The
reduction potentials Of many other compounds were determined to estimate their

electron affinities.

124

125
Several other studies were undertaken in order to determine other factors
affecting electron capture. The responses of several steroids under EI, CI, ECNI, and
ECD conditions were compared. The use of different modulating gases for steroids
also was studied. ECNI responses Of steroids were compared to UV data. Finally, the
day-to-day reproducibility of ECNI spectra on the JEOL 505 was assessed.

Determination of Optimum Instrumental Conditions f

In studies with ECNI-MS, the control of instrumental parameters, especially
source temperature and pressure, is paramount to sensitivity and reproducibility (4,5).
Gas chromatographic and mass specual conditions were optimized for two
representative ketosteroids, l-androsten-3,17-dione and l,4-androstadien-3,ll,l7-
trione. The former is a low-response compound whose mass specu'um consists of only
one peak, that representing [M-I-I]‘. The latter steroid is a high-response compound
whose mass spectrum consists of only one peak, that representing M".

W. Because electron capture depends on a collision between
a thermal electron and the analyte of interest, source temperature plays an important
role in determining the rate of this reaction. Temperature also affects the reagent gas
reactions which result in the formation of the thermal electrons. In general, the lower
the source temperature, the greater the ECNI response. This is true for "conventional"
ECNI compounds (highly halogenated compounds) such as octafluoronaphthalene
(OFN), but does not hold true for steroids (see Figure 4.1). In this experiment,
chromatograms from full-scan analysis rather than SIM were compared. OFN is a
fairly sensitive compound and is about 22000 times more sensitive than Sat-androstan-
3-one. The ECNI mass spectrum of OFN consists of two peaks at mlz 272 (M") and
mlz 238 ([M-34]'). Usually, as the temperature increases, so does the amount of

fragmentation; however, the opposite is true for OFN. As is typical for highly

126

A
93

Peak Area (Arbitrary Units)

 

40.00

 

 

errrrirrlrtrrrrrrrlirrmr

 

 

0.00 JJIIITIIIWI1UIII‘UIIIIVIII1JIIIIYITJJIIHIJTIIIJW]

100.00 150.00 200.00 250.00 300.00 350.00
Temperature (00)

(b)

150.00

100.00

Peak Area (Arbitrary Units)
8

 

rrlrrrrrrlrrrrrrrLtlrrrrrrrrrlrrrrrrtrrl

 

 

0.00 T1111IIIIIUTII1UIIT11UUIIfTIIIUTIJITIITIIIIIII1II]

100.00 1 50.00 200.00 250.00 300.00 350.00
Temperature (0C)

Figure 4.1: ECNI response as a function of source temperature. The source pressure

was held constant at 2 x 10‘5 torr. (a) The response of octafluoronaphthalene

decreases with increasing source temperature. (b) The response of 4-androsten-
" 3,6,17-trione is maximized at a source temperature Of 200 °C.

127

halogenated compounds, the sensitivity decreases as the temperature increases. This is
true for OFN, as shown in Figure 4.1a. In comparison, the data for 4-androsten—3,6,17-
uione (Figure 4.1b) shows an optimum temperature of 200°C. In general, steroids
have greater sensitivity at higher temperatures, but the Optimum varies from compound
to compound. For example, the optimum temperature for the detection of 1,4-
androstadien-9a-fluoro-l6a-methyl-3,l 1,17-trione (oxidized dexamethasone) is at
150°C.

WW. Pressure/temperature studies were
performed at four source temperatrues and three pressures of methane gas in order to
determine the best combination of temperature and pressure for maximum sensitivity.
For 1,4-androstadien-3,l 1,17-trione, a high-response compound that has a peak for the
molecular anion at mlz 298, the optimum source temperature was relatively high
(250°C) with a low reagent gas pressure (1 x 10'5 torr). For l-androsten-3,l7-dione, a
low-response compound which has a [M-H]' anion at m/z 285, the best conditions
were a low temperature (150°C) coupled with a high reagent gas pressure (4 x 10"5
torr). See Figure 4.2 for a graphical representation of these data. Both high- and low-
response compounds were used in this study; therefore, a compromise temperature and
pressure was used. The source temperature was set at 200°C, and the reagent gas
pressure was held at 2 x 10‘5 torr.

Also of interest is the relationship among the peak areas of individual
components of the three—compound internal standard. If these ratios are constant over
a range of temperatures and pressures, then the internal standard probably will
compensate for small variations in instrumental conditions over a series of analyses.
The ratio of areas of the ion at mlz 298 to the ion at mlz 285 were compared using
various temperatures and pressures (see Figure 4.3). The ratio was most nearly alike at

150°C, but the ratios at 200°C are within the error bars. Therefore, run-tO-run

128

 

 

 

13

E 200- (a)-

I)

>\

I...

(U

L...

:1:

..Q

L—

5}; 100- ..

(U "

93 —-O— P=1x10-5Torr \

< ---{3--— P=2x10-5Torr

x

(U "“A" P=4x10'5Torr

(D

0; o ‘ I I 1
100 200 300

Temperature (00)

 

800 r (b)

[M - HI- = 285
600 - A‘s
‘\ '—.'— P =1 x10-5 Torr
‘\\ -~D-~ P =2x 1&5 Torr
400 " A~~~ '“fi” P=4x10-5Torr

 

200 "

PeakArea (Arbitrary Units)

 

 

 

I .
100 200 300

Temperature (00)

Figure 4. 2: ECNI response as a function of source temperature and pressure. (a) The
optimum conditions for the analysis of a low-response compound such as l- androsten-
3,17-dione were at a source temperature Of 150 °C and a source pressure of 4 x 105
torr. (b) The optimum conditions for the analysis of a high-response compound such
as l ,4-androstadien-3, ll, 17- uione were at a source temperature of 250 °C and a source
pressure of l x 10‘5 torr.

129

 

 

1
:1 * P-1x10‘5Torr
080 r + P-2x10-5Torr
' s C] P-4x10-5Torr
.9 ‘
‘5 _
“3 I
_ :5\
0.40 - '
. "
EA;
0.00 I I I I I I I I I T I I I I I I I I I I I I I I I I I I I I I I I I
150.00 200.00 250.00 300.00

Temperature (00)

Figure 4.3: The ratio of the peak areas of the ion at mlz 298 (from 1.4-androstadien-
3,11,17-trione) to ion at m/z 285 (from l-androsten-3,l7-dione) as a function of source
temperature and pressure.

variations in pressure can be compensated by comparing the analyte area to that Of an
internal standard.

Wm. The composition of the reagent gas can also
influence sensitivity. A preliminary study of reagent gas composition used methane
and ammonia gases; a more involved study with many gases will be presented later.
For the high-response steroids (those that exhibit a molecular anion), there was no
difference in sensitivity between the two gases. However, for compounds with a [M-
I-I]' ion, such as l-androsten-3,17-dione, sensitivity was enhanced by using ammonia
as the reagent gas. Methane was chosen as the reagent gas because it is relatively

nonselective for the compounds analyzed.

 

130

W. Conventionally, the reagent gas enters the source
perpendicular to the introduction Of analyte through the GC. Perhaps if the reagent gas
and the GC effluent were introduced in the same direction, the sensitivity might be
improved. The reagent gas line was attached to the make-up line in the GC. One
advantage was a more stable baseline of pressure. The chromatography was not
improved, however. Sensitivity was slightly increased for high-response compounds,
but not at all for low-response compounds. The make-up line inlet is no longer being
used.

WWW. Normally, the ionization voltage for ECNI is
set by JEOL at 200 eV for the 505 instrument. A modification in the circuit allowed
control of this parameter. When the voltage was varied between 100 and 200 eV, but
no difference in peak areas was observed. Therefore, the default ionization voltage of
200 eV was used. The ionization current has two possible values on the JEOL 505,
100 and 300 ILA. Unquestionably, 300 HA is the recommended setting for ECNI,
because at 100 p.A ions can barely be observed.

mama. ECNI has a narrow range of linearity because Of a limited
concentration of thermal electrons. "Overloading" the sample on a GC column causes
flat-topped reconstructed mass chromatogram peaks under ECN I conditions. A study
was undertaken to confirm that the amount of steroid used in the internal standard fell
within the linear range for these compounds. l,4-Androstadien-3,l 1,17-trione was
linear in the region of 0.4 to 20 ng/ttL (r2 = 0.999). If the point at 20 ng/ttL was
ignored, 4—androsten-3,6,l7-trione was linear between 0.3 and 3 ng/uL. In other
words, the concentrations Of steroid used for the internal standard are within the linear
range. In order to assure linearity for the other compounds tested, their peak area
should be smaller than the peak area of the largest chromatographic peak in the

internal standard.

 

131

W. Three methods Of data acquisition are available using the
JEOL 505: full scan, magnetic field switching selected ion monitoring (MFSIM), and
electric field switching SIM (EFSIM). All three techniques gave the same. corrected
response factors for the same compound analyzed. In a comparison of MFSIM vs. full
scan (100-400 u, scan rate = 0.9 s), two compounds were tested. The mass specu'um of
5a-androstan-3-one has only one significent peak, namely that representing [M-l]';
l,4-androstadien-3,l7-dione has three peaks, those representing [M-H]’, [M-2H]' and
[M-CH3]". In each case, MFSIM produced data that had lower coefficients of variance
(cv). Each compound was spiked with the three-component internal standard (as
described below) and analyzed five times. After division by the response of the
internal standard, the cv of the full scan data was 15-20%, while the cv of the MFSIM
analyses was under 10%. The advantage is that under SIM conditions, the signal-to-
noise ratio is increased sufficiently so that smaller peaks can be detected. Another
advantage of SIM over full scan is the amount of memory needed to store the data. A
typical SIM file has about 130 blocks as compared to 5000 to 10000 blocks (depending
on analysis time and noise levels) for a full scan file. Typically, free space on the
computer is between 100,000 to 150,000 blocks. A back-up tape has only 180,000
blocks. To make data acquisition and storage much more feasible, SIM was used for
all analyses. With respect to sensitivity, EFSIM and MFSIM are not significantly
different. MFSIM, however, is the better documented of the two; therefore, it will be
used for future analyses. Full scan analyses were extremely important in this study
because they were used to determine the retention times of the compounds as well as
to verify the peaks chosen for analysis.

Multiplier. To improve sensitivity, the ion multiplier voltage can be
increased, but this also increases the noise. At the nominal ion multiplier voltage of

1000 V, the signal was too weak to observe. The ion multiplier was increased to 1200,

132
1400, and 1600 V. The 1400 V setting was chosen as a compromise between
increased signal and increased noise.

W. GC was used as the method of choice for sample
introduction. Use of the direct-insertion probe (DIP) was studied in an attempt to
shorten the analysis time without affecting reproducibility. To quantify the responses
of steroids, a three-component internal standard was used. Upon introduction by DIP,
these cOmpounds desorb simultaneously. Assuming the steroid of interest will
probably desorb at the same time, there would be four compounds competing for the
same thermal electrons. This could introduce variables not found in GC analyses. For
example, a compound could have a diminished electron capture response because it
cannot compete as well for electrons as a coeluting compound. The reproducibility of
the technique is on the order of that of a GC inlet, with a cv of about 10%. However,
using the DIP, the reconsu'ucted ion peak shape was very poor. Determining the area
under the reconstructed ion curve was difficult because Often it was impossible to
judge where the peak began and ended.

Wine. One of the most important contributions to reproducibility
is control of the injection volume. After trying many different injection techniques,
the best one is the solvent-push technique. This involves placing about 1 uL of ethyl
acetate in the syringe followed by about 1 1.11. of air, and finally 1 uL of sample. The
sample volume should now be read by placing one end of the liquid at 0 Ill. and
reading the other value. The liquid should be just at the end of the barrel (but not in
the tip of the needle) before injection. Injection should be a slow, smooth motion.
Also, it is easier to read the volume injected from a S-uL syringe, which is calibrated
to 0.1 uL rather the more commonly-used lO—ttL syringe calibrated to 0.2 uL.

WW. With the splitless GC injector, it
is recommended that one start well below the boiling point of the lowest-boiling

analyte or below the boiling point of the solvent. At first, the analysis was started at

 

133

50°C, which is below the boiling point of the solvent. Time was wasted in two ways.
First, it takes time for the temperature to ramp up, even at 40 degrees per minute.
Secondly, it takes a significant time to cool the GC oven back to 50 degrees because
heat loss is slow during the last 30 degrees. Under £1 conditions, the chromatography
improved when the temperature program was started at 140' or 160 degrees instead of
at 50 degrees. At starting temperatures higher than 160°C, the chromatography
degraded again. The best sensitivity was Obtained when starting at 140°C. This
temperature program does not work for all compounds. For example, the EI-MS
sensitivity of methyl stearate was greater starting at 50 degrees rather than 140
degrees.

Under ECN I conditions, the best chromatography and sensitivity were obtained
using 140 degrees as the starting temperature. Also, if one holds the temperature at
140 degrees for a minute, a bit longer than the time needed for the purge valve to open,
the sensitivity is also improved.

WW1. As mentioned in the previous section,
the steroids tail much worse under ECNI conditions than under EI. Reviewing twenty-
nine randomly-selected papers in the literature was revealing, as shown in Table 4.1.
About 60% of the compounds tail. The type of instrument and the column does not
seem to have an effect: for example, PAHs tail on both nonpolar and slightly polar
columns. Tailing seems to be an inherent trait of ECNI. As shown by Others (10),
interactions with free radicals on the ion source surface result in severe tailing.

W. For most of the preliminary work, two DB-l
30-m capillary columns with 0.25-micron packing were used. One had an inner
diameter of 0.25 mm and the other, 0.32 mm. Tailing was noted on both columns,
especially for the later-eluting substances. Sometimes the tailing was so extensive that

the later-eluting compounds were completely lost in the noise. Tailing accounts for

Table 4.1: Chromatography Under ECNI Conditions

Compound
Bromobenzenes
Chlordane
Bromobenzenes
Perfluorocarbons
Fluoranil

Chloranil

PAH

PAH

PFB-TMS of Dopamine
PFB of Diclofenac
HFB of Clebopride

Methylated Dihydrophyridine

Calcium Antagonist
Phenothiazines
Mepirodipine

I-IFIP Of Quinolinic Acid
s-Triazines

PFB of Histamine

Abscisic Acid Methyl Ester

Bisu'ifluoromethylbenzyl
of Quinoxaline

I-IFB of Quinoline
Lorazepam

PFB of Indomethacin

HFB of Estradiol

MO-TMS of Corticosteroids

MO-TMS of Corticosteroid
Acetates

Trifluoromethylbenzoyl
of Steroids

Pregnane Fatty Acyl Esters
Oxidized Dexamethasone
TMS of Dexamethasone

Column

DB-l
DB- 1
SE-54
SIS-54

SE-54

OV-l
DB-l

134

Instrument
Finnigan 3300

Nerrnag RIO-10C

Finnigan 4000
Finnigan 4510

VG 7070 E-HF
VG 7070 E-HF

HP 59858

Finnigan 3200
Finnigan 3300
Finnigan 4500
Finnigan 4021
JEOL DX-303

JEOL D-300
JEOL DX-303
Finnigan 3200
HP5890

Nermag R10- 10C
Finnigan TSQ-70

Finnigan 4500

JEOL DX-303
VG 12-250

Finnigan TSQ-46

VG 70—70E
HP 5998A
HP 5998A

Finnigan 4500

VG 7070H
TSQ-70

0V-1701 HP 5985

Peak Shape
Slight tailing
Gaussian
Fronting
Tailing
Tailing
Gaussian
Slight Tailing
Tailing
Gaussian
Tailing
Gaussian
Tailing

Tailing
Tailing
Gaussian
Severe tailing
Slight tailing
Gaussian
Gaussian

Tailing
Severe tailing
Gaussian
Tailing
Gaussian
Gaussian

Tailing
Gaussian
Slight tailing
Gaussian

Ref.

00

10
10
11
12
13
14
15
16

17
18
19
20
21
22
23

24

26
27
28
29

30

’31

32
33

135

some of the problems with precision because it is more difficult to determine where the
peaks begin and end. Both of these columns developed a contaminant that coeluted
with 4-androsten-3,6,l7-trione, a component of the internal standard.

The source of contamination was sought by testing other columns. Another
DB-l column was tested for contamination. This DB-l 30—m x 1.0-um x 0.25-mm
inner diameter column was fairly new and had only been used for ABA (not also for
service work like the other two columns mentioned previously). In Figure 4.4, the
comparison between the old column (0.25-um packing) and the new column (1.0-um
packing) is evident. Not only does the new column not have the contamination, it also
miminrizes the tailing. Unfortunately, the 30-m column, with a thicker film, also
doubled the retention time. The 15-m column keeps most analysis times below 10

minutes, an important consideration in replicate analyses.

Experimental
Materials

Commercially-available steroids were purchased from Steraloids (Wilton, NH)
or Sigma (St. Louis, MO). Quinones were purchased from Aldrich (Milwaukee, WI).
Many cholestanes, monocyclics, and bicyclics were a gift from Dr. William Reusch of
the Department of Chemistry, Michigan State University. The other steroids were
Obtained from the Steroid Reference Collection (Queen Mary College, London, D. N.
Kirk, Curator). HPLC-grade ethyl acetate was from Mallinckrodt (Paris, ~KY) or J. T.
Baker (Phillipsburg, NJ). All samples were stored in silanized vials.
Internal Standard

To minimize variances between replicate analyses, a three-component internal
standard was added to each sample. Androstanes were chosen because pregnanes and
cholestanes have longer retention times. Compounds with three different responses

were used. The standard was made of 2525 ng/uL 1-androsten-3,l7-dione, 14.2 ng/uL

136

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6 Retention Time (min) 7

(a)

300.2

285.2 I‘

1150 1300 1450
Sean number
Retention Time (min)
12 13 11:1

(b)

300.2 A

298.2

285.2 A

1‘ _ f - fl 1 I .
2300 ' ‘ 2600 3000

Sean number

Figure 4.4: Reconstructed mass chromatograms of the major ions of the three
components of the internal standard. (a) Note the tailing on the DB-l 30-m x 0.32-mm
x 0.25-pm column. (b) The tailing is minimized on a DB-l 30—m x 0.25-mm x 1.0-
pm column, but the retention time is doubled.

 

137

l,4—androstadien-3,11,17-uione, and 17.7 ng/uL of 4-androsten-3,6,17-trione. To
prepare a sample for analysis, 50 11L of the internal standard stock solution was added
to 500 pl. of the sample solution. If the analyte of interest coeluted with one of the
internal standards, a two-component internal standard was used.
Gas Chromatography-Mass Spectrometry

ECNI mass spectral data was obtained on a JEOL JMS-AXSOSH mass
spectrometer (Peabody, MA) interfaced to a Hewlett Packard 5890 gas chromatograph
(Palo Alto, CA). The following parameters were used: source temperature, 200°C;
ionization current, 300 M; electron energy, 200 eV; modulating gas, methane; source
pressure (measured at the source housing), 2 x 10'5 torr; conversion dynode voltage,
10 keV; and accelerating voltage, 3 keV. A J&W 15-m x 0.25-mm id x 1.0—um DB-l
column (Folsom, CA) was directly inserted into the mass spectrometer source. Helium
was used as the carrier gas. The splitless injector was held at 260°C and the transfer
line at 260°C. For steroids, the GC oven was held at 140°C for 1 min, then ramped to
260°C at 40°lmin, followed by a 4°lmin ramp to 290°C. For quinones, the GC oven
was ramped at 40°lmin from 50 to 160°C, then at 15°lmin to 270°C, and finally at
4°lmin to 290°C. For cyclohexanones and other compounds that elute at less than
130°C, the following program was used: the GC oven was ramped at 10°/min between
50 and 140°C, then at 40°lmin to 270°C, and finally at 4°lmin to 290°C. Selected ion
monitoring (SIM) in the magnetic field switching mode (with an open collector slit)
was used to analyze the most abundant ions in each sample. To increase the signal, the
electron multiplier was held at 1400 V, regardless of the age of the multiplier.

Electron impact spectra were obtained to check the composition and purity of
each sample. Either the JEOL 505 or the HP Mass Selective Detector (MSD) Model
5970 was used. The JEOL was used under identical conditions as above, except the

ionization current was lowered to 100 11A. The scan rate was 1.0 s for a mass range

 

138

from 50 to 500 u. The MSD was connected to a HP 5890 GC. Source temperature
was 200°C, interface temperature was 280°C. and the mass range was 50 to 500 u.
Gas Chromatography

All GC chromatograms were obtained on a Varian 3700 gas chromatograph
(Walnut Creek, CA) that had been modified to accept megabore (0.53-mm inner
diameter) columns. The following instrumental conditions were used for all modes of
operation: injection temperature, 260°C; detector temperature, 360°C; flow rate, 15
mm of nitrogen; 15-m x 0.53-mm id x 1.5-um DB-l column from J&W. The
temperature program for steroids started at 200°C and ramped at 5°/min to 280°C with
a 5-minute hold at the upper temperature. The temperature program for other
compounds varied. The integrator used was a HP 3393A. For FID analyses, the range
was 10—11 amp/mV with an instrumental attenuation of 1. When using the ECD mode,
the range was changed to 10 amp/mV. the output to negative, and the attenuation to 32
through 128 depending on sample response and the integrity of the detector. The
make-up gas was adjusted to 40 mL/min total flow.

The response for 1,4-androstadien-3,l l,l7-trione was Optimized on the Varian
GC. The detector temperature should be about 50°C above the highest temperature of
the ramp program. In this experiment, the detector temperature was varied from 300 to
380°C in steps of ten degrees. The relative responses at each temperature were similar.
No advantage of higher or lower temperature was observed, so a temperature of 360°C
was used. The optimal flow rate was also determined. The make-up flow was
adjusted to a total flow varying between 12 and 86 mL/min (measured using a
bubblemeter at the ECD outlet). Three different compounds with three vastly different
responses were used as probes: l,4-androstadien-3,11,17-nione, l,4-androstadien-
3,17-dione, and Sa-androstan-17-one. In each case, the response was highest at 12
mm and decreased almost linearly until 55 mL/min. Any higher value of flow

resulted in almost a constant response. The responses to flow track each other;

139

therefore, the relative responses should be independent of flow. Flow rates under 20
mL/min are difficult to achieve in practice because the separation suffers. Therefore, a
total flow of 30 mL/min was chosen.
Sample Preparation

Samples were prepared at concentrations ranging from 0.01 ng/uL to 2 ng/uL,
depending on the electron-capture response. The ECD was used to estimate electron-
capture response and concentration needed for mass spectrometry. All samples were
analyzed by GC-EI-MS to determine structure and purity. The concentrations of
impure samples were determined using the GC-FID and the effective carbon number
technique described below.

The samples were analyzed by full scan ECNI-MS to determine the major ions.
Then, the internal standard was added to the sample. After replicate (n = 3-5) SIM
analyses of the most abundant peaks, the sum of all the ion current monitored was
divided by the ion current for the internal standard.
PCC of Selected Steroids

Some of the steroids that were not available from other sources were prepared
using chemical oxidation. The procedure used is described in Chapter 5 under the
experimental procedure for mouse plasmas. The oxidized steroids were analyzed by
E1 to confirm structures. Those with sufficient yields were used in the ECNI response
study. Effective carbon number (ECN) procedures described below were used to
determine concentration in every case save one. Oxidized dexamethasone was
synthesized in sufficient quantity and recrystallized to give a highly accurate primary
standard. Those compounds ptepated by the PCC method an: listed in Table 4.2.
Preparation of Derivatives

WW These derivatives were
prepared based on the method of Wilson and co-workers (34). About 100-150 ug of

steroid were transferred to a silanized reacti-vial (Pierce) and dissolved in 50 LIL of

140
benzene. Then, 50 pl. of trifluoroacetic acid anhydride (for the TFA derivative) or
heptafluorobutyric anhydride (for the HFB derivative) were added, and the mixture
was heated at 60°C for 30 min. After the reagents were evaporated under nitrogen, the

residue was dissolved in 50 uL benzene.

Table 4.2: Steroids Prepared by Chemical Oxidation

Starting Material Product
Dexamethasone 1,4-Androstadien-3,11,17-trione-90t-F-16a-methyl
Flumethasone 1,4-Androstadien-3,1 1,17-one-90t,6a-F- 16a-methyl
Fludrocortisone 4-Androsten-3,11,17-trione-90t-F
Fluoromethalone 1,4»Androstadien-3,1 1,17—trione-9a—F-6a—methyl
Fluoxymesterone 4cAndrosten-9ot-F- 170t-methyl-17B-ol-3,1 l-diol
6B—Hydroxydexamethasone 1.4-“Androstadien-3,6,l 1,17-tetraone-9a-F—l60t-

me y

Cholestan-Sa,6a-epoxy-3-ol Cholestan-Sa,60t-epoxy-3-one
4—Androsten- 16a,17B—diol-3-one 4—Androsten-3, 16,17-trione

l,4-Pregnadien-1 1,17,21-triol- 1,4.Androstadien-3,1 1,17-dione-60t-F
3,20-dione-6ot-F

4-Pregnen-1 1-ol-9a-F-3,20-dione 4-Androsten-90t-F-3,l 1,17-trione

W. The procedure for this derivatization is
based on the method suggested by the Pierce Catalog (35). To 100 ug of steroid in a

silanized l-dram vial was added 0.1 mL Florox. The reaction mixture was heated at
65° for 1 h. After cooling, the solvent was evaporated under nitrogen. Then 1 mL
cyclohexane and 1 mL water were added to the vial and vortexed. The cyclohexane
layer was removed to another vial and dried with N a2804.

W (pentafluorophenylsilyl ethers) will selectively react depending
on the conditions used (36—37). A 10:1 flophemesyldiethylamine:flophemesyl chloride

 

141
solution will react with unhindered secondary hydroxy groups and tertiary 17B-
hydroxy groups. This reagent will not react with a 113-hydroxy group. To about 100
[lg of steroid in a reacti-vial, 20 pi of the 10:1 mixture plus 20 pl. of pyridine were
added and vortexed.

MS. These derivatives (38) were made by adding 100 ILL of 100 rig/Ill.
methoxyamine-HCI (Aldrich) in dry pyridine to about 40 pg of steroid. After heating
at 60°C for 4 h, the solvent was evaporated under nitrogen until glassy (about 30—45
min). Then zoopL of Sylon BTZ (Supelco) were added, and the solution was heated
to 80°C for 24 h.

IMS, These derivatives were synthesized using the method of Kilts and co-
workers (39). In a reacti-vial, 10 [lg of potassium acetate and 10 ug of steroid were
pre-dried. After 50 uL of BSTFA (ND.bis(trimethylsilyl)trifluoroacetamide) were
added, the mixture was heated at 80°C for 5 h. After the solvent was evaporated under
nitrogen, the residue was reconstituted in 50 11L hexane.

mm. The acetylation is based on the method of Wortic and co-workers
(40). To the sample was added 50 IL of pyridine and 130 11L of acetic anhydride.
The reaction mixture was heated for 2 h at 60°C. After evaporation under nitrogen, the
residue was dissolved in 50 1.1L of ethyl acetate.

Determination of Concentration Using Effective Carbon Number

Many compounds obtained for this study did not have accurately-known
concentrations due to the presence of impurities or incomplete synthesis in this
laboratory. To detemrine concentration, the GC-FID was used as a carbon counter.
The environment that the carbon group is in will determine its contribution to the total
carbon number of the molecule, commonly referred to as "effective carbon number"
(ECN). For example, an aromatic carbon will contribute 1.0 to the ECN, but a

carbonyl carbon will not contribute anything. Steinberg and co-workers (41) have

142
developed ECN for a variety of substituents. Scanlon and Willis (42) added several
derivatives to the list. Table 4.3 summarizes the findings of both studies.
In preliminary work with the ECN concept, a standard curve was made with
standards that had the same ECN as the analyte with unknown concentration. This is a

tedious process, and many compounds of interest do not have pure standards available.

Table 4.3: Contributions to the Effective Carbon Number

Atom Type ECN Contribution

C Aliphatic 1.0

C Aromatic 1.0

C Olefinic 0.95

C Acetylenic ' 1.30

C Carbonyl 0.0

C Nitrile 0.3

O Ether -1.0

0 Primary alcohol 06

0 Secondary alcohol 075

O Tertiary alcohol, ester -0.25

Cl 2 or more on a single
aliphatic carbon -0.12 each

Cl On olefinic carbon 0.05

N Amine as in alcohols
H-C-O-TMS (alcohol) 3.69-3.78
COZ-TMS (acid) 3.0
CH=N-O-TMS (silyl oxime) 3.3
CH=N-O—CH3 (methoxirne) 0.92-1.04

The ECN concept has been very underutilized, but it has been used in the
quantification of PAHs (43,44), steroids (45), TMS of alcohols, and MO-TMS of
carbohydrates in complex mixtures without the use of an authentic standard for each

compound. In general, internal standard(s) with accurately-known ECN were added to

 

143
the sample. To determine the concentration, the following calculations were

performed. The relative weight response factor, F(R-wt), is defined by (42):

MW of compound x ECD-ref.
F(R-wt) =

 

MW of reference x ECN-comp.
The weight of the compound thus will be given by:

area counts for ref. x weight of compound

 

weight of compound =
area counts for compound x F(R-wt)

If the amount injected is known, the resulting concentration is easily calculated. This
method has an accuracy of 2-3% (46).

The choice of reference compound is very important. Its ECN should be
accurately known, independent of response tables. The reference compound also
should be very pure and should be chromatographically resolved from the
compound(s) of interest. Naphthalene, with an ECN of 10, has been used in one
impressive study (46) and will be used for this work.

Table 4.4: Estimated Concentration Using the ECN Method

Compound. Actual Estimated Percent
Concentration (ng/uL) Concentration (ng/uL) Difference
1,4-A-3,11,17-trione 188 180 4.0
1,4—A-3,ll,l7-trione 104 100 ’ 3.9
4—A-3,6,17-trione 177 180 2.0

*A=Androstane parent molecule.

A 434 ng/ttL solution of scintillation-grade naphthalene (99+% pure, Aldrich)
was made in ethyl acetate. 100 11L of this stock solution was added to 400 uL of the

sample solutions. All samples were analyzed in triplicate on the Varian GC-FID with

 

144
Table 4.5: Concentrations Estimated by ECN Method

Compound Estimated Concentration (ng/uL)
PCC A4-16,17-ol-3-one 39
C5-3-one-4,4-Me 319 :l: 4
C5-3-one 45 :l: 5
aA-2,4-bromo—3,17-one 303 :1: 10
A5-17-one 390 i 2
C5-3,4-one 61
C5-3,7-one-4,4-Me 27 a: l
aA-l 1-one 630 :l: 10
OLA-1,6,17-one 179 i 20
aA-11,17—one 3300 :l: 600
0tA-3-F-7,11,17-one 550 :t 40
0tA-17-F-3-one 370 i 20
aA-2,17-one-9-F 111 i 20
PCC P1t4-11,17,21-triol-3,20.dione—6a-F 23 a: 0.1
PCC P4-11-ol-9a-F-3,20-dione 21 i 2
aC-4-one-5,6-epoxy 430 :1: 10
(lP-3,20-one-4B,5B—epoxy 390 :l: 10
Isophorone oxide 4900 :l: 100
PCC C—5,6—epoxy-3-ol 37 i 3
Compound 3 20 :l: 0.4
Compound 4 36 2t 1
Compound 5 19 i 0.6
Compound 6 550 i 20
Compound 9 940 21:8
Compound 12 68 :l: 8
Compound 14 51 :l: 1
Compound 22 32 :l: 1

 

145
a DB-l megabore column. The temperature program started with a 2-min hold at
50°C, then a ramp at 10°/min to 280°C and another 5-min hold.
The procedure was attempted on compounds whose concentrations were
accurately known. From the results of this study (Table 4.4), the accuracy of this
method was found to be about 2-4%. The list of compounds whose concentrations was

determined by this technique is in Table 4.5.

ECNI Response Studies of Various Classes of Compounds
Preliminary ECD Study

A preliminary response study was performed on the Varian GC with electron
capture detection. The results are presented in Table 4.6. l,4-Androstadien-3,11,17-
trione was clearly the compound with the highest response. The absence of either an
l-one or an ll-one decreased the response by two orders of magnitude. The 4wen-3,6-
dione structure had a response one order of magnitude less than 1,4-androstadien-

3,11,17-trione. The other compounds tested had poor responses (<50 arbitrary units).

Table 4.6: Relative ECD Responses of Steroids

Exp. Rel. Lit. Rel.
Compound Response Response (2)
5a-Androstane 1 1
5a-Androstan-3—one 4.3 «-
50t-Androstan-l7-one 3.2 ---
Sa-Androstan-3,l7-one 1.3 10
4»Androsten-3,11,17-trione 18 21
l,4nAndrostadien-3,17-dione 123 450
2-Androsten-17-one 6.1 ---
5a-Androstan-3,1 l,l7-trione 26 «-
4-Androstene-3,1 1,17-trione 340 24800
l,4-Androstadien-3,1 1,17-trione 15000 53500
4—Androsten-3,6, l7-tlione 2600 «-
4—Pregnen-3,20—dione 36 2220
5a-Pregnan-3,1 1,20-tlione 45 «-

146

The results from this study were compared to those obtained by Lovelock and
co-workers (2). The trends in response are the same, but the absolute values differ
greatly, especially for medium-response compounds like 4-pregnen-3,20—dione. This
may be due to differences in the electron-capture mechanism. Lovelock used a 3I-I-foil
while modern electron capture detectors are equipped with °3Ni-foils. Also, the
Lovelock study used 5% methane in argon as the carrier. This carrier gas is far
superior in its ability to facilitate electron capture processes than the nitrogen carrier
gas used in this study.

ECNI Responses of K etosteroids

Ketosteroids with cross or linear rt-lt conjugation, as well as steroids not so
highly conjugated, were analyzed to determine their ECNI response. The effect of
halogen substituents in steroids with and without extensive conjugation was also
studied. Epoxides can increase the ECNI response; therefore, the relative responses of
epoxides within a steroid nucleus were determined. The results from these response
studies are presented in Tables 4.7-4.9. All molar responses are corrected for internal
standard and reported relative to that of Sa-androstan-3-one. The major ions are those
monitored by SIM; they are listed in order of intensity.

Some functional groups do not contribute to the response; methyl groups and
cyano groups are two examples. An ethynyl group on the same atom as a hydroxyl
does not affect the response. For example, contraceptive steroids with the 170t-
ethynyl, l7B-ol functionality are not amenable to ECNI analysis without the use of
fluorinated derivatives.

An unfunctionalized steroid skeleton, e.g., Sa-androstane, has an extremely
low electron-capture cross section. The presence of isolated n-electron functional
groups (such as carbonyls, olefins, and epoxides) enhances electron capture by more
than a hundred-fold, but the effective relative response is still low. Relative to 50t-

androstan-3-one, these responses range from 0.5 to 1.2.

147

Table 4.7: ECNI Relative Responses of Steroids

Compound
Androstanes
aA
01A-3-one
aA-17-one
orA-l 1-one
aA2-17.one
A5-17-one
A4-3—one
A4’15-3-one
A3t5-17-one
(IA-3,17-Ofle
Ai-3,17-one
A-3, 17-one

A4v6 -3,17-one
A1495-3,17-one
A4-3,16,17-one
A4-3,6,17-one
nor-aA2t7-1,4-one-14a-Me
AL 4 -,3 17-one
aA-3,11,17-one
A4 3,11,17-one
A5-3,11,17-one
A1-4-311,17—one
aAWli)-3,17—one
A14 9(11)-3,17-one
aA16 17-CN
«A4 17a-ethynyl-17B—ol-3—one

Pregnanes
aP-3,20—one
P4-3,20-one
P‘-3,1 1,20-one
P4-3,6,20-one

Chglestanes

C4: 6

C4 6- 3- Cl
aC-2-one
aC-3-one
C1-3-one
C5—4-one

C5- 3-one

C5- 3-one-4, 4-Me
C5- 3 7 -one-4 ,4-Me
C4-3-one-2, 2-Me
C4-3-one
C-3,6-one
C4-3,6-one

I"A-3—one has been assigned a relative response of 1.0; n = 3 to 5.

Mam—Inns

[M-2H]'
[M-Hl'
[M-Hl'
[M-Hl'
[M-Hl'
[M-Hl'
[M-Hl'
[M-Hl'
[M-Hl’
[M-Hl'

[M-H] ,[M- “CI-I3]
[M- H]

[M-ZH] ,[M-16]’
[M-HT

M‘

[M-H]
[M-CH3] ,M'
[M-H]

lM-Hl'

[M-Hl'
[M-Hl'
gM-Zm'JMJQ'

[M- H] .2[M’r0 -H]
[M-H] [M+02'Hl
[M-H]
[M-Hl'

MW

0.0040 :1: 0.0008
1.0 :l: 0.07
1.2 i 0.1
0.95 :t 0.22
1.1 :t 0.2
0.21 i 0.03
1.2 i 0.1
0.94 :t 0.09
0.67 :1: 0.07
1.4 i 0.1
0.98 i 0.11
1.2 :l: 0.1 .
18 :l: 2 ."
300 :l: 50
27 :t: l
300 :1: 15
100 :l: 6
2.1 :1: 0.1
0.89 i 0.11
3.2 :l: 0.3
0.24 :t 0.01
360 i 40
0.59 :l: 0.1
150 :l: 20
1.2 i 0.1
0.37 :t 0.06

’5." .

l.2:l:0.1
1.4i0.3
9.1 21:0.5
280:1:23

0.58 :l: 0.19
0.47 :l: 0.03
0.35 :1: 0.03
0.12 :l: 0.01
0.27 :l: 0.02
0.05 :t 0.003
0.91 :l: 0.11
0.17 :1: 0.05
0.22 i 0.04
7.2 :1: 0.8
3.5 :1: 0.5
0.25 :t 0.02
1.6 :t 0.5
280 :1: 24

148

Table 4.8: ECNI Relative Responses of Epoxysteroids

Compound

Androstanes
aA-3—one
aA-laza-epoxy-3-one
aA-9a,11a—epoxy-3-one
A4 -3-one
A4- l6a,l7a-epoxy-3-one
aA-3-one- l7-ol acetate
A-4B,SB-epoxy-3-one-
17-ol acetate
A1-4B,SB—cpoxy-3-one-
17-ol acetate
A1'4-3,17-dione
orA-3,17-dione
A4- 3, 17- dione
A-4a,5a-epoxy-3,l7-dione

Pregnanes
txP-3,20-one
P-4a,5a-epoxy-3,20—one
aP-16a,17a-epoxy-3,20-dione
P4-16a,17a-epoxy-3,20-dione
P4- 813- -l,9-epoxy-3 20-dione
P4-3,11,20-trione
aP- 16a,17a-epoxy-

3,1 1,20-trione

Cholestanes

C-Sa,6a-cpoxy

C-3-cyano-5a,6a-epoxy

C-3- Cl- ~50t,6a-epoxy

C-Sor ,-6a~epoxy ~4-one
5.6B-epoxy-4-one
4,5[3-epoxy-3 -one

- 1a,2a-epoxy-3-one

*A-3-one has been assigned a relative response of 1.0; n = 3 to 5.

[M-Hl'. [M-CH3]
[M-H ]

[M-Hl'

M‘

[M'Hl'

M-

[M'H] '

[M'H] '

[M'H]- 9M-9 [M'OZ] -
M-

[M-Hl’

[M-Hl'
[148]', [M-H]', M'

[M-Cll'. M'. [M-Hl'

[MP1]

0.13
6.8
6.4
6.
5.
9.
5

HHHHHHH
O‘F‘OOOOO
NO‘F‘a‘bc
Wu—

1
2
2
8

149
Table 4.9: ECNI Responses of Halogenated Steroids

 

Compound Maierlens W5:
ctA-3-one [M-H]‘ 1.0 i 0.1
aA-3-one—17a-F [M-HT - 0.41 a; 0.02
aAl6-3-one-17a-F [M-H]' 1.6 :t 0.08
orA-3-one-1701,17B-F [M-H]' 2.1 :l: 0.1
OLA-2,17-one [M-H]' 3.5 :l: 0.3
aA-2,17-one-901-F [M-H]' 0.75 :l: 0.05
aA-11,17-one [M-H]' 3.8 :1: 0.5
aA-11,17-one-30t-F [M-H]‘,[M-F]' 0.09 :l: 0.01
OLA-7,11,17-one-3a-F [M-HF-ZH]',[M-H]' 4.6 i: 0.4
aA—3,l7—one [M-H]‘ 1.0 :l: 0.05
aA-2a-Br-3,17-one [M-Br]', Br’ 0.35 :l: 0.05
orA-2a,4a-Br-3,17-one [M-ZBr]', Br“ 0.80 i 0.37
aA—2a-I-2,l7-one [M-IT, I' 0.35 i 0.05
A4-3,11,17-one-9a-F [M-I-IF]',M',[M-HF—CH3]' 2600 :1: 300

A4-l7B-ol-3,1l-one-9a-F-l7a-Me [M-HF]',M', [M-HF-H201' 3000 a: 150
2.14-3.1 1,17-one-901-F-16a-Me [M-HF]',M‘,[M-HF-CH3]‘ 4100 3: 420
A1.4-3,1 1,17-one-9a,6a-F-1601-Me [M-HF]',M',[M-HF-CH3]' 5400 :t 250
A1v4-3,11,17-one-9a-F-6ot-Me [M-HF]',M',[M-HF—CH3]‘ 3900 a: 470
A1v4-3,6,1 1,17-one-9a-F-16a—Me [M—I-IF]',M‘,[M-HF-CH3]' 7400 a: 500
*A-3-one has been assigned a response of 1.0.

150

When two or three isolated rt-functions are present in a steroid, little
enhancement of ECNI response is observed. Even if two such functions are
conjugated, as in the common 4—en-3-one moiety, the relative response remains low.
For example, note the responses of 501-androstan-3,11,17-trione (with a relative
response of 0.89), 4-pregnen-3,20—dione (1.4), and 4-androsten-16a,1701-epoxy-3-one
(1.8).

The introduction of a fourth unconjugated n-function appears to give a modest
enhancement of response, as observed for 4-androsten-3,ll,17-trione (3.2) and 4-
pregnen-3,11,20-trione (9.1). Interestingly, the (rt-diketone, 4-androsten—3,16,l7-
trione, has an unusually high response (27) relative to similar compounds. This
characteristic of a-diketones is expected from their low reduction potential (for
3.3.6.6-tetrarnethyl-1,2-cyclohexanedione, 151/2 = -0.71 V vs. SCE (47)). This
provides a useful reference for evaluating the results for a,B-epoxyketones discussed
below.

From the data in Table 4.8, it is clear that simple mB-epoxyketones have ECN I
responses at least an order of magnitude greater than those of their corresponding “~3-
unsaturated ketones. For example, 5a—androstan-10r,2a-epoxy-3-one has a relative
response 25 times greater than that of 1(5a)-androsten-3,l7-dione. When compared to
other cyclic ethers, epoxides are exceptionally reactive with various nucleophilic
reagents. Indeed, organic chemists have described this behavior as "carbonyl-like,"
just as cyclopropane is referred to as "olefin-like." Clearly, this is a helpful analogy
for this work, because ct,B-epoxyketones then become "ct-diketone-like," and the
enhanced response is expected.

The greatest ECNI responses result from polyunsaturated ketosteroids that have
more extensive conjugation than is present in the previous cases. The ECNI responses
of compounds with cross conjugation differ from those with extended (or linear)

conjugation. For instance, the linearly-conjugated dienone, 4,6-androsten-3,l7-dione,

151
has nine times the response of its cross-conjugated isomer, 1,4—androstadien-3,l7-
dione. Extending the 4-en-3—one conjugation by the introduction of a carbonyl group
at C-6 results in a dramatic enhancement of response, as noted for 4-androsten-3,6,20-
trione (relative response of 300), 4-pregnen-3,6,l7-trione (280), and 4-cholesten-3,6-
dione (280). For these linearly-conjugated steroids, the substitution at the C-17
position does not affect the overall ECN I response. The enedione moiety in the 4-en-
3,6-dione compounds is a vinylog (a system extended by the placement of an allyl
group between the functional groups, as shown in Figure 4.5) of an a—diketone, which

has been previously noted for its enhanced electron-capture response.
0 0
II II
(a) —" C _ C ——
O O
b u n

Figure 4.5: Structure (b) is a vinylog of structure (a).

An intriguing aspect of this study is the synergistic effect that cross-conjugated
or homo conjugated rt-functions have on the higher-response groupings such as 01-
ketoepoxides and linearly-conjugated enediones. Examples of this synergistic effect
include the C-1 double bond in 1,4,6—androstatrien-3,17-dione (300) and l-androsten-
4B,5|3-epoxy-3-one-17-ol acetate (200). The epoxide in 4,6-cholestadien-1a,20t-
epoxy-3-one (58) is another example. This synergistic effect is particularly dramatic
in cases of 1,4-dien-3-one cross-conjugated systems combined with a 9(11)-double
bond or a C-11 carbonyl function, such as 1,4,9(11)-androstat1ien-3,l7—dione (150)
and 1,4-androstadien-3,l 1,17-trione (360). The presence of a carbonyl group at C-ll
enhances the response greater than the corresponding double bond.

{.4

152

Because similar ECNI responses were observed for the 4-en-3,6—dione and the
l,4-dien-3,11-dione systems in androstanes, we attempted to estimate the contribution
of a polar substituent on the overall response by using the Taft equation (48).
Analogous to the Hammett equation, the Taft equation applies linear free-energy
relationships to aliphatic systems. Using a modified Taft treatment, the effects of a
polar substituent, designated by 0*, were calculated assuming that the inductive effects
would be transmitted to C-3 by all possible bond paths in the steroid. Also, it was
assumed that each bond would attenuate the effect of any substituent by a factor of 0.5
per carbon atom (48). Ethylene bonds were assumed to attenuate the effect by a factor
of 0.508. Thus, the net inductive effect was calculated for all reasonable bond
connections by the following equation:

20: = 0"(a1-l- bm + on + ...)
where a, b, and c are the attentuation factors for either single or double bonds and l, m,
and n are the number of bonds in that pathway. By this treatment, the inductive effect
for the 4-en-3,6-one system was 3.4 times greater than that for the l,4-dien-3,l l-dione
system. As shown in Table 4.7, the relative responses of 4-androsten-3,1 1,17-trione
(300) and l,4-androstadien—3,l 1,17-trione (360) are similar. Therefore, the marked
influence of an ll-carbonyl appears to be an example of a through-space nonclassical

conjugation effect on the A-ring.

ECNI Fragmentation of Ketosteroids

The ECNI spectra of the 1,4-dien-3-one steroids include a peak for a [M-CH3]'
ion as well as a M" or [M-HT ion with the exception of 1.4—androsmdien-3,ll,17-
trione whose spectrum consists of only a peak for M". In all of these compounds, the
18-methyl group is ideally oriented for fragmentation. After capture of an electron and
loss of the 18-methyl group as shown in Figure 4.6, a very stable phenolate ion is
formed. There is one exception to this pattern. Loss of the 18-methyl group from

153
1,4-androstadien-3,l l,l7-trione should be less favorable because the electron-rich
oxygen atom of the 11-carbonyl group crowds or has a nonbonding interaction with
the C-1 hydrogen which, as part of the phenolate ring, will have a negative charge.

This crowding is more severe in the [M-15]' phenolate ion than in the M" parent ion.

\ \ \

.
+ e . . C H e .
O i G) 3Q. ! 3 6:9.

Figure 4.6: Proposed mechanism for the formation of a phenolate ion in 1,4-dien-3-
one steroids.

The ECNI spectrum for 4-androsten-3,11,l7-trione is interesting. The base
peak represents [M-2]'; a less intense peak represents [M-l6]'. The formation of these
unusual ions may involve an interaction between the ll-carbonyl group and the A-
ring. The ll-carbonyl oxygen is close to C-1 and may be able to extract a hydrogen
atom. The resulting radical at C-1 may, in turn, induce fragmentations leading to
formations of these ions (see Figure 4.7).

ECNI Responses of Epoxysteroids

For enhanced ECNI response, an epoxide group must be alpha to an
electrophilic group. An isolated epoxide group, such as the one in 501,601-
epoxycholestane, has a poor response. If epoxide and carbonyl groups are located at
opposite ends of a molecule, as in 4-androsten-l6a,l7a-epoxy-3-one, no enhancement
of ECN I response occurs. However, by positioning an electrophilic group, such as a
chlorine, a cyano, or a carbonyl, in proximity to or in conjugation with an epoxide, an
enhanced response is observed. Results for substituted cholestanes are reported in
Table 4.8. The combination of an epoxide in the 9,11-position with a carbonyl

function at C-3, as in androstan-9a,11a-epoxy-3-one, also showed enhanced response.

154

.283
-:.:.m-=2m8e§-v «o 8.58% qum 05 E «:25 -87.): e5 -82. Co 53.58 05 e8 Ewan—Boa 38qu ”5. «Sufi

 

 

 

155
In general, a,B-epoxyketones have an enhanced response at least an order of
magnitude greater than that of the equivalent enones. For example, 5a-pregnen-4a,5a—
epoxy-3,20-dione has a response 15 times greater than that of 4-pregnen-3,20-dione.
Other functional groups in the molecule have a little effect on the overall response. This
is illustrated by the fact that Sa-pregnan-16a,17a-epoxy-3,20—dione and 4-pregnen—
16a,l70t-epoxy-3,20-dione have almost the same response.

ECNI Responses of Halogenated Steroids

If a halogen is added to an already low-response steroid (see Table 4.9), the
relative response will decrease further. For example, the presence of a 17—fluorine atom
in 5a-androstan-3-one decreases the response. Another electrophilic group near the
fluorine will increase the response slightly, as in those of 5a-androstan—l7a,17B-
difluoro-3-one vs. Sat-androstan- 17 a-fluoro-3-one.

Varied responses have been noted for or-halo ketones. B0th Sa-androstan-Za-
bromo-3,l7-dione and 5a-androstan—2a-iodo—3,17-dione have 0.35 times the response
of the unhalogenated 5a-androstan-3,17-dione. The major peak in the ECNI spectrum
in each case is [M-halogen]'; the halogen anion is a minor contributor (< 5%). Even in
the E1 spectrum, [M-halogen]+ is the highest mass peak. Two processes must be
occuring in these compounds: the loss of a halogen followed by the capture of an
electron. Because the fragmentation preceeds the electron capture, the efficiency is
probably not as good as that of resonance or dissociative capture processes, and this
could contribute to the lowered responses.

In contrast, fluorine atoms in some highly-conjugated systems induce impressive
responses. A fluorine added alpha to an ll-carbonyl group in a cross-conjugated
system (e.g., l,4-androstadien-90t-fluoro-16a-methyl-3,11,17-trione) increases the
response by a factor of 10. However, a second fluorine substituent located at C-6 does

not significantly affect the relative response. In the case of a less conjugated system,

156
such as 4-androsten-9a-fluoro-3,l1,17-trione, 9a-fluorine substitution increases the
relative response by a factor of 200. It is likely that 1,4-androstadien-l60t-methyl-
3,11,17-trione has such powerful electron-capturing conjugated n—functions that the
addition of a fluorine atom has a smaller relative contribution to the response than the
addition of a 9a-fluorine to 4-androsten-3,ll,17-trione. The Compound with the best
response of all those tested in this study is 1,4-androstadien-90t-fluoro-l6a—methyl-
3,6,11,17-tetraone. This compound has all the functions associated with a high
response, including a cross-conjugated 1,4-dien-3,11-dione system and a 901-fluorine as

well as a linearly-conjugated 4-en-3,6-dione.

 

 

o I ,CHs e'b' / )Krcrl,
00° ” /
HF
o 0 /

 

Figure 4.8: Fragmentation of 4—androsten-9a-fluoro-16a-methyl-3,l1,17-trione.

An explanation for the different ECNI responses of a—haloketones lies in the
orientation of the halogen. If the halogen is axial, as in the ketosteroids containing the
9a-fluorine, the carbon:halogen bond overlaps the rt-electron orbital of the carbonyl
group at 011, providing a higher electron-capture response. In addition, the [M-HFT'
and [M-HF-CH3]’ fragments are exceptionally stable anions. For the loss of HF, it is
assumed that the hydrogen is from the C-12 position on the other side of the carbonyl
group (see Figure 4.8). This type of HX loss has its basis in the well-characterized
F avorslcii reaction (49). In the Favorskii reaction, a-haloketones in the presence of base
lose hydrogen from one side of the carbonyl and halogen from the other. A carbanion is
formed which rearranges to an ester. After the loss of the l8-methyl group, the
resulting anion has the very stable 4,8,10-t1ien-3-one structure (see Figure 4.8). In

contrast, an equatorial halogen, such as that in 5a-androstan-201-bromo-3,17-dione, does

157
not overlap well with the adjacent C-3 carbonyl group, and provides little or no
enhancement of response.
ECNI Responses of Deri vatized Steroids

Most steroids must be derivatized before GC—ECNI—MS analysis because they
are not inherently electrophilic and/or are thermally labile and cannot survive the GC
intact. The oxidation methodology has been compared to a variety of commonplace GC
derivatives for the analysis of dexamethasone and 50t-androstan-3B,l7B-diol.

Dexamethasone has a thermally-labile 17-substituent. TMS and MO-TMS
derivatives have been often used in assay development of dexamethasone under ECNI
conditions (28, 33). These two derivatives were compared to underivatized
dexamethasone as well as the chemical oxidation product of dexamethasone, 1,4-
androstadien-9a-f1uoro-l6a-methyl-3,l1,17-trione. Because the underivatized
dexamethasone does not survive the GC intact, thus decreasing sensitivity, it was
analyzed by direct probe. The other derivatives were analyzed by GC-MS to insure that
any electrophilic by-products would be separated from the analyte. Internal standard
was added to all samples before analysis.

As shown in Table 4.10, the tetra-TMS and MGTMS derivatives have about the
same ECNI response as the underivatized dexamethasone. Therefore, the advantages
gained using these derivatives are merely 'chromatographic; no additional electron-
capture properties are added to the molecule. In contrast, the oxidized dexamethasone
has a response 40 times greater than unoxidized dexamethasone. Not only are the
chromatographic properties improved, but the electron-capture properties were also
enhanced.

The chemically-oxidized dexamethasone was also compared to more
conventional electron—capturing derivatives, such as TFA, I-IFB, and flophemesyl, in a
separate study (see Table 4.11). All three of these derivatives have significant

fragmentation. Only in the case of the flophemesyl derivative is the most abundant

158
peak indicative of the parent ion ([M-504]' corresponds to the loss of the 17-
substitution). The response of the TFA derivative was disappointing, on the order of a
low-response ketosteroid. The HFB and flophemesyl derivatives had a large enough
response to be beneficial for assay development; however, chemical oxidation has a
response an order of magnitude higher. Therefore, for compounds that can be oxidized

to highly-electrophilic species, chemical oxidation is the best derivative.

Table 4.10: Relative ECNI Responses of Dexamethasone Derivatives

Derivative Relative Response“
None (DIP) 1.0 i 0.2
Tetra-TMS 1.3 :I: 0.1
MO-TMS 1.5 :1: 0.06
Chemical oxidation 41 2t 5

*Underivatized dexamethasone has been assigned a relative response of 1.0.

The chemical oxidation protocol does not work as well for compounds that are
oxidized to low-response species. For example, 5a—androstan-3,l7-diol oxidizes to Sc-
androstan-3,17-dione, a fairly low-response compound. The oxidized product has a
response 500 times that of 5a-androstan-3,l7-diol (see Table 4.12), but only because
the hydroxylated compound is lost on the injection liner or the column packing. The
fluorinated derivatives all provide better response than chemical oxidation. Mono- and
di-TFA derivatives have the TFA substituent as their most abundant peak; however, the
response is fair. The ions at mlz 178 and mlz 194 in the HFB and TFA derivatives are
most likely from wall reactions because of their non-Gaussian mass chromatograms.
The mono-flophemesyl derivative has a better response, but is very difficult to
synthesize. The Florox derivative has a relative response of 1700 which makes it a

good candidate for assay development, but the pentafluorobenzyl moiety is the most

159

Table 4.11: ECNI Relative Responses of Halogenated Derivatives of

Dexamethasone

Compound Maior Ions

[M-101]‘, [M-2HFT',

 
   
 

[M-HFI"

Hzposucnpz-cfir5
coggznalzcars

[Ni-504]", [M-336l"
[Mel “-Cstln, [M-Cerl-

..cna

[M-HF]",M",[M-HF-CH3]'

 

[M-80]",[M-1}', [OCOCF3]"

[000031313 [M-COC3F7]‘.

Relative Response

0.94

260

120

4200

0.05

10

80

30

*Sa-Androstane-SJ 7-dione has been assigned a relative response of 1.0.

160
- Table 4.12: ECNI Responses of Derivatives of 5a-Androstane-3,l7-diol

Relative Response*

 

 

0.0018 i 0.0003

   
   

1.0 i 0.27
[OCOCF3]-,[2COCF3]-, 8.1 i 2.1
[M‘Hl'
[OCOCF3]-,IZCOCF3]-, 31 i“. 8
[M-OCCF3]',[M-H]‘
[M-HF]",[M-2HF]", 2100 4..- 200
[COCgF7]-,[OCC3F5]-
[M-HF]",[M-2HF]", 5500 a 1500

[COC3F7]',[OCC3 F6]-

[OCHC6F6]‘,[OCH205F6]', 1700 i" 300
[178]",[M-HF]"

OSKCHslchFS

[1241",[06115]: 65
M".[06F4l"

H-

16

*Sa-AndrostaneeSJ 7-dione-has been assigned a value of 1.0.

161
abundant anion. All steroids treated with Florox would have this anion in their
spectrum; this would make for a non-selective assay. By far the best derivatives were
the mono- and di-HFB derivatives, which combine both high responses and abundant
peaks indicative of the parent molecule (in this case, [M-HF] ').

For dexamethasone and other highly conjugated 'steroids, the best ECNI
derivative is achieved through chemical oxidation. The highly fluorinated derivatives as
well as the non-electrophilic silylated derivatives all have poor responses,
corresponding merely to enabling the molecule to survive the GC intact. However, for
steroids without electrophilic groups, such as androstan-3 [3,17B-diol, the HFB derivative
is the best choice for optimum sensitivity.

ECNI Spectra of Bicyclics

Bicyclic compounds with cub-unsaturation were available courtesy of Dr.
Reusch. The compounds were either bicyclononanes (fused five- and six-membered
rings) or bicyclodecanes (two fused six-membered rings). The compounds will be
referred to as compound 1, compound 2, etc. The structures of these compounds are
presented in Figure 4.9. Many of these compounds were of interest because features not
found in steroids were available.

The spectra of the bicyclic compounds are slightly more complicated than those
of ketosteroids. More fragmentation is noted, and the losses are not all simple. The
major ions found in the spectra of these compounds are listed in Table 4.13.

Eight of the twenty-six compounds display a molecular anion. Most of these
compounds have more than one conjugated group, such as compounds 12 and 19.
Compound 15 also has a triple bond; [M-2H]' and [M-H20]‘ ions are also noted due to
fragmentation involving the hydroxyl group. The [M-2]' ion appears due to thermal
degradation and is often present in the E1 spectra of alcohols. Compound 5 and 6 are
similar, except for a carbonyl or an epoxide group. Compound 5, with the epoxy group,
has M' as the base peak in its spectrum. The base peak of compound 6 is an oxygen

162

 

Figure 4.9: Structures of the bicyclic compounds analyzed in this dissertation.

163

 

 

Figure 4.9 (cont'd).

164

Table 4.13: ECNI Mass Spectra of Bicyclic Compounds

Compound“

“macro-auto:—

uuuunnununwuuunh
sun—owaaamauuuc

26

[Anion]', (Relative Intensity)

[M-Hl'. (100); [M-CO-Hl'. (15)

[M-Dl'. (100); [M-27l'. (26); [M-D-Hl'. <22)

[M-Hl'

[M-Hl'. (100); [M-2Hl’. <13); [M+CH3]'. (20)

M“, (100)

MO]: (100); M: (8)

[M-H]', (100)

[M-Hl'. (100)

[M-Hl'. (100); [M-ZHT. <49); [NI-3H]: (22); [M-CHgl'. <29)
[M-Hl‘. (100); [M-C6H301'. (29)

[M-Hl'. (100); [M-ZHT. (23); [M-H-CH3l’. (19); [M-2H-CH31'. (48)
M‘, (100)

M‘.(100); W-Hl‘. (75); [M4411 (31); [M4513 27

[M-Hl'. (100); [M-ZHT. (84); [M-Hzol'. (32); [M-Hzo-CHzl'. (18)
M'. (100); [M-Hl'. <50); [M-ZHI‘. <55); [M-H20l'. (6)
[M-Hl‘. (100); [M-propenyll'. (17>; [M-56l'. (28)
[M-C3H4]', (51); [M-C3H5]', (100)

[M-H]', (100); [M-C3H7]‘, (65); [M-78]‘, (14)

[M-C3H41'. (100); [M-C3Hsl'. (99); M'. (3)

[M-Hl‘. (100)

[M-Hl'. (100)

M21100); 1M+c11'. (19)

[M-Hl'. (100)

[M-Hl‘. (100)

M', (100)

lM-Hl‘. (100); [M-461'. 7s

*The compounds are illustrated in Figure 4.9.

165
containing adduct, M' being only a minor contributor. The addition of a halogen group
forces resonance capture as opposed to [M-H]' formation (compare compounds 21 and
22); however, a significant [M+Cl]‘ peak is also noted. This unusual ion may be due to
a contaminant.

Compounds that have ECN I spectra consisting of the [M-H]' anion, tend to be
simple structures with two carbonyl groups and little or no other substituents.
Compounds 3, 7, 8, 21, 23, 20, and 24 all are in this category. Many compounds have
spectra with [M-HJ' as the base peak, but also produce significant [M—2H]' peaks as
well as other peaks resulting from fragmentation or adduct-formation. These
compounds, 4, 9, ll, l4, 16, 18, and 26, tend to have an enone structure with another
carbonyl group three bonds away. Some of these compounds have other interesting
fi'agments. Compounds with methyl substitutions, like 9 and 11, exhibit [M-CH3J‘ or
[M—H-CH3]' ions in their spectra. The spectrum of compound 4 has an adduct ion at
[M+CH3]'; however, the structurally similar compound 3 does not produce this adduct.
In molecules with propene chains, the chain is sometimes lost, as in compounds 16 and
17. Some ring fissure is also possible. Compound 10 has a loss of C6H30 from one
ring (see Figure 4.10). This must be a very facile loss because it is not observed for any

other bicyclic compounds.

0
138

\I'

Figure 4.10: Fragmentation of CgHgO from Compound 10 results in an ion at mlz 138.

O

166
ECNI Responses of Bicylic Compounds

The relative ECNI responses of bicyclics are presented in Table 4.14. The ions
monitored are also noted in this table. Structures of the bicyclic compounds are shown
in Figure 4.9.

The size of the bicyclic rings have no effect on the response. Both
bicyclononanes and bicyclodecanes with two carbonyl groups have the same response
(cf. compounds 1 and 7).

Several functional groups do not affect the response. As shown for other classes
of compounds, the addition of a methyl group does not affect response. Deuterium
substitution in compound 2 decreases the response slightly. One site of 01,13-
unsaturation, such as in compound 11, does not increase the response for bicyclics. For
the compounds which also have a lactone ring moiety (compounds 14-26), the presence
of a double bond in conjugation with a carbonyl group increases the response by a
factor of ten. If the double bond is not conjugated with the carbonyl, the response is
similar to the compound without a double bond.

Compound 6, which has a carbonyl bonded to a ring incorporating a Bflvdouble
bond, has an enhanced response over a comparable cub-unsaturated compound.
Compound 12, with an additional site of unsaturation, has a response 30 times greater
than compound 6.

A linearly conjugated enedione has a far better response than a linearly-
conjugated dienone. Compound 12, an enedione, has a response 60 times greater than
compound 13 with its dienone.

The presence of an epoxide enhances the relative ECN I response. In compound
5, the spiroepoxide is in conjugation with a double bond. The relative response is 2000,
the largest response for any compound in this set.

A neighboring group is important for bicyclic compounds. Compounds 10 and
11 are similar, except that 10 has a side-chain double bond located three bonds away

167

Table 4.14: ECNI Relative Responses of Bicyclic Compounds

Compound

Response‘

Ions Monitored

[M-Hl'
[M-Dl‘

[M-Hl'

[M-Hl‘

M-

M-

[M-Hl'

[M-Hl'

[M-Hl'. [M-ZHI'. [M-CH31‘
[M-Hl'. [M-ZHl'

[rd-161‘. [M-Hl'

M', [M-16]'

[M-44]', [M-HT, M'
[M-ZHl‘. [M-Hl'. [M-Hzol'
M', [M-2H]'

[M-Hl‘

[M-Hl'

[M-Hl‘

[M—41]', M'

[M-Hl'. [M-581'

[M-Hl'

M-

[M-Hl'

[M-Hl'

M', [M—2H]‘

[M-Hl'. [M-441‘

*Compound 1 has been assigned a relative response of 1.0.

Relative

1.0 :l: 0.2
0.67 :l: 0.06
1.6 :l: 0.1
1.1 :l: 0.06
2000 :l: 400
41 i 4

1.0 :l: 0.2
0.46 :l: 0.06
1.9 :l: 0.2
6.7 :l: 0.9
1.3 :l: 0.07
1500 :l: 200
17 :l: 0.9
7.9 :l: 0.2
22 :l: 5

1.8 :1: 0.2
0.49 :l: 0.11
0.53 :l: 0.06
330 :l: 30
1.1 :t 0.1
0.84 :l: 0.04
7.3 :1: 0.8
0.14 :l: 0.04
3.6 :l: 0.6
36 :1: 5

4.4 :l: 0.7

168
from the B—carbon of an enone. Compound 10 has a response seven times greater than
compound 11.

A triple bond substituent one carbon removed from the ring has a response
greater than a double bond. This response is larger than for the unsubstituted
bicyclodecanedione. A comparable increase in response for a triple bond is not noted
for steroids; for example 4-androsten-17a—ethynyl-17B-ol-3-one does not show an
increase in response. ‘

Compounds 16-20 have a propenyl or propyl substitution at C-6. If the only
double bond is in the C3 chain, as in compounds 16-18, the response is similar and low.
If two double bonds are present, one in the propyl chain and the other alpha to the
carbonyl, the response increases by a factor of 300. This increase in response may be
due to the proximity of the propyl double bond to the cub-unsaturation in the ring.

Addition of a halogen and a methyl group to compound 21, as in compound 22,
increases the relative response by a factor of eight. From other studies, it is known that
the methyl group does not affect the response; therefore, all increases in response are
probably due to the chlorine substitution alpha to the carbonyl group.

ECNI Responses of Cyclohexanones

Cyclohexanones, being single-ring compounds, are not able to internally
distribute excess energy and thus are not as sensitive to ECNI as steroids.
Cyclohexanone, with one carbonyl group, is one-tenth as sensitive as its steroid
counterpart, Sa-androstan-3-one. The responses of cyclohexanones are presented in
Table 4.15.

Linear free-energy relationships such as the Hammett equation have been used
in ECNI-MS to compare substituents on aromatic compounds based on the log 0f the
ratio of the major ions. One comparison involves the ratio of a substituent fragment to

the molecular anion (50). Taft 0* values are used in aliphatic linear-free energy

169
Table 4.15: Relative ECNI Responses of Cyclohexanones

Compound MaiQLlQns RelatimResponse:
_ Cyclohexan-l-one [M-H]' 1.0 :l: 0.05
Cyclohexan-l-one-4—methyl [M-H]' 0.89 :l: 0.06
2-Cyclohexen-1-one [M-Hl' 0.91 :l: 0.09
Cyclohexan-l,3-dione [M-H]' 1.7 :l: 0.7
Cyclohexan-l,3,5-trione-2,2,4,4,6,6-methyl M',[M-CH3]' 310 :l: 46
Cyclohexan-Z-one—1,1-dimethyl-3-benzylidine M' 70 :l: 15

*Cyclohexan-l-one has been assigned a relative response of 1.0; n= 3 to 4.

relationships. However, if one calculates the Taft number for the substituted
cyclohexanones, there is no correlation with the ECNI responses. For example, the
20"“ value for cyclohexan-l,3-dione is 0.877, and its ECNI relative response is 1.7. For
2,2,4,4,6,6-hexamethyl-cyclohexan-1,3,5-trione, the Taft number increases to 1.75,
while the relative response increases to 310. In the case of cyclohexan-Z-one-1,1-
dimethyl-3-benzylidine, the Taft number falls to 0.318, but the response increases to 70.
Other factors must be involved in generating the ECNI response and not just the polar
effects measured by the Taft equation.
ECNI Mass Spectra of Quinones

The ECNI-MS relative responses of quinones were of interest because quinones
are biologically important, easily available in halogenated and non-halogenated forms,
and their electron affinities, reduction potentials, and UV values known fiom the
literature. Quinones are instrumental in electron-transfer reaction in biological systems
(51) Ubiquinone (also known as coenzyme Q10), found in mitochondria, is important
because it is widely-distributed, lipid-soluble, and participates in oxidation-reduction
reactions involving the release and uptake of protons. A number of quinones, including
plastoquinone, are the electron carriers between photosystems I and 11 during

photosynthesis. Vitamin K1, a 1,4-naphthoquinone analog, helps maintain the

170
coagulative properties of blood. In this work, simple quinones, such as benzoquinones,
naphthoquinones, and anthraquinones, were studied.

In the EI mass spectra of quinones, the molecular ion often is accompanied by
an abundant ion at [M+2H]+. This ion probably corresponds to the reduction of the
quinone to the dihydroquinone, possibly in the ion source (52). The [M+2H]+ ion is
usually present for compounds with high redox potentials and is more pronounced in
ortho quinones than para quinones. In FAB-MS, this ion is also observed, but it is
dependent on the choice of sample matrix (53). Other prominent ions in E1 include the
losses of CO and 2C0.

In ECNI spectra (see Table 4.16), the most abundant peak is usually M". Very
little fragmentation is noted and no losses of C0 are present in any quinone. Prominent
ions at [M+H]’ and [M+2H]" are also present in these spectra. These ions probably
correspond to the dihydroquinone formation noted in El spectra. The peak at [M+H]' is
most likely the result of proton abstraction from the dihydroquinone. Typically, if
[M+2]+ is enhanced in El spectra, it will also be enhanced for ECNI spectra. For
example, anthraquinone has a very small [M+2]+ peak in El, whearas
phenanthraquinone has a [M+2]+ ion peak the same size as the molecular ion. In ECNI-
MS, the spectrum of anthraquinone has a [M-Hl' ion at a relative abundance of 20%.
The phenanthraquinone molecule, which exhibited a larger hydroquinone ion in its EI
spectrum, also was found to have a larger [M+H]’ ion (relative abundance of 33%) in
ECNI.

Adduct ions are much more common in quinone spectra than in the spectra of
ketosteroids. Oxygen adducts are present in half of the compounds tested. Most of
these oxygen adducts also have a [M+O+l4]' ion, probably from additional adduct
formation with CH2 from the reagent gas. 1,4-Benzoquinone has only the [M+14]’
adduct peaks without any contribution from [M+O]'. Interestingly, this adduct is the
base peak.

171

Table 4.16: ECNI Spectra of Quinones

Compound
1 ,4-Benzoquinone

Tetrachloro- l ,4—
benzoquinone

1,2-Naphthoquinone

[Anion], (Relative Intensity)
[Ml-141'. (100); [M+H]'. (38); M'. (67)
[M-Cll'. (100); [M-ZCll'. (29)

M'. (100); [M+H]'. (99); [M+0]'. (60); [M+0+14]'. (24)

1,4-Naphthoquinone M', (100); [M+2H]', (42); [M+O]', (22); [M+O+14]', (10)
2,3-Dichloro-1,4- M', (100); [M-Cl]', (17)

naphthoquinone

2-Methyl-1,4- M’, (100); [M+2H]‘, (38); [M+O]', (22); [M+O+14]', ( 12)
naphthoquinone

9,10-Anthraquinone M', (100); [M+H]’, (20)

1-Chloro-9,10- M', (100); [M-Cl]', (50)

anthraquinone

B-Mcmyl-9.10 M'. (100); W+H]’. (20); [M+O]'. (3)

anthraquinone

9,10-Phenanthraquinone M', (100); [M+H] ', (33); [M+O]', (76); [M+O+14]’, (37)
Benzophenone M', (100); [M+H]‘, (32); [M-2H]', (25); [M+O]', (5)

Relative Responses of Quinones

The conjugated structure of quinones account for their enhanced ECNI
responses. Several quinones studied have chlorine substituents which also contribute to
their ECNI responses. Full scan ECNI, SIM ECNI, GC-ECD, EA, E1 ,2, and UV data
are compared in Table 4.17.

EQNLSIM. Quinones were analyzed by SIM in the same manner as the steroids;
that is, the major ions were monitored. SIM is a poor method for the comparison of the
responses of quinones because the spectra are so complicated that the monitoring of a

few ions is not representative of the response of the whole molecule. After several

172

 

8985 23 as... oeoeofioneom

on 2.7.85. o3 Sod e3 283355525

0:83.555

an :5 Ned o3 -Sdiaaza

2553855

5 as :5 Ed 2.3.8253

am new 3.7.3.? 92 :5 2 .o 25 823352.25

05523553:

SN ...: and Rd 3 -..._-_.£.o2-~

veep—385393:

as 2 mac 3. {Essences

3 2.33 :2 3 3 . 3 85:85:32-}

an 8.7.8.? e3 Rod Ed oeoeeeoefiszfi

25556352

am 5185+ was :5 sad 2 3153839

g 3.33 as was 1: a 3 Rod oeos=§=om4._

a e an E SC 8:88: 3 25 ...cm .3... 252.55
2.5 .5 E am 93 am new 8:883— .28

85:30 ..e com—.8 e5 22.—zoo do 8288: 3:53. “2... 2%...

173

$2 .35» 362 so; 32.. .m .3 a Si 5.38:8 38.3 o redefine a: s .< .3on
3m .uowuom 68m .3552: 8 6559 856 5 3:350 803 ~3on 2F .8ng 885208 ...: T e 05 mm seam—388
.8 com: 533%“ 2:. .38 2a 205 5255.6 come 05 ..8 8:028? coma:— ofifi 025 muse—.3635; .8.“ «88% >3 A8

.39. .33 dew EEC .s .m .m .350.“ 88“. 33:22— :33.5. 263.2“: A3

.Avmv 3:282 Sec 8% .0255 553m 3

.55—8 Boa—$08 Tma a .23 saw—wefiego mam mm «Emma Hogméetom a no 38:3 8.6 now By
.832— 025 he 95 “moms. .«o mercies. com 382% 3

28... e. .8.

3.288 :6 033—.

174
attempts at SIM, the method was abandoned, and full scan ECNI was used for the
comparison study.

Wes. Full scan ECNI studies resulted in much more accurate
quantitation and much better correlation with electron affinities. Several uends are
noted for these quinones. The unsubstituted parent quinones have relative responses in
the order l,4-benzoquinone < 1.2-naphthoquinone ~ 9,10—phenanthraquinone ~ 9,10-
anthraquinone < l,4-naphthoquinone. Usually, ortho quinones have higher redox
potentials (and presumably higher ECNI responses) than para quinones (55), but this is
opposite to the naphthoquinone data. The electron affinities of these parent molecules
are anthraquinone < 1,4-naphthoquinone < 1,4.benzoquinone. The molecular orbital
energies of the 1t"' LUMOs increase for quinones in the order benzoquinone <
naphthoquinone < anthraquinone. The negative charge is presumed to be concentrated
on the oxygen, rather than spread over the whole molecule; therefore, EAs for these
three compounds are not very different (54). The ECN I responses do not follow this
trend of electron affinities.

The number of chlorine atoms in a quinone correlates with the magnitude of
ECNI response. The addition of one chlorine in the l-position of anthraquinone does
not affect the ECNI relative response at all. Two chlorines added to the 2- and 3-
positions of 1.4-naphthoquinone increase the response by a factor of five. Tetrachloro-
l,4—benzoquinone has a relative reSponse almost 100 times greater than the parent
quinone. The addition of more chlorines will increase the response, but the nature of
the parent quinone determines which compound has a higher response. For example,
the two-ring 2,3-dichloro-l,4-naphthoquinone has a higher response than the one-ring
tetrachloro-lA—benzoquinone. This does not agree with the trend expected from
electron affinity data.

ECD. The trend in ECD responses mimics the trend in ECNI full scan

responses with the exception of tetrachloro-l,4-benzoquinone and benzophenone.

175
Even the magnitude of the responses is similar. For quinones, ECD is a good indicator
of ECNI relative response.

W- Electron affinity is a fair indicator of ECD and full scan
ECNI response, with the exceptions of 1.4-benzoquinone and tetrachloro-l,4-
benzoquinone. The discrepancies in response might be explained by the
chromatographic properties of these two compounds. Both have much poorer peak
shape on a DB-l column than condensed quinones like naphthoquinone and
anthraquinone. Tetrachloro—l,4-benzoquinone has especially bad tailing. The mass
spectra for this compound are hard to replicate even under electron impact conditions.
The stability of these two compounds in dilute solution may be suspect.

For compounds where both are available, the electron affinity and the reduction
potential data show the same trend. Even though the reduction potential data was
obtained in an aprotic solvent, the Chen and Wentworth (56) equation relating electron
affinity to reduction potential does not work for these quinones.

1.1!. Benzoquinones generally have three absorption bands, attributed to 1: -)
1t*, 1t -) n", and n -) It“ transitions. The spectra of condensed quinones are more
complex because both quinoid and benzenoid absorptions may be present. Four
absorptions are possible, corresponding to benzenoid 1t —) 1t*, quinoid 1t -9 1t*, a
second quinoid 1t -) 1t*, and a n -) If“. Not all of these are found in every quinone.
The first x -) rt“ band was used for comparison in Table 4.17.

The UV data do not correlate with the EA data very well. Perhaps the electron
occupies different orbitals in these two cases. The ECD and ECNI data do not
correlate well to UV data either.

Relative Responses of Arene Epoxides
Although steroid epoxides have excellent responses, not all epoxides have such

useful ECNI responses. The arene epoxides (see Table 4.18) are examples of low-

176
response epoxides. The lowest response is found for compound A, which has a
response approximately one-half that of the steroid 5a-androstan-3-one.

A few conclusions can be drawn from this study. Methyl groups do not
contribute to the relative response, as shown in the comparison of compounds A and B.
The tetraphenyl-substituted compound C has a slightly greater response. Substitution
of a carbonyl group for a phenyl group, as in compound D, slightly enhances the
response.

ECNI Responses of M ycotoxins and Trichothecines

Mycotoxins are fungal metabolites found as contaminants of various grains,
feeds, milk, and peanut butter (57). These multi-ringed compounds often have
extended conjugation and epoxide groups. All mycotoxins were acetylated to facilitate
GC analysis. The responses are shown in Table 4.19.

The trichothecines DON and DOM-1 differ by a double bond or an epoxy
group in the same position, yet they have similar responses. In steroids, the epoxide-
containing molecule exhibited higher responses than comparable molecules containing
a double bond In DON, however, the epoxy group is not located near a carbonyl
group, so the response is not enhanced. Structural features other than the epoxide must
be responsible for enhanced response. For example, zearalenone, which does not
contain an epoxide, but has extended conjugation in the form of an aromatic ring, has a
better relative response.

In conclusion, these mycotoxins have enhanced ECNI responses due to
extended conjugation, not because of contributions from epoxides. Even with the
extended conjugation, the response is an order of magnitude less than that of a highly

conjugated steroid like l,4-androstadien-3,11,l7-trione.

177
Table 4.18: ECNI Relative Responses of Arene Epoxides

Cemmund Manama mac“
0
P5 1 3<CH3 [M-CHOJ' [M—CH 1' 1.0 i 0.04
H Ph A [M—HJ' 3
Ph 0 H
H>L3<Ph B [M-H]‘,[M-Ph]‘ 1.3 i 0.1
0
Ph Ph [M-Phl'M' 5.0 :t 0.3
Ph>LA<Ph C

[101]',M', 8.1 :l: 1.0

o o
H C
3 WK CH3 D [M+O-CHO]‘
H30
*Compound A has been assigned a relative response of 1.0; n = 3-5.

Table 4.19: ECNI Relative Responses of Mycotoxins and Trichothecines

Command MaiQLIQns KW
'30”! 9“

*.

    

nae

 

 

 

.3 cu, [193]',[M-H]‘ 9.61:0.8

e.,...
Deoxynivalenol acetate

-ON

 

[M-OAc-H]",[M-20Ac-2H]',

 

 

 
  

°~ .. [M-3OAc-3H]',[M-H]' 12 i 0.6
DOM-1 acetate .
uac ’§ ° ’; 3°"
W, [119]',[182]',[226]',
(ca3)2cucn2 ‘oooca, [M-OAC-H]"M' 16 i 1

 

   

; “a
. éuzooccna
T2-Tox1n acetate

w [M-Ac-H]‘,[M-Ac]’,[M-H]‘ 37 .4: 6
I!
no / °

11
Zearalenone acetate

*A-3-one has been assigned a relative response of 1.0; n = 3-4.

178
Other ECNI Studies
[M-HI' Formation: An Indication of Low Response?

As shown in Table 4.7, the ECNI mass spectra of steroids are very simple,
usually consisting of either the molecular anion or simple losses of groups such as H,
CH3, or HP. For ketosteroids, the presence of a [M-I-I]' ion usually indicates low
relative response (value less than 10). From studies with deuterium-labelled steroids,
it was determined that the hydrogen is lost from the tit-carbon to the carbonyl. _

Proton abstraction is thought to be due to an ion-molecule reaction with OH”,
02, or 0' present in the source. In one case, the ECNI spectrum of the methyl ester of
abscisic acid (ABA) shows a [M-I-I]’ ion and an ion at mlz 141 only if oxygen is
present in the source (22). On the JEOL 505 mass spectrometer, the ECN I mass
spectrum of ABA always had a contribution from [M-Hl‘ as well as the peak at mlz
141. This suggests that oxygen is always present in the JEOL 505 source. At times it
was possible to achieve conditions on the TSQ 70 when these ions were not present in
the spectrum of ABA. When the [M-H]' ion was not present in the ABA spectrum, the
spectra of both l-androsten-3,l7-dione and 2-androsten-17-dione still exhibited only
the [M-H]' ion. Two possible conclusions can be drawn. Perhaps proton abstraction
by oxygen in the source is not the mechanism responsible for the [M-H]" ion formation
in steroids. Alternatively a trace of oxygen in the source may be sufficient for proton
abstraction from low-response steroids but not for abstraction from ABA.

To determine the position of proton abstraction, several steroids were labelled
with deuterium alpha to the carbonyl group. About 1-3 mg of steroid was dissolved in
1 mL CH3OD (KOR Isotopes, Cambridge, MA). A small chunk of sodium was added,
and the reaction proceeded for 1, 6, or 24 h. One hour seemed sufficient for the
labelling procedure. The reaction was stOpped with the addition of 1 mL of methylene
chloride. One ml. of water was added and vortexed. The methylene chloride layer

179
was extracted and dried with NaZSO4. A small amount of this layer was dried under
nitrogen and reconstituted in acetonitrile for ECNI-MS analysis.

ECNI-MS analysis of deuterated 50t-androstan-l7-one proved that the major
peak had shifted to [M-2]'. Therefore, the hydrogen lost is alpha to a carbonyl in a
steroid without double bonds. If more than one carbonyl is present in a molecule, it is
not known from which group the hydrogen is abstracted. Proton abstraction alpha to a
carbonyl was also demonstrated for low-response bicyclic compounds through the
analysis of the deuterated bicyclic compound 2.

A possible mechanism for the formation of more stable anions by the loss of a
hydrogen is shown in Figure 4.11. After loss of a hydrogen, resonance structures can
be drawn, showing stabilization of the ion. Perhaps this type of stabilization is
necessary to elicit any type of response from these compounds.

The general rule that [M-H]- formation indicates a low-response ketosteroid is
not true for the epoxysteroids listed in Table 4.8. As shown in Tables 4.20 and 4.21,
there is no correlation between ECNI response and molecular anion formation. For
example, pregnan-4a,50t—epoxy-3,20-dione and pregnan-l6a,l7a—epoxy-3,20—dione
have similar responses, but the spectrum of the former consists of only M", while the
latter produces [M-l-I]'. For epoxysteroids, the position of the epoxide within the
steroid nucleus is the only consideration for [M-I-I]' formation, not the electron-capture
cross section. Following electron capture, the resulting radical anion may react by
opening the epoxide ring with relief of ring strain, as shown in Figure 4.12. At this
point, two rearrangements are possible. If a hydrogen is available on the B-carbon, it
will be abstracted preferentially, resulting in an [M-H]' anion. However, if a hydrogen
is not available on the B—carbon, as in the 4.5-epoxides, rearrangment involving
cleavage of the carbon-carbon bond between the B-carbon and R2 may occur (Figure
4.12). Because R2 is part of a ring structure, it will remain attached, resulting in

retention of the M" anion.

180

$28885: 3588
E .5958 3:358 a 9 SEE sumac»... a we 32 05 .3 :25 038m 958 a .«e cougar—8 05 ..8 Emmaanoofi 388$ ":6 253m

181

Table 4.20: Steroid Epoxides with Molecular Anions

mud KW
A-4a,5a-epoxy-3,l7-dione 39
A-4B,SB-epoxy-3-one-l7-ol acetate 26
A1-4B,SB-epoxy-3-one-17-ol acetate 200
P-4a,5a-epoxy-3,20-dione 21
C-4B,5B-epoxy-3—one 9.2
*A-3-one has been assigned a relative response of 1.0.

Table 4.21: Steroid Epoxides with [M-Hl' Ions

mm W’
A-la,2a-epoxy-3-one 25
A—90t,11a-epoxy-3-one 5
A4-16a,17a-epoxy-3,l7.dione 1.8
aP—16a,170t-epoxy-3,20—dione 47
P4-16a,17a-epoxy-3,20-dione 57
aP—160t,17a-epoxy-3,l 1,20-trione 63
C-Sa,6a-epoxy-4-one 61
C4»6-1a,2a-epoxy-3-one 58

’A-3-one has been assigned a relative response of 1.0.

182

R‘=Horaringunit
R2=Horaringunit

R3 = a ring unit

f/°\

:0—-—c—(:—c—R3

R1 R2

O'

9.. |
:9-c=c—c—R3

 

 

R1 R2
If112:}, IfR2=ring unit
-H°
o
e» u 9,, ll
‘9—C=C C R3 :9—c=c—c—R3<°‘2>
l J. )

ORZ (or 3)

Figure 4.12: Proposed mechanisms for the capture of an electron in an a-ketoepoxide.
After the opening of the epoxide rin to relieve strain, two rearrangements are possible.
If a hydrogen is available on the g-carbon to the epoxide, a [M-H]‘ anion will be
formed. Otherwise, the M' anion will be formed after the breakage of the ring.

183

Size Comparison Study

Three categories of compounds were analyzed for the size comparison study:
steroids, bicyclics, and cyclohexanones. A comparison of the three types of compounds
is presented in Table 4.22. Each of these compounds had only one functional group or,
in the case of bicyclics, two isolated functional groups. The relative responses are
indicative of the size of the molecule. Steroids have a greater relative response because
they can distribute the excess energy imparted by the capture of an electron more

effectively.

Table 4.22: Size Comparison Study

Compound ECNI Relative Response"
Cyclohexanone 1.0
Compound 8 2.0
Sa-Androstan-B-one 10

*Cyclohexanone has been assigned a relative response of 1.0.

Comparison of Reduction Potentials to ECNI Responses
Ketosteroids

Chen and Wentworth (56) correlated half-wave reduction potentials (E1 [2) in
aprotic solvents to experimental gas-phase electron affinities (EA) using the expression
EA = E1 ,2 + 2.49 :i: 0.2 eV. Therefore, the trend in E1 ,2 data should mimic the trend in
electron affinity data.

It is expected that compounds with higher elecuon affinities (more positive
reduction potentials) should have higher ECNI responses. This is generally true for
ketosteroids (see Table 2.10). Compounds with E1 ,2 of less than -2.0 V have low
responses; the higher-response compounds all show reduction potentials of more than -

2.0. No exact correlation between ECNI response and reduction potential is possible.

184

This is not an unusual occurrence; other researchers have noted that EA values and
ECNI responses do not correlate well for a series of polyaromatic hydrocarbons (58).
Bicyclic Compounds

The electrochemical reduction potentials are tabulated in Table 2.11. A half-
wave potential of -2.35 V is the value where the solvent reduces. Many bicyclics tested
had reduction potentials lower than this value.

Most of the reduction potentials and relative ECN I responses track each other
(see graphical representation in Figure 4.13). For example, compounds 11 and 15 have
the same E1 ,2 and the same ECNI response. Compound 25 had a lower reduction
potential and a lower response than these two.

Many dissimilarities exist between reduction potential and ECNI response.
Compounds 20 and 26 have the same reduction potential, but thier ECNI responses
differ by one order of magnitude. As was observed for ketosteroids, the reduction

potential of bicyclics is not a very good indication of ECNI response.

 

400

(a
o
O
4 r‘

200 ‘

Relative Response

.5

O

O
l

 

 

 

 

 

-3.00 -2.00 -1.00
Reduction Potential (V)

Figure 4.13: Half-wave reduction potentials of ketosteroids as a function of ECNI
relative response. All responses are relative to androstan-3-one.

185
Comparison of UV Spectra to ECNI Response

Field (59) has suggested that electron affinities (and therefore ECNI responses)
are affected by the same structural features that affect the wavelength of light
absorption. In light absorption, an electron is promoted fi'om an occupied to an
unoccupied orbital, as in n -) 1t* and rt" -) 1t* transitions. In electron capture, the
electron enters the LUMO, which should be the same orbital that an electron would be
promoted to in absorption. Of course, the absolute energies in these two cases are not
the same, but the same trends in structural features should be noted. The UV spectra of
many steroids in methanol and ethanol have been recorded (see Table 4.23). The
maximum wavelength of spectra recorded in either ethanol or methanol can be directly
compared (60).

With the exception of the epoxides, all compounds that have a low response
(relative response factor less than 3) all have maximum wavelengths of absorption
below 243 nm. Epoxides are exceptions to the rules in [M-H]' formation and
electrochemical trends.

As was observed for electrochemical reduction potentials, there are no trends
between maximum wavelength and ECNI response. In some cases, there is a
correlation, but it does not hold true for several n0table exceptions. The steroids with
the largest km”, 4,6-androstadien-3,17-dione, has an ECNI relative response of only 18.
The steroid with the largest relative response, 1,4-androstadien-3,11,17-trione, has a
km” of only 238 nm.

The trends in Am” do not correlate to the trends in electrochemical reduction
potential (an indication of electron affinity values). In conclusion, UV spectra can be

used only as a very rough guess of the ECNI relative response.

186
Table 4.23: UV Spectra of Steroids

Name e Ref."I ECNI Response“
mar (L-cm'lmole‘l)
Androstanes
5a-A-3-one < 220 a 1.0
50t-A-3,l7-one < 220 a 1.4
5a-A-2a-Br-3,17-dione 244 11800 b 0.35
5a-A-20t-I-3,17-dione 242 14800 b 0.35
Sa-A-Za,4a-Br-3,17-one < 220 a 0.80
5a-A-3,11,17-one < 220 a 0.89
A3-5-17-one 241 18000 b 0.67
A4J5-3-one 240 17170 b 0.94
A1-3,17-one 230 10700 a 0.98
A4-3,17-dione 239 16600 a 1.2
A4-3,1 l,l7-trione 238 15400 a 3.2
A1v4-3,17-dione 243 15400 a 2.1
A4’5-3,l7-dione 283 26100 a 18
A1’4-3,11,17-trione 238 14500 c 360
A4-3,6,17-trione 253 10800 b 300
Pregnanes
P4-3,20-dione 241 ~ 16400 a 1.4
P4-3,6,20-trione 252 10900 a 280
P4-16B,17-ex-3,20-dione 240 17100 a 57
P-16a,l7-ex-3,20-dione < 220 a 24
Cholestanes
04.6 235 24000 b 0.47
5a-C-3-one < 220 a 0.27
5B-C-4B,5-ex-3-one < 220 a 9.2
04-3-6116 241 16700 a 0.91
C5-4,4-Me-3-one < 220 a 1.7
C4-3,6-dione 250 7870 c 280

*References are the following:
(a) Data from reference 61; methanol as solvent.
(b) Data fiom reference 62; ethanol as solvent.
(c) From this work; methanol as solvent; see Chapter 2 for experimental details.

“Androstan-3-one has been assigned a relative response of 1.0

187
Reproducibility

Even with careful control of the instrumental parameters that directly affect the
ECNI ionization process, the reproducibility of spectra is a major problem. The
variability in ECNI spectra has been minimized by Stemmler and Hites (63) on an
HPS985 quadrupole GC-MS by careful tuning of the drawout and focus lenses. In the
Stemmler and Hites study, after tuning with perfluorotributylamine to set the mass axis
and peak widths, the spectrum of DFI'PP was measured at a source temperature of
250°C and a pressure of 0.55 torr (measured by a capacitance manometer) to meet the
following criteria: mlz 442 (>60%), m/z 365 (>10%), m/z 275 (<10%), and m/z 167
(<20%). The drawout and ion focus lenses were adjusted to meet these criteria. Over a
period of one year, the relative standard deviations for the ion abundances were less
then 10%, with the exception of the weak ion at mlz 275.

The effect of tuning parameters on the reproducibility of ECN I spectra obtained
on the JEOL 505 were studied. The source temperature was held at 200 °C and the
source pressure (measured at the source housing) was 2 x 10'5 torr. About 0.4 ng of
DFI'PP (Ultra Scientific, Hope, R1) was injected, and ions at m/z 442, 365, 275, and 167
were monitored using magnetic field switching SIM. The areas under the peaks were
used to calculate relative intensities. Replicate injections at the same set of conditions
had a cv of less than 1.8%.

"Perfect" conditions were determined by the optimum tuning of
perfluorokerosene (PFK), the usual tuning compound. The voltages for various lenses
were varied one at a time. The spectrum of PFK does not change markedly with the
change in lens potential, but the spectrum of a test compound like DFI'PP does. Two
settings different from the "perfect" setting were used for each parameter. Voltage
measurements corresponding to these settings are not available. In all cases, the base

peak was at mlz 442 (M"); all data were reported relative to this base peak.

188

The intensity of the mlz 275 ion (corresponding to [M-C6F5]') was usually small
and undependable. Variations from "perfect" conditions uielded ion intensities between
20 and 100%. This ion also caused troubles in the Stemmler and I-Iites study; in fact, it
was not part of the claim for good reproducibility over the span of a year. The 27 5 ion
will not be studied further.

The relative abundance of the ions at mlz 365 ([M-C6H5]‘) and mlz 167 ([C6F5]'
) varied depending on the parameter. Deviations fiom "perfect" were calculated; any
relative standard deviation greater than 10% was considered to be significant.

Focus. On the JEOL 505 source there are two focus lenses, right and left. Each
lens was varied separately. For all positions, there was a significant (20-30%) variation
from "perfect" for mlz 365 and a large (60-90%) variation for mlz 167. As in the case of
the Stemmler study, the adjustment of the focus lens is significant for the determination
of ion abundances.

W. The electron energy is typically set at 200 eV; for this
experiment, it was varied between 50 and 100 eV. Minor differences were noted at 50
eV for mlz 167. The Stemmler study also noted small differences with the variation of
electron energy.

W. The literature study revealed that the emission current had an
effect on certain compounds. On the JEOL 505, the ionization current has two settings.
The emission current is a function of the ionization current setting and the position of
the filament (a misaligned filament will cause the emission current to be. higher than
normal). The lower setting of ionization current resulted in a great loss of sensitivity.
For these two reasons, variation of ion abundance due to emission current was not
studied.

Beadle; The change in repeller potential had a minor effect on ion abundances.

This was also the case for the literature study.

189

m. Variations in the deflector potential had a minor effect on the mlz 167
ion abundance. Other ions were not affected.

W. The voltage on the accelerating lens can be adjusted. The
differences were important only for the mlz 167 ion. At very high adjustments of the
accelerating voltage, the ions may shift almost an entire mass unit; therefore, any
measurements at these voltages would obviously result in skewed abundances.

02131291:- The octapole determines the peak shape more than any other lens
potential. Changes to the octapole lowered ion abundances in almost every case. This
could be due to the distortion of the peak shape so that the correct mass value was
obliterated.

DFTPP was monitored at a source pressure of 2 x 10'5 and a temperature of 200
°C for a three-month period. The peak at mlz 442 was always the most abundant. The
average relative abundance of the ion at mlz 365 was 51% with a cv of 15%. The peak
at mlz 167 usually did not meet the tuning criteria. Adjusting the focus to meet tuning
criteria often decreased the signal to levels that were not sensitive enough for ECNI
analyses. The relative abundance of m/z 167 under conditions favorable to ECNI
usually was between 20 and 30%. This high value could also be attributed to the
reagent gas pressure used; 0.55 torr was used in the literature study, but there is no
method of measuring the CI volume pressure directly on the JEOL 505. The average
abundance for the ion at mlz 167 was 25%with an rsd of 40%.

Reagent Gas Study

The choice of reagent gas is very important for enhanced responses under ECNI

conditions. The actual responses are dependent on the type of compound (64»66). In

general, heavier mass gases have the best sensitivities.

 

190
A test mixture consisting of the following steroids was used in this study:

androstan-3-one at 189 ng/ttL

l-androsten-3,17-dione at 214 ng/pL

1,4-androstadien-3,11,17-trione at 2.48 ng/pL

4-androsten-3,6,17—trione at 7.90 ng/uL

oxidized dexamethasone at 0.62 ng/uL (determined by ECN).
Each sample was analyzed in replicate at two different gas pressures, 2 x 10'5 and 1 x
10'5 torr, measured at the gauge on the source housing. The only gases that showed
significant differences between the two pressure levels were methane and carbon
dioxide, both of which were enhanced at higher source pressures. Helium was used
only at a source pressure of 1 x 10'5 torr due to problems with regulation. Surges in
pressure are normal for helium; therefore, helium is more likely to overpressurize the
instrument when utilizing higher soru'ce pressures. All gases, with the exception of
helium, were used fiom the cylinder without purification. Helium was passed through a
heated oxygen scrubber before use.

Different steroids respond differently to various reagent gases. In general, 1-
androsten-3,17-dione and androstane-3-one, both of which have only one peak in their
spectra, corresponding to [M-H]', behave in a similar fashion to various reagent gases.
Both 1,4-androstadien-3,11,l7-trione and 4-androsten-3,6,l7-trione, both of which have
M" as the only peak in the ECNI spectrum, also track each other. Oxidized
dexamethasone, which dissociates plus produces a small molecular anion, behaves as a
steroid that produces M" exclusively. Because the behavior changes with reagent gas,
it is difficult to choose an internal standard. An ideal internal standard must not alter its
response under a variety of conditions. This is not possible when dealing with different
reagent gases. In fact, if one wants to prove the superiority of a reagent gas over
another for a particular compound, one only needs to choose the correct internal

standard. For example, the relative responses for 4-androsten-3,6,17-trione using

191
different reagent gases differs widely depending on the internal standard chosen.
Remember, the same data is used for all comparisons presented. If androstan-B-one is
chosen as the internal standard, the relative response of 4-androsten-3,6,17-trione is: air
< ammonia < helium < methane < carbon dioxide < isobutane < nitrogen. If the internal
standard is changed to 1,4-androstadien-3,1 1,17-trione, the relative responses shift to:
nitrogen < helium < ammonia < methane < carbon dioxide ~ isobutane < air. In this
case, all gases except nitrogen had approximately the same response, with air generating
the lowest response. In the previous case, air produced the best response. Caution must
be exercised in interpreting any data in the literature that compares relative responses of
reagent gases using an internal standard (for example, data in reference 64). Other
literatme studies have compared the responses under different reagent gases without the
use of an internal standard. There is no method of compensating for variations in
response from such that factors change between analyses, such as cleanliness of the

source.

Table 4.24: Reagent Gas Study‘

Compound Air N113 / C02 He Isobutane N2 Methane
A-3-one 0.18 2.8 0.25 0.12 1.2 0.20 1.1
A1-3,17-one 0.18 2.0 0.31 0.078 0.86 0.26 1.0
A1'4-3,11,17-one 0.69 68 21 0.80 230 170 170
A4-3,6,17-one 0.40 20 9.6 0.24 ’ 90 36 70
Ox. dexamethasone 4.7 420 140 9.1 1400 770 1000

*A1-3,l7-one analyzed using methane ECNI was assigned a relative response of 1.0.

Without the use of internal standard, the relative responses for seven reagent
gases are shown in Table 4.24. The trend for low-response compounds is helium.< air ~
nitrogen ~ carbon dioxide < methane ~ isobutane < ammonia. For high-response
steroids, the trend is helium ~ air < carbon dioxide < ammonia < nitrogen < methane <

isobutane.

 

 

192

Several differences between the high- and low-response steroids are noted.
Ammonia is a much better reagent gas for low-response steroids than is methane. Even
though ammonia does have a better Kf (electron thermalization rate, see Table 1.2), the
increase in response for this certain set of compounds must be due to the ease of proton
abstraction. Nitrogen is better for high-response steroids than for low. Isobutane is
slightly better than methane as a reagent gas, probably because it has a higher mass.
Isobutane. may be worth pursuing in assay development. Helium is a poor choice for
both low- and high-response compounds. For helium, both Kf and collisional
stabilization are poor, resulting in loss of electron capture ability.

Carbon dioxide has been mentioned in the literature as comparable or superior to
methane as a reagent gas (65). This was not found to be true for the analysis of steroids,
even though relatively pure (99.99%) carbon dioxide was used.

Air, with its oxygen content, should enhance the formation of [M-H]' ions in
compounds which form this ion primarily. This was not observed for either androstan-
3-one or l-androsten-3,17-dione. The presence of Oz in the source must not be the
cause of the [M-HT formation, even though the possibility of abstraction by OH' was
not ruled out by this experiment.

Fragmentation patterns also varied using different reagent gases. In this
experiment, oxidized dexamethasone was the only compound studied that produces
observable fi'agments. As shown in Table 4.25, fragmentation patterns are dependent
on the nature of the reagent gas (<l% means that the peak was not detected). Under all
reagent gas conditions, the peak at mlz 310 ([M-HF] ') was the most abundant. The peak
at mlz 295 ([M-HF-CH3]') varies greatly with reagent gas. This peak is enhanced when
using air or isobutane. The molecular anion at mlz 330 is typically small at the source
temperature used (ca. 200°C); however, some enhancement was noted when using

ammonia as the reagent gas.

193

Table 4.25: ECNI Spectra of l,4-Androstadien-90t-fluoro-16a-methyl-3,ll,l7-
trione (Oxidized Dexamethasone) Using Different Reagent Gases

mlz Air NH3 C02 He Isobutane N2 Methane
295 34 l 1 1 1 7.2 30 16 16

310 100 100 100 ' 100 100 100 100
330 <1 * 2.4 0.81 <1 0.64 0.51 0.47

* <1 means that the peak was not detected.

Comparison of Ionization Techniques

ECNI is a selective technique for electrophilic compounds. The response under
ECNI conditions was compared to the other commonly used GC-MS techniques of E1
and methane CI as well as to GC-ECD for both electrophilic and non-electrophilic
compounds.

Samples were made in concentrations amenable to the technique in question.
The internal standard was l-androsten-3,l7-dione, added to each sample at a
concentration of 165 ng/uL regardless of the concentrations of the other compounds.
All compounds were analyzed by full scan in each ionization technique to choose the
most abundant peaks for further SIM analysis. SIM was chosen over full scan for the
comparison analysis because any quantitation of these compounds would normally be
done using this technique.

Under full scan EI conditions (see Table 4.26), all compounds tested had almost
identical molar responses. All steroids should be ionized and detected by E1 in a similar
fashion. Striking differences were noted when the SIM of M+ ions of each compound
was performed. The differences in response may be due to variations in mass spectra.
The EI spectra of 1,4-androstadiene-3,11,l7-trione and 3,5-cholestadiene exhibit M+ as
the base peak. The spectra of 1-androsten-3,l7-dione and 4-pregnen-3,6,l7-trione
produce M"' in 90% relative abundance. The spectrum of androstane-B-one has a 60%

relative abundance of M+. The differences in abundances of the molecular ion do not

194
account for the differences in the El full scan and El SIM responses. These differences

could result from the relative amount of fragmentation in the entire mass spectrum.

Table 4.26: Comparison of Various Ionization Techniques

Ionization Method
Compound EI Full Scan El SIM CI SIM ECNI SIM GC-ECD
A-3-one 1.3 d: 0.2 1.6 i 0.3 1.1 i: 0.2 2.9 :t 1.3 0.3

A1-3,17-one* 1.0 3: 0.09 1.0 :t 0.09 1.0 :t 0.1 1.0 :l: 0.4 1.0

A1v4-3,1 1.176116 0.88 3: 0.12 0.73 :l: 0.08 0.56 3: 0.18 370 3; 70 240
P4-3,6,20—one 0.75 i 0.07 0.36 :l: 0.04 0.39 :t 0.06 320 :l: 50 130
C35 1.1 :l: 0.05 1.6 :t: 0.08 0.68 3: 0.07 0.43 :l: 0.05 0.22

*This compound was the internal standard.

The relative response of the MH" ion monitored under CI conditions had a
similar response to the M+iion monitored in El, with the exception of 3,5-cholestadiene.
CI is based on the collision of the analyte with a gas molecule and then transfer of a
proton. There is no dependence of this mechanism on structure; therefore, no
differences were noted in the CI response. The exception is 3,5-cholestadiene whose CI
response is lower than its El response. Under CI conditions, [MH-ZH]+ ions are also
prominent in the spectrum of 3,5—cholestadiene. This fragmentation may account for its
lowered CI response.

Unlike CI responses, ECNI responses are dependent on the structure of the
compound. ECNI responses for 1,4.androstadien-3,11,l7-t1ione and 4-pregnen-3,6,20-
trione were higher than their corresponding E1 or C1 responses by three orders of
magnitude. Androstan-B-one, a low-response compound with [M-Hl' as its prominent
ECNI peak, did not show any enhancement of response. 3,5-Cholestadiene has two
ECNI peaks, one at [M-H]‘, the other at [M+Oz]‘. Even so, the electron capture

processes are so inefficient that the ECNI response was lower than that of E1 or CI.

195
GC-ECD responses are enhanced similarly to those found for ECNI, but the
actual numbers differ. The poorest responses for androstan-3-one and 3,5-cholestadiene
were obtained using GC—ECD.
Of the three mass spectral ionization techniques tested, ECNT is superior for
electrophilic compounds. For compounds that have poor ECN I responses, similar (or at

times better) relative responses could be obtained by either E1 or C1.

Conclusions

ECNI-MS is a selective technique for certain steroids. "Low-response" steroids,
that is, those with no functional groups, isolated functional groups, or only one site of
unsaturation, have responses on the order of those obtainable by E1 or C1. For enhanced
ECNI response within a steroid nucleus, the following structural features are important:
a cross-conjugated system with a neighboring group, such as in l,4—dien-3,11-dione; a
linearly-conjugated system, such as 4-en-3,6-dione; a fluorine atom alpha to a carbonyl
group in a highly-conjugated steroid; or an epoxide alpha to a carbonyl group. Usually,
the presence of a [M-H]'- peak indicates a low-response compound, except for
epoxysteroids. In epoxysteroids, if a hydrogen on a B—carbon is present, an [M-H]’ ion
will be preferentially formed.

The reduction potentials in acetonitrile can be assumed to be indicative of the
trends of electron affinity. In a series of ketosteroids, a half-wave reduction potential
more positive than -2.0 V indicates a high-response compound, except in the cases of
epoxysteroids. UV maximum wavelengths have been postulated as correlating to ECNI

responses; however, UV data can only be taken as a very rough estimate of ECNI
response.

196

Table 4.27: Comparison of ECNI Relative Response to Literature Values

ECNI Response ECNI Response

Compound Experimental" Literature"
A4-3,17-one 1.2 1
A4-3,6,17-one 300 . 350
A1-4-3,17-one 2.1 1
A4-3,11,17-one 3.2 6
A1v4-3,1 1,17-one 360 350
A1-4-3,11,17-one-9a—F-l6a-Me 4100 700
A1v4-3,l1,17-one-90t,60r-F—160t-Me 5400 700
Alf-3,1 1,17-one-9a-F—60t-Me 3900 700
A4-3,1 1,17-one-9a-F 2600 525
P4-3,20-one 1.4 1
P4-3,11,20-one 9.1 4

*A-3-one has been assigned a relative response of 1.0
”A4-3,17-dione has been assigned a relative response of 1.0; from reference 1.

Several of the compounds used for the ECNI response study were the chemical
oxidation products of steroidal drugs. The responses of the chemically oxidized
derivatives were compared to other electron-capturing derivatives. For compounds that
have extensive conjugation, chemical oxidation is the best derivative; for steroids that
Oxidize to low-response moieties, the HFB derivative is far superior.

Steroids, being large molecules, capture electrons more efficiently than one- or
two-ring compounds. In general, the relative ECNI response of steorids is ten times
greater than that of cyclohexanones and five times greater than that of bicyclic
compounds.

Results from this ECNI study were compared to literature values in Table 4.27.
The correlation is very good, except for steroids with fluorine substitutions. The
literattue study was done on a different mass spectrometer (quadrupole HP5985) under
different conditions. This comparison shows that these responses hold under different

instrumental conditions.

References

1.

399:5!“

9°

10.

ll.
12.

13.
14.

15.
16.

17.
18.
19.
20.
21.
22.

Her, G. R.; Watson, J. T. Biomed. Mass Spectrom. 1986, 13, 57.

Loveéocgk, J. E.; Simmonds, P. G.; Vandenheuvel, W. J. A. Nature (London) 1963,
I97, 4 .

Koshy, K. J. Chromatogr. 1976, 126, 641.

Dougherty, R. C. Anal. Chem. 1981, 53, 625A.

Stemmler, E. A.; Hites, R. A.; et al. Anal. Chem. 1988, 60, 781.

Greaves, J.; Bekesi, J. G.; Roboz, J. Biomed. Mass Spectrom. 1982, 9, 406.

Erhardt-Zabik, 8.; Watson, J. T.; Zabik, M. J. Biomed. Mass Spectrom. 1990, 19,
101.

Buser, H.-R. Anal. Chem. 1986, 58, 2913.
Begley, P.; Foulger, B.; Simmonds, P. J. Chromatogr. 1988, 445, 119.

Sears, L. J.; Campbell, J. A.; Grimsrud, E. P. Biomed. Mass Spectrom. 1987, 14,
401.

Stemmler, E. A.; Buchanan, M. V. Org. Mass Spectrom. 1989, 24, 94.

Low, G. K-C.; Batley, G. E.; Lidgard, R. O.; Duffield, A. M. Biomed. Mass
Spectrom. 1986, 13, 95.

Hunt, D. F.; Crow, F. W. Anal. Chem. 1978, 50, 1781.

Leis, H. J .; Gleipach, H.; Nitsche, V.; Malle, B. Biomed. Mass Spectrom. 1988,
I 7, 437.

Robinson, P. R.; Jones, M D.; Maddock, J. J. Chromatogr. 1988, 432, 153.

Umeno, Y.; Matsushima, E.; Maniwa, M.; Marunaka, T. J. Chromatogr. 1988,
434, 123.

Ishikawa, Y.; Suzuke, O.; Hattori, H. For. Sci. Intern. 1990, 44, 93.
Teramura, T.; Kobayashi, S.; Higuchi, S. J. Chromatogr. 1990, 528, 198.
Heyes, M. P.; Markey, S. P. Anal. Biochem. 1988, 174, 349.

Huang, L. Q.; Mattina, M. J. I. Biomed. Mass Spectrom. 1989, I8, 828.
Payne, N. A.; Zirrolli, J. A.; Gerber, J. G. Anal. Biochem. 1989, I78, 414.

Heath, T. G.; Gage, D. A.; Zeevaart, J. A. D.; Watson, J. T. Org. Mass Spectrom.
1990. 25 . 655.

197

23.

24.

25.
26.

27.
28.

29.

30.
31.
32.

33.
34.
35.
36.
37.
38.

39.

41.

42.
43.

198

Murray, 8.; Gooderham, N. J .; Boobis, A. R.; Davies, D. S. Carcinogenesis 1989,
10, 763.

Makino, U.; Tasaki, Y.; Ohta, S. Hirobe, M. Biomed. Mass Spectrom. 1990, I9,
415.

Koves, E. M.; Yen, B. J. Anal. Toxicol. 1989, I3, 69.

Dawson, M.; Smith, M. D.; McGee, C. M. Biomed. Mass Spectrom. 1990, I9,
453.

Gaskell, S. J .; Brooks, P. W. Intern. J. Mass Spectrom. Ion Phys. 1983, 48, 241.

Midgley, J. M.; Watson, D. G.; Healey, T.; Noble, M. Biomed. Mass Spectrom.
1988, 15 , 479.

Watson, D. G.; Midgley, J. M.; McGhee, C. N. J. Rapid Commun. Mass
Spectrom. 1989, 3, 8.

Murray, S.; Watson, D. J. steroid Biochem. 1986, 25 , 255.
Evershed, R. P. Goad, L. J. Biomed. Mass Spectrom. 1988, I6, 169.

Ragganich, K. A.; Heath, T. G.; Watson, J. T. J. Am. Soc. Mass Spectrom. 1990,
, 41.

Girault, J .; Istin, B.: Fourtillan, J. B. Biomed. Mass Spectrom. 1990, I9, 295.
Wilson, A.; Mason, P. A.; Fraser, R. J. steroid Biochem. 1976, 7, 611.
Pierce Catalog, Rockford, IL, 1989, p. 172.

Morgan, E. D.; Poole, C. F. J. Chromatogr. 1974, 89, 2.

Poole, C. F.; Zlatkis, A.; et al. Lipids, 1980, 15 , 7.

Courtesy of Beverly Chamberlain, Metabolic Profiling Laboratory, Michigan
State University.

Kayganich, K.; Watson, J. T.; Kilts, C.; Ritchie, J. Biomed. Mass Spectrom. 1990,
I9, 341. .

Wotiz, H. H.; et al. Biochem. Biophys. Acta 1961, 53, 449.

Sternberg, J. C.; Gallaway, W. 8.; Jones, D. T. L. In Gas Chromatography;
Brenner, N.; Callen, J. E.; Weiss, M. D., Eds; Academic Press: New York, 1962,
p. 231. -
Scanlon, J. T.; Willis, D. E. J. Chromatogr. Sci. 1985, 23, 333.

Tong, H. Y.; Karasek, F. W. Anal. Chem. 1984, 56, 2124.

Tong, H. Y.; Karasek, F. W. Anal. Chem. 1984, 56, 2129.

45.
47.
48.

49.
50.
51.

. 52.

53.
54.

55.
56.
57.

58.

59.

61 .
62.
63.

65.

199
Edwards, R. W. H. J. Chromatogr. 1978, I53, 1.
Jorgensen, A. D.; Picel, K. C.; Starnoudis, V. C. Anal. Chem. 1990, 62, 683.
Leonard, N. J.; Laitinen, H. A.; Mottus, E. H. J. Am. Chem. Soc. 1953, 75, 3300.

Kirk, D. N.; Hartshom, M. P. Steroid Reaction Mechanisms, Elsevier:
Amsterdam, 1968.

Kende, A. S. Organic Reactions 1960, II, 261.
Trainor, T. M.; Vouros P. Anal. Chem. 1987, 59, 601.

Conn, E. E.; Stumpf, P. F.; Bruening, G.; Doi, R. H. Outlines of Biochemistry;
Wiley: New York, 1987.

Zeller, K.-P. In The Chemistry of Quinoid Compounds Part 1; Ed., S. Patia;
Wiley: New York, 1974, p. 231.

Phillips, L. R.; Bartrness, J. B. Biomed. Mass Spectrom. 1989, 18, 878.

Heinis, T.; Chowdhury, S.; Scott, S. L.; Kebarle, P. J. Am. Chem. Soc. 1988, 110,
400. '

Morton, R. A. Biochemistry of Quinones; Academic Press, New York, 1965.
Chen, E. C. M.; Wentworth, W. E. J. Chromatogr. 1981, 217, 151.

Brumley, W. C.; Sphon, J. A. In Applications of Mass Spectrometry in Food
Science; J. Gilbert, Ed.; Elsevier. UK, 1987, 141.

Hsu, C S.; Dechert, G. J.; Rose, K. D. 38th ASMS Conference on Mass
Spectrometry and Allied Topics, Tucson, AZ, 1990, 0. 595.

Field, F. H. "Chemical Ionization: Negative Aspects," 28th ASMS Conference on
Mass Spectrometry and Allied Topics, May 1980, p. 2.

Silverstein, R. M.; Bassler, G. C.; Morrill, T. C. Spectrometric Indentification of
Organic Compounds; Wiley: New York, 1981.

Neudert, W.; Ropke, H. Atlas of Steroid Spectra; Springer-Verlag: Berlin, 1965.
Dorfman, L. Chem. Rev. 1953, 53, 47.
Stemmler, E. A.; Hites, R. A. Biomed. Mass Spectrom. 1988, 15, 659.

Huang, S. K.; Despot, K. A.; Sarkahain, A.; Bierl, T. W. Biomed. Mass Spectrom.
1990, 19, 202.

Sears, L. J.; Grimsrud, E. P. Anal. Chem. 1989, 61 , 2523.
Gregor, I. K.; Guilhaus, M. Int. J. Mass Spectrom. Ion Process. 1984, 56, 167.

Chapter 5: Methodology for the Determination of Steroids in Biological Fluids

The determination of steroids by chemical oxidation followed by analysis of
the oxidized extract by GC-ECNI-MS was pioneered in this laboratory (1-3).
Dexamethasone in plasma and 6B—hydroxycortisol in mine have been successfully
determined. Due to this success, other researchers have asked our assistance in the
analysis of steroids in biological samples.

Dr. Stephen Bromley of the Department of Zoology at Michigan State
University had been studying the regeneration of limbs of the newt, Notophthalmus
viridensens. In previous work using radiolabelled compounds (4), he found that
cortisol and corticosterone accumulate in the regenerating portions of the limb; the
maximum accumulation occurred at seven days. Aldosterone, testosterone, glucose,
cholesterol, and leucine do not accumulate in newt limbs, but esu'adiol is found. To
verify that corticosteroids such as cortisol, corticosterone, and 18-
hydroxycorticosterone are found in higher concentrations in regenerating portions of
newt tissue, a GC-MS method was desired. Because the corticosterones in this study
oxidize to poorly electrophilic products, other derivatization methods were
investigated.

Dr. Clinton Kilts of the Department of Psychiatry at Duke University has
collaborated with this laboratory in the past (3) and has decided to continue the
collaboration. His interests include the pharrnacokinetics of dexamethasone in the
dexamethasone suppression test (DST), a method for distinguishing Cushing's
syndrome and a probable indicator of endogenous depression. The RIA method
currently used for the determination of dexamethasone may suffer fi'om problems with
cross-reactivity with 6B—hydroxydexamethasone, the most common metabolite of

dexamethasone, especially at low dexamethasone concentration.

200

201

Dr. Pamela Fraker of the Department of Biochemistry at Michigan State
University has been studying the effects of endogenous and synthetic corticosterones
on the immune system of mice. The fluorescent-labelling technique currently used in
her laboratory to quantify steroids did not work with prednisolone, but the chemical
oxidation methodology was sensitive and selective enough for this determination.

Along with applying the chemical oxidation methodology to new problems,
attempts were made to improve the methodology. Another oxidant, tetra-n-

propylammonium perruthenate, was compared to pyridinium chlorochromate.

Alternative Oxidation Methods

Pyridinium chlorochromate (PCC) has been the oxidation reagent of choice for
the chemical oxidation procedure. Advantages of PCC include fairly efficient yields
and easy clean up. Problems with the reagent are long reaction times and toxicity
(PCC is a suspected carcinogen). Recently, tetra-n-propylammonium perruthenate
(TPAP) was proposed as a mild oxidant for alcohols (5,6). This reagent converted
primary and secondary alcohols into their corresponding aldehydes and ketones,
generally with short reaction times and very high yields. For example, Six-androstan-
17B-ol-3-one was converted into 5a-androstan-3,l7-dione in 99% yield in 1.5 h.
Other functional groups such as silyl ethers, indoles, pyrroles, acetals, sulfones, esters,
epoxides, double bonds, lactones, and halides did not react with TPAP. The TPAP
reaction occurs at room temperature with reaction times between ten minutes to one
hour. The reagent works well for small quantities (on the order of one milligram).

The TPAP oxidation method was compared to the PCC oxidation method for
5a-androstan-l7flcol-3-one, 5a-androstan-3B,l7B—diol, and prednisolone. The first
steroid was chosen because it was efficiently converted to the diketone by TPAP (5).
Prednisolone was chosen to determine if the TPAP could effectively oxidize a 17-

substituted steroid.

202

For the TPAP oxidation, about 0.1 eq steroid (0.05 mmol) and 1.5 eq N-
methylmorpholine N-oxide (Aldrich, Milwaukee, WI) were dissolved in 2 mL/mrnol
dry HPLC-grade methylene chloride (Mallinckrodt, Paris, KY). The N-
methylmorpholine N-oxide was the co-oxidant. To improve the rate and efficiency of
the reaction, 500 mg/mmol of powdered 4 A molecular sieves (Aldrich) were added.
Griffith and Ley (5) have noted that in dichloromethane some oxidations did not go to
completion; the addition of 10% acetoniuile overcame this problem. Addition of
acetonitrile was not necessary in this case. The methylene chloride mixture was stirred
at room temperature for 10 min, and then 5 mole percent TPAP was added. After
stirring at room temperature for 1.5 h (6.5 h for prednisolone), the mixture was filtered
through a ZOO-mg silica solid-phase extraction column (Burdick & Jackson) with
methylene chloride as the eluent.

For the PCC oxidation, methodology developed for the determination of
prednisolone was used. The reaction was allowed to proceed 1.5 h for the androstanes
and 6.5 h for prednisolone.

TPAP oxidation of prednisolone had very disappointing results. Two
compounds were observed after the reaction, one being the starting material (34%
recovery) and the other being 1,4-androstadien-3,11,17-trione, the expected oxidation
product in a 40% yield.

Table 5.1: Comparison of PCC and TPAP Oxidations

Compound Percent Yield
PCC of Androstan-3-one-l7B—ol 84
TPAP of Androstan-3—one-17B-ol 92
PCC of Androstan-3l3,17B-diol 50

TPAP of Androstan-3B, l7B-diol 44

 

 

203
The yields for the other model compounds are summarized in Table 5.1.
Steroids with one hydroxyl group appear to be more efficiently oxidized than those
with two hydroxyl groups. The yields for both oxidation methods are about the same;

consequently, there is no advantage to switching to the TPAP oxidant.

Determination of Corticosteroids in Newts

To quantify the amount of corticosteroids in newt tissue and plasma, sensitive
mass spectral techniques are necessary due to the small sample size. The chemical
oxidation methodology that was previously used effectively for synthetic
corticosteroids in this laboratory will not work for the corticosteroids of interest here.
Both cortisol (4-pregnen-l lB,17a,21-triol-3,20-dione) and corticosterone (4-pregnen-
118,21-diol-3,20—dione) oXidize to the same product, 4»androsten-3,11,17-trione. This
steroid has a fairly low ECNI response; nanograms of compound are necessary for
detection. 18-Hydroxycorticosterone (4-pregnen-1 10, l 8,21-triol-3,20-dione) does not
react with PCC. Methoxyarnine-trimethylsilyl (MO-TMS) and trimethylsilyl (TMS)
derivatives have been used extensively for the analysis of steroids using GC and GC-
MS techniques, and were used for the analysis of corticosterones.

Many procedures for the TMS derivatization of steroids have been published.
Three of these methods were attempted using a relatively large amount (ca. 50 ug) of
cortisol and corticosterone. Efforts to prepare the TMS -ether-TMS -enol derivative
using a mixture of MSTFAzTMIS (N-methyl-n-trimethylsilyltrifluoroacetamidezN-
trimethylsilylimidazole) (7) gave poor results. If 100 ttL of MSTFAzTMCS (TMCS =
trimethylchlorosilane) (100:2) are used at 60°C for 10 min (7), both cortisol and
corticosterone react. Upon analysis of the reaction mixture by GC-EI-MS, many peaks
are noted; the primary chromatographic peak corresponded to the tri-TMS moiety. If
50 [LL BSTFA with 10 ug potassium acetate as the catalyst are reacted with the steroid

for 5 h at 80°C (3), two major products are observed for both cortisol and

204
corticosterone. These products were identified as the di- and tri-TMS moieties of
cortisol and corticosterone. Because TMS derivativization resulted in multiple
products, it was discontinued.

MO-TMS derivatives were prepared based on the urinary steroid procedure
from the metabolic profiling lab (8). To the steroid residue were added 100 LIL of
methoxyamine-HCl (100 mg/mL in dry, redistilled pyridine). After heating at 60 °C
for 4 hr, the sample was dried under a stream of nitrogen for 45 min (until glassy).
Then 200 11L Sylon BT'Z (Supelco, Bellefont, PA) were added and heated at 80 °C for
24 h. Sylon BTZ is a 3:2:3 mixture of BSA (N,O-bis(trimethylsilyl) acetamide),
TMCS, and TMIS. A Lipidex column was used to separate the products from the
excess reagents. In preliminary studies, the Lipidex clean-up step contaminated the
sample further. Several other clean-up procedures were then attempted, including the
use of a C-18 solid-phase extraction, a methanol/hexane extraction, and a hexane
exu'action. The hexane procedure resulted in the best extraction efficiency, even for
low levels of sample.

The MO-TMS derivatization process was applied to real samples. Newt tissue
(from a limb) and newt blood were extracted using C-l8 solid-phase extraction
columns and the mouse plasma extraction procedure described later in this chapter.
Before extraction, each sample was spiked with the internal standard, dexamethasone.
GC chromatograms of the MO-TMS products of the real samples are shown in Figure
5.1. After analysis by GC-EI-MS, no evidence of ions indicative of dexamethasone,
cholesterol, cortisol or corticosterone were found. Two explanations are possible.
One is that the extraction procedure was not very efficient. The other is that the MO-
TMS derivative is not very sensitive to detection by GC-EI-MS.

At this point, all further research ceased because Dr. Bromley was too busy

with other commitments.

205

5.778

_ t
8.662
12.517

10.701

 

(b)

 

 

Figure 5.1: GC Chromatograms of (a) newt plasma and (b) newt limb.

The Determination of Dexamethasone in Human Plasma

The dexamethasone suppression test (DST) has been shown to differentiate
endogenous depression and depression caused by neurotic, reactive, or
characterological factors (9). In this test, a patient is given a l-mg dose of
dexamethasone. Plasma cortisol levels are determined at 8, 16, and 24 h following
administration. If the plasma cortisol levels are suppressed for a 24-h period, the
patient is classified as non-endogenously depressed; however, if the plasma cortisol
levels return to normal after 3.4 h, the patient is diagnosed as being endogenously
depressed. Studies have indicated that the bioavailability of dexamethasone is
important in interpreting the DST (10). The DST may be improved by simultaneous

monitoring of the plasma cortisol and dexamethasone levels.

206

Plasma dexamethasone concentrations for the DST are in the 0.1 to 4 ng/mL
range. RIA has been successfully used in this range; however, the Kilts' group wished
to independently validate their RIA method. A major concern was at the end of the
24-h test, when dexamethasone levels are low, the levels for the major metabolite of
dexamethasone, 6B—hydroxydexamethasone, may be interfering with the RIA analyses.

In this study, RIA methods were compared to the chemical oxidation method.
After the concentration of practice samples was validated, values obtained by GC-
ECNT-MS were compared to those obtained by RIA for a large set of samples to
determine coefficients of variance. Finally, the amounts of dexamethasone were

determined for a 24-h collection of plasma for a normal patient undergoing the DST.

Experimental

W. The general scheme for the extraction and oxidation of
human plasma samples is shown in Figure 5.2. Steroids were extracted from plasma
by James Ritchie who works for Dr. Kilts at Duke University. All samples were
spiked with 2.5 or 5 ng/mL of flumethasone (Sigma, St. Louis, MO). Sep-Pak ZOO-mg
solid-phase extraction cartridges (Waters, Milford, MA) were preconditioned as
follows. Firsr, 5 mL 8 M urea were applied to react with the active sites. Then, 5 mL
methanol were washed through the column. Organic solvents such as methanol open
up the hydrocarbon chains and increase the surface area available for interaction with
the analyte as well as remove residues from the packing material that may interfere
with the analysis (11). If this step is not properly done, poor recoveries of analyte due
to reduced retention on the column and interference peaks are possible. Next, 10 mL
water were placed on the column. This step removes excess methanol as well as
prepares the surface for the sample. To be most effective, this wash should be as close
as possible to the sample in polarity, ionic strength, and pH. After the sample was

applied, it was washed with 4 mL water to remove any polar material. A 38%

207

Condition C18 Sep-Pak
l
Add spiked plasma sample

l

Wash
l

Extract
l

Evaporate under nitrogen
l
Oxidize for 6 hours at 60 0C
J.
Condition silica solid phase extraction column
l
Add oxidation mixture

l

Wash
l

Extract
l

Evaporate under nitrogen

l

Reconstitute

Figure 5.2: Sample preparation scheme for the determination of dexamethasone or
prednisolone in plasma.

208
methanol/62% water wash (4 mL) was used to remove polar endogenous steroids and
other polar interferences. The dexaruethasone was eluted with 4 mL of 70%
methanol/30% hexane into a silanized l-dram vial. The extracts were dried under
nitrogen, frozen, and shipped to MSU overnight under dry ice. All samples were
frozen at -4°C until use.

demign. To each residue, 1 mL methylene chloride (HPLC grade,
Mallinckrodt, Paris, KY) was added, and the mixture was vortexed. About 10 mg
anhydrous sodium acetate (J. T. Baker, Phillipsburg, NJ) were added to buffer the acid
liberated during the oxidation. PCC (Aldrich, Milwaukee, WI) and Celitem‘ analytical
filter aid (Johns-Manville, Lompoc, CA) were ground together in a 1:2 ratio with a
mortar and pestle. The reaction, with PCC as the oxidant, occurs on Celite, a solid
support. An excess of the PCC/Celite mixture was added to each vial. The reaction
mixture was stirred and heated to 60°C for 6 h.

W. The ZOO-mg silica solid-phase extraction
columns (Burdick & Jackson, Muskegon, MI) were conditioned with several column
volumes of methylene chloride. Because they are more easily wetted with the
consumption of significantly less solvent and no loss of efficiency, ZOO—mg silica
columns were substituted for the 500—mg columns used in previous work. The
oxidation mixture was added and then the column was washed with 1.5 mL methylene
chloride. The oxidized dexamethasone was eluted with 4 mL of 5% acetone in
methylene chloride. After evaporation under nitrogen, the sample was reconstituted in
75 uL of ethyl acetate.

Wong. ECNI mass spectral data were obtained on a JEOL JMS-
AX505H mass spectrometer (Peabody, MA) interfaced to a Hewlett Packard 5890 gas
chromatograph (Palo Alto, CA). The following parameters were used: source
temperature, 150°C; ionization current, 300 uA; electron energy, 200 eV; modulating

gas, methane; source pressure, 2 x 10'5 torr; and accelerating voltage, 3 keV. A J&W

209

lS-m x 0.25-mm x 1.0-um DB-l column (Folsom, CA) or a J&W lS-m x 0.25-mm x
0.25-um DB-1701 column was directly inserted into the mass spectrometer ionization
chamber. Helium was used as the carrier gas. The GC oven was held at 140°C for 1
min, then ramped to 260° at 40°lmin, followed by a 4°lmin ramp to 280°C. The
splitless GC injector was held at 260°C, and the u'ansfer line at 260°C.

The mass spectrometer was operated in the selected ion monitoring (SIM)
mode to analyze the plasma extracts. Ion crurents at mlz 310.2 (corresponding to [M-
I-IFT for oxidized dexamethasone) and 328.2 (corresponding to [M-HFT for oxidized

flumethasone) were monitored using magnetic field switching SIM.

Changes to the Oxidation Methodology

In previous collaborative work with the Kilts' group analyzing dexamethasone
in plasma, several parameters were different than in the work now undertaken.
Previously, the plasma samples were sent unextracted to MSU; now, samples are
extracted at Duke. The internal standard, GC column, and GC-MS instrument were
also changed. To be certain that the methodology worked before analyzing real patient
samples, practice standard curves and unknowns were extracted and sent to MSU for
further analysis.

Stable isotope dilution is the premier technique for the quantitation of
biological compounds by MS. Not only are stable isotope-labeled analogs useful as
tracers in the body (12), they are the best choice for internal standards (13). Evidence
suggests that stable isotope dilution does not make an assay more accurate; however,
the precision is improved (14). The structure of the analyte and standard are the same;
therefore, they behave identically under extraction and chromatographic analyses.
Correct choice of isotopically-labelled standards is important to avoid errors with loss
of label and problems with equilibration of the analyte in the biological sample. In
previous studies, the internal standard was (1,2,3,4,10,19-1306, 19,19,19-2H3)

210
dexamethasone from Dr. J. P. Freeman of the National Center for Toxicological
Research in Jefferson, AR. The major problem encountered with this internal standard
was that it was not isotopically pure, resulting in a higher limit of detection. In this
work, it was decided to change to a structurally-similar analog, flumethasone, which
has one more fluorine group than dexamethasone.

Other changes from the previous procedure include the type of column and,
more importantly, the type of mass spectrometer. Because the JEOL 505 at one point
did not have the sensitivity and also had problems with hot/cold spots in the uansfer
line, the Finnigan TSQ-70 triple quadrupole mass spectrometer was used in the single
quadrupole mode for previous work on dexamethasone plasma samples. Since then,
the sensitivity of the JEOL 505 has been improved by cleaning the flight tube,
realigning the magnet, and using one particular source for ECNI (the source designated
as the non-collision cell source). The detection limit for the JEOL 505 is not as good
as that for the TSQ-70. For the JEOL 505, the detection limit is 3 pg (S/N = 4; m/z
310) for a pure sample of oxidized dexamethasone. When using the TSQ-70, the
detection limit is 160 fg (SIN = 7, mlz 310). The higher detection limit for the JEOL
505 was not a problem for this assay because the interferences of the matrix are greater
than 3 pg. ‘

Preliminary work with authentic samples of oxidized flumethasone and
oxidized dexamethasone on a DB-l 15-m x 0.25-mm x 1.0-um column resulted in the
overlap of the minor isotope peak of flumethasone with the major isotope peak of
dexamethasone (see Figure 5.3). The most intense peak in the ECNI mass spectrum of
oxidized dexamethasone is at mlz 310. Unfortunately, oxidized flumethasone also
gives a small contribution from mlz 310 (see Figure 5.4); therefore, even in the
analysis of the blank, a peak representing oxidized dexamethasone was present.

Quantitation will be affected because the limit of detection (LOD) will be higher.

211

9 0 RETENTION TIME (MIN)

328.2 M

w Vi

- A A A A

 

 

310.2

 

 

 

 

 

V V v v V ' I I V V V

1900
SCAN NUMBER

Figure 5.3: On a DB-l 15-m x 0.25-mm x 1.0-um column, the minor isotope peak of
oxidized flumethasone (top mass chromatogram) overlaps with the major isotope peak
of oxidized dexamethasone (bottom mass chromatogram).

Different columns were used in an attempt to separate the oxidized
flumethasone isomer peak from the oxidized dexamethasone peak A longer DB-l
column did not work. Neither did a slightly more polar DB-5 30-m x 0.25-mm x 0.25-
pm column (Hewlett Packard, Avondale, PA). Polar columns such as OV-17 (SO-m x
0.32-mm x 0.25-um, Quadrex, New Haven, CT) and DB-1701 (lO-m x 0.2-mm x 0.15-
pm, Chrompack, Raritan, NJ) were tried unsuccessfully. Changing the internal
standard to fludrocortisone would not have interfered with the dexamethasone
analysis. However, the Kilts' group was unwilling to use to fludrocortisone because
the extraction procedure would also have to be changed. They were willing to accept
the higher LOD rather than redo the extraction procedure.

The fragment ion peak at mlz 310 in the ECNI mass spectrum of flumethasone
cannot be explained by any logical loss, so MS/MS was used to elucidate the source of
this fragment. The TSQ-70 triple quadrupole mass spectrometer was operated by Tim

212

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a o
( ) 1 0: 328]
80 ‘ f
g so i
j r
3 348 i
o 40 4 i‘
>
3; ‘ i
t r
‘3‘ 20 l 310 :20 '_
l 313 C
0 hr A_A A -4 n :A ..g afi AV..I-:-...' A .-.nll_l ll ‘ -.--a l f A AA A7 A f v
260 280 300 320 340 350
mlz
(b) 1001
t 3 10 r
g so ‘ T
I
< 60 I r
g l
a: 40 ;
l
j 295 ’_
20 p
l 330 .
; a- . ALI AAA 5- - A; L _llkl:1 11 l A- ‘ J 41 ll Al‘f - e b
260 280 300 320 340 360
mlz
100
(C) i 324
so ;
so I 3 .
-I 309 >-
i r
40 1 .
2 1 ~
:3 : ’
as 20 j x 10
0 iv A A .1 A r H Li 1‘ 1 1.1 I“ L11 rr'Ll _ L l 214 -1 l r 111. r r ‘
260 280 300 320 340 360

mlz

Figure 5.4: ECNI mass spectra of (a) l,4-androstadien—6a,9a-difluoro-160t-methyl-
3,11,17-trione (oxidized flumethasone), (b) l,4-androstadien-9a-fluoro-16a-methyl-
3,11,17-trione (oxidized dexamethasone), and (c) l,4-androstadien-9a-fluoro-16a—
methyl-3,6,11,17-tetrone (oxidized 6B—hydroxydexamethasone).

213
Heath in the ECNI mode with ammonia as the modulating gas and argon as the
collision gas.

A full scan spectrum of oxidized flumethasone has a peak of 16.4% relative
intensity at m/z 310 and 13.8% at mlz 330 under the conditions used on the TSQ-70.
The peak at m/z 330 is also an isotope peak of mlz 328 but, as an isotope peak, the
percent intensity should have been only 2.4%. The daughter spectrum of the
molecular ion of oxidized flumethasone included peaks at m/z 328, 313, 302, and 293.
No peak at mlz 310 or 330 is derived fiom the molecular ion. The parent of peak at
mlz 310 in the spectrum of oxidized flumethasone is mlz 330. The peak at m/z 350 is
also enhanced in the ECNI spectrum of flumethasone. The daughters of m/z 350
include mlz 330, 315, 310, 286, and 267. The peak at m/z 350 is not just an isotope
peak of the molecular anion at mlz 348, if the relative intensities are compared (see

Table 5.2).

Table 5.2: Isotope Peaks of the Molecular Anion of Flumethasone

mlz % Relative Intensity % Relative Intensity Difference (%)
from Spectrum Calculated

348 27.16 --- ---

349 5.60 5.63 0.5%

350 1.41 0.59 140%

Obviously, m/z 349 is an isotope peak, but the peak at m/z 350 has a higher abundance
than it should. The peak at m/z 350 could correspond to 4-androsten-6a,90t-difluoro-
l6a-methyl-3,l l,l7-trione, which has one less double bond than oxidized
flumethasone. This coeluting electrophilic compound could easily fragment to [M-
ZHFT represented at mlz 310. The contaminant could be a by-product of the PCC
reaction, or the precursor could be a contaminant in commercially-available

flumethasone.

214
The Contamination Problem

In order to validate the method, practice samples consisting of 1 mL of steroid-
stripped plasma spiked with 5 ng/mL flumethasone as the internal standard and 0, 0.1,
0.75, and 5.0 ng/mL dexamethasone were analyzed using the oxidation methodology.
The ratio of the peak area of the selected ion current profile at mlz 310.2
(corresponding to oxidized dexamethasone) to that at mlz 328.2 (corresponding to
oxidized flumethasone) was nearly the same for all samples. In fact, the average of the
peak ratios for all samples was 0.23 i 0.05, which suggests that there is no difference
between the spiked samples. Obviously, all samples were contaminated with similar
amounts of either dexamethasone or another compound with peaks at mlz 310 and 295
in its mass spectrum, a highly unlikely combination. This contamination could be
from a variety of sources including: (1) the use of commercially-available steroid-
stripped plasma, (2) the commercial flumethasone used as the internal standard may
have residual dexamethasone, (3) the rinses used by the Kilts' group in the extraction
step may be contaminated, (4) cross-contamination of samples may occur during
spiking or extraction, (5) the oxidation reagents and the rinses used in this laboratory
may be contaminated, and (6) some type of interconversion was happening to change
flumethasone to dexamethasone in the plasma sample. Option (6) is highly unlikely.
Option (4) is plausible, only if some of the samples were contaminated, not all of them.
The other possible sources of contamination were addressed in the next few
experiments.

When the reagents (PCC, sodium acetate, Celite, and methylene chloride) were
oxidized alone, no dexamethasone (or anything else) was observed. Authentic samples
of flumethasone were oxidized in high concentrations, but no dexamethasone
contamination was noted.

The next set of samples from Duke University consisted of the following

spiked in 1 mL plasma: blank (no steroids added), 5 ng flumethasone, 5 ng

215

dexamethasone, and both flumethasone and dexamethasone. The steroid-stripped
plasma was thought to be the source of the problem, so the samples were extracted
from plasma that had never been touched by dexamethasone, i. e., plasma donated by
Jim Ritchie himself. The contamination was noted in all samples, which proves the
internal standard is not the source of the problem. The contamination had to occur
during the plasma extraction or the PCC clean-up steps, but other PCC reactions that
were being used for different experiments did not have contamination problems.

The extraction step in question did not occur in this lab. To prove that the
plasma extraction step indeed was causing the problem, a set of plasma samples that
had already been extracted as well as the corresponding unextracted plasma samples
were prepared at Duke University. The unextracted plasmas were worked up in this
laboratory as well as a set of blood-bank plasmas that was spiked and extracted in this
laboratory. All samples extracted in this laboratory were not contaminated, but those
prepared in the Kilts' lab were. After the Kilts' group learned that the source of
contamination was in the extraction step, they decided to use freshly-prepared

solvents. Thus, the contamination problem was solved.

Practice Standard Curves

Now that the contamination problem was solved, a practice standard curve and
five unknowns were prepared to ascertain the viability of this technique. To prepare
the standard curve, samples containing 0, 0.1, 0.75, and 5.0 ng/mL of dexamethasone
were prepared in duplicate and analyzed by GC-ECNI-MS in duplicate. Before
extraction, each of the points on the standard curve as well as each of the unknowns
had been spiked with 5 ng flumethasone. The standard curve was linear with the
equation of the line y = 0.0795 + 0.150x with a correlation coefficient (r2) of 0.999.
The limit of detection (LOD) has been defined as yB + 3 S]; where ya is the value of
the y-intercept and SB is the standard deviation of the blank. If this definition for LOD

216
is used, the probability of either a false positive or a false negative is 7% (15). The
LOD is 0.25 ng/mL, which is unacceptable because patient samples will have
concentrations as low as 0.1 ng/mL. The high LOD made it difficult to distinguish
between the blank values and the 0.1 ng/mL values on the standard curve; this made
quantification of the unknowns difficult (see Table 5.3).

Table 5.3: Unknowns from Practice Standard Curve

Sample ECNI-MS Estimate RIA Estimate
A 2.3 ng 3 ng
B 0toO.l ng 0.15 ng
C 1.8 ng 0.70 ng
D 0t00.1ng 0.11ng
Patient 2.3 ng 1.9 ng

In order to improve the DOD, the amount of internal standard was decreased by
half, to 2.5 ng/mL. The final volume of the reconstituted reaction mixture was
decreased from 100 uL to 75 uL. The standard curve for this practice set was
excellent, with a correlation coefficient of 1.000. Because only one blank value was
used, the LOD cannot be determined. However, there was no agreement between the
values for dexamethasone determined by ECNI-MS as compared to the real values and
values obtained RIA analysis (see Table 5.4).

Table 5.4: Unknowns from Practice Standard Curve with Less Internal Standard
(all concentration values are in ng/mL)

Sample Real Concentration RIA Estimate ECNI-MS Estimate
E 0.60 0.66 1.0
F 0.25 0.25 0.63

G 2.1 3.0 2.6

217

Another practice standard curve with unknowns was prepared This time, six
blanks were analyzed in order to determine the LOD more precisely. The rest of the
samples were prepared in duplicate and analyzed by GC—MS in duplicate. The
standard curve was not as linear as in the previous analysis (correlation coefficient of
only 0.990), but the LOD has decreased to 0.11 ng/mL. This is a lower LOD than
previously reported for the chemical oxidation of dexamethasone using an isotopically-
labelled internal standard (3). The values determined by ECNI-MS were much closer
to the RIA values (see Table 5.5) than previous practice sets; therefore, the assay was

validated.

Table 5.5: Unknowns from Practice Standard Curve

Sample RIA Concentration (ng/mL) ECNI-MS Concentration (ng/mL)
17a 0.80 0.90 i 0.18 (n = 2)
18b 1.8 2.0 i0.1(n=2)
19c 0.34 0.41 i 0.12 (n = 2)

Patient 1.44 1.08

Dexamethasone Precision Study

Now that the standard curve was linear and the values for the practice samples
agreed with those obtained by RIA, a precision study was undertaken. In this study, a
six-point standard curve was prepared in duplicate. To determine a precise LOD
value, the six blank samples were prepared. Six unknown samples at four different
concentration. values were analyzed by both GC-MS and RIA. This was a double-
blind study; the operator for each technique did not know either the concentration
values or which samples were in which pool. For ECNI-MS, the standard curve was
linear (see Figure 5.5) if the 0.3 ng/mL points were rejected. These points were
discarded due to cloudy solutions, probably occurring from faulty PCC clean-up

218
procedures. The equation of the line was y = 0.0646 + 0.299x with a correlation

coefficient of 1.000. The LOD (yB + 3 83) was 0.096 ng/rnl.

 

Relative Response

 

 

 

 

0.0 l ‘ I ‘ 7 ‘ ' '
0 1 2 3 4

ng dex/mL plasma

Figure 5.5: Standard curve for the determination of dexamethasone in human plasma.
The equation of the line was y = 0.0646 + 0.299x with a correlation coefficient of
1.000. The LOD (YB + 3 S3) was 0.096 ng/mL.

The values for each set of data obtained by RIA as well as the means and
coefficients of variation (cv) are presented in Table 5.6. The results from the chemical
oxidation study are in Table 5.8. It was concluded that three samples were not
extracted and/or oxidized efficiently because no response for oxidized dexamethasone
and a very diminished response for the internal standard were noted. To improve the
variances for the ECNI-MS data, statistical outlier tests (16) were applied to each pool
of data. Dixon's r11 statistic is a test for a single upper outlier, x“, in a normal sample

with 32 unknown. The test statistic is

X " -
xn‘7‘2

219
where xn is the upper outlier, x“ is its next nearest neighbor, and x2 is the second
lowest value . The Dixon r22 statistic tests for an upper outlier pair, xn_l, xn, in a

normal sample with s2 unknown. The test statistic is

X -X _
T: n “2 (2)

xn-x2

 

where xn,2 is the third highest value. To test for a lower and upper outlier pair, x1, xn,
in a normal sample with the mean and s2 unknown, the test statistic is
T = internally studentized range = gill—— (3)

The appropriate tabulated significance levels can be found in reference 16. Using the
test statistic in equation 3, the upper and lower values for pool #1 were rejected. The
test statistic in equation 2 was used to reject the upper value in pool #2 and the lower
value in pool #3. The two upper outliers in pool #4 were rejected using equation #1.
The modified means and standard deviations are found in Table 5.8.

When the RIA and GC-MS data are compared to the true values (Table 5.7),
several conclusions can be drawn. At low levels of dexamethasone, RIA values were
closer to the real value; however, the GC-MS data were better at high values. In
general, the coefficient of variance was lower for RIA than for GC-MS. This may be
due to the fact that no extraction steps are necessary for the RIA determination, while
two are necessary for the oxidation technique. These two extractions were done in two
different labs, which may contribute to the variance. The samples spiked with 0.45
ng/mL dexamethasone as well as 0.15 ng/mL 6B-hydroxydexamethasone resulted in
high values for both the RIA and GC-MS determinations. At first, this was attributed
to a mistake in the spiking step; however, further evidence suggests. that the

interference of 6B-hydroxydexamethasone is to blame, as will be explored further in

the next section.

220

Table 5.6: Dexamethasone Precision Study-RIA Values for Replicate
Analyses of the Indicated Sample Pools (all concentrations in ng/mL)

Pool #1 Pool #2 Pool #3 Pool #4
0.10 0.38 1.46 6.99
0.08 0.35 1.37 . 8.49
0.13 0.40 1.34 11.63
0.09 0.37 1.43 28.47
0. 12 0.40 1.37 6.08
0. 14 0.37 1.67 6.24
n 6 6 6 6
mean 0.11 0.38 1.42 1 1.3
sd 0.02 0.02 0. 14 8.6
cv (%) 22 5.1 9.7 76

Table 5.7: Comparison of Dexamethasone Concentrations Experimentally
Obtained by RIA and ECNI-MS. All results expressed in ng/mL.

Real Value RIA Results GC-MS Results GC-MS Results
DB-ll Column DB-l701 Column
0.15 0.11 i 0.02 0.31 :l: 0.04 0.26 :l: 0.08
0.45 0.38 :l: 0.02 0.42 i 0.04 0.42 :1: 0.05
1.20 1.4 120.1 1.1 3:02 1.1 :l:0.1
0.45 +0.15 11 14 2.7 :02 0.481013

6B-OH dex

221

Table 5.8: Dexamethasone Precision StudquCNI-MS Values for Replicate
Analysis of the Indicated Sample Pools (all concentrations in ng/mL)

Pool #1 Pool #2 Pool #3 Pool #4
, 0.28 0.57 1.13 3.9
0.30 0.44 0.91 4.1
0.37 0.44 1.37 2.8
0.39 0.37 1.04 2.9
0.16 1.13 2.6
0.45 2.3
n 5 4 6 6
mean 0.28 . 0.45 1.0 3.1
sd 0.07 0.07 0.3 0.6
cv (%) 24 16 28 21

After statistical outlier tests

n 4 3 5 4
mean 0.31 0.42 1.1 2.7
sd 0.04 0.04 0.2 . 0.2

cv (%) 12 8.5 14 7.3

222
The Interference of 6 B-Hydroxydexamethasone

The reproducibility of the GC—MS method, although not as precise as the RIA
method, was deemed sufficiently satisfactory to continue the project. At this point, the
importance of the suspected interference from the 6|3-hydroxydexamethasone on both
the RIA and GC-MS assays was stressed. Six samples containing 0.25 ng/mL
flumethasone and 0.2 ng/mL 6B-hydroxydexamethasone (and no dexamethasone)
underwent analysis by both RIA and the oxidation ECNI-MS methods. By RIA, no
response from dexamethasone was noted; however, by GC-MS on the DB-l‘ column
previously used for all analyses, there was a response.

An authentic sample of 6B—hydroxydexamethasone, a gift from the Kilts' group,
was oxidized by PCC. 6B«Hydroxydexamethasone was synthesized by Dr. E. Cook of
Research Triangle Institute (Research Triangle Park, NC). After oxidation for 6 h and
the usual clean-up procedure, the oxidized product, l,4-andmstadien-9a—fluorine—16a-
methyl-3,6,11,17-tetrone (MW = 344) had the spectrum shown in Figure 5.4 (c). The
peak at mlz 309 corresponds to [M-HF-CH3]'; the peak at mlz 310 is the isotope peak.
Unfortunately, mlz 310 also corresponds to the peak monitored in the analysis of
oxidized dexamethasone. On a DB-l lS-m x 0.25-mm x 1.0-um column, the oxidized
dexamethasone and the oxidized 6B-hydroxydexamethasone chromatographic peaks
overlap.

Several other columns were used in an attempt to separate oxidized
dexamethasone from oxidized 6B-hydroxydexamethasone. The two. compounds
overlapped on a longer DB-l column, 30m x 0.25-mm x 0.25-um (J&W, Folsom,
CA). In previous work (3), it was stated that the two compounds separated on a
slightly more polar DB-S column. When a 30-m x 0.25-mm x 0.25-um DB-S column
(Hewlett Packard, Avondale, PA) was used, the two compounds separated by 0.04

minutes, but because both compounds have alpha and beta isomers at the 16-methyl

223
position, this will not be enough resolution. On a DB-l701 (15-m x 0.25-mm x 0.25-
ttm, J&W, Folsom, CA) column, the two separated by more than a minute.

When a DB-1701 column was used to analyze the six samples containing only
flumethasone and 6Bohydroxydexamethasone, only a small contribution from
dexamethasone was present. The dexamethasone level was 0.074 ng/mL. The LOD
for this analysis was 0.059 ng/mL; therefore, the level of dexamethasone in the 6B-
hydroxydexamethasone samples is slightly above the detection limit. Perhaps there is
residual dexamethasone in the authentic sample of 6B-hydroxydexamethasone.

The precision study was redone using the DB-1701 column. Again, all samples
were analyzed randomly. The standard curve again was very good, with an equation
of y = 0.0320 + 0.188x and a correlation coefficient of 0.999. The limit of detection
was 0.059 ng/mL. The peak at mlz 324.2, corresponding to the [M-HFJ' peak for
oxidized 6B-hydroxydexamethasone, was also monitored to determine the contribution
from this metabolite. In general, there was very little to no contribution from m/z 324.
From the data summarized in Table 5.7, the coefficient of variance for this GC-MS
analysis is higher than that from the previous analysis, but the accuracy is much better.
In fact, the value for pool #4 is almost the same as the real value. From the
contribution at m/z 324, one suspects that there was more than 0.15 ng/mL 6B-
hydroxydexamethasone spiked into these samples; the amount was probably 10-100

times as much.

Post-Dexamethasone Nighttime Study

The oxidation methodology was used to validate the RIA values from a 24—
hour post~dexamethasone study. Of special interest were the values at low
concentrations of dexamethasone. At long collection times, the amount of
dexamethasone should decrease and the relative amount of 6l3-hydroxydexamethasone

increase. It is suspected that the RIA results may not be valid at this point.

224

A dose of dexamethasone was given to a normal male aged 22 years. Blood
was drawn every half hour for the first two hours, then every 15 min for the next seven
hours. Eight hours after this point, the last sample was drawn.

For the oxidation methodology, each sample was spiked with 2.5 ng of
flumethasone no matter what the starting volume of plasma (some were below 1 mL).
A standard curve consisting of 6 blank samples and two each of 0.1, 0.3, 0.5, 1.0, and
3.0 ng/mL dexamethasone was prepared. Each point on the standard curve except for
the blanks were analyzed in duplicate; selected patient samples also were analyzed in
duplicate.

The peak at mlz 324 corresponding to oxidized 6B-hydroxydexamethasone was
monitored along with the peaks at mlz 310 for oxidized dexamethasone and at mlz 328
for oxidized flumethasone. Only about 40% of the patient plasma samples had any
contribution from 6B—hydroxydexamethasone, and there was no pattern as to which
samples contained the metabolite. The signal for the metabolite was weak in all
instances. Perhaps better results for 6B.hydroxydexamethasone could be obtained if
the extraction and oxidation steps were also optimized for the determination of this
metabolite.

As shown in Table 5.9, all values obtained from the GC-MS method were
higher than those obtained from RIA. No reasons for the discrepancy are inherently

obvious; therefore, another sample set was analyzed.

Dexamethasone Precision and Patient Study

Another sample set was analyzed in response to the poor correlation of the
post-dexamethasone nighttime study RIA data to GC-MS results. All plasma samples
were l-mL volumes. The standard curve was prepared as before, with the addition of
a 4.8 ng/mL point. A standard curve for 6B—hydroxydexamethasone, consisting of
plasma samples spiked with 0.2, 0.4, 0.8, and 1.5 ng/mL analyte, was also analyzed.

225

Table 5.9: Post-Dexamethasone Nighttime Study. Experimentally Determined
Values for Dexamethasone in Plasma by Indicated Methods at the

Indicated Times.
Sample Number Time Drawn RIA Conc. GC-MS Conc.
. ng/mL ng/mL
l 1844 23:30 <0.02 3.43
1 1845 23:00 <0.02 0.005
11846 23:30 2.51 4.18
11847 24:00 3.89 4.54
11848 00:30 2.64 4.59
11849 01:00 3.81 4.41
11850 01:15 2.65 5.35
11851 01:30 2.35 -----
11852 01 :45 1.58 3.59
11853 02:00 2.18 4.17
11854 02: 15 2.19 2.66
11855 02:30 3.28 4.23
11856 02:45 2.52 4.03
11857 03:00 2.56 4.01
11858 03:15 2.55 5.12
1 1859 03:30 2.26 3.48
l 1860 03:45 2.08 3.65
11861 04:00 1.89 3.22
11862 04:15 2.15 3.43
11863 04:30 1.98 3.58
11864 04:45 2.06 3.61
1 1865 05:00 2.00 2.96
11866 05:15 1.54 3.13
1 1867 05:30 2.25 3.39
11868 05:45 2.13 4.34
11869 06:00 1.90 2.93
11870 06:15 1.69 3.12
11871 05:30 1.65 3.45
11872 05:45 2.63 4.69
11873 07:00 1.57 2.71
11874 ‘ 07:15 2.48 -----
11875 07:30 0.81 2.25
11876 07:45 2.21 3.11
11877 08:00 2.08 0.14
11878 16:00 0.63 5.16

226
The precision study consisted of three pools of eight samples each. The patient study
consisted of 16 samples. All samples were extracted on the same day. The analysis of
all samples by GC—MS also occurred on one day. The standard curve for oxidized
dexamethasone was linear, with a correlation coefficient of 0.993. The LOD for this
analysis was 0.034 ng/mL.

Table 5.10: Precision Study for Dexamethasone. Values for the Replicate
Analysis of the Indicated Sample Pools Expressed in ng/mL.

Pool #1 Pool #2 Pool #3
0.24 1.46 0.52
0.31 1.53 0.40
0.20 1.65 0.44
0.27 1.60 0.52
0.25 1.55 0.51
0.27 1.49 0.37
1.66
1.40
n 6 8 6
mean (ng/mL) 0.26 1.54 0.46
sd 0.03 0.09 0.06
cv (%) 13 5.6 13
Spiked (ng/mL) 0.34 1.91 0.58
RIA (ng/mL) 0.33 2.32 0.68

The results for the precision study are presented in Table 5.10. The samples
with poor extraction efficiencies were not reported. Without using any statistical
outlier tests to reject data, the precision of the GC-MS data was far better than that of
the study summarized in Table 5.7. Correlations with the spiked value and the RIA
values were not as good. On the average, the RIA values were higher than the spiked
values, but the GC—MS values were lower than the spiked values.

The patient study results did not follow a consistent pattern when compared to
the RIA (see Table 5.11). Sample 22729 had problems with overlapping
chromatographic peaks which may account for its low GC-MS value. In general,

227
plasmas with high RIA values of dexamethasone (above 1 ng/mL) had lower GC-MS
values. At lower levels of dexamethasone, the RIA values were lower than the GC-
MS values. In the precision study analyzed on the same day, all the GC—MS values

were lower than their corresponding RIA values.

Table 5.11: Dexamethasone Patient Study. Comparison of RIA and ECNI-MS

Experimental Results.
Sample Number RIA Concentration GC-MS Concentration
ng/mL ng/mL ‘
22733 2.56 -----
22734 1.68 -----
22735 2.50 -----
22736 1.82 1.60
22737 1.91 1.77
22739 3.09 1.73
22740 2.02 1.36
22743 2.24 1.99
22749 0.86 0.86
22752 0.84 -----
22755 0.75 0.70
22759 0.70 0.85
22764 0.45 ----
22767 0.39 0.47
22773 0.21 0.32
22776 0.15 -----

A rough 6B-hydroxydexamethasone curve was analyzed (see Figure 5.6). The
concentration of the standards were low as compared to the amounts found in the
patient samples. Many spiked standards had poor extraction efficiencies and were
subject to poor chromatographic peak shape. The correlation for the points that did
work was not good (r2 = 0.974). In Table 5.12, the estimated concentrations of 6B-

hydroxydexarnethasone in patient samples was tabulated. Values range from 0.6 to 1.6

228

 

 

 

 

0.03
0
2
o a
$0.02-
o
I a
o
2
E 0.01“
o
. T
0.00 fi u . 1 v r . 1 r
0.0 0.2 0.4 0.6 0.8 1.0

nglmL

Figure 5.6: Standard curve for the determination of 6B-hydroxydexamethasone in
human plasma. The equation of the line was y = 0.0605 + 0.0199x with a correlation
coefficient of 0.974.

Table 5.12: Estimated 6BJIydroxydexamethasone Concentrations for Patient

Samples
Sample Number 6B-Hydroxydexamethasone Concentration (ng/mL)
22737 1.12
22739 1.03
22740 1.41
22743 1.67
22749 1.06
22759 0.61

22767 0.66

229
ng/mL. No correlation of 6B-hydroxydexamethasone concentration with
dexamethasone concentration was noted. This assay is not ready for simultaneous

detection of dexamethasone and its major metabolite, 6B-hydroxydexamethasone.

Conclusions

Many changes have been made to this assay, but it still remains a valid method
of determining dexamethasone in human plasma. The JEOL 505, after improving the
sensitivity, worked just as well as the TSQ—70 for the ECNI-MS analyses. The use of
flumethasone rather than an isotopically-labelled internal standard resulted in better
detection limits. To avoid any overlap with the, major metabolite of
dexamethasone, 6B-hydroxydexamethasone, a polar column such as a DB-l701
column must be used.

Using this method, the level of the 6B—hydroxydexamethasone was low. If this
compound is to be quantitated in conjunction with dexamethasone, the assay must be
improved. Perhaps better choices of the temperature and pressure of the mass
spectrometer source may increase the sensitivity of the peak at mlz 324. The 38%
methanol in water wash used to wash steroids more polar than dexamethasone off the
column also elutes 6B-hydroxydexamethasone (17). 6fi-Hydroxydexamethasone was
more effectively extracted if a 15% methanol wash was used instead (17). The 5%
acetone in methylene chloride eluent used after the PCC oxidation was optimized for
dexamethasone, but not tested for 6B—hydroxydexamethasone. In six hours,
dexamethasone is converted to its oxidized product in a 77% yield, but the optimum
oxidation time of 6B—hydroxydexamethasone has not been determined.

In the first precision study comparing RIA and the GC-MS method proposed
here, the coefficient of variance is lower for RIA, but in most cases, the accuracy is
usually better for the GC-MS method. It has been suggested (18) that a rsd of about

16% should be sufficient fer the mass spectral analysis of trace components in food

230
and drugs. The reproducibility study meets this criterion only if the outliers are
ignored. Also, if possible, a calibration curve should cover the limit of quantitation to
one order of magnitude above trace level. The correlation coefficient of this line
should be no less than 0.94 ( 18). In this work, all standard curves had correlation
coefficients of 0.990 or above. .

In the second precision study, the coefficient of variance of the GC-MS method
decreased, but the accuracy was not as good. The values obtained by GC—MS were
lower than the spiked values or the values obtained by RIA.

The analysis of actual patient data gave very different results than the RIA
assay. In the first study, the GC—MS results were higher than the RIA results. In the
second study, the GC-MS results were lower at high concentrations of dexamethasone
and higher at low concentrations of dexamethasone. The lack of correlation between
results fi'om RIA and GC-MS analyses for steroid hormone such as testosterone,
progesterone, estradiol, and cortisol has been documented (19). Immunoassays were
usually positively biased in comparison with the GC—MS value. This observation was

generally true for the second patient study.

Determination of Prednisolone in Mouse Plasma

Dr. Pamela Fraker's group is interested in the effects of the immunosuppression
of prednisolone on the mouse immune system. Prednisolone is commonly used for the
treatment of inflammation due to injury or a disease such as arthritis, allergy and
asthma. While it is well documented that pharmacologically-active glucocorticoids
cause thymic atropy due to induction of apoptosis (programmed cell death) in
immature T-cells (cells that direct cell-mediated immunity), the effects of prednisolone
on normal B-cell (cells that synthesize and secret antibodies to introduced antigens)
development is virtually unknown. Using an in-vitro murine bone marrow culture

system, it was found that 10‘8 M, 10'7 M, and 10'6 M prednisolone caused 36%, 73%,

231

and 85% inhibition of the bone marow B-cell response to the T-cell independent
antigen, trinitrophenol-lipopolysaccharide (TN P-LPS). Modest elevation of plasma
prednisolone (2 ng/mL, determined by the method described below) in mice 10 days
after implantation of a prednisolone pellet caused splenic and thymic atrophy, as well
as decreased white blood cell counts. In addition, flow cytometric data revealed that
there was a 30% decrease in proportion of 3220 and surface immunoglobin-bearing
cells. This was accompanied by a 60% reduction in bone marrow response to TNP-
LPS. Taken together, these data indicate that low levels of prednisolone significantly
alters bone marrow B-cell function by depletion of B-lineage cells. Further, this
depletion appears to be caused in part by prednisolone induction of apoptosis (20).

I-IPLC has been the method of choice for the determination of prednisolone in
clinical samples (21-25). Using a silica HPLC column, prednisolone and prednisone
could be easily separated. Detection limits are on the order of 4-10 ng/mL. RIA has
been used for the determination of prednisolone; however, interferences from other
steroids, especially cortisol and prednisone, are problematic (25). GC-MS techniques
have been used to determine levels of prednisolone after preparation of the MO-TMS
derivative (26, 27). The chemical ionization method is not as sensitive as the ECNI
method. The latter can detect 0.25 ng of pure material, but no information ‘was
available for detection in a biological matrix.

Originally, the collaborators roughly estimated the amount of prednisolone in a
typical sample of mouse plasma to be between 10-20 ng. This was a very high
estimate; actual values were between 1-5 ng per sample. The sample size is small;
depending on how quietly the mouse is bled, 30-150 11L of plasma can be obtained.
The amount of prednisolone in these samples is below the detection limit of current
HPLC methods. The ECNI-MS method using MO-TMS derivatives may work, but
MO-TMS derivatives can be difficult to prepare and suffer losses due to the two-step

reaction procedure and possible formation of multiple products.

232

The oxidation of prednisolone followed by analysis by GC-ECNI—MS was
chosen for the determination of prednisolone in mouse plasmas. The oxidation
methodology is based on converting prednisolone to l,4-androstadien-3,11,17-trione
using the PCC oxidant. A pure sample of this compound has a detection limit of 170
pg (mlz 298, S/N = 11) on the JEOL AX505 mass spectrometer. The ECNI spectrum
of this compound consists of only one peak at mlz 298. Unfortunately, another
corticosteroid drug, prednisone, is also oxidized to the same product under the same
conditions (see Figure 1.7), making it impossible for the assay to distinguish between
the two corticosteroids. To be pharmacologically active, prednisone must first be
converted to prednisolone by the reduction of the ll-carbonyl group in the liver (28).
This interconversion is thought to occur rapidly (29), so the bioavailability of the two
drugs is thought to be similar. There is no evidence that prednisolone is converted to
prednisone in viva; however, the possibility cannot be disregarded in assay
development. The inability to distinguish between prednisolone and prednisone was
communicated to the Fraker group. Because they were interested in the total amount
of prednisolone dosed, the distinction between the two corticosteroidal drugs in viva
was not important.

Isotopically-labelled prednisolone was not available for the internal standard,
so a structurally-similar compound was substituted instead. A corticosteroid drug was
used because there is no possibility of its occurring naturally in the mouse. Because
oxidized dexamethasone coelutes with l,4-androstadien-3,11,17-trione, on a DB-l
column, it was not used. Flumethasone had no coelution problems, so it was chosen as

the internal standard.

Experimental
The extraction procedure is based on the work of Kayganich and co—workers

(3) with a few modifications. The manufacturer of the C-18 solid-phase extraction

233

columns was changed to Burdick & Jackson because these columns were already
available. For the PCC clean-up step, 200-mg silica solid-phase extraction columns
were used rather than the 500mg columns because the smaller amount of silica are
more easily wetted with less solvent and less time with no loss of efficiency. In the
initial plasma extraction step, the selective wash (30% methanol in water) and the
eluent (38% methanol in hexane) were changed to the more general water wash and
methanol eluent to recover all steroids. For the same reason, ethyl acetate was
substituted for the 5% acetone in methylene chloride eluent in the PCC clean-up step.

111mm. ECNI mass spectral data were obtained on a JEOL JMS-
AX505H mass spectrometer (Peabody, MA) coupled with a Hewlett Packard 5890 gas
chromatograph (Palo Alto, CA). The following parameters were used: source
temperature, 200°C; ionization current, 300 HA; electron energy, 200 eV; modulating
gas, methane; source pressure, 2 x 10‘5 torr; and accelerating voltage, 3 keV. A J&W—
15 m x 0.25-mm id x 0.25-tun DB-1701 column (Folsom, CA) was directly inserted
into the mass spectrometer ionization chamber. Helium was used as the carrier gas.
The GC oven was held at 140°C for 1 min, then ramped to 260° at 40°lmin, followed
by a 4°lmin ramp to 280°C. The splitless injector was held at 260°C, and the transfer
line at 260°C.

Selected ion monitoring (SIM) was used to analyze the plasma extracts. Ion
currents at mlz 298.2 (corresponding to M’ for oxidized prednisolone) and 328.2
(corresponding to [M-HF]' for oxidized flumethasone) were monitored using magnetic
field switching SIM.

WW. Mouse plasma samples were spiked with 3 ng
of flumethasone in 10 uL of methanol. The ZOO-mg C18 solid-phase extraction
columns were conditioned with 5 mL of 8 M aqueous urea, followed by washes with 5
mL of methanol and 10 mL of distilled water. The plasma samples were applied,
followed by 4 mL of distilled water. Prednisolone was eluted with 4 mL of methanol.

234

W. Following the removal of the methanol under
nitrogen, 1 mL of methylene chloride, approximately 10 mg of anhydrous sodium
acetate, and an excess of pyridinium chlorochromate finely ground with Celite (about
1:3 w/w) was added to the dried residue. The reaction mixtures were stirred and
heated to 60°C for 6 h.

WW. Silica solid-phase extraction columns (200 mg)
were conditioned with several column volumes of methylene chloride. After the
addition of the reaction mixtures, the columns were washed with 2 mL of methylene
chloride. The oxidation product was eluted with 4 mL of ethyl acetate. 'After
evaporation under nitrogen, the residue was reconstituted in 75 [LL of ethyl acetate; 4-6
mL of this solution were required for analysis by GC-ECNI-MS.

MW. To blank mouse plasma was added 3 ng
of the internal standard, flumethasone. Following this, additions of 0, 1, 5, 10, and 20
ng of prednisolone were made. Each standard was prepared in duplicate. The

extraction and oxidation steps were the same as above.

Preliminary Study

In order to verify the extraction and oxidation steps for the determination of
prednisolone in mouse plasma, spiked mouse plasma samples and one dosed mouse
sample were analyzed. Four aliquots of mouse plasma (110 uL) were spiked with no
steroids, 55 ng prednisolone, 11 ng prednisolone, or both 11 ng flumethasone and 11
ng prednisolone. The real mouse sample was spiked with 11 ng flumethasone only.
All samples were analyzed with a nonpolar DB-l column (IS-m x 0.25-mm x 1.0—um,
J&W, Folsom, CA). All samples were easily detected by this method. Using a one-

point estimation, the concentration of the real sample was 17 ng.

235

Mouse Plasma Sample Set #1

Two sets of mice were studied for the immunosuppression effect. The first set
did not result in as great an effect as the second. The first set of mouse plasmas was
used to assess the oxidation methodology. _

Blank mouse samples from other experiments were spiked with 0, 10, 20, and
50 ng of prednisolone. All samples, included those to be analyzed, were spiked with
7.5 ng of flumethasone. After analysis by GC-ECNI-MS with a DB-l column, the
results are as summarized in Table 5.13, column 3. Mice numbers one through eight

were controls; the rest were dosed with prednisolone.

Table 5.13: Determination of the Amount of Prednisolone in Mouse Plasma

Sample Set #1
Mouse # Amount of Plasma Amount of Prednisolone Amount of Prednisolone
uL ng, DB-l column ng, DB-1701 column
1 65 1 0.15
2 100 <LOD NA
3 90 <LOD <LOD
4 100 1 1 NA
7 65 <LOD <LOD
8 100 4 0.9
9 40 <LOD 0.1
12 60 <LOD <LOD
13 85 6 3.9
15 100 ' <LOD 0.05

The standard curve was fairly linear with a correlation coefficient of 0.994.
The problem with the analysis was the real samples were at very low concentrations
where there were no points on the standard curve. The standard curve needed some

improvement with respect to reproducibility and detection limits.

Possible Interferences
The blank value for 1,4-androstadien-3,11,17-trione was higher than expected.

At very low concentration values, the contamination in the blank at the retention time

236

of 1,4-androstadien-3,l 1,17-trione becomes very important. Possible biological
interferences were investigated. Corticosterone, an endogenous steroid, oxidizes to 4-
androsten-3,1 1,17-trione, which has a molecular weight of 300. Under ECNI
conditions, 4-androsten-3,11,17-trione is a low-response compound. Usually, the mass
spectrum of low-response compounds is dominated by a peak at [M-H] '. In this case,
[M-H2]‘ is the most intense peak in the spectrum. This peak occurs at m/z 298, the
same mass as oxidized prednisolone. On a nonpolar DB-l column, these compounds
coelute. Corticosterone has a level of 10 ug/dL in mouse plasma. Interestingly, the
peak area of the blank corresponded to the response expected from about 10 [lg/(1L of
4-androsten-3,1 1,17-trione.

To remove the corticosterone contamination, two methods are possible. One is
to use a selective extraction procedure by choosing the washes, eluents, and column
type of the plasma extraction step and/or PCC clean-up step to exclude corticosterone.
This should be possible because corticosterone and prednisolone differ by a 1,2-double
bond and a 17-hydroxy group. However, optimizing the chromatography can be a very
time-consuming procedure; in fact, several different extraction procedures attempted
were unsuccessful. The other, easier method of overcoming the contamination
problem is to separate 4—androsten-3,11,17-trione (oxidized corticosterone) and 1,4-
androstadien-3,11,17-trione (oxidized prednisolone) chromatographically on the GC.
More polar columns should separate these closely-related compounds with better
resolution. A 50-m OV-17 column (50% phenyl/50% methyl) of unknown (and
therefore questionable) history did not separate the two steroids, but reasonable
separation was obtained on a 10-m DB-1701 (14% cyanopropylphenyl/86% methyl)
column. For the majority of the mouse samples, background from corticosterone was
not a problem. The high blank value still persisted.

Other endogenous steroids in the mouse were examined as possible

interferences. The biosynthesis of adrenocorticosteroids is shown in Figure 5.7.

237

Progesterone, deoxycorticosterone, 17 a—hydroxyprogesterone, and ll-deoxycortisol
all oxidize to 4-androsten-3,17-dione (MW = 286), which has a different retention time
and mass spectrum than 1,4-androstadien-3,11,17-trione. Cortisol and corticosterone
oxidize to 4-androsten-3,ll,17-trione, which has been proven not to cause the
interference. Aldosterone oxidizes to 4-androsten-3,11,17,18-tetrone, which has a
molecular weight of 314, so this is not a potential interference. 18-
Hydroxycorticosterone could possibly oxidize to 4-androsten-18-ol-3,11,17-trione. If
this compound loses a water molecule, it could have a peak at m/z 298 in its ECNI
mass spectrum. To check this possibility, a sample of 18-hydroxycorticosterone
(Ikapharrn, Ramat-Gan, Israel) was oxidized using PCC for 6 h. The molecular weight
of the reaction product was 305, possibly corresponding to the loss of the group at C-
17 from the starting material. 18-Hydroxycorticosterone is not an interference.

After the reanalysis of the spiked prednisolone samples and the mouse sample
set #1 using the DB-1701 column, the standard curve did not improve in linearity, but
the LOD was now 0.04 ng. The results from this study are in Table 5.13, column 4.

Lower concentration values can now be detected.

Mouse Plasma Sample Set #2

The second set of mouse samples consisted of eight control mice (#9-16) and
eight mice dosed with prednisolone (#1-8). This sample set was considered to be the
one which proved the immunosuppression theory. For the standard curve, blank
plasma samples were spiked with 0, 1, 5, 10, and 20 ng of prednisolone. All samples
were spiked with 3.0 ng of flumethasone before extraction. Each point on the standard
curve and each of the actual samples were analyzed by GC-ECNI-MS in duplicate.
The standard curve had an equation of y = 0.0535 + 0.0604x with a correlation
coefficient of 0.998 (see Figure 5.8). The LOD was 0.3 ng. The results from this

study are summarized in Table 5.14.

238

 

 

 

Figure 5.7: The biosynthesis of andrenocorticosteroids. Adapted from Onen, J. M.;
Neuhaus, O. W. Human Biochemistry, Mosby: St. Louis, 1982, p. 618.

239

Table 5.14: Determination of the Concentration of Prednisolone in Mouse Plasma

Sample Set #2
Mouse # Amount Plasma Amount Prednisolone Concentration Prednisolone
uL ng ng/mL
1 95 7.4 78 :1: 2
2 75 0. 34 4.5 :l: 0.2
3 30 <LOD <LOD
4 135 <LOD <LOD
5 40 0.055 1.4 :l: 0.6
6 100 0.81 8.1 :l: 3.5
7 55 <LOD <LOD
8 85 <LOD <LOD
9 120 <LOD <LOD
10 45 <LOD <LOD
1 1 45 <LOD <LOD
. 12 114 <LOD <LOD
13 165 <LOD <LOD
14 90 <LOD <LOD
15 95 <LOD <LOD
16 65 0.064 0.98 :I: 0.12

Of the eight mice dosed with prednisolone, only four gave a response above
background, with an average of 4.7 :l: 3.4 ng/mL plasma (disregarding the contribution
from mouse #1). The large standard deviation could occur from differences in
bioactivity of prednisolone between individual mice. Mouse #1 had an abnormally
high concentration of prednisolone (78 ng/mL) in its bloodstream. Most of the control
mice had no prednisolone in their bloodstream, with the exception of mouse #16. This
could be a false positive, a problem inherent in the design and execution of the assay.

Four more mouse plasma samples were analyzed separately. These samples
were drawn 42 hours after prednisolone pellet implantation. Another standard curve,
spiked as previously described, was analyzed along with the four samples. One of the
samples did not have a measureable amount of prednisolone. The average amount .of
prednisolone in the remaining three samples was 8.3 i 0.9 ng/mL. This result was low
as compared to the experimenters' expected value of on the order of 100 ng/mL.

240

 

 

 

 

1.5
8
g 1.0
O.
U)
G)
(I:
d)
.2
g 0.5
0)
(I
000 I—' I 1 I T 1 Y I I I U I T I I 1 I T r
o 4 a 12 16 20

ng prednisolone

Figure 5.8: Standard curve for the determination of prednisolone in mouse plasma.
The equation of the line is y = 0.0535 + 0.0604x with a correlation coefficient of
0.998. The LOD was 0.3 ng.

Conclusions

Even though this assay cannot distinguish between prednisolone and
prednisone, it can detect prednisolone at levels found in mouse plasma. The limit of
detection of this assay was between 0.3 and 0.5 ng/mL, which is just slightly better
than competing HPLC methods that can distinguish between prednisolone and
prednisone.

The oxidation methodology for the determination of prednisolone could be
improved by improving the LOD. An isotOpically-labelled internal standard may help.
More selective plasma extractions and/or more selective PCC clean-ups could decrease
the biological background sufficiently to allow more sensitive detection of
prednisolone. The yield of the PCC reaction also could be improved. After a 6-h
reaction time, the yield is only 47%. Perhaps longer reaction times or. a different

oxidant are needed.

References

10.
11.
12.

13.

14.
15.

16.
17.
18.
19.
20.

21.
22.
23.

Her, G. R.; Watson, J. T. Anal. Biochem. 1985, 151, 292.
Watson, J. T.; Kayganich, K. Biochem. Soc. Trans. 1989, I 7, 254.

Kayganich, K. A.; Watson, J. T.; Kilts, C.; Ritchie, J. Biomed. Mass Spectrom.
1990, 19, 341.

Bromley, S. C. J. Exp. Zool. 1977, 201, 101.
Griffith, W. P.; Ley, S. V. Aldrichemica Acta 1990, 23, 13.

Griffith, W. P.; Ley, S. V.; Whitcombe, G. P.; White, A. D. Chem. Soc., Chem.
Commun. 1987, 1625.

Sample, R. H. B.; Baenziger, J. C. In Analytical Aspects of Drug Testing; D.
Deutch, ed.; Wiley: New York, 1988, pp. 247-272.

Courtesy of Bev Chamberlin.

Carroll, B. J. et al. Arch. Gen. Psychiatry 1981, 38, 15.
Lowry, M. T.; Meltzer, H. Y. Biol. Psychiatry 1987, 22, 373.
McDowall, R. D. J. Chromatogr. 1989, 492, 3.

Esteban, N. V.; Yergey, A. L.; Liberato, D. J.; Loughlin, T.; Loriaux, D. L.
Biomed. Mass Spectrom. 1988, 15 , 603.

Gaskell, S. J. In Methods of Biochemical Analysis, Volume 29; D. Glick, ed.,
Wiley: New York, 1983, pp. 385-434.

Garland, W. A.; Barbalas, M. P. J. Clin. Pharmacol. 1986, 26, 412.

Miller, J. C.; Miller, J. N. Statistics for Analytical Chemistry; Wiley: New
York, 1986.

Barnett, V.; Lewis, T. Outliers in Statistical Data; Wiley: New York, 1984.
Kayganich, K. A. Ph. D. Dissertation, Michigan State University, 1989.
Cairns, T.; Siegmund, E. G.; Stamp, J. J. Mass Spectrom. Rev. 1989, 8, 93.
Dikkeschei, L. D.; et al. J. Clin. Immunoassay 1991, I4, 37.

Voetberg, B. J.; Garvy, B. A.; Mayer, H. K.; Watson, J. T.; Fraker, P. J.
Endocrinol" submitted.

Frey, F. J .; Frey, B. M.; Benet, L. Z. Clin. Chem. 1979, 25, 1944.

Rocci, M. L.; Jusko, W. J. J. Chromatogr. 1981, 224, 221.

Cheng, M. H.; Huang, W. Y.; Lipsey, A. 1. Clin. Chem. 1988, 34, 1897.
241

24.
25.
26.
27.

28.

29.

242
Vichyanand, P. et al. J. Allergy Clin. Immunol. 1989, 84, 867.
Colbum, W. A.; Buller, R. H. J. Pharmacokin. Biopharm. 1975, 3, 329.
Matin, s. 13.; Amos, B. J. Pharm. Sci. 1978, 67,923. '

Houghton, E.; Teale, P.; Dumasia, M. C.; Wellby, J. K. Biomed. Mass
Spectrom. 1982, 9, 459.

Frey, B. M.; Walker, C.; Frey, F. J.; de Week, A. L. J. Immunol. 1984, 133,
2479.

Gambertoglio, J. G.; et al. Kidney Intern. 1982, 21, 621.

Chapter 6: Conclusions

One possible method for the analysis of steroids is by chemical oxidation
followed by analysis by GC-ECNI-MS. In this work, the ECNI characteristics of
various cyclic molecules was studied, in addition to improving the oxidation
methodology for steroidal drugs.

Electron capture techniques improve the sensitivity of certain compounds. For
enhanced ECNI response within a steroid nucleus, the following structural features are
important: a cross-conjugated system with a neighboring group, such as 1,4-dien-3,11-
dione; a linearly-conjugated system, such as 4-en-3,6-dione; a fluorine atom alpha to a
carbonyl group in a highly-conjugated steroid; or an epoxide alpha to a carbonyl
group. The trends in ECD response mimic those for ECNI; however, the actual values
differ. Both high- and low-response ketosteroids have similar E1 and CI responses.
For low-response steroids, the ECNI response is on the order of the corresponding EI
response.

Size considerations are also important. The same structural features that
increase the response in steroids also increase the response for cyclohexanones and
bicyclic compounds. The response for steroids is greater than those for either of these
classes of compounds because steroids, being larger molecules, can more effectively
distribute excess energy imparted by electron capture.

Usually, the presence of a peak for [M-H]' indicates a low-response compound,
except for epoxysteroids. For saturated ketosteroids, the hydrogen alpha to a carbonyl
is lost. In the case of epoxysteroids, an [M-H]' peak will be present if a hydrogen on
the carbon beta to the epoxide is available.

The reduction potentials in acetonitrile can be assumed to be indicative of the

trends of electron affinity. In a series of ketosteroids, a half-wave reduction potential

243

244
more positive than -2.0 V indicates a high-response compound, except in the cases of
epoxysteroids.

The choice of reagent gas can enhance the electron-capture response.
Ammonia selectively increases the responses of compounds with [M-H]' peaks. A
better choice for reagent gas for ECNI studies is isobutane, which is nonselective and
has a better response than methane.

PCC oxidation is the method of choice for the chemical oxidation of steroids;
alternative oxidative methods were also investigated. The TPAP oxidation had a
comparable yield and reaction time as the PCC oxidation. Electrochemical oxidation
had a very poor yield due to fouling of the electrode. Subsequent experiments that
place the electrode directly into the mass spectrometer source also did not work.

An improvement of the oxidation technique may be possible using HPLC-
electrochemisu'y-thermospray mass spectrometry. Using this technique (1), the
oxidation products of oxymethalone were analyzed. Not many steroids were amenable
to this technique; for example, testosterone did not oxidize at all. Perhaps the highly
conjugated steroids like dexamethasone or prednisolone could be successfully
analyzed by this technique.

The oxidation methodology was extended to the analysis of other real-world
samples. Dexamethasone in human plasma was analyzed and compared to RIA
values. The GC-MS values did not correlate with the RIA values and also tended to be
less precise. Using a structurally-similar internal standard, the limit of detection was
0.03 ng/mL, which is better than the limit of detection (0.25 ng/mL) obtained in a
study using an isotopically-labelled internal standard.

The oxidation methodology is a good method for determining prednisolone
content of mouse plasma if prednisolone is not present. The LOD is 0.51 ng/mL,
which is on the order of competing methodologies such as HPLC. For the samples

analyzed, the dosed mice had about 2 ng/mL prednisolone in their bloodstream.

245

Based on the ECNI response studied and what is known about the oxidation
methodology as well as other electron-capturing derivatives, predictions can be made
about the extension of ECNI methodologies to other compounds. Some of these
predictions are based on preliminary work contained in this dissertation.

The presence of epoxides increases the relative responses of ketosteroids.
Many trichothecine molecules also contain epoxide groups. However, the presence of
an epoxide group does not seem important for the overall ECNI responses of
uichothecines. The utility of epoxides for enhanced response may be confined to the
steroid nucleus and is not as applicable to other biological molecules.

Sensitive assays are needed for anabolic steroids because of problems with
abuse. Most anabolic steroids are not amenable to oxidation techniques because they
are analogs of testosterone. Testosterone and its analogs oxidize to 4-androsten-3,17-
dione, a compound with poor electron-capture response. Other anabolics have a 17a-
methyl group, which prevents the oxidation at this position. The only anabolic that
seems promising for assay development is fluoxymesterone. This steroid can be
oxidized to a compound with the 1,4-dien-9a-fluoro-3J l-dione system, which has
excellent electron-capture properties. Other anabolic steroids must be derivatized for
ECNI analysis. Based on the ECNI studies of 5a—androstan-3B,l7B-diol, HFB
derivatives are probably the best choice.

Many steroidal drugs containing fluorine can be oxidized to structures for
which assay development seems promising. The oxidized products of flumethasone,
fluorandrenolone, and fluocinolone will have the l,4-dien-6a,90t-fluoro-3,11—dione
system. Triamcinolone will oxidize to a steroid containing the 1,4-dien-9a-fluoro-
3.11-dione system as well as two adjacent carbonyls at C-16 and C-17 in the D ring.
Preliminary studies of the oxidation of triamcinolone have suggested that the PCC
conditions described in this dissertation do not work well; perhaps either a different

oxidant or more stringent conditions are necessary.

 

246
Other candidates for advantageous assay development are steroids with the 5-
en-3-ol moiety. This functionality oxidizes to 4-en—3,6-dione, which, regardless of the
other substituent on the molecule, captures an electron readily. Compounds with the
5-en-3-ol structure include sterols such as jervine (an azasteroid), cholesterol, and

ergosterol (5,7,22-cholestatrien-3-ol).

 

References

1. Getek, T. A.; Tabor, J. E.; Morgan, M. M. 39th ASMS Conference on Mass
Spectrometry and Allied Topics, Nashville, TN, 1991.

247

 

"I7'11lll'lillllfllllll'lf