IMPROVING THE SEPARATION OF DRUG ISOMERS USING CHEMICAL MODIFIERS IN HIGH-FIELD ASYMMETRIC WAVEFORM ION MOBILITY SPECTROMETRY

By

MICHAEL SHENMING WEI

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2019

© 2019 Michael Shenming Wei

To my friends and family. Thanks for dealing with me when I needed it.

ACKNOWLEDGMENTS

I would not be able to present the following research without the advice and support of my friends, family, and colleagues. First and foremost, I want to thank my advisor Dr. Rick Yost for all of the guidance, opportunities, advice, and stories that I got to experience over my time at UF. I would also like to thank the members of my graduate committee: Dr. Kari Basso, Dr. Benjamin Smith, Dr. Timothy Garrett, as well as my previous committee member Dr. Nicholas Polfer. Each motivated and inspired me along my time as a graduate student. Thanks also go out to Dr. Jodie Johnson, whom I learned so much about mass spectrometry instrumentation and troubleshooting from. I want to extend recognition to Dr. George Dubay at Duke University, who first talked mass spectrometry to me and inspired me to pursue a doctorate in analytical chemistry.

I want to thank my fellow Yost group members, most notably Robin Kemperman, whom I shared many hours with talking about science and bantering about other topics.

Robin has been a critical colleague in my high-field ion mobility research and a great friend. Additional thanks go out to Louis Searcy, Kevin Davis, and my undergrad

Christopher Gongar, whom have each helped me out in a huge way during my time in the Yost Group. I also want to give my gratitude to Dr. Elizabeth Dhummakupt for her mentorship during my first year, Dr. Michael Costanzo for his mentorship and collaboration during my second and third, and Dr. Jared Boock for also being a nerd like me (:smileyface:).

I want to thank my family for supporting me throughout the years. To my father, I want to give my thanks for the advice and help on so many little life things – car maintenance, cooking, insurance, etc. To my mother, I want to give my thanks for

4

always welcoming me back home with open arms and a hearty meal. To my sister, I want to give my thanks for being someone that I can share stories and experiences with about college.

Finally, I want to thank my friends, who challenged me intellectually and with whom I spent many long night gaming with. In particular, I would like to recognize

Rodger Zou, Wayne You, and Cullen Wallace. Rodger was instrumental in helping me prepare for my qualification exam presentation and continues to inspire me to learn new skills. Wayne administers the online chat program that has helped me feel connected with so many college friends that I would otherwise not be able to interact with. Cullen has helped me mature so much as a person. From Cullen, I came to truly appreciate the value of keeping one’s eyes on the goal, of seeking out information to make the best possible decisions, of dedication when times get tough, and of practice making perfect.

5

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 4

LIST OF TABLES ...... 8

LIST OF FIGURES ...... 9

LIST OF ABBREVIATIONS ...... 11

ABSTRACT ...... 12

CHAPTER

1 INTRODUCTION ...... 14

Ion Mobility and High-Field Ion Mobility Spectrometry ...... 14 Principles of High-Field Asymmetric Waveform Ion Mobility Spectrometry ...... 14 FAIMS Instrumentation ...... 17 Modifiers for Improving FAIMS Performance ...... 20 Ionization and Mass Analysis ...... 21 Electrospray Ionization ...... 21 Quadrupole Ion Trap Mass Spectrometry...... 23 Overview of Research...... 24

2 EFFECTS OF SOLVENT VAPOR MODIFIERS FOR THE SEPARATION OF OPIOID ISOMERS IN FAIMS-MS ...... 32

Introduction ...... 32 Materials and Methods...... 34 Chemicals ...... 34 FAIMS-MS Instrumentation and Methods...... 34 Solvent Vapor Procedure ...... 35 Results and Discussion...... 36 FAIMS-MS Separation of Opioids ...... 36 Aprotic Solvents Versus Protic Solvents ...... 41 Conclusions ...... 44

3 FAIMS SEPARATION OF ANABOLIC ANDROGENIC STEROID EPIMER PAIRS USING CATION MODIFIERS ...... 57

Introduction ...... 57 Materials and Methods...... 61 Chemicals ...... 61 FAIMS Instrumentation and Methods ...... 62 Results and Discussion...... 63

6

Separation of Testosterone & Epitestosterone using Group 1 Cations ...... 63 Effect of Multimer Cation Complexes on FAIMS Separations ...... 65 Separation of Androsterone & Trans-androsterone ...... 68 Conclusions ...... 68

4 RESOLVING ANDROSTERONE ISOMERS AND ISOBARS USING CATION- MODIFIED FAIMS-MS ...... 77

Introduction ...... 77 Materials and Methods...... 79 Chemicals ...... 79 FAIMS Instrumentation and Methods ...... 80 Results and Discussion...... 81 Resolving Androsterone Epimers ...... 81 Resolving Non-Epimer Isomers and Isobars ...... 83 Effects of DF on CF Peaks ...... 87 Effects of Cation Size on CF Peaks ...... 90 Cation Concentration Effects ...... 92 Conclusions ...... 93

5 CONCLUSIONS AND FUTURE WORK ...... 106

LIST OF REFERENCES ...... 111

BIOGRAPHICAL SKETCH ...... 116

7

LIST OF TABLES

Table page

2-1 Average CF peak widths and standard deviations for carrier gas compositions...... 47

4-1 Ionic radii and electronegativities for the cation species used for the separation of androsterone isomers and isobars ...... 96

8

LIST OF FIGURES

Figure page

1-1 Graph representing an idealized waveform and ion motion for FAIMS ...... 26

1-2 Effects of applying a compensation voltage to the FAIMS waveform ...... 27

1-3 Example FAIMS analysis of morphine [M+H]+ ions ...... 28

1-4 Illustration of ion mobility behaviors with increasing electric field strengths ...... 28

1-5 Three-dimension and cross-section representations of planar, cylindrical, and micromachined chip-based FAIMS cell geometries ...... 29

1-6 Schematic depiction of an ESI source ...... 30

1-7 Mathieu stability diagram illustrating stable ion trajectories for a 3D quadrupole ion trap ...... 30

1-8 2D linear ion trap representation depicting the quadrupole rod assembly, direction of ion flow, and voltage application ...... 31

2-1 Chemical structures for four opioids including molecular formulae and weights ...... 48

2-2 Schematic for solvent vapor addition apparatus ...... 49

2-3 FAIMS spectra for a morphine solution and corresponding mass spectra acquired using dry nitrogen and acetonitrile vapor in nitrogen ...... 50

2-4 Plot of CFs with increasing DFs for a mixture of morphine and norcodeine ...... 51

2-5 Plot of CF shifts with increasing acetonitrile vapor concentration in nitrogen for four opioid solutions and related FAIMS spectra ...... 52

2-6 Plot of CF shifts with increasing vapor concentration for the m/z 286 ion from a morphine solution ...... 53

2-7 Plot of peak width with increasing vapor concentration for the m/z 286 ion from a morphine solution ...... 54

2-8 Plots showing CF shifts with respect to vapor concentration for acetonitrile, acetone, ethyl , methanol, and water for opioid standard solutions ...... 55

2-9 Plots of CF shifts with addition of vapor from a series of alkyl and a series of alcohols for morphine and norcodeine ...... 56

9

3-1 Chemical structures for two anabolic steroid epimer pairs, including molecular formulae and weights ...... 71

3-2 FAIMS spectra for the [M+H]+ ions of testosterone and epitestosterone and related mass spectra ...... 72

3-3 FAIMS spectra for the [M+Na]+ ions of testosterone and epitestosterone ...... 73

3-4 Overlaid FAIMS spectra for Group 1 cation adducts of testosterone ...... 73

3-5 FAIMS spectra for the [M+Na]+, [2M+Na]+, and [3M+Na]+ ions of testosterone and related mass spectra ...... 74

3-6 FAIMS spectra for the [3M+Li]+ ions of testosterone and epitestosterone and related mass spectra ...... 75

3-7 FAIMS spectra for the [M+K]+ ions of androsterone and trans-androsterone ..... 76

4-1 Chemical structures for five androsterone isomers and isobars, including molecular formulae and weights ...... 97

4-2 Chemical and 3D structures of the epimeric steroids androsterone, etiocholanolone, and trans-androsterone ...... 98

4-3 FAIMS spectra for the [M+K]+ ions of androsterone isomers and isobars at DF 250 ...... 99

4-4 FAIMS spectra for the [3M+Li]+ and [M+Li]+ ions of androsterone isomers ...... 100

4-5 FAIMS spectra for the [M+Na]+ ions of androsterone isomers and isobars at DF 250 Td...... 101

4-6 Plot of CFs with increasing DFs for the [M+K]+ ions of androsterone isomers and isobars ...... 102

4-7 Plot of CFs with increasing DFs for the [M+Na]+ ion of androsterone ...... 103

4-8 Plot of CF shifts with increasing cation size for monomer adduct ions of androsterone at DF 250 Td ...... 104

4-9 Plot showing the change in CF peak signal intensities at various cation concentrations ...... 105

10

LIST OF ABBREVIATIONS

AC Alternating Current

CV or CF Compensation Voltage / Compensation Field

DC Direct Current

DV or DF Dispersion Voltage / Dispersion Field

ESI Electrospray Ionization

FAIMS High-Field Asymmetric Waveform Ion Mobility Spectrometry

FWHM Full-Width at Half Maximum

GC Gas chromatography

IMS Ion Mobility Spectrometry

Kh Ion mobility coefficient at high electric field

LC Liquid chromatography

[M+H]+ Protonated Monomer Ion

[M+Na]+ Sodiated Monomer Ion

MS Mass Spectrometry

MS/MS Tandem mass spectrometry m/z Mass-to-charge ppm Parts per million

Rs Resolution

Td Townsends (unit)

V Volts (unit)

[2M+Na]+ Sodiated Dimer Ion

11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

IMPROVING THE SEPARATION OF DRUG ISOMERS USING CHEMICAL MODIFIERS IN HIGH-FIELD ASYMMETRIC WAVEFORM ION MOBILITY SPECTROMETRY

By

Michael Shenming Wei

May 2019

Chair: Richard A. Yost Major: Chemistry

Ion mobility spectrometry is an emerging technology that offers a complementary separation step and a number of advantages over traditional separation techniques.

High-field asymmetric waveform ion mobility spectrometry (FAIMS) is a variation of conventional ion mobility spectrometry that utilizes an asymmetric alternating electric field to filter ions by the difference between their ion mobilities at high and low fields as they traverse down an analytical cell. FAIMS separations can be performed rapidly, typically requiring an analysis time of ten to a hundred milliseconds. In addition, FAIMS can be operated at atmospheric pressure and is significantly more portable than many other separation methods, making it attractive for use as a standalone chemical detector or as a complimentary separation step in chromatography and/or mass spectrometry. FAIMS has been applied across a variety of fields ranging from chemical agent detection in security and forensics to biomedical applications.

However, FAIMS generally exhibits lower resolving power compared to more traditional separation techniques. As an emerging technology, innovations in electronics, instrumentation, and methodology have dramatically improved the

12

sensitivity and resolving power of FAIMS since its inception and continue to be explored in order to further improve performance. Recently, the controlled addition of various chemical modifiers to FAIMS methods have been shown to significantly enhance resolving power, enabling the separation of small isomeric compounds.

The work presented in this dissertation investigates and characterizes effects of chemical modifier addition in FAIMS methods coupled with mass spectrometry toward the separation of isomers from several drugs of abuse. Addition of solvent vapor to the

FAIMS carrier gas is utilized to improve the separation of several opioid isomers. A series of alcohols and alkyl acetates are utilized as solvent vapors to explore FAIMS peak shifts and changes in resolving power. A summary and conclusion about the potential applications of FAIMS with chemical modifiers is provided at the end, along with future research directions for follow-up investigations.

13

CHAPTER 1 INTRODUCTION

Ion Mobility and High-Field Ion Mobility Spectrometry

Principles of High-Field Asymmetric Waveform Ion Mobility Spectrometry

High-field asymmetric waveform ion mobility spectrometry (FAIMS) is an atmospheric-pressure gas-phase separation technique that utilizes an asymmetric alternating electric field to filter ions by the difference between their ion mobilities at high and low fields as they are carried down a separation cell. FAIMS was developed in the

1990s and is based on similar principles used to separate gas-phase ions in conventional ion mobility spectrometry (IMS).1–3

When an ion in a gas is subjected to an electric field, it will move in the direction of field lines with a drift velocity vd equal to the product of its ion mobility coefficient K and the strength of the electric field E, as described in Equation 1-1:

푉 푣 = 퐾 × 퐸 = 퐾 × (1-1) 푑 퐿 where the electric field E can be described as the voltage difference V across a gap length L. The ion mobility coefficient K is dependent on the size and charge of the ion.

When the applied electric field strength is low (about 200 V/cm), the mobility coefficient

K is independent of the electric field strength E. In conventional drift tube IMS, ions are pulsed into the drift tube and can be separated from each other by differences in their drift velocities. As long as the electric field is held constant, these differences in drift velocities correspond to differences in ion shape and size.

14

However, when the applied field strength exceeds a certain threshold (about

10,000 V/cm), K becomes dependent on E in a nonlinear manner.1,2,4 This function is commonly approximated using Equation 1-2:

퐸 퐸 2 퐸 4 퐾 = 퐾 [1 + 훼 ( )] = 퐾 [1 + 훼 ( ) + 훼 ( ) + ⋯ ] (1-2) ℎ 0 푁 0 2 푁 4 푁 where E is the magnitude of the applied electric field (V/cm), N is the gas number

3 density (molecules/cm ), Kh is the high-field mobility, K0 is the low-field mobility, and

α(E/N) is the alpha function. The alpha function represents an even power series with

2,5–9 coefficients α2, α4, etc. The units for the E/N ratio are commonly reported as either

V•cm2 or Townsends (Td), where 1 Td = 10-17 V•cm2. It should be noted that the physical mechanisms behind nonlinear mobility behavior at high electric field strengths is not currently known. As a result, the high-field dependence of Kh is approximated using the alpha function.

The most basic FAIMS spectrometer consists of two parallel plates with an alternating electric field applied orthogonal to the direction of gas and ion flow.1 A graphical represtation of an idealized FAIMS separation and waveform is presented in

Figure 1-1. The alternating electric field is generated by an asymmetric waveform composed of a shorter high-voltage portion and a longer lower voltage portion of the opposite polarity. The asymmetric waveform is designed such that the sum of the voltage-time products is equal to zero, as described in Equation 1-3:

푉1푡1 + 푉2푡2 = 0 (1-3) where V1 is the voltage of the higher electric field portion of the waveform, V2 is the voltage of the lower field portion, and t1 and t2 are their applied times respectively. By

15

convention, V1 is used to describe the magnitude of the FAIMS separation and is refered to as the dispersion voltage or dispersion field (DF).

As ions travel orthogonal to the applied electric field, they will experience two different mobilities within one waveform cycle, resulting in two different velocities as the ions alternate back and forth between the electrodes. For example, during the higher field portion of the waveform, the velocity that ions move towards one electrode is described with Equation 1-4:

푉 푣 = 퐾 퐸 = 퐾 1 (1-4) 1 ℎ,푉1 1 ℎ,푉1 퐿

Equation 1-5 describes the distance that these ions travel during the higher field portion:

푑1 = 푣1푡1 = 퐾ℎ,푉1퐸1푡1 (1-5)

Under the lone influence of the DF, an ion must experience little to no net displacement as a carrier gas flow pushes it through the FAIMS cell in order to be successfully transmitted. However, because of the nonlinear mobility dependence on electric field at high field strengths, most ions will be displaced towards one of the electrodes after one cycle of the waveform. Over time, these ions will be displaced further towards the electrode until they strike it and are eliminated. To successfully transmit ions that would normally be eliminated, a direct current offset voltage can be applied to the waveform to restore the ion’s net displacement to an acceptable range; this offset voltage is referred to as the compensation voltage or compensation field (CF) and illustrated in Figure 1-2.

Since different ions have different Kh, a mixture of ions can be selectively transmitted through a FAIMS cell by scanning a range of CF values. An example FAIMS analysis of protonated and sodiated morphine ions is presented in Figure 1-3.

16

The trends between changes in high-field ion mobility Kh and increasing electric field strength can be divided into three behavior categories: Type A, Type B, and Type

C. Figure 1-4 illustrates the general trend for each behavior type. For ions exhibiting

Type A behavior, an ion’s mobility increases with increasing field strength. For Type C behavior, an ion’s mobility decreases with increasing field strength. For Type B behavior, an ion’s mobility initially increases with field strength, but subsequently reverses to become decreasing with field strength. In FAIMS, disparate ions are separated based on differences between their high-field mobilities. It should be noted that nonlinear high-field mobilities typically only differ from linear behavior by 1-10%.

FAIMS Instrumentation

The FAIMS cell consists generally of two electrodes separated from each other to form a gap of uniform height across the length of the electrodes. Ions enter this gap, whereupon they are separated by the mechanisms previously described. The successfully transmitted ions exit the FAIMS cell for detection. FAIMS has been applied in standalone devices as well as an interfaced technique with mass spectrometry (MS), where it either adds an additional separation step or serves as an alternative to a more time-consuming technique (e.g. gas and liquid chromatography).

Since the technique’s inception in 1993, several different geometries have been developed for the FAIMS electrodes. Figure 1-5 illustrates several notable geometries for the context of this work. Buryakov et al. developed the first and most basic FAIMS cell geometry, which consisted of two parallel flat plate electrodes (Figure 1-5a).1 FAIMS conducted with this geometry is referred to as planar FAIMS or differential mobility spectrometry (DMS) and continues to be widely employed for a variety of applications.3,7,10–18

17

In 1995, Carnahan et al. demonstrated that FAIMS could be achieved using two concentric cylinder electrodes, one inside the other (Figure 1-5b). Using curved electrode geometries was found to improve ion transmission through the cell by electrostatic focusing due to the nonuniform electric field created in the gap.19 However, the same nonuniform field can also lead to a loss in resolving power, as the strength of the electric field varies at different points between the electrodes.3 This means that for any particular CF value during a cylindrical FAIMS separation, uncontrolled variations in the electric field strength result in a wider range of mobility differences and negatively affects CF peak shape. Nonetheless, the improvement in ion transmission makes curved FAIMS geometries competitive with planar geometries, and development of both geometries continue to this day.

One particular variation of the cylindrical geometry changes the direction of gas and ion flow from parallel with the concentric electrodes into perpendicular to the electrodes (Figure 1-5c). First, the ions are flowed perpendicularly through the outer electrode by an entrance hole and subseqeuently pass over or under the inner rod electrode while undergoing FAIMS separation in the gap. Finally, the ions leave by another hole machined into the outer electrode. This perpendicular cylindrical geometry allows for easier interfacing between an ionization source and a mass spectrometer, while still providing the same general trade-offs as concentric cylindrical FAIMS.

Thermo Fisher Scientific (San Jose, CA) uses the perpendicular cylindrical geometry in its FAIMS systems and is currently the only commercial supplier of curved FAIMS geometries.

18

Since 2000, several micromachined planar FAIMS geometries have been developed for portable standalone detection of chemical agents and biomedical applications, capitalizing on a number of advantageous features.20,21 The most important of these features is the ability to perform FAIMS separations at atmospheric pressure, resulting in no requirement for vacuum pumping and significantly increasing the portability of the technique. Portable FAIMS systems also apply significantly lower voltages than larger systems (e.g. 250 V vs 5,000 V) in order to reduce the size of the power supply and waveform generator. In order to retain the high electric field strengths needed for FAIMS separations, micromachined geometries significantly shrink the size of the analytical gap compared to larger systems (e.g. 0.035 mm vs 2 mm). However, the dramatically reduced size of the micromachined FAIMS gap also dramatically reduces ion transmission. Owlstone Inc. (Cambridge, UK) addresses this problem in their micromachined FAIMS system with an array of parallel planar channels, where each channel functions as a miniature planar FAIMS cell (Figure 1-5d). The array combines the ion transmission through each FAIMS channel to improve overall transmission while retaining low voltage requirements. In addition to its standalone device, Owlstone Inc. also markets micromachined FAIMS systems that interface with mass spectrometers.

A number of different waveform types have been explored for FAIMS.3,22–24 While the ideal asymmetric waveform for FAIMS would be a rectangular wave as illustrated in

Figures 1-1 and 1-2, in practice an approximation using the sum of two sine waves is most commonly used. This is primarily due to practical limitations in electronics. The voltage potential shifts from the high-field to the low-field portions of the waveform need

19

to be generated in an extremely short timeframe, which typically requires high power comsumption and associated problems with heat dissipation. Instead, the vast majority of FAIMS waveform generators utilize a bisinusoidal wave described by Equation 1-6:

휋 2 sin 휔푡 + sin(2휔푡 − ) V(푡) = DV [ 2 ] (1-6) 3 where ω is frequency, t is time, and DV is the dispersion voltage – i.e. the peak voltage of the high-field portion. All FAIMS spectra acquired in this research used an Owlstone micromachined FAIMS system and bisinusoidal waveform.

Modifiers for Improving FAIMS Performance

FAIMS has found applications across a variety of fields, including homeland security, environmental monitoring, biomedical applications, and research disciplines such as proteomics and metabolomics.11,12,15,25–28 In some applications, FAIMS is even able to separate isomeric species, which has been demonstrated for proteins, lipids, and various small molecules.16,18,29–32 However, FAIMS generally exhibits lower resolving power than separations observed in conventional IMS or in chromatography.

As an emerging technology, a number of modifications unrelated to cell geometry and waveform have also been explored to improve FAIMS resolving power.6,9,25,29,31–36

In particular, the addition of chemical modifiers into the carrier gas has been shown to dramatically improve the resolving power and resolution between FAIMS analyte peaks.6,9,31,37,38 For example, addition of 1.5% methanol vapor improved the resolution between ortho- and para-phthalic acid isomers to 18.3, up from 0.67 in dry nitrogen.6 This improvement is theorized to result from ions dynamically clustering and declustering with neutral solvent molecules as the electric waveform alternates between

20

low-field and high-field regimes. The effects do not appear to follow a predictable trend and must be empirically determined for combinations of sample analyte and vapor compound, although previous work suggests a trend of greater CF shifts for an analyte ion added in comparison to dry nitrogen as larger members of a vapor compound class are, e.g. from methanol to 1-butanol.25 Small organic solvent molecules are the most commonly used vapor modifiers for this kind of work and include acetonitrile, methanol, isopropanol, and ethyl acetate. Water vapor has been explored, but is not commonly used for vapor modified FAIMS due to low vapor pressure and issues with uncontrolled arcing.6,37

Ionization and Mass Analysis

Electrospray Ionization

Electrospray ionization (ESI) is a liquid-phase ambient ionization technique that produces a charged aerosol by applying a high voltage potential to a liquid sample. This technique allows the ionization of analyte molecules directly from liquid solution to gas- phase ions. The development of ESI enormously expanded the range of compounds that could be analyzed using mass spectrometry, which previously struggled with the ionization of nonvolatile and high-weight molecules such as proteins. In addition, ESI simplified the ionization of liquid chromatography (LC) eluent and allowed LC to be interfaced with mass spectrometry on a broad range. As of now, ESI is one of the most versatile and widely utilized ion sources on commercial mass spectrometers, often including one or more heated drying gases to assist in desolvation.

To generate ions in ESI, the liquid sample is flowed through a thin metal capillary

(~0.1 to 0.5 mm) that is electrified to a high voltage potential relative to a counter- electrode (~3 to 6 kV). For this work, the counter-electrode will either be the inlet of the

21

mass spectrometer or the entrance cap of the FAIMS system. The combination of the high voltage applied to the liquid and the narrow size of the capillary results in a high electric field at the tip of the capillary. A Taylor cone forms at the capillary tip due to competition between Coulombic repulsion and the cohesive force of surface tension.

When the repulsive forces exceed surface tension, a fine jet produces the aerosol of small charged droplets that are accelerated towards the counter-electrode. As these droplets shrink from solvent evaporation, the charge density of the droplet increases until a critical threshold that is referred to as the Rayleigh limit. At this threshold,

Coulombic repulsion once again exceeds the surface tension, causing the droplet to split in smaller stable droplets. These droplets contrinue to undergo desolvation and subsequent divisions as they travel until they eventually release fully desolvated gas- phase analyte ions. An illustration of the ESI process is presented in Figure 1-6.39

The exact release mechanism of analyte ions from the charged droplets is not currently known; the two most widely accepted mechanisms are the the charge residue model (CRM) and the ion evaporation model (IEM).39 The CRM proposes that once the charged droplet is small enough, any remaining solvent molecules evaporate and leave behind a single charged analyte ion. In contrast, the IEM proposes that the electric field on the droplets becomes strong enough to eject analyte ions with a solvation shell of a few molecules, which are subsequently lost as the cluster enters the mass spectrometer inlet. More recently, several newer models have been proposed to explain the release mechanism, including the chain ejection model for polymers, “neutralizing the counter ion”, and “separating the ions”.39,40

22

Quadrupole Ion Trap Mass Spectrometry

Ion trap mass spectrometry broadly describes mass analyzers that are capable of trapping and storing ions for extended periods of time during mass analysis. Several examples of ion trap mass analyzers include ion cyclotrons, orbitraps, and quadrupole ion traps. Quadrupole ion trap mass analyzers utilize oscillating quadrupole electric fields to trap and store ions in stable trajectories; they are fundamentally similar in operation to that of quadrupole mass filters.

The first quadrupole ion trap consisted of a hyperbolic ring electrode and two cap electrodes, one on each opening of the ring; this configuration is referred to as the three-dimensional (3D) quadrupole ion trap.41 The oscillating hyperbolic electric fields are generated by applying an alternating current (AC) potential on the ring and direct current (DC) potential to the cap electrodes. Ions become confined in three-dimensional space and follow stable trajectories described by the Mathieu equations; stable ion motions must satisfy certain combinations of the parameters a and q that are described in Equations 1-7 and 1-8:42

푒 −푈 푎푧 = −2푎푟 = −16 ( ) ( 2 2) (1-7) 푚 푟0 휔

푒 푉 푞푧 = −2푞푟 = 8 ( ) ( 2 2) (1-8) 푚 푟0 휔 where e is the charge on the ion, m is the mass of the ion, U is the amplitude of the DC potential applied to the ring electrode (if any), V is the amplitude of the AC potential applied to the ring electrode, r0 is the radius of the ring electrode, and ω is the frequency of the AC potential. Solutions to the Mathieu equation that result in stable trajectories can be plotted to generate the Mathieu stability diagram and is presented in

23

Figure 1-7. Ramping the amplitude of the AC potential destabilizes ions in the axial direction, eventually leading to ions being ejected from the trap in order of mass.

Two-dimensional linear quadrupole ion traps were developed in 2002 and offered a number of advantages over the 3D quadrupole ion trap.43–45 The 2D linear ion trap consists of four hyperbolic rods arranged in a quadrupole configuration reminiscint of a quadrupole mass filter with two ion lenses at each end. Opposite pairs of rods are connected electrically, and AC potentials are applied to each pair. Additional DC potentials are applied to the entrance and back lenses. Ions are introduced into the trap through one of the ion lenses and subsequently confined radially by satisfying the

Mathieu equation and axially by the DC potentials applied the two lenses. Mass analysis is done by ejecting ions radially through exit slots machined into one pair of rods. An illustration of the 2D linear ion trap is presented in Figure 1-8.

In comparison with 3D ion traps, 2D linear ion traps feature higher ion capacities and trapping efficiencies because of larger trap volume and less space charging.46,47 In addition, 2D linear ion traps featured higher detection efficiencies by placing an ion detector at each exit slots. In contrast, 3D ion traps only featured one ion detector at one exit hole, as the other hole was used exclusively for ion introduction. All mass spectra in this research were acquired using a Thermo Scientific LTQ XL 2D linear ion trap mass spectrometer.

Overview of Research

This dissertation presents the research applying FAIMS-MS towards the separation of small isomeric drug molecules and investigating the effects of chemical modifiers on the FAIMS separation of these compounds. Chapter 2 applies FAIMS-MS with vapor modifiers to separate isomers of morphine and investigates the effects of

24

various solvent vapors on the FAIMS separation of isomers. Chapter 3 investigates the separation of anabolic steroid epimers by monitoring their cation adduct species.

Chapter 4 characterizes the effects observed when using cation-modified FAIMS for the separation of five androsterone isomeric and isobaric compounds. Lastly, Chapter 5 summarizes the results of the aforementioned research, provides conclusions about the application of FAIMS with chemical modifiers, and offers potential future directions for follow-up investigations.

25

Figure 1-1. Graph representing an idealized waveform and ion motion for FAIMS. The asymmetric waveform (top) is applied to one electrode, while the other is held at ground. Ions oscillate between the electrodes and will be eliminated by striking an electrode (red & purple dotted lines) unless the ions have no net displacement (green dotted line). Adapted from Costanzo 2015.48

26

Figure 1-2. Effects of applying a compensation voltage (CV) to the FAIMS waveform. Changes in the voltage-time product (striped blue & orange areas) can allow previously eliminated ions to be transmitted through the FAIMS cell (purple dotted line), while simultaneously causing other ions to be eliminated (red & green dotted lines).

27

Figure 1-3. Example FAIMS analysis of morphine from neat standard solutions. Mass spectra at various CF values confirm the transmission of morphine [M+H]+ (m/z 286) and [2M+Na]+ (m/z 593) ions at CF 2.0 Td.

Figure 1-4. Illustration of ion mobility behaviors with increasing electric field strengths. Kh represents the ion’s mobility at high-field and K0 represents the ion’s mobility at low-field. Adapted from Purves et al.2

28

Figure 1-5. Three-dimension and cross-section representations of planar (a), cylindrical (b, c), and micromachined chip-based FAIMS cell geometries (d). The dotted- line arrows indicate the direction of carrier gas and ion flow. Example connections for ground and the high-voltage waveform are provided for each geometry.

29

Figure 1-6. Schematic depiction of an ESI source. The inlet of the mass spectrometer serves as the counter-electrode in this example. Adapted from Konermann et al.39

Figure 1-7. Mathieu stability diagram illustrating stable ion trajectories for a 3D quadrupole ion trap. Ions with stable trajectories have az and qz values within the shaded region. Most quadrupole ion traps do not apply a DC offset to the AC potential and thus operate along the az = 0 line. Adapted from March, R.E.49

30

Figure 1-8. 2D linear ion trap representation depicting the quadrupole rod assembly, direction of ion flow, and voltage application. Adapated from LTQ XL Hardware Manual.50

31

CHAPTER 2 EFFECTS OF SOLVENT VAPOR MODIFIERS FOR THE SEPARATION OF OPIOID ISOMERS IN FAIMS-MS

Introduction

Over the last three decades, the abuse of and addiction to prescription opioid analgesics has become an national health crisis. In 2016, more than 63,000 people died from drug overdoses in the United States, with about 66% of drug overdose deaths involving opioids.51–53 The modern opioid crisis can generally be divided into two facets: the use of illicit opioids such as heroin, and the misuse/abuse of prescription opioids.

Regarding the first, the death rate due to heroin increased 20% from 2015 to 2016. In addition, the death rate due to illicit synthetic opioids – commonly fentanyl or related compounds – increased 72% over the same time period, outpacing the death rate due to heroin for the first time in recent history.51 Regarding the second, it is reported that

21-29% of patients who are prescribed opioids for chronic non-malignant pain misuse them, and 40% of drug overdose deaths result from misuse/abuse of prescription opioids.51–53 The abuse of prescription opioids can also exacerbate illicit opioid usage, as 80% of heroin addicts report starting with prescription opioids before transitioning to heroin.52,53

Screening for opioids typically starts with immunoassays for initial high- throughput detection and limited identification of compound class. Gas chromatography- or liquid chromatography-tandem mass spectrometry (GC- or LC-MS/MS) can be subsequently used for confirmation and more specific compound identification.

However, the separation of opioid isomers is still challenging for GC- and LC-MS/MS methods, typically requiring long sample pretreatment and chromatography times.54,55

Ion mobility is an emerging technology that offers a complementary separation step and

32

a number of advantages to traditional GC- and LC-MS/MS methods. IMS operates at the time scale in between chromatography (minutes) and mass spectrometry (µs to ms).

As an additional separation step, it can reduce chromatography time and increase the throughput of these methods. Finally, it dramatically improves the separation of isomers.

Recently, several groups separated opioid isomers using high-field asymmetric waveform ion mobility spectrometry (FAIMS) coupled to mass spectrometry using minimal or no sample pretreatment.14,56

In this study, I investigated the effects of several protic and aprotic solvent vapor modifiers on the separation of the [M+H]+ ions of four opioids: morphine, hydromorphone, norcodeine, and codeine; three of these four compounds are isomers

(morphine, hydromorphone, and norcodeine). Isomeric compounds cannot be resolved using only a single stage of mass spectrometry, thus necessitating an additional separation technique such as FAIMS. Morphine and codeine are naturally occurring opioids found in opium poppies that are widely used as analgesics and as precursor materials to many semi-synthetic opioids (e.g. hydromorphone, oxycodone, hydrocodone, etc.). In addition, several natural and synthetic opioids are metabolized to morphine by the liver, such as codeine and heroin. Hydromorphone, commonly known under its brand name Dilaudid, is an analgesic several times more potent than morphine; mistaking hydromorphone for morphine can result in overdose and death.57

In contrast to morphine and hydromorphone, norcodeine is a metabolite of codeine and has relatively little opioid effect by itself. Norcodeine was included in this study because of its isomeric relationship, but with different chemical properties from morphine and morphine-derivatives.

33

Materials and Methods

Chemicals

Morphine, hydromorphone, norcodeine (each molecular weight 285.3), and codeine (molecular weight 299.3) standards were purchased from Cerilliant Corporation at a stock concentration of 1.0 mg/mL in methanol. The structures for the four opioids used in this work are presented in Figure 2-1. Sample solutions were prepared for each individual standard at a concentration of 10 µg/mL in methanol with 0.1% formic acid. A mixture of 10 µg/mL codeine and 10 µg/mL norcodeine was also prepared for DF optimization experiments. Relatively high analyte concentrations (10 µg/mL) were used in this work for method development. Solutions were infused directly into the electrospray ionization source at a flow rate of 5 µL/min with no chromatographic or other separation step prior to ionization. Analyte ions were detected in positive-ion mode with the spray voltage set to 5.0 kV.

For solvent vapor addition, LC-MS grade methanol, water, acetonitrile, and ethyl acetate were purchased from Fisher Scientific, and high purity (≥99%) ethanol, 1- propanol, 2-propanol, propyl acetate, and n-butyl acetate were purchased from Sigma-

Aldrich.

FAIMS-MS Instrumentation and Methods

Experiments were conducted using a modified commercial Owlstone UltraFAIMS system interfaced to a Thermo Scientific LTQ XL linear ion trap mass spectrometer. The inlet capillary of the chip cap was modified to improve introduction of solvent vapor. The

UltraFAIMS chip cell consists of 21 parallel channels of paired gold-coated electrodes.

Each channel is 4.62 mm long and has an analytical gap size of 100 µm. The separation path length is approximately 700 µm.

34

FAIMS separation was performed at DFs ranging from 10 to 250 Tds; the highest attainable DF was limited by temperature set for the heated capillary on the mass spectrometer and corresponds to a DF of ~260 Td. The CF was scanned from -5 to 5

Tds over 60 seconds. FAIMS data was extracted from the mass spectra for each opioid using Thermo Xcalibur Qual Browser and processed in Microsoft Excel, R V3.3.2,

RStudio V1.0.153, and MS Convert V 3.0.9134.

We define resolution for FAIMS with Equation 2-1:

|퐶퐹2 − 퐶퐹1| 푅푠 = (2-1) 퐹푊퐻푀푎푣푒 where Rs is the resolution, CF1 and CF2 are the CF peak values in Td corresponding to the analytes of interest 1 and 2, respectively, and FWHMave is the average peak width of analyte 1 and 2 at 50% of the full height in Td.

Solvent Vapor Procedure

A homebuilt apparatus for introduction of solvent vapors into the Owlstone uFAIMS cell was constructed for these experiments (Figure 2-2). A dry nitrogen supply

(output pressure ~120 psi) was divided into two “channels”, and the gas flow through each channel was regulated with a mass flow controller (MKS Instruments). Each mass flow controller was calibrated for nitrogen gas and could be operated from 0.5 to 5.0 liters per minute. Solvent vapor was generated in one channel by bubbling dry nitrogen through ~300 mL of liquid solvent using a sparger attachment in a sealed 1L HPLC bottle at room temperature (~24 °C). Solvated nitrogen exited the bottle through the bottle cap and was then mixed with dry nitrogen from the second channel. The mixed gas was finally introduced into uFAIMS chip via the sweep gas port of the Thermo LTQ.

35

For these experiments, the combined gas flow from both mass flow controllers was optimized for the analytes and held constant at 2 L/min. The amount of solvent vapor introduced to the uFAIMS cell was varied by changing the ratio of solvated nitrogen flow to dry nitrogen flow while maintaining the 2 L/min combined flow. Solvent vapor concentration was calculated by measuring the loss in mass of the solvent HPLC bottle after a known amount of dry nitrogen was bubbled through the solvent. We assume that the change in mass is entirely due to evaporated solvent, and that there is no other loss of mass from the HPLC bottle. The final calculated concentrations are presented as molar fractions in parts per million (ppm) of moles of solvent molecules per total moles of nitrogen and solvent. Different vapor concentration ranges can be calculated for different solvents due to differences in vapor pressure at room temperature. More volatile solvents such as acetone produce higher vapor concentrations than less volatile solvents such as water.

Results and Discussion

FAIMS-MS Separation of Opioids

FAIMS and mass spectra for morphine are presented in Figure 2-3. Figure 2-3a is obtained when dry nitrogen is used as the carrier gas, and Figure 2-3b is obtained when acetonitrile vapor in nitrogen is used. When using dry nitrogen, the major ions that appear in the mass spectra for the tested opioids are the [M+H]+, [M+Na]+, [2M+H]+, and

[2M+Na]+ ions, with [M+H]+ as the base peak. Upon addition of solvent vapor, the signal intensities for [M+Na]+, [2M+H]+, and [2M+Na]+ ions decrease while the signal for the

[M+H]+ ion increases. In some cases, the ion corresponding to the [M+H+solvent]+ can be observed (Figure 2-3b). For addition of aprotic solvent vapor, the [M+Na]+, [2M+H]+, and [2M+Na]+ ions drop below the detection limit of the mass spectrometer.

36

The DF was optimized for these experiments by scanning across the values 10 to 250 Td using either a mixture of codeine and norcodeine or a mixture of morphine and norcodeine. For both sample sets, FAIMS spectra were acquired across the specified DF range with a carrier gas composition of either dry nitrogen or 8700 ppm acetonitrile vapor in nitrogen. The results of these scans for morphine and norcodeine are shown in Figure 2-4. When dry nitrogen is used as the carrier gas, the CF peaks for morphine, norcodeine, and codeine shift towards positive CF values as the DF strength is increased. However, their CF peaks for the [M+H]+ ions are not resolved at any DF strength from 10 to 250 Td (Figure 2-4a and 2-4b). On the other hand, when acetonitrile in nitrogen was used as the carrier gas, the CF peaks for morphine, norcodeine, and codeine shift in the opposite direction, i.e. towards negative CF values as the DF strength is increased. This negative CF shift appears for all tested solvent vapors, corresponding to dramatic changes in the high-field mobility as a result of solvent molecule interactions with analyte ions. In addition to this difference in CF shifts, the

[M+H]+ ions for morphine and norcodeine begin to resolve from each other starting at a

DF strength of 160 Td. At a DF of 250 Td, the CF peaks for morphine and norcodeine are baseline separated with a resolution of ~2.6 (Figure 2-4c). Since the best separation between morphine and norcodeine was observed at a DF of 250 Td, all further data were acquired using this DF strength.

These observations suggest a change in mobility behavior from Type A to Type

C. In this case, ions of small molecules are believed to have Type A behavior (higher mobility with increasing field) under dry nitrogen conditions. Upon introduction of solvent vapor, we assume that a “shell” of solvent molecules forms around the ion. As the ion

37

cluster oscillates during the FAIMS separation, the solvent shell is stable enough to survive the oscillations and “expands” during the high-field portions of the waveform, which decreases the overall mobility of the ion cluster because of a larger collisional cross section. During the low-field portions of the waveform, the ion cluster “contracts” as it returns to a rested state. The ion cluster is disrupted prior to mass analysis, most likely by turbulence due to the vacuum upon entering the mass spectrometer. As a result, the ion changes to Type C behavior (lower mobility with increasing field).

The shifts in CF peaks with increasing acetonitrile vapor concentration for the

[M+H]+ ions of all four opioids are shown in Figure 2-5a. The CF peak shifts show analogous behavior to each other with increasing acetonitrile vapor concentrations.

Separation between opioids can be observed even at the lowest acetonitrile vapor concentrations (3000 ppm). FAIMS spectra corresponding to 5800 ppm acetonitrile vapor concentration show near baseline resolution between morphine and norcodeine, as well as partial resolution between morphine and codeine (Figure 2-5b). Mass spectrometry allows the separation of codeine from morphine because of differences in molecular weight, thus complementing the FAIMS separation. Significant overlap between the CF peaks of morphine and hydromorphone was observed for all tested concentrations of acetonitrile vapor.

The CF peaks for the [M+H]+ ions of the opioids shift negatively upon introduction of any tested solvent vapor; this behavior for the [M+H]+ ion of morphine is shown in

Figure 2-6. Interestingly, we observed that the CF shifts resulting from addition of certain aprotic solvents appear to follow a trend-line that is relatively independent of the specific solvent used. As shown in Figure 2-6, for a DF strength of 250 Td and a vapor

38

concentration of about 3000 ppm, the CF peak for the [M+H]+ ion of morphine transmits at -1.2 ± 0.1 Td for each of the three solvents acetonitrile, ethyl acetate, and propyl acetate. When the solvent vapor concentration is about 8000 ppm, the morphine CF peak transmits at -2.4 ± 0.1 Td when using acetone, acetonitrile, and ethyl acetate.

Similar results were observed for the opioids hydromorphone, norcodeine, and codeine, as will be discussed later. In contrast, addition of protic solvent vapors produced distinct

CF shifts for different solvents (except 1- and 2-propanol) and do not follow the solvent- independent trend-line observed with aprotic solvent vapors. These results suggest that for the addition of vapor from small aprotic solvents, the same clustering mechanism occurs for ions of an opioid, which results in the same differential mobilities for a given vapor concentration. It should be noted that the addition of n-butyl acetate vapor does not produce CF shifts that follow the trend exhibited by other aprotic solvents for any of the tested opioids. We hypothesize that this deviation reflects a critical threshold in one or more chemical properties (e.g. gas-phase proton affinity) that results in alternative clustering behavior to the observed trend. The molecular volume of solvent molecules, relative gas-phase basicity between opioid ions and solvent molecules, and the Gibbs free energy of different ion-solvent cluster conformations are currently being investigated.

The magnitude of the shift increases with solvent vapor concentration until about

10000 ppm vapor in the carrier gas for aprotic solvents. When the solvent vapor concentration is above this range, the rate of CF shift begins to level off and may start reversing, which can be seen for acetone vapor concentrations of 13000 ppm and higher. This behavior is consistent with previously published results.15,25,31 We assume

39

that a “saturated solvent shell” forms around the ion at sufficiently high vapor concentrations. At this point, the maximum CF shift is reached for that analyte and solvent. It should be noted that the maximum CF shift and reversal exhibited in the literature occurs over a very broad vapor concentration range, from 200 to 20000 ppm solvent vapor in the carrier gas. This broad range may be accounted for by differences in the FAIMS configuration (e.g. total gas flow, FAIMS cell temperature, curved versus planar geometries).

The effects of solvent vapor on CF peak widths were also investigated and are shown in Table 2-1 and Figure 2-7. For all tested solvents, the peak widths for the

[M+H]+ ions of opioids narrow upon addition of solvent vapor and significantly improve the resolution of different ions in the FAIMS separation. The average full width at half maximum (FWHM) peak width for opioid CF peaks is about 0.84 Td when using dry nitrogen. The FWHM decreased to an average minimum of 0.39 Td (standard deviation

= ±0.06 Td) with increasing vapor concentration for all solvents except water; adding more vapor after the minimum FWHM was reached did not appear to significantly change the peak width. It should be noted that adding water vapor produced less narrow peaks than the other solvents tested, with an average FWHM of 0.70 Td.

However, the trend-line for water in Figure 2-7 suggests that the FWHM will continue to narrow if higher concentrations of water vapor were added, possibly also reaching the same minimum peak width as the other tested solvents.

Our observations of peak narrowing with increasing vapor concentration were consistent with previous work from our lab.25 We hypothesize that the peak narrowing results from the formation of ion clusters with a complete solvent “shell” resulting from

40

deliberate addition of enough solvent vapor. As a result, the CF peak and its width are determined by only the mobility differences between the complete solvent “shell” of the ion at high- and low-fields. Increasing the vapor concentration allows the formation of larger complete shells, which produce larger CF shifts, but the formation of only complete shells keeps the variability of ion clusters low and CF peak width narrow. We also hypothesize that the opposite phenomenon occurs in cases where CF peaks broaden resulting from either very low concentrations of solvent vapor or “dry” nitrogen contaminated with uncontrolled amounts of vapor; a larger variety of ion clusters with incomplete solvent shells are formed, resulting in larger variability of ion clusters and a wider CF peak.

Aprotic Solvents Versus Protic Solvents

The CF shifts corresponding to opioid separation using acetonitrile, acetone, ethyl acetate, methanol, and water are shown in Figure 2-8; plots of the CF shifts for every tested solvent are included in the supplementary material. Opioid separation was more improved using aprotic solvents than using protic solvents. For most aprotic solvents, the CF peak for morphine can be nearly baseline resolved from that of its isomer norcodeine and partially resolved from that of the analogue compound codeine.

Exceptions to this trend occur with addition of propyl acetate vapor, which was unable to resolve the CF peaks of morphine and codeine, and with addition of n-butyl acetate vapor, which resulted in significant overlap between the CF peaks for morphine and norcodeine in addition to the inability to separate codeine. For protic solvents, the major

CF peaks for each opioid were unresolved. CF peak intensity decreases significantly when adding vapor from water, 1-propanol, and 2-propanol.

41

Condensation of the solvent can be a problem for solvent vapor addition FAIMS because it causes irreproducibility in the vapor concentration and could result in hardware damage. In this work, problems due to condensation were encountered most frequently when adding vapor from water, 1-propanol, 2-propanol, and n-butyl acetate; condensate from these solvents was observed in the gas tubing even for low vapor concentrations. However, some of the CF peaks for the [M+H]+ ion of opioids appear to have improved resolution at higher vapor concentrations. For example, the resolution between the [M+H]+ ions of hydromorphone and morphine appears to improve at higher vapor concentrations of aprotic solvents such as acetone. One method that could enable the addition of higher vapor concentrations would be heating the solvated gas channel all the way to the FAIMS cell. The higher temperature would result in higher vapor pressure for the solvent, allowing for higher possible vapor concentrations with increased reproducibility.

A lower intensity CF peak for morphine (7-9% of base peak) can be observed in the range of 0.5 Td to 2.0 Td upon addition of most of the tested solvent vapors; this lower intensity peak appears to behave differently between morphine ions and hydromorphone ions. For 5800 ppm acetonitrile in nitrogen, the minor CF peak is about

11 times more intense for morphine than for hydromorphone across the same CF range

(1.8 × 103 vs. 0.2 × 103 signal counts, respectively). When using 9700 ppm methanol in nitrogen, the signal intensities for these minor peaks are similar (4.4 × 103 counts for morphine vs. 1.5 × 103 for hydromorphone), but the transmission CF values are different between morphine and hydromorphone (CF 1.33 Td vs. 0.85 Td, respectively). Other solvent vapors that produce notable signal for these peaks are acetone, ethyl acetate,

42

ethanol, and 2-propanol. These minor peaks appear reproducibly with solvent vapor addition, although mass spectra taken across these peaks do not reveal a difference in ion species from the dominate CF peak corresponding to [M+H]+. Based on our previous work and reports in the literature for low-field mobility separations, we hypothesize that the minor peaks correspond to multimer ions for each opioid that fragment to monomers upon introduction to the high vacuum environment of the mass spectrometer.2,10,58

Figure 2-9 shows the CF shifts corresponding to the separation of morphine and norcodeine using a series of alkyl acetates and a series of alcohols. As noted previously, the CF shifts for morphine appear to follow a trend-line and deviate from the trend for addition of n-butyl acetate; the same behavior is observed for hydromorphone, norcodeine, and codeine for the series of alkyl acetates. For example, addition of ethyl acetate and propyl acetate vapors also show the independent trend-line behavior for the

[M+H]+ ion of norcodeine as observed for the same ion of morphine. The CF shifts for the [M+H]+ ion of norcodeine resolve from those for the morphine [M+H]+ ion starting at a vapor concentration around 1800 ppm. In a similar manner, the CF shifts for norcodeine with n-butyl acetate also deviate from its smaller alkyl acetate trend and also show increasing separation from morphine at a vapor concentration around 2800 ppm.

For alcohols, the CF shifts for the four tested opioids did not resolve from each other in a significant manner. Figure 2-9b shows the CF shifts for morphine and norcodeine from addition of alcohol vapors. While the CF shifts from addition of ethanol and 1-propanol vapors initially appear similar to the trend-line behavior observed with addition of ethyl

43

acetate and propyl acetate vapors, the CF shifts from addition of methanol and 2- propanol do not appear to follow any trend in common with the other alcohols.

Previous work from our lab suggested that solvents with larger molecular volumes produce CF shifts with greater magnitude.25 Our work on the vapor addition of a series of alcohols appears to be consistent with the previous work, as the magnitude of CF shifts for opioid ions appears to increase as the size of the alcohol increases.

However, we observed the opposite trend when adding vapor for a series of alkyl acetates; as the size of the alkyl acetates increases, the magnitude of CF shifts decreases. This opposite pattern suggests that there are factors in addition to molecular volume that affect the magnitude of CF shifts when adding solvent vapor to FAIMS. It is likely that the clustering mechanism that causes the solvent-independent trend-line observed with aprotic solvent vapors also results in the opposite pattern observed for a series of alkyl acetates. Because the intermolecular interactions between molecules of aprotic solvents are weaker than those of protic solvents, the shell made from larger aprotic solvent molecules may be broken more easily during the FAIMS separation. This would produce a smaller “expanded” ion cluster during the high-field portion and lower magnitude CF shifts relative to the case where a smaller aprotic solvent were added. In contrast, protic solvent molecules are able to create stronger interactions via gas-phase hydrogen bonding and allow the shell to remain intact. Larger protic molecules produce a bigger “expanded” ion cluster during the high-field portion and larger magnitude CF shifts as a result.

Conclusions

Solvent vapor modified FAIMS significantly enhances the selectivity and separation of isomers in mass spectrometric analyses. Selection of the solvent vapor

44

that produces optimal separation is heavily dependent on the target analytes. As a result, solvent vapor modified FAIMS is best employed for targeted analyses. We have investigated and reported how various aprotic and protic solvents can affect the separation of several opioid compounds, including the isomers morphine, hydromorphone, and norcodeine. These opioid isomers could not be separated from each other using dry nitrogen carrier gas in FAIMS. Addition of vapor from small aprotic solvents produced the best separation between opioids, which also appeared to produce CF shifts that follow a trend-line independent of the specific identity of the solvent. Identification of the chemical or physical properties that account for the trend- line behavior may improve our fundamental understanding of high-field ion mobility and are currently being investigated.

Addition of solvent vapor also produces several other analytical benefits in addition to enhanced selectivity, including increased analyte transmission through the

FAIMS cell, higher signal intensities, and narrower peak widths, each as observed previously in our lab. Increasing vapor concentration initially broadens CF peak widths, followed by peak narrowing after a sufficient concentration of solvent vapor has been reached. At high enough vapor concentrations, peak widths appear to approach a minimum width that is hypothesized to result from mobility differences as the solvent shell of the ion cluster expands and contracts during the FAIMS separation. Peak widths reach a minimum when enough solvent vapor is present to form a complete solvent shell. Larger CF shifts can be obtained with increasing vapor concentration even after peak widths reach a minimum. Increasing the vapor concentration allows the formation

45

of larger complete shells, which produce larger CF shifts, but the formation of only complete shells keeps the variability of ion clusters low and CF peak width narrow.

The performance of FAIMS depends on the geometry and instrumentation of the system. Increasing the residence time of analyte ions in the FAIMS cell can lead to improved resolving power and could reduce the amount of solvent vapor required to produce separation. Developing a robust and reliable solvent vapor addition system is critical for reproducible vapor modified FAIMS separations. Adequate temperature control of the solvent vapor reduces the risk of condensation in the system, which causes inaccuracy in vapor concentration measurements and analyte signal intensities.

Future research efforts aim at elucidating the changes in solvent vapor modified FAIMS separations for structurally similar molecules.

46

Table 2-1. Average CF peak widths and standard deviations for carrier gas compositions. Average peak widths and standard deviations are also presented for all aprotic solvents and alcohols because of similarity between their respective constituents. Carrier Gas Average Peak Standard Width (Td) Deviation (Td) Dry Nitrogen 0.84 ± 0.11

Water 0.70 ± 0.09

Aprotic Solvents 0.41 ± 0.07

Acetonitrile 0.35 ± 0.02

Acetone 0.50 ± 0.04

Ethyl Acetate 0.44 ± 0.05

Propyl Acetate 0.36 ± 0.04

n-Butyl Acetate 0.40 ± 0.04

Alcohols 0.37 ± 0.05

Methanol 0.38 ± 0.04

Ethanol 0.32 ± 0.03

1-Propanol 0.40 ± 0.06

2-Propanol 0.37 ± 0.04

47

Figure 2-1. Chemical structures for four opioids including molecular formulae and weights. The structures for three isomers of morphine are shown in (a), and the structure for codeine is shown in (b). Codeine was used as a chemically similar control for this work.

48

Figure 2-2. Schematic for solvent vapor addition apparatus. The dry nitrogen supply is split into two channels. Solvent vapor is generated by bubbling the dry nitrogen through liquid solvent using a sparger attachment in a sealed bottle. Solvated nitrogen exits the bottle through the bottle cap and then mixes with dry nitrogen from the second channel before being introduced into the FAIMS cell as the carrier gas. Gas flow through each channel is regulated using a mass flow controller.

49

Figure 2-3. FAIMS spectra for [M+H]+ at m/z 286 from a morphine solution and corresponding mass spectra acquired using either (a) dry nitrogen as the carrier gas or (b) 8700 ppm acetonitrile vapor in nitrogen as the carrier gas. The ion at m/z 327 corresponds to the cluster of [M+H]+ with one molecule of acetonitrile ([M+H+CH3CN]+).

50

Figure 2-4. Plot of CFs with increasing DFs for a mixture of morphine and norcodeine. The graph in (a) shows the CF peak value with increasing DF strength. Morphine and norcodeine begin to resolve with a DF strength of about 160 Td. The FAIMS spectra shown in (b) and (c) correspond to dry nitrogen and 8700 ppm acetonitrile at a DF of 250 Td, respectively.

51

Figure 2-5. Plot of CF shifts with increasing acetonitrile vapor concentration in nitrogen for the m/z 286 ions from four opioid solutions, along with the FAIMS spectra for the opioid ions at an acetonitrile vapor concentration of 5800 ppm. The plot in (a) corresponds to the CF peak values with acetonitrile vapor concentration. The FAIMS spectra in (b) correspond to the data points indicated by the red line. All data were acquired at a DF of 250 Td.

52

Figure 2-6. Plot of CF shifts with increasing vapor concentration for the m/z 286 ion from a morphine solution. Addition of any solvent vapor shifts the peak from positive CF values towards negative CF values. Most aprotic solvent vapors (acetonitrile, acetone, ethyl acetate, & propyl acetate) produce the same CF shifts for a given vapor concentration.

53

Figure 2-7. Plot of peak width with increasing vapor concentration for the m/z 286 ion from a morphine solution. The average full width at half maximum (FWHM) for solvent vapors excluding water is 0.39 Td with a standard deviation of 0.06 Td (red lines).

54

Figure 2-8. Plots showing CF shifts with respect to vapor concentration for acetonitrile (ACN), acetone, ethyl acetate (EtOAc), methanol (MeOH), and water (H2O) for the m/z 286 ion from opioid standard solutions. FAIMS spectra corresponding to specific vapor concentrations (red lines) are presented on the right of their respective solvent. All data were acquired at a DF of 250 Td.

55

Figure 2-9. Plots of CF shifts with addition of vapor from (a) a series of alkyl acetates and (b) a series of alcohols for the m/z 286 ion from individual solutions of morphine and norcodeine. Similar CF shift behavior can be observed for both morphine and norcodeine ions. All data were acquired at a DF of 250 Td.

56

CHAPTER 3 FAIMS SEPARATION OF ANABOLIC ANDROGENIC STEROID EPIMER PAIRS USING CATION MODIFIERS

Introduction

Anabolic androgenic steroids (AAS, or anabolic steroids) are the most common class of performance-enhancing drugs and are drugs that attract significant national and international media attention. Anabolic steroids act upon the androgen receptor, which produces various anabolic and androgenic effects depending on the specific steroid and tissue localization. Anabolism refers to the retention of nitrogen in the body and is linked to increased protein synthesis. Androgenism refers to the development of the male reproductive system and secondary sex characteristics, such as deeper voice and changes in hair growth or pattern. Typically, steroids with higher anabolic character bind more weakly to the androgen receptor than steroids with higher androgenic character.59–61 Anabolic steroids with higher anabolic character are more desirable for doping than those with higher androgenic character, but both varieties of anabolic steroids are used as performance-enhancing drugs.

Testosterone was first isolated and synthesized in 1935. Over the next several decades, various synthetic analogues of testosterone were developed by pharmaceutical companies for therapeutic uses, and some anabolic steroids are still currently used for the treatment of osteoporosis, cardiovascular disease, and male hormone dysregulations.59 At the same time, anabolic steroids began to be used in competitive athletics, starting in strength-intensive sports. In 1954, Dr. John Ziegler learned about the use of anabolic steroids by the Soviet team at the world weightlifting championships in Vienna and would later encourage the use of anabolic steroids by US weightlifters.60 One of the biggest scandals about athletic usage of anabolic steroids

57

was the state-sponsored doping program in East Germany that was uncovered after the fall of the Communist government.62 The International Olympic Committee banned the use of anabolic steroids in the Olympics in 1975, although they continue to be used and occasionally result in a public scandal. These scandals about elite athletes who were caught doping give the misleading impression that performance-enhancing drug abuse, including anabolic steroid abuse, is largely a problem at the professional sport level.

In actuality, the majority of anabolic steroid users are not elite athletes, but noncompetitive bodybuilders.60,61,63 In 2016, it was reported that there were 3 million anabolic steroid users in the United States alone.61 The most common reasons for a person to use anabolic steroids are to increase muscle mass and decrease body fat. It is estimated that about 3% of men worldwide have used anabolic steroids at some point; male gym attendees are reported to have even higher usage rates at 15-25%.61,64

About one-third of anabolic steroid users report becoming dependent on the drugs.60,61,65 The abuse of anabolic steroids can lead to organ damage, endocrine dysregulation, psychological changes, and various blood complications.60,61,66

Aberrations in reproductive organs and infertility are the most common problems reported by chronic users of anabolic steroids, as even most steroids with higher anabolic character exhibit androgenic side-effects with prolonged use.59,61,63 Many side- effects of anabolic steroid use are reversible after the user stops taking the drugs.

However, anabolic steroid use among athletes and bodybuilders typically involves maintaining higher-than-normal concentrations of one or more anabolic steroids for cycles lasting several weeks.59,61,65 In some cases, anabolic steroids are used in conjunction with other supplements such as creatine or opioids, leading to more serious

58

permanent problems.60 In addition, the long term use of anabolic steroids can exacerbate reversible side-effects and increase the chance of dependence.

The United States government attempted to restrict the availability of anabolic steroids with the Anabolic Steroid Control Acts of 1990 and 2004, which regulated anabolic steroids as a Schedule III controlled substance, forbidding their sale or distribution without a prescription. The 2004 Act added anabolic steroid precursors to the list of Schedule III controlled substances. However, companies dodge regulation and continue to distribute anabolic steroids by developing “designer steroids” and marketing them as nutritional or dietary supplements.59,60,63 In addition, the Internet has dramatically enhanced the availability of designer steroids and the ability of customers to purchase them.

Designer steroids are anabolic steroids that are specifically designed to dodge regulations and evade detection. These designer steroids are often novel compounds that are synthesized from various androgen precursors that were previously developed but abandoned in the 1960s by pharmaceutical companies.60,63 Many derivatives of testosterone have aimed at maximizing the anabolic properties of the molecule while minimizing the androgenic ones, although no derivative has completely dissociated those two properties. Designer steroids pose an even greater health risk than traditionally anabolic steroids because of the complete lack of toxicity or safety testing in animals or human. For example, the designer steroid methasteron was sold as an unregulated dietary supplement and was later reported to cause severe hepatotoxicity and renal failure.63 In addition, evidence suggests that other nutritional supplements are being contaminated by designer steroids that are not listed in the ingredients, whether

59

intentionally or unintentionally.63 As a result, bystanders not intending to ingest anabolic steroids could be at risk of androgenic and/or toxic side-effects when consuming these supplements. Proper regulation of designer steroids is also challenging, as drug manufacturers will switch to a different androgen structure and develop a new designer steroid to replace ones detected by screening labs.

Improving the throughput and accuracy of screening tests would enable regulators to stay on top of illicit anabolic steroids, help physicians treat patients who are suffering from anabolic steroid side effects, and potentially reduce the profitability of manufacturing designer steroids. Gas chromatography- or liquid chromatography- tandem mass spectrometry (GC- or LC-MS/MS) is commonly used for confirmation of initial screening and for compound identification. However, the separation of isomers is still challenging for GC- and LC-MS/MS methods, typically requiring long sample pretreatment and chromatography times. Ion mobility is an emerging technology that offers a complementary separation step and a number of advantages to traditional GC- and LC-MS/MS methods. IMS operates at the millisecond time scale and conveniently fits in between the time scales for chromatography (minutes) and mass spectrometry

(µs to ms). As a result, it can increase the separation power of the method without increasing the overall analysis time, thus increasing the throughput of these methods.

Finally, it can dramatically improve the separation of isomers. Several groups have utilized ion mobility coupled to mass spectrometry to improve the separation of anabolic steroid isomers.67–69

In this work, I investigate the separation of anabolic steroid epimers using the high-field ion mobility technique FAIMS. For reference, epimers are defined as isomers

60

that differ only by the stereochemistry around a single chiral center. For example, testosterone and epitestosterone differ from each other only by the orientation of the hydroxyl group at the C17 position. Two epimer pairs were used for the work in this project: testosterone & epitestosterone, and androsterone & trans-androsterone. Cation modifiers were utilized in conjunction with FAIMS to enhance the resolution between epimers. Acetate salts for various Group 1 metal cation species were added to sample solutions to promote the formation of adduct ions with the steroid analytes. Ultimately, this strategy aims at enhancing the resolution of steroid epimers and isomers with no increase in analysis time or sample preparation. Here I report on the FAIMS separation of each epimer pair, as well as characterize several phenomena observed for cation modified FAIMS.

Materials and Methods

Chemicals

Testosterone, epitestosterone (each molecular weight 288.3), androsterone, and trans-androsterone (each molecular weight 290.4) standard solutions were purchased from Cerilliant Corporation at a stock concentration of 1.0 mg/mL in methanol. The structures for both anabolic steroid epimer pairs used in this work are presented in

Figure 3-1. Formic acid (99.7%) was purchased from Fisher Scientific. , , and salts were purchased in solid form also from

Fisher Scientific. acetate and cesium acetate salts were purchased in solid form from Sigma-Aldrich. Sample solutions were prepared for each individual standard at a concentration of 10 µg/mL in methanol with 0.1% formic acid or 10 µg/mL cation acetate salt added. Relatively high analyte concentrations (10 µg/mL) were used in this work for method development. Solutions were infused directly into the electrospray

61

ionization source at a flow rate of 5 µL/min with no chromatographic or other separation step prior to ionization. Analyte ions were detected in positive-ion mode with the spray voltage set to 5.0 kV.

FAIMS Instrumentation and Methods

Experiments were conducted using a commercial Owlstone UltraFAIMS system interfaced to a Thermo Scientific LTQ XL linear ion trap mass spectrometer. The

UltraFAIMS chip cell consists of 21 parallel channels of paired gold-coated electrodes.

Each channel is 4.62 mm long and has an analytical gap size of 100 µm. The separation path length is approximately 700 µm.

FAIMS separation was performed at DFs ranging from 10 to 250 Tds; the highest attainable DF of 260 Td corresponds to an electric field of 59 kV/cm (590 V divided by a gap distance of 100 µm) normalized to gas density N at temperature 61 °C. The CF was scanned from -5 to 5 Tds over 60 seconds. FAIMS data were extracted from the mass spectra for each opioid using Thermo Xcalibur Qual Browser and processed in Microsoft

Excel, R V3.3.2, RStudio V1.0.153, and MS Convert V 3.0.9134 0.

We define resolution for FAIMS with Equation 3-1:

|퐶퐹2 − 퐶퐹1| 푅푠 = (3-1) 퐹푊퐻푀푎푣푒 where Rs is the resolution, CF1 and CF2 are the CF peak values in Td corresponding to the analytes of interest 1 and 2, respectively, and FWHMave is the average peak width of analyte 1 and 2 at 50% of the full height in Td.

62

Results and Discussion

Separation of Testosterone & Epitestosterone using Group 1 Cations

Protonated ions ([M+H]+) are the most common analyte ion of interest in positive ion mode mass spectrometry. Shown in Figure 3-2 are the FAIMS spectra for the

[M+H]+ ions for testosterone and epitestosterone from sample solutions with 0.1% formic acid added, as well as the mass spectra for each steroid epimer for the observed

CF peak. The FAIMS spectra for both epimers are very similar, with both steroids having a CF peak at 1.7 Td for the [M+H]+ ion (Figure 3-2a). The mass spectra across the CF peaks at 1.7 Td reveal the same m/z species and very similar relative intensities for both steroids. Figure 3-2b shows the mass spectrum across the CF peak at 1.7 Td for testosterone, where we can see high signal for the [M+H]+ (m/z 289), the [M+Na]+

(m/z 311), the [2M+H]+ (m/z 577), and the [2M+Na]+ (m/z 599) ions. The [M+H]+ and

[2M+H]+ ions are the most intense species in the mass spectrum, with the [M+Na]+ ion at ~40% relative intensity to the base peak and the [2M+Na]+ ion at ~10%. The same m/z ions can be seen in Figure 3-2c for epitestosterone and at similar relative intensities. As a result, we are not able to resolve the epimers testosterone and epitestosterone from each other using their [M+H]+ ions in FAIMS. However, the notable signal intensities for the sodiated adducts despite no addition of sodium ions served as the starting point for our investigation of cation modified FAIMS.

Figure 3-3 shows the FAIMS spectra for the [M+Na]+ ions of testosterone and epitestosterone from sample solutions with 10 µg/mL sodium acetate added. For testosterone, we observed two CF peaks in the FAIMS spectra: one peak at 1.7 Td and a second peak at 2.6 Td. Both CF peaks are of comparable intensity. The CF peak at

1.7 Td appears overlaps with the peak for the [M+H]+ ion shown previously, and the

63

mass spectrum across this peak shows predominantly the [M+Na]+ ion (m/z 311) with low signal for the other ion species described previously. On the other hand, the peak at

2.6 Td does not have an equivalent feature in the [M+H]+ FAIMS spectrum. The mass spectrum across the CF 2.6 Td peak for testosterone shows high signal intensities for the [M+Na]+ and [2M+Na]+ ions. When we overlay the FAIMS spectra for the [M+Na]+ ion of both steroid epimers, we can see that the CF peak at 2.6 Td also does not appear for epitestosterone. As a result, we are able to resolve testosterone from its epimer using the [M+Na]+ ion. The mass spectrum for the CF peak at 1.7 Td for epitestosterone is similar to the spectrum for the same peak for testosterone. The differences between the mass spectrum at CF 1.7 Td for sample solutions with added sodium acetate and those with added formic acid suggest that anabolic steroids have higher cation affinities than proton affinities.

We investigated the FAIMS spectra for cation adducts of testosterone formed with various Group 1 metal cations. Because the FAIMS spectrum for the [M+Na]+ ion of testosterone showed a CF peak that was unique between it and epitestosterone, we wanted to determine if other unique peaks could be observed for larger cations. Figure

3-4 shows the overlaid FAIMS spectra for monomer cation adducts ([M+X]+, where X is the cation) of testosterone from sample solutions with the corresponding acetate salt added. For all the tested cation adducts, the CF peaks at 1.7 Td and 2.6 Td can be observed in all cases. FAIMS spectra for the [M+K]+, [M+Rb]+, and [M+Cs]+ ions show a significant CF peak at ~0.8 Td. However, [M+Li]+ and [M+Na]+ show some signal across the CF from 0.0 to 1.2 Td, suggesting that all the tested cation adducts transmit at 0.8

Td to some degree. Aside from the CF peak at 0.8 Td, no new unique peaks were

64

observed for different Group 1 cations. Absolute signal intensities for monitored cation adducts decrease as the size of the cation increases. On the other hand, the relative intensities of the CF peaks at 2.6 Td and 1.7 Td decrease relative to the peak at 0.8 Td with increasing cation size. We hypothesize that these observations result from three factors. First, each CF peak corresponds to the transmission of a specific ion cluster of testosterone and cation, and that the CF value does not vary with larger or smaller cations. Second, the CF peak at 0.8 Td occurs for all tested testosterone cation adducts, but becomes more pronounced with larger cations because of decreasing signal intensities of the peaks at 1.7 Td and 2.6 Td. Finally, the ion cluster that is transmitted through FAIMS at CF 0.8 Td becomes more stable than the clusters at 1.7

Td and 2.6 Td as the size of the cation increases. Mass spectra across the 0.8 Td peak show high signal for the trimer cation adduct ([3M+X]+, where X is the cation) in addition to the [M+X]+ and [2M+X]+ ions. This suggests that the [3M+X]+ trimer ions become more stable than the monomer or dimer ions as the size of the cation increases, producing a higher relative signal at CF 0.8 Td as compared to 1.7 Td or 2.6 Td.

However, the trimer ion is not stable enough to remain intact before mass analysis and fragments to yield [M+X]+ and [2M+X]+ ions in the mass spectrum. We assume that the most likely location for this fragmentation is the entrance to the high-vacuum region of the mass spectrometer.

Effect of Multimer Cation Complexes on FAIMS Separations

Further investigations were conducted to determine what ion species could account for the CF peak at 2.6 Td for testosterone. Figure 3-5 shows the FAIMS spectra for the [M+Na]+, [2M+Na]+, and [3M+Na]+ ions of testosterone from a single sample solution with sodium acetate added, as well as the mass spectra across observed CF

65

peaks. For the FAIMS spectrum of the [M+Na]+ ion shown in Figure 3-5a (blue trace), we can see the two intense CF peaks at 1.6 Td and 2.6 Td that were described previously, as well as some signal in the CF range from 0.0 to 1.2 Td. The mass spectrum across the CF peak at 1.6 Td shows signal predominantly for the [M+Na]+ ion of testosterone (Figure 3-5b). In contrast, the mass spectrum across the CF peak at 2.6

Td shows significant signal for both the [M+Na]+ ion and the [2M+Na]+ dimer ion. The

FAIMS spectrum of the [2M+Na]+ ion in Figure 3-5a (green trace) reveals two CF peaks, one centered at ~0.8 Td and a second that overlaps completely with the [M+Na]+ peak at CF 2.6 Td. This result supports the previously proposed idea that CF peaks correspond to the transmission of specific ion clusters through the FAIMS. In the case of sodiated adducts, the CF peak at 2.6 Td appears to correspond with transmission of the

[2M+Na]+ dimer ion through the FAIMS, which subsequently fragments prior to mass analysis. The FAIMS spectrum of the [3M+Na]+ ion shows this phenomenon once again with the CF peak at 0.8 Td. The mass spectrum across the CF 0.8 Td peak shows significant signal for the [3M+Na]+ trimer ion in addition to the [M+Na]+ and [2M+Na]+ ions.

It should be noted that an ion at m/z 352 was observed in the mass spectra across the CF peaks at 1.6 Td and 2.6 Td (Figure 3-5b). This ion appears to be chemically related to testosterone because its FAIMS spectrum closely resembles the spectrum for the [M+Na]+ ion of testosterone. The m/z 352 ion also appears in the mass spectrum for epitestosterone. We assume that this ion corresponds to a cluster of the steroid cation adduct with solvent molecules that results from incomplete desolvation in the heated transfer capillary of the mass spectrometer.

66

The FAIMS spectra for multimer cation adducts of epitestosterone were also investigated and compared to those for testosterone. In general, the absolute signal intensities for multimer cation adducts were significantly higher for epitestosterone than those observed for testosterone. In addition, certain multimer cation adducts produced unique CF peaks between the epimers. Shown in Figure 3-6 are the FAIMS spectra for the [3M+Li]+ ions for epitestosterone and testosterone from sample solutions with lithium acetate added, as well as the mass spectra for each steroid epimer across the CF range from 2.5 to 3.5 Td. The FAIMS spectrum for the [3M+Li]+ ion of testosterone in

Figure 3-6a shows only one CF peak at 0.1 Td, which is representative in general of the

FAIMS spectra for the trimer cation adduct for testosterone. In contrast, the FAIMS spectrum for the [3M+Li]+ ion of epitestosterone shows two CF peaks, one at 0.1 Td that overlaps with the peak for testosterone, and a unique peak at 2.9 Td (Figure 3-6a). The mass spectra taken across the CF range from 2.5 to 3.5 Td for epitestosterone (Figure

3-6b) and testosterone (Figure 3-6c) show that only epitestosterone has significant signal intensity for the [3M+Li]+ ion (m/z 871). As a result, we are able to resolve epitestosterone from testosterone using the [3M+Li]+, complimenting the resolution of testosterone using [M+Na]+ that was shown previously. Both testosterone and epitestosterone show high signal intensities for the [2M+Li]+, which precludes the use of the lithiated dimer ion for resolving the epimers. Interestingly, the mass spectrum for testosterone across the CF range show notable signal for m/z 327. This ion was identified as a cluster of the lithiated monomer ion with an additional methanol molecule

+ + + ([M+CH3OH+Li] ). The relative signal between the [M+CH3OH+Li] ion and the [M+Li] varies from analysis to analysis, which is believed to result from the same incomplete

67

desolvation described previously. Future experiments will need to be conducted to

+ optimize the ionization source and/or to determine if the signal for [M+CH3OH+Li] ion can be produced reproducibly.

Separation of Androsterone & Trans-androsterone

FAIMS spectra for androsterone and trans-androsterone showed similar features and patterns as those for testosterone and epitestosterone. However, in contrast to testosterone and epitestosterone, we were able to achieve resolution between androsterone and trans-androsterone by monitoring only a single cation adduct species for both epimers. Figure 3-7 shows the FAIMS spectra for the [M+K]+ ions of androsterone and trans-androsterone from sample solutions with 10 µg/mL potassium acetate added. The FAIMS spectrum for androsterone shows two CF peaks at 0.4 Td and 1.9 Td, while the spectrum for trans-androsterone shows two CF peaks at 0.4 Td and 2.7 Td. In this case, we are able to resolve androsterone from trans-androsterone in

FAIMS by their unique CF peaks in the range from 2.0 to 3.0 Td for the [M+K]+ ion species. The resolution between the unique CF peaks is fairly high, with a Rs of ~1.9.

The CF peaks at 0.4 Td do not resolve between the epimers, although mass spectra across the peak show similar transmission of dimer and trimer ions as observed with testosterone and epitestosterone.

Conclusions

In this work, we have demonstrated how cation-modified FAIMS is capable of enhancing the resolution of anabolic steroid epimers. This enhancement is possible with no increase in analysis time, no additional instrumentation (beyond the FAIMS system), and potentially with no change in sample preparation. Because of the naturally high background concentration of sodium, cases where steroid epimers and isomers can be

68

well-resolved by their sodiated adduct ions would not necessarily require any change in sample preparation. Validation experiments would need to be conducted in order to determine the reproducibility of FAIMS peaks if no additional sodium is used.

Selection of what cation to incorporate into a FAIMS method for enhanced separation depends on the analytes of interest and must be determined empirically.

From our work with Group 1 metal cations, smaller cations up to potassium produced the best balance between the absolute signal intensities for CF peaks and the features in the FAIMS spectra. Rubidium and cesium adduct ions produced CF peaks with much lower signal-to-noise ratios, which could complicate the reproducibility of the method and the quantification of analytes. Initial research on the FAIMS separation of anabolic steroids using Group 2 and transition metal cations is both promising and challenging because of the greater number of potential adduct ions that could be used to resolve isomers.

Additional experiments will be needed to be determine the stability of the electrospray ionization source when using cation modifiers and optimize cation concentration. It is well known that high concentrations of cation salts cause ionization suppression in electrospray ionization. All data presented in this work were acquired using direct infusion of samples into the electrospray source. We observed that the signal stability of our instrument seemed to suffer after several days of infusing solutions with 10 µg/mL of acetate salt, although rinsing the electrospray source through with blank solutions high in aqueous content recovered signal stability. Initial experiments suggest that good signal-to-noise ratios for the CF peaks of anabolic steroid adduct ions can be achieved with cation concentrations significantly lower than 10 µg/mL, which

69

would help maintain the stability of the method and decrease the frequency of cleaning the instrument.

Future research efforts would also aim at elucidating the structures of the specific ion clusters that are transmitted at different CF peaks. Because the fundamental mechanisms that influence high-field ion mobility are not well understood, FAIMS methods must be developed and optimized on a case-by-case basis. As stated previously, the selection of cation that produces optimal separation of steroid epimers must be determined empirically. Characterizing the structures of ion clusters in high electric fields will benefit FAIMS methods by reducing the amount of time needed to develop and optimize them.

70

Figure 3-1. Chemical structures for two anabolic steroid epimer pairs, including molecular formulae and weights. Structure pairs are divided into (a) testosterone and epitestosterone, and (b) androsterone and trans- androsterone.

71

Figure 3-2. Spectra for the [M+H]+ ions (m/z 289) from 10 µg/mL testosterone and epitestosterone sample solutions showing overlapping CF peaks at 1.7 Td. FAIMS spectra for each steroid are shown in (a). Mass spectra at CF 1.7 Td for (b) testosterone and (c) epitestosterone show the same ion species with similar relative intensities.

72

Figure 3-3. FAIMS spectra for the [M+Na]+ ions (m/z 311) of testosterone and epitestosterone. Testosterone shows a CF peak at 2.6 Td that does not appear for epitestosterone. No signal is observed for the [M+H]+ ion (m/z 289) of testosterone at CF 2.6 Td. As a result, testosterone can be resolved from epitestosterone using sodium adduct ions, but not vice versa.

Figure 3-4. Overlaid FAIMS spectra for Group 1 cation adducts of testosterone. CF peaks can be observed at 0.8 Td, 1.7 Td, and 2.6 Td, with the relative intensity of the 1.7 Td and 2.6 Td peaks decreasing as the size of the cation increases. All spectra were acquired at a DF of 250 Td.

73

Figure 3-5. FAIMS spectra for the [M+Na]+ (m/z 311), [2M+Na]+ (m/z 599), and [3M+Na]+ (m/z 887) ions of testosterone, as well as mass spectra across each observed CF peak. The FAIMS (a) and mass spectra (b) suggest that specific CF peaks correspond to the transmission of specific multimer ions.

74

Figure 3-6. FAIMS spectra for the [3M+Li]+ ions (m/z 871) of testosterone and epitestosterone showing a unique CF peak for epitestosterone at 2.9 Td. FAIMS spectra for each steroid are shown in (a). Mass spectra at CF 2.9 Td for (b) epitestosterone and (c) testosterone show that only epitestosterone has significant signal for the [3M+Li]+ ion. The [M+Li+CH3OH]+ ions observed in the mass spectrum for testosterone most likely result from a cluster formed with solvent from sample solutions.

75

Figure 3-7. FAIMS spectra for the [M+K]+ ions (m/z 329) of androsterone and trans- androsterone. Both epimers have CF peaks that overlap at 0.4 Td, but each also has a unique peak in the range from 1.0 to 3.0 Td. The unique CF peaks allow both epimers to be resolved from each other.

76

CHAPTER 4 RESOLVING ANDROSTERONE ISOMERS AND ISOBARS USING CATION- MODIFIED FAIMS-MS

Introduction

Many strategies to avoid detection and regulation have been employed by manufacturers of designer steroids, including distributing steroids as nutritional supplements, synthesizing steroids that are not specifically banned, and evading detection by mimicking endogenous anabolic steroids.59–61,63 In particular, mimicking endogenous steroids has proven to be a popular and challenging strategy for current screening methods. A well-known problem faced by drug screening labs is whether the testosterone concentration for an athlete is indicative of testosterone doping, as the endogenous level of testosterone can vary considerably between individuals.59,70

Screening labs accredited by the World Anti-Doping Agency address this problem by monitoring the testosterone/epitestosterone ratio as well as the 13C/12C isotope ratio of testosterone.59 The requirement for relatively large amounts of injected analyte imposed by isotope ratio mass spectrometry has also lead to the monitoring of the metabolites androsterone and etiocholanolone. All of these methods are further complicated by the variation in endogenous steroids among individuals, by the use of masking steroids that have no biological activity but appear to maintain endogenous steroid ratios, and by prohormone designer steroids that are metabolized into analogues or isomers of endogenous anabolic steroids. For example, 5α-dihydrotestosterone is a potent agonist of the androgen receptor but is an isomer of androsterone. Separation and identification of isomers and isobars would improve our ability to detect designer steroids that attempt to evade detection by resembling endogenous anabolic steroids.

77

Previous work from our group has explored the separation of several anabolic steroid isomers, including four epimers of androsterone, using conventional drift tube ion mobility spectrometry.58,68 For reference, epimers are defined as isomers that different only by the stereochemistry around a single chiral center. For the [2M+Na]+ dimer ions, the drift time peaks for the 3α-hydroxy androsterone epimers (androsterone and etiocholanolone) can be baseline resolved from the peaks observed for the 3β- hydroxy androsterone epimers (trans-androsterone and epietiocholanolone). However, the peaks between each 3α- or 3β-hydroxy androsterone epimer pair overlap significantly for the [2M+Na]+ ion, requiring the use of other cation adducts to achieve separation.68 The high resolution between the [2M+Na]+ ions of androsterone and trans- androsterone was confirmed to result from differences in the collisional cross sections of the most stable gas-phase adduct configuration for each steroid.58 The most stable adduct configuration at low electric fields in a partial vacuum (our drift tube ion mobility separation occurs at fields from 9.6 to 18.6 V/cm at 4 Torr, approximately 1/200 of an atmosphere) may not be representative of the case in high-field mobility separations at ambient pressure (FAIMS field of ~58000 V/cm at 1 atmosphere), prompting us to investigate the system using FAIMS.

In this work, I characterize the FAIMS separation for the cation adducts of five androsterone isomers and isobars: androsterone, trans-androsterone, etiocholanolone,

5α-dihydrotestosterone, and oxabolone (Figure 4-1). Androsterone, trans-androsterone, and etiocholanolone are metabolites of testosterone and epimers to each other. These androsterone epimers are all less potent anabolic steroids than testosterone. However, they are still used by bodybuilders and, as a result, are monitored by drug screening

78

labs. The mean endogenous concentration for androsterone and etiocholanone are about 100 times higher than for testosterone and epitestosterone.69 In contrast to the androsterone epimers, 5α-dihydrotestosterone (5α-DHT, or simply DHT) is a potent androgen and binds to the androgen receptor (AR) several times more strongly than testosterone. Endogenously, DHT is a metabolite of testosterone and is both a metabolite and intermediate in the conversion of androsterone. DHT is an isomer to androsterone, with the same molecular formula but a different chemical structure.

Oxabolone is a designer steroid derived from nandrolone and is isobaric to androsterone. For reference, isobaric compounds are compounds with the same nominal mass but different molecular formulae and therefore different exact masses. On a mass spectrometer with unit mass resolution, isobaric compounds cannot be resolved from each other by mass analysis alone. Because previous work demonstrated that the

FAIMS separation of steroid isomers is improved between cation adducts, I conducted this work focusing on adduct ions for several cation species commonly used or encountered in mass spectrometry methods. Specifically, I investigated the effects of

DF, cation size, and cation concentration on the presence and separation of CF peaks.

Materials and Methods

Chemicals

Androsterone, trans-androsterone, etiocholanolone, and 5α-dihydrotestosterone

(chemical formula C19H30O2, each molecular weight 290.4) standard solutions were purchased from Cerilliant Corporation at a stock concentration of 1.0 mg/mL in methanol. Oxabolone (chemical formula C18H26O3, molecular weight 290.4) standard was purchased as a solid also from Cerilliant Corporation. Lithium acetate, sodium acetate, potassium acetate, and salts were purchased in solid form

79

also from Fisher Scientific. LC-MS grade methanol solvent was also purchased from

Fisher Scientific. Sample solutions were prepared for each individual steroid at a concentration of 10 µg/mL in methanol with either 0.1 µg/mL, 0.5 µg/mL, or 1.0 µg/mL cation acetate salt added. Solutions were infused directly into the electrospray ionization source at a flow rate of 5 µL/min with no chromatographic or other separation step prior to ionization. Analyte ions were detected in positive-ion mode with the spray voltage set to 5.0 kV.

FAIMS Instrumentation and Methods

Experiments were conducted using a commercial Owlstone UltraFAIMS system interfaced to a Thermo Scientific LTQ XL linear ion trap mass spectrometer. The

UltraFAIMS chip cell consists of 21 parallel channels of paired gold-coated electrodes.

Each channel is 4.62 mm long and has an analytical gap size of 100 µm. The separation path length is approximately 700 µm.

FAIMS separation was performed at DFs ranging from 10 to 250 Tds; the highest attainable DF of 260 Td corresponds to an electric field of 59 kV/cm (590 V divided by a gap distance of 100 µm) normalized to gas density N at temperature 61 °C. The CF was scanned from -5 to 5 Tds over 60 seconds. FAIMS data were extracted from the mass spectra for each opioid using Thermo Xcalibur Qual Browser and processed in Microsoft

Excel, R V3.3.2, RStudio V1.0.153, and MS Convert V 3.0.9134 0.

We define resolution for FAIMS with Equation 4-1:

|퐶퐹2 − 퐶퐹1| 푅푠 = (4-1) 퐹푊퐻푀푎푣푒

80

where Rs is the resolution, CF1 and CF2 are the CF peak values in Td corresponding to the analytes of interest 1 and 2, respectively, and FWHMave is the average peak width of analyte 1 and 2 at 50% of the full height in Td.

Results and Discussion

Resolving Androsterone Epimers

The structural formulae and 3D structures for the epimers androsterone, trans- androsterone, and etiocholanolone are presented in Figure 4-2. Androsterone and trans-androsterone differ from each other only in the stereochemistry of the hydroxyl moiety at the C3 position, while androsterone and etiocholanolone differ from each other only in the conformation of the ring at the C5 position. Previous work demonstrated that the epimers androsterone and trans-androsterone (in Chapter 3) can be resolved from each other in FAIMS by monitoring the [M+K]+ ions of each compound and served as the starting point for the investigations in this work. Figure 4-3 shows the overlaid FAIMS spectra for the [M+K]+ of all five anabolic steroids of interest, as well as the mass spectra across the CF peaks in the range from 1.5 to 3.5 Td for the three androsterone epimers. The FAIMS spectra for all five steroid ions show CF peaks in two general ranges: one peak centered at 0.4 Td, and another peak in the range from 1.5

Td to 3.5 Td. The CF peaks at 0.4 Td overlap significantly and cannot be used to resolve any of the androsterone isomers or isobars. However, the CF peaks between

1.5 and 3.5 Td show separation between androsterone isomers. As previously reported, androsterone and trans-androsterone are well-resolved from each other. The mass spectra between the CF peak at 1.9 Td for androsterone and the peak at 2.7 Td for trans-androsterone (Figure 4-3b and 4-3d, respectively) reveal significantly higher relative intensities for multimer adducts of androsterone as opposed to trans-

81

androsterone. In addition, potassiated adduct ions for trans-androsterone are formed in much lower relative abundance than sodiated adduct ions, suggesting that the trans- androsterone molecule has much higher affinity for sodium cations than potassium cations. The CF peak at 2.1 Td for etiocholanolone partially resolves from the peak for androsterone at 1.9 Td and from the peak for trans-androsterone at 2.7 Td. The mass spectrum for etiocholanolone at CF 2.1 Td is similar to that for androsterone, although the relative intensities of potassiated adduct ions compared to sodiated ions are lower than those observed for androsterone (Figure 4-3c).

The partial resolution of etiocholanolone from trans-androsterone matches expectations from the low-field drift tube mobility separation of the two epimers.

However the resolution between androsterone from etiocholanolone is dramatically improved using FAIMS in contrast to their drift tube mobility separation. Similar relative intensities for cation adducts between androsterone and etiocholanolone suggest that the change in ring conformation at the C5 position produces notable changes in the high-field mobilities of each steroid but does not affect the ability of each steroid to form multimer adduct ions. This result is somewhat surprising when considering that the ring conformation change at the C5 position results in a molecule with a more chair-like structure (Figure 4-2b) as opposed to the relatively planar structures for androsterone and trans-androsterone (Figures 4-2a and 4-3c). From the theoretical structures of androsterone and trans-androsterone dimer ions from previous work and the 3D structures of the three androsterone epimers, we hypothesize that the stereochemistry change of the hydroxyl moiety at the C3 position produces ion cluster structures that

82

have greater mobility differences than those produced from the change in ring conformation at the C5 position.

Resolving Non-Epimer Isomers and Isobars

The FAIMS resolution between androsterone epimers was enhanced when monitoring the [M+K]+ ions. However, the [M+K]+ ions for DHT and oxabolone could not be resolved from the androsterone epimers. In Figure 4-3a, the FAIMS spectrum for

DHT shows CF peaks at 0.5 and 2.8 Td. However, the CF peak at 2.8 Td for DHT overlaps significantly with the peak at 2.7 Td for trans-androsterone. The CF peaks for potassiated multimer ions of DHT also overlapped with those for trans-androsterone.

Interestingly, the mass spectrum for DHT across the CF peak at 2.8 Td closely resembles that for trans-androsterone, with relatively weak signal for the [M+K]+ ion and lower intensities for multimer ions. This close resemblance in mass spectra between

DHT and trans-androsterone was observed for most cation adduct species that were monitored in this work and suggest that both steroids form very similar ion cluster structures during FAIMS separations. The FAIMS spectrum for the [M+K]+ ion of oxabolone shows CF peaks at 0.5 and 1.8 Td, and the CF peak at 1.8 Td overlaps with the peak for androsterone at 1.9 Td. As previously shown, the CF peaks at 0.5 Td for the [M+K]+ ions of all five tested steroids overlap significantly and do not allow us to resolve any of the compounds. It should be noted that the CF peak for oxabolone at 1.8

Td does allow it to be resolved from the peak for DHT at 2.8 Td. However, better separation between oxabolone and the androsterone isomers in general can be achieved using other cation adducts species that will be shown.

Lithiated adduct ions displayed the best FAIMS separation of DHT from the other androsterone isomers in addition to dramatically better resolution between the CF peaks

83

for each androsterone epimer. Figure 4-4 shows the FAIMS spectra for the [M+Li]+ and

[3M+Li]+ ions of DHT and the three tested androsterone epimers. From Figure 4-4a, two

CF peaks for the [3M+Li]+ trimer ion of DHT can be observed at 0.5 Td and 1.6 Td, in contrast to one peak each for androsterone at 3.3 Td and for etiocholanolone at 2.5 Td.

The FAIMS spectra for oxabolone is not shown in Figure 4-4a because the overall signal intensity for the [3M+Li]+ ion of oxabolone was extremely low and cannot be distinguished reliably from instrument noise; low signal for the multimer ion adducts of oxabolone is typical and will be discussed later in this chapter. Both CF peaks for DHT are highly resolved from the peaks for androsterone and etiocholanolone. For the CF peak at 1.6 Td, the Rs was calculated to be ~2.0 between DHT and etiocholanolone and

~3.5 between DHT and androsterone. FAIMS spectra for the [3M+Li]+ also show near baseline resolution between the epimers androsterone and etiocholanolone, with a Rs of

~1.5. Unfortunately, the CF peak for trans-androsterone at 1.8 Td overlaps with the peak for DHT at 1.6 Td, preventing us for relying exclusively on the [3M+Li]+ for the separation of androsterone isomers. Nonetheless, the trans-androsterone CF peak at

1.8 Td still resolves well from the peaks for androsterone and etiocholanolone, with a Rs of ~3.1 between trans-androsterone and androsterone and a Rs of ~1.5 between trans- androsterone and etiocholanolone. The resolution of DHT from trans-androsterone can also be achieved using the CF peak at 0.5 Td, which is normally not possible for peaks in the range from 0.0 to 1.0 Td. In this case, only DHT shows significant signal for the

[3M+Li]+ ion species in the CF 0.0 to 1.0 Td range and can be baseline resolved from androsterone and etiocholanolone. It should be noted that the CF peak at 0.5 Td allows us to resolve DHT from trans-androsterone, but not vice versa.

84

Fortunately, the [M+Li]+ monomer ion also proves to be useful in the separation of the four tested androsterone isomers, especially for the resolution of trans- androsterone from DHT. From Figure 4-4b, we can see that most of the androsterone isomers have two CF peaks: one at about 1.8 Td, and a second peak in the range from

2.0 to 3.0 Td. The CF peaks for [M+Li]+ ions at 1.8 Td overlap for all androsterone isomers. However, the CF peaks in the range from 2.0 to 3.0 Td show at least partial separation from each other. In particular, the FAIMS spectrum for trans-androsterone shows a CF peak at 2.3 Td that resolves from the peak at 2.8 Td for DHT with a Rs of

~1.3. The CF peak at 2.3 Td for the [M+Li]+ appears to be unique for trans-androsterone and can be used for identification in the same manner as the 0.5 Td peak for the

[3M+Li]+ ion of DHT. Thus, we are able to resolve each isomer of androsterone from each other by monitoring the lithiated monomer and trimer adduct ions.

Separation of oxabolone from the other steroid compounds used in this work was achieved by the FAIMS separation of their sodiated adduct ions. Figure 4-5 shows the

FAIMS spectra for the [M+Na]+ ions for all five androsterone isomers and isobars, along with the FAIMS spectra for the sodiated monomer, dimer, and trimer adducts of oxabolone. From Figure 4-5a, we can see that the FAIMS spectra for androsterone, trans-androsterone, etiocholanolone, and DHT have CF peaks across two ranges: one range from -0.5 to 1.0 Td, and another range from 1.0 to 3.5 Td. The number of peaks and their centers varies across each range for the androsterone isomers described, but the peaks for any pair of isomers overlap extensively. In contrast, the [M+Na]+ ion for oxabolone shows a unique CF peak at 1.4 Td, along with a smaller peak at 2.4 Td and notable signal in the range from 0.0 to 1.0 Td. The CF peak at 2.4 Td overlaps

85

completely with those described previously, but the unique peak at 1.4 Td can be used to resolve oxabolone from all of the other androsterone isomers and isobars. The mean resolution value for the oxabolone CF peak versus the most intense peak for each steroid isomer across the range 1.0 to 3.5 Td was calculated to be ~2.1 with a standard deviation of 0.5. The lowest Rs value was calculated to be ~0.8 and corresponds to the less intense peak for etiocholanolone at CF 1.8 Td, while the highest Rs value was calculated to be ~2.9, corresponding to the less intense peak for androsterone at CF 2.8

Td. The signal for oxabolone in the CF range from 0.0 to 1.0 Td overlaps with the CF peaks for the other steroid compounds, although it does suggest that there is another

CF peak for the [M+Na]+ ion in that range that may “split” off from the peak at 1.4 Td at higher DF values. The phenomenon of CF peak splitting will be discussed in the following section.

The FAIMS spectra for the sodiated multimer adducts of oxabolone may also be useful in the separation of androsterone isomers and isobars. From Figure 4-5b, we also noted that the signal intensities for the [2M+Na]+ and [3M+Na]+ ions of oxabolone are dramatically lower than the intensity of its [M+Na]+ signal. The highest signal intensity for the [2M+Na]+ ion was observed at CF 1.0 Td and constituted less than 2% of the highest signal observed for the [M+Na]+ ion. The signal intensities for the

[3M+Na]+ ion of oxabolone were similar to the lithiated trimer ion and could not be distinguished reliably from instrument noise. This trend appears for oxabolone cation adducts in general, where the monomer cation adduct signal is significantly more intense than the signals for dimer and trimer adducts. Methods aiming to deconvolute

86

the FAIMS spectra of isobaric androsterone species could take advantage of this trend to prevent oxabolone signal from interfering with those for other steroids.

Effects of DF on CF Peaks

The effect of DF on the CF shifts was also investigated for the five androsterone isomers and isobars. Figure 4-6 shows a plot of the CF peaks with the best separation for the [M+K]+ ions of all the tested steroids as the DF is increased, as well as the representative FAIMS spectra for each steroid at DF 250 Td. From Figure 4-6a, we can see that there are two trends in CF shifts for each steroid. One trend is for CF peaks that split off from a central peak at a DF around 130 Td and shift toward larger positive

CF values. The other trend is for CF peaks that split but remain in the CF range from

0.0 to 0.5 Td. Two patterns of lines are displayed in the CF range 0.0 to 0.5 Td: one set of solid colored lines with data markers (x) at each tested DF, and another set of dashed colored lines with no data markers. The set of colored lines with data markers correspond to the CF value of local maxima that are assumed to be peaks observed in actual experimental data. The set of dashed lines with no data markers are artificial and represent the theoretical CF values where the peaks observed at higher DFs are thought to transmit. In many cases, distortions in the shape of the central CF peak suggest that the two peaks observed in the steroid FAIMS spectrum have begun splitting from each other, but the lack of a local maximum prevented us from assigning a

CF value. Nevertheless, the CF peaks between 0.0 and 0.5 Td that were actually observed in our experiments greatly resemble those shown in Figure 4-6b and did not resolve between the steroids at any DF.

For the CF peaks that shift toward larger positive CF values, we can see that the

5 steroids appear to begin resolving at a DF of 150 Td. The separation between these

87

CF peaks improves with higher DFs, although trans-androsterone and DHT remain unresolved at DF 250 Td as shown in the previous section and do not appear to gain resolution with increasing DF. The CF peak at 1.9 Td for the [M+K]+ ion of oxabolone also overlaps with the peak for androsterone as previously shown, although we noted that the lower relative signal intensity of the peak at 1.9 Td versus the peak at 0.4 Td occurs at every tested DF and runs contrary to the signal intensity behavior for the other steroids. In addition, the absolute signal intensities for the CF peaks of oxabolone are uncharacteristically low among its monomer ions. These two factors limit the usefulness of resolving oxabolone from the other steroids using the [M+K]+ ion and the impact that oxabolone has on the CF peak at 1.9 Td for androsterone. At DFs higher than 250 Td, we assume that the CF peaks at 1.9 Td for androsteorne and oxabolone will continue to overlap, although the signal intensity contribution of oxabolone ions will become negligible compared to that from androsterone.

The CF peak splitting that results from increasing DFs has neither been reported in the literature nor previously observed by our group. We characterized this CF peak splitting using sodiated adduct ions of androsterone. Figure 4-7 shows a plot of each CF peak observed for the [M+Na]+ ion of androsterone with increasing DFs, as well as the

FAIMS spectra for sodiated monomer and multimer ions of androsterone at DFs 150,

210, and 250 Td. Similar to the case with potassiated adduct ions, multiple CF peaks appear to split from a central peak and resolve from each other as the DF is increased from 150 Td. In Figure 4-7a, the red dashed lines in the DF range from 150 to 190 Td with no data markers are artificial and represent the theoretical CF values where the peaks observed at higher DFs are thought to transmit. For the [M+Na]+ ion of

88

androsterone, three CF trends appear in the FAIMS spectra with increasing DF. One trend results in the CF 0.3 Td peak at DF 250 Td and shifts in the CF range from 0.0 to

0.5 Td. The second and third trends shift toward in larger positive CFs with increasing

DF and result in the peaks at 2.2 Td and 2.8 Td. The two trends that shift toward larger

CFs were observed to split from each other at a DF of 190 Td.

At DF 250 Td, mass spectra across each observed CF peak revealed notable signal for sodiated multimer ions, suggesting that some of the peaks result from the transmission of a multimer species that fragments prior to mass analysis. In Figure 4-

7b, the FAIMS spectra for the [2M+Na]+ and [3M+Na]+ ions of androsterone show overlap with the CF peaks at 0.3 and 2.2 Td for the [M+Na]+ and supports the idea that transmission of the larger sodiated multimer ions are primarily responsible for those peaks. Because the FAIMS spectra for both the [2M+Na]+ and [3M+Na]+ ions closely resemble each other for both CF peaks, we hypothesize that two different ion cluster structures of the sodiated trimer ion are resolved in FAIMS. One ion cluster structure results in the CF peak at 0.3 Td, and the structure results in the peak at 2.2 Td. Note that the CF peak at 2.8 Td only appears in the FAIMS spectrum for the sodiated monomer ion. The mass spectrum across the peak shows ion signal predominately for the [M+Na]+ ion. The lower resolution for the trimer ion CF peak at 2.2 to the monomer ion peak at 2.8 relative to the other trimer ion peak at 0.3 is surprising, as one would initially assume that the ion cluster structures for monomer ions in FAIMS should be significantly different that the structures for trimer ions and produce larger differences in

CF peaks. This phenomenon may indicate two divergent mechanisms of ion cluster formation for androsterone and is currently under investigation.

89

Effects of Cation Size on CF Peaks

The effects of cation size and electronegativity on observed CF peaks during

FAIMS was investigated for monomer adduct ions of protonated, lithiated, sodiated, potassiated, and ammoniated species. Table 4-1 lists the cation species used in this work with the size of their ionic radii and electronegativities. The value for the ionic radius and electronegativity of ammonium was taken from Whiteside el al., 2011.71 The atomic radius of hydrogen is presented instead of the radius of a proton (H+), as the size of the electron cloud constitutes nearly the entire volume for all the other cations used in this work and should still be a significant factor around the ionizing hydrogen for protonated adduct ions. Figure 4-8 shows the plot of observed CF peak values for monomer adduct ions of androsterone with various cations as a function of ionic radii, along with the FAIMS spectrum for the [M+Na]+ ion at DF 250 Td. The three different markers on the plot in Figure 4-8a correspond to each CF peak in the FAIMS spectrum in order of most intense to least intense and are illustrated as an example in Figure 4-

8b. From the plot of CF shifts with increasing cation size, the [M+H]+ ion shows a CF peak at about 1.9 Td. As the size of the cation increases from hydrogen (53 pm) to lithium (76 pm), the FAIMS spectrum for the [M+Li]+ shows a new CF peak at about 2.8

Td in addition to the peak at 1.9 Td. This pattern occurs again as the cation size increases from lithium to sodium (102 pm), which has a FAIMS spectrum for the

[M+Na]+ ion that shows a new CF peak at about 0.3 Td as well similar peaks to those from [M+Li]+. However, no new CF peaks were observed for cations larger than sodium; the FAIMS spectrum for potassiated (138 pm) monomer ions of androsterone retain CF peaks in similar ranges as those observed for [M+Na]+, while the spectrum for ammoniated (146 pm) monomer ions actually lose the peak at 2.8 Td.

90

We noted that the magnitude of CF shifts does not appear to change with increasing cation size. For example, the center values of the CF peaks around 1.9 Td for all monomer cation adducts do not appear to exhibit either an increasing or decreasing trend with larger cation size. On average, those peaks have a mean CF value of 1.9 Td and a standard deviation of ± 0.2 Td. However, the FWHMave for those peaks is 0.5 Td. This suggests that most of the variations in the peak center lie within the average width of the peak and are due to random fluctuations during data collection.

The same behavior also appears to be true for the other two groups of CF peaks around

0.3 Td and 2.8 Td. The effect of electronegativities on the CF peaks of androsterone produced very similar results.

While the size of the cation does not appear to affect the magnitude of CF shifts, we observed that certain CF peaks do not appear in the FAIMS spectra of monomer adduct ions for small cations. From Figure 4-8a, three groups of CF peaks can be seen across all tested cations, which are listed as follows in order from most intense to least intense: one group around 1.9 Td, a second group around 2.8 Td, and the third group around 0.3 Td. As previously stated, a CF peak around 1.9 Td can be consistently observed for all monomer cation adducts of androsterone tested in this work. However, the peak at 2.8 Td is only observed for monomer ions when the cation is larger than or equal to lithium (76 pm). In a similar manner, the peak at 0.3 Td is only observed when the cation is larger than or equal to sodium (102 pm). This result suggests that each of the aforementioned CF groups correspond to their own common ion cluster structure, and that the only dependence on cation size for these structures is whether or not the cation is too small to allow formation. Because the magnitude of CF shifts does not

91

appear to change with respect to cation size, this result also suggests that the factors in

FAIMS that each of these common ion cluster structures experience do not produce significant mobility differences with respect to the size of the cation.

Cation Concentration Effects

Initial experiments investigating the effects of cation concentration on the CF peaks were conducted for sodiated and lithiated adduct species of androsterone.

Lithiated adducts were chosen because they produced the best overall FAIMS separation of the steroids used in this work, while sodiated adducts were chosen because their monomer and multimer ions are often the most intense ion species observed in the mass spectra regardless of the steroid and cation acetate salt added into sample solutions. For example, Figure 4-9a shows the mass spectrum (no FAIMS separation) of an androsterone sample solution with 1.0 µg/mL lithium acetate salt added. Despite no addition of extra sodium into the sample solution, the highest signal intensity observed in the mass spectrum corresponded to the [2M+Na]+ ion. The relative intensities for lithiated ions were generally lower than those for sodiated ions but still comparable between monomer and multimer species.

Figure 4-9b shows the FAIMS spectra for the [2M+Li]+ and [2M+Na]+ ions for androsterone solutions, each with 1.0 µg/mL of its respective cation acetate salt added.

The plot in Figure 4-9c depicts the change in signal intensities for the most intense CF peaks observed in each FAIMS spectra from the addition of cations at three concentrations, as well as the linear regression equation and coefficient of determination for each data set. Cation concentrations were converted from µg/mL to

µM in order to more accurately reflect the concentration of cation added to sample solutions. Relatively high cation concentrations were investigated in this preliminary

92

work for method development and reproducibility. Each signal intensity was determined from its own FAIMS spectrum, and three measurements were used to calculate the average signal and standard deviation at each concentration. For [2M+Li]+ ions, we observed a positive linear trend between the signal intensity of the CF peak at 2.6 Td and the concentration of added lithium ions. However, the signal intensities for

[2M+Na]+ ions do not show the same positive linear trend. In Figure 4-9c, we observed that the signal for the CF peak at 2.0 Td slightly decreased as the concentration of added sodium ions increased over an order of magnitude. It should be noted that the low value for the coefficient of determination for the case with sodium ion addition suggests that there is not truly a negative linear trend between signal and added concentration. Instead, it is much more likely that the signal intensity for the sodiated dimer CF peak did not change with sodium concentration because the tested range was significantly lower than the background concentration of sodium derived from materials and instrumentation. Experiments conducted by our lab measured the background sodium concentration in our methanol solvent to be about 150 µM, which is ten times more concentrated than the amount of sodium ion added to sample solutions in this work.

Conclusions

In this work, we characterized the FAIMS separation of five isomers and isobars of androsterone using the adduct ions formed with cations that are commonly utilized in mass spectrometry. Special focus was given to the effects on the FAIMS spectra resulting from changes in DF, cation size, and cation concentration. Overall, we observed that some cation adducts for some steroids produced unique CF peaks that could be used for resolution of different steroids. Mass spectra across many of the

93

observed CF peaks reveal notable signal intensities for larger multimer ions in addition to signal for the monomer, suggesting that some of these CF peaks correspond to the transmission of the larger multimer adduct and fragmentation of the adduct prior to mass analysis. Other CF peaks appeared uniquely for specific monomer or multimer cation adducts. Increasing the DF of the FAIMS separation generally improved the separation of the five steroids, although we noticed that the CF peaks observed at DF

250 Td appeared to have resulted from peak splitting from the central peak that appears at lower DFs. The peak splitting phenomenon was explored using sodiated adduct ions of androsterone and may indicate the presence of multiple ion cluster structures for a particular cation adduct that are stable enough to be transmitted through at different

CFs. Cation size was not found to affect the magnitude of CF shifts, although certain CF peaks were not observed for certain adduct ions with smaller cations. Preliminary experiments with cation concentration show positive linear response when adding larger concentrations of cation except in the case of sodium. Addition of sodium did not appear to produce any change in CF peak signal for sodiated adducts; this most likely results from the background concentration of sodium ions being at least ten times higher than the additional sodium ions added to sample solutions used for this work.

Future research efforts aim at characterizing changes in the FAIMS spectra from lower concentrations of cations, investigating effects from lower analyte concentrations, and integrating FAIMS for cation adducts into an existing steroid screening method for validation. We assume that the relative signal intensity and presence of certain CF peaks are concentration dependent; limiting the amount of cation or analyte is expected

94

to reduce the relative intensity of those peaks and corresponds to hindering the formation of the associated ion cluster structure.

Multiple examples from our work suggest that ion structures more complex than bare analyte multimer adducts are the principle species that actually undergo high-field mobility separations. Characterization of these structures will be critical to advancing our understanding of ion mobility at high electric fields and may be possible by coupling

FAIMS with techniques that provide structural information, such as conventional ion mobility spectrometry or infrared spectroscopy. Theoretical modeling and simulations of the high-field mobility separations of these five steroid compounds can validate the experimental characterization and improve the development of future methods.

95

Table 4-1. Ionic radii and electronegativities for the cation species used for the separation of androsterone isomers and isobars. The effective ionic radius of ammonium was taken from Whiteside et al., 2011.71 Electronegativity Adduct Species Ionic Radius (pm) (Paulings Scale)

Hydrogen* (H) 53* 2.20

+ Lithium Ion (Li ) 76 0.97

+ Sodium Ion (Na ) 102 0.91

+ Potassium Ion (K ) 138 0.73

+ Ammonium Ion 146 0.77 (NH4 ) *The atomic radius is used for hydrogen instead of the radius of a proton.

96

Figure 4-1. Chemical structures for five androsterone isomers and isobars, including molecular formulae and weights. Structures are divided into (a) epimers of androsterone, (b) an isomer of androsterone, and (c) an isobar of androsterone. Stereochemistries at the C3 and C5 positions are labeled for each epimer in (a).

97

Figure 4-2. Chemical and 3D structures of the epimeric steroids (a) androsterone, (b) etiocholanolone, and (c) trans-androsterone. 3D structures were adapted from ChemSpider Online Database.72

98

Figure 4-3. FAIMS spectra for the [M+K]+ ions of androsterone isomers and isobars at DF 250. FAIMS spectra for all five steroids are shown in (a). Mass spectra for CF peaks between 1.5 and 3.5 Td are shown for (b) androsterone, (b) etiocholanolone, and (c) trans-androsterone.

99

Figure 4-4. FAIMS spectra for the (a) [3M+Li]+ and (b) [M+Li]+ ions of androsterone isomers. Spectra of the [3M+Li]+ ions allows resolution of androsterone and etiocholanolone from trans-androsterone/DHT, while spectra of the [M+Li]+ ions allows the resolution of trans-androsterone from the other isomers. All spectra were acquired at a DF of 250 Td.

100

Figure 4-5. FAIMS spectra for the sodiated adduct species of oxabolone. The FAIMS spectra for [M+Na]+ monomer ions of oxabolone and androsterone isomers at DF 250 Td are shown in (a). (b) FAIMS spectra for monomer and multimer sodiated adduct ions of oxabolone show dramatically lower signal intensities for the multimer species.

101

Figure 4-6. Plot of CFs with increasing DFs for the [M+K]+ ions of androsterone isomers and isobars and overlaid FAIMS spectra. The graph of CF peak value with increasing DF strength is shown in (a). The dashed lines from DF 90 to 170 Td represent hypothesized CF values. (b) FAIMS spectra for the [M+K]+ ions of the five steroids shown previously in Figure 4-3a.

102

Figure 4-7. Plot of CFs with increasing DFs for the [M+Na]+ ion of androsterone and corresponding FAIMS spectra. The graph of CF peak value with increasing DF strength is shown in (a). Patterned lines and data markers represent different CF peaks observed in the FAIMS spectra at each DF (b). The red dashed lines represent hypothesized CF values.

103

Figure 4-8. Plot of CF shifts with increasing cation size for monomer adduct ions of androsterone at DF 250 Td. The data markers shown by the graph in (a) represent different CF peaks as shown in (b) and are presented in order from most intense to least intense.

104

Figure 4-9. Plot showing the change in CF peak signal intensities at various cation concentrations. (a) Mass spectra for an androsterone sample solution with 1.0 µg/mL (15 µM) show significant signal for sodiated adduct ions. (b) The CF peak signals for each cation adduct at each cation concentration added were used to generate the plot in (c).

105

CHAPTER 5 CONCLUSIONS AND FUTURE WORK

This research presented in this dissertation investigates the effects of chemical modifiers on the sensitivity, resolving power, and peak shifts for high-field asymmetric waveform ion mobility spectrometry (FAIMS). In particular, FAIMS with chemical modifiers was coupled with mass spectrometry (FAIMS-MS) and applied to several drugs of abuse for the purpose of separating isomers of these drugs in a rapid manner.

For each study, FAIMS performance changes are characterized using a series of structurally related chemical modifiers. Because FAIMS is a comparatively young analytical technique, many fundamental aspects that affect ion separation remain unknown. The studies described above explore trends in the hopes of enabling new fundamental studies on high-field ion mobility, aiding the development of innovations in

FAIMS, and expanding the market of routine users who utilize ion mobility separations.

Chapter 2 details the application of solvent vapor modified FAIMS-MS for the separation of several opioids and morphine isomers. The abuse and addiction of prescription opioid analgesics has become a national health crisis over the last three decades, with 66% of all drug overdose deaths in 2016 involving opioids in the United

States. A rapid screening method that is capable of resolving different opioids and their isomers would greatly assist the prevention of opioid misuse and the treatment of overdose patients. FAIMS-MS analysis using dry nitrogen carrier gas was unable to resolve ions from morphine isomers. Addition of organic solvent vapor to the carrier gas enables the resolution of some of these isomers, with aprotic solvents producing the best improvement. Plotting the CF peak shifts with respect to solvent vapor concentration reveals that aprotic solvent vapors generally produce the same CF shifts

106

for a given vapor concentration, while protic solvent vapors do not produce this behavior. This result suggests one or more chemical properties are common between small aprotic solvents that produce the same high-field mobility behavior for a given vapor concentration. Aprotic solvents with large molecular weights and protic solvents do not share those common properties.

Chapter 3 describes the use of cation adducts in FAIMS for the separation of anabolic steroid epimers. Anabolic steroids are primarily thought of as performance enhancing drugs that are used by athletes seeking to gain an unfair competitive advantage. However, the majority of anabolic steroid users are not actually professional athletes but noncompetitive bodybuilders who use the steroids recreationally. The proliferation of designer steroids has challenged the regulation of anabolic steroids in both the professional and recreational scenes. Designer steroids are novel compounds that are often synthesized from androgen precursors and are intended to evade regulations and screening tests. A rapid method that is capable of separating anabolic steroid isomers would not only benefit athletic screening labs but would also improve the characterization of unregulated steroids being sold to bodybuilders. FAIMS-MS analysis of the [M+H]+ ions for the epimers testosterone and epitestosterone was not able to resolve the two isomers from each other. However, monitoring the [M+Na]+ ions in the FAIMS spectra for both steroids revealed a unique CF peak for testosterone.

FAIMS spectra acquired for larger Group 1 cation adducts of testosterone show that a total of three CF peaks can be observed across the various cation adducts tested. In addition, these three peaks do not vary in CF value even as the specific cation changes.

This results suggests that each CF peak corresponds to the transmission of a specific

107

ion cluster structure and that the high field mobility of each structure does not significantly change between different cations. Some of the previously described CF peaks appear to result from the transmission of a multimer cation adduct that undergoes fragmentation before mass analysis. Epitestosterone can be resolved from testosterone in FAIMS by monitoring its lithiated trimer ion ([3M+Li]+). Androsterone and trans- androsterone follow similar trends as the testosterone epimers, suggesting that these behaviors can be expected for anabolic steroids in general.

Chapter 4 provides a detailed investigation on the FAIMS effects of cation modifiers for the separation of five androsterone isomers and isobars. Of the five steroids used in this work, two were epimers of androsterone (trans-androsterone and etiocholanolone), one was an isomer of androsterone (dihydrotestosterone), and the final one was isobaric to androsterone (oxabolone). Several of the trends observed in

Chapter 3 were explored in more detail, with special focus on effects resulting from changes in DF, in cation size, and in cation concentration. Androsterone, trans- androsterone, and etiocholanolone could be baseline resolved from each other using the lithiated trimer ion in FAIMS. Two CF peaks were observed for the lithiated trimer ion of dihydrotestosterone; both peaks could be baseline resolved from those of androsterone and etiocholanolone, but only one peak could be used to resolve dihydrotestosterone from trans-androsterone. Increasing the DF of FAIMS separations improved the resolution of CF peaks between the five steroids, but CF peaks were also observed to split with increasing DFs. This result suggests that there may be multiple stable configurations for a particular ion cluster that transmit through the FAIMS. Cation size was not found to affect the magnitude of CF shifts, although it appeared that certain

108

CF peaks do not appear when the size of the cation is too small. Initial cation concentration experiments were conducted that show positive linear response when adding larger concentrations of cation except in the case of sodium. Addition of sodium did not produce any change in CF peak signal for sodiated adducts, althought this most likely resulted from the naturally high background concentration of sodium in solvents and on labware.

Future studies can be divided between methodology modifications learned from current studies and fundamental investigations of the observed phenomenon.

Regarding the first, several hardware modifications should be implemented in order to improve the quality of the FAIMS separation. Temperature control of the carrier gas and solvent vapor is important for acquiring reproducible CF peaks, since variations in temperature affect the ion-neutral interactions in the FAIMS gap. Introducing solvent vapor cooled below ambient temperature could be an elegant way to maintain reproducible vapor concentrations and prevent condensation in the gas lines. Increasing ion residence time in the FAIMS gap should improve CF peak widths; doubling the path length of the Owlstone micromachined FAIMS chip is expected to improve peak widths while mostly retaining the higher ion transmission intrinsic to the original design. Finally, application of a rectangular asymmetric waveform should be pursued in order to improve the resolving power of FAIMS and the quality of empirical data that can be used for fundamental investigations. All current commercial FAIMS instruments utilize a bisinusoidal waveform due to the challenges associated with generating a high- frequency, high-voltage asymmetric wave with modern electronics. However, a

109

rectangular wave is still considered to be the optimal shape for maximizing FAIMS resolutions.

Regarding the second, molecular modeling and simulations of the ion-clusters in the high electric fields described in this work should be pursued. Characterizing the structures of ion-clusters provides insight into the physical parameters that contribute to the high-field mobility coefficient and into the mechanisms that occur when solvent vapor molecules are added into FAIMS. Investigating the collisional cross-sections for

CF peaks using conventional ion mobility spectrometry may provide a method for experimental confirmation of modeling and simulations. Overall, achieving a more in- depth understanding of the fundamentals of high-field ion mobility behaviors will be essential to expanding the applications of FAIMS.

110

LIST OF REFERENCES

(1) Buryakov, I. A.; Krylov, E. V.; Nazarov, E. G.; Rasulev, U. K. Int. J. Mass Spectrom. Ion Process. 1993, 128 (3), 143–148.

(2) Purves, R. W.; Guevremont, R.; Day, S.; Pipich, C. W.; Matyjaszczyk, M. S. Rev. Sci. Instrum. 1998, 69 (12), 4094–4105.

(3) Schneider, B. B.; Nazarov, E. G.; Londry, F.; Vouros, P.; Covey, T. R. Mass Spectrom. Rev. 2016, 35 (6), 687–737.

(4) Yousef, A.; Shrestha, S.; Viehland, L. A.; Lee, E. P. F.; Gray, B. R.; Ayles, V. L.; Wright, T. G.; Breckenridge, W. H. J. Chem. Phys. 2007, 127 (15).

(5) Guevremont, R. J. Chromatogr. A 2004, 1058 (1–2), 3–19.

(6) Rorrer, L. C.; Yost, R. A. Int. J. Mass Spectrom. 2011, 300 (2–3), 173–181.

(7) Sinatra, F. L.; Wu, T.; Manolakos, S.; Wang, J.; Evans-Nguyen, T. G. Anal. Chem. 2015, 87 (3), 1685–1693.

(8) Cumeras, R.; Figueras, E.; Davis, C. E.; Baumbach, J. I.; Gràcia, I. Analyst 2015, 140, 1391–1410.

(9) Schneider, B. B.; Covey, T. R.; Nazarov, E. G. Int. J. Ion Mobil. Spectrom. 2013, 16 (3), 207–216.

(10) Campbell, J. L.; Yang, A. M. C.; Melo, L. R.; Hopkins, W. S. J. Am. Soc. Mass Spectrom. 2016, 27 (7), 1277–1284.

(11) Tsai, C.-W.; Tipple, C. A.; Yost, R. A. Rapid Commun. Mass Spectrom. 2018, 32 (7), 552–560.

(12) Baker, P. R.; Campbell, J. L.; Duchoslav, E.; Hunter, C. AB SCIEX Tech. Appl. Note, RUO-MKT-02-4802-A 2017, 2–7.

(13) Schneider, B. B.; Londry, F.; Nazarov, E. G.; Kang, Y.; Covey, T. R. J. Am. Soc. Mass Spectrom. 2017, 2151–2159.

(14) Liu, C.; Gómez-Ríos, G. A.; Schneider, B. B.; Le Blanc, J. C. Y.; Reyes-Garcés, N.; Arnold, D. W.; Covey, T. R.; Pawliszyn, J. Anal. Chim. Acta 2017, 991, 89–94.

(15) Beach, D. G. J. Am. Soc. Mass Spectrom. 2017, 1–13.

(16) Bowman, A. P.; Abzalimov, R. R.; Shvartsburg, A. A. J. Am. Soc. Mass Spectrom. 2017, 1–10.

111

(17) Willems, J. L.; Khamis, M. M.; Mohammed Saeid, W.; Purves, R. W.; Katselis, G.; Low, N. H.; El-Aneed, A. Anal. Chim. Acta 2016, 933, 164–174.

(18) Šala, M.; Lísa, M.; Campbell, J. L.; Holčapek, M. Rapid Commun. Mass Spectrom. 2016, 30 (2), 256–264.

(19) Guevremont, R.; Purves, R. W. Rev. Sci. Instrum. 1999, 70 (2), 1370–1383.

(20) Miller, R. .; Eiceman, G. .; Nazarov, E. .; King, A. . Sensors Actuators B Chem. 2000, 67 (3), 300–306.

(21) Costanzo, M. T.; Boock, J. J.; Kemperman, R. H. J.; Wei, M. S.; Beekman, C. R.; Yost, R. A. Int. J. Mass Spectrom. 2016.

(22) Shvartsburg, A. A.; Tang, K.; Smith, R. D. J. Am. Soc. Mass Spectrom. 2005, 16 (1), 2–12.

(23) Papanastasiou, D.; Wollnik, H.; Rico, G.; Tadjimukhamedov, F.; Mueller, W.; Eiceman, G. A. J. Phys. Chem. A 2008, 112 (16), 3638–3645.

(24) Prieto, M.; Tsai, C.-W.; Boumsellek, S.; Ferran, R.; Kaminsky, I.; Harris, S.; Yost, R. A. Anal. Chem. 2011, 83 (24), 9237–9243.

(25) Rorrer, L. C.; Yost, R. A. Int. J. Mass Spectrom. 2015, 378, 336–346.

(26) Zhao, Y. Y.; Liu, X.; Boyd, J. M.; Qin, F.; Li, J.; Li, X.-F. J. Chromatogr. Sci. 2009, 47 (1), 92–96.

(27) Covington, J. A.; van. der Schee, M. P.; Edge, A. S. L.; Boyle, B.; Savage, R. S.; Arasaradnam, R. P. Analyst 2015, 140 (20), 6775–6781.

(28) Hebert, A. S.; Prasad, S.; Belford, M. W.; Bailey, D. J.; McAlister, G. C.; Abbatiello, S. E.; Huguet, R.; Wouters, E. R.; Dunyach, J.-J.; Brademan, D. R.; Westphall, M. S.; Coon, J. J. Anal. Chem. 2018, 90 (15), 9529–9537.

(29) Shvartsburg, A. A.; Danielson, W. F.; Smith, R. D. Anal. Chem. 2010, 82 (6), 2456–2462.

(30) Shvartsburg, A. A.; Bryskiewicz, T.; Purves, R. W.; Tang, K.; Guevremont, R.; Smith, R. D. J. Phys. Chem. B 2006, 110 (43), 21966–21980.

(31) Purves, R. W.; Ozog, A. R.; Ambrose, S. J.; Prasad, S.; Belford, M.; Dunyach, J.- J. J. Am. Soc. Mass Spectrom. 2014, 25 (7), 1274–1284.

(32) Schneider, B. B.; Nazarov, E. G.; Covey, T. R. Int. J. Ion Mobil. Spectrom. 2012, 15 (3), 141–150.

112

(33) Schneider, B. B.; Covey, T. R.; Coy, S. L.; Krylov, E. V.; Nazarov, E. G. Anal. Chem. 2010, 82 (5), 1867–1880.

(34) Shvartsburg, A. A.; Seim, T. A.; Danielson, W. F.; Norheim, R.; Moore, R. J.; Anderson, G. A.; Smith, R. D. J. Am. Soc. Mass Spectrom. 2013, 24 (1), 109– 114.

(35) Blagojevic, V.; Koyanagi, G. K.; Bohme, D. K. J. Am. Soc. Mass Spectrom. 2014, 25 (3), 490–497.

(36) Wang, Y.; Wang, X.; Li, L.; Chen, C.; Xu, T.; Wang, T.; Luo, J. Rapid Commun. Mass Spectrom. 2016, No. May, 1914–1922.

(37) Beekman, C. Ion Mobility Mass Spectrometry as a Rapid Separations Technique for Isomeric, Metabolomic, and Clinical Analysis. Ph.D. Dissertation, University of Florida, 2015.

(38) Kafle, A.; Coy, S. L.; Wong, B. M.; Fornace, A. J.; Glick, J. J.; Vouros, P. J. Am. Soc. Mass Spectrom. 2014, 25 (7), 1098–1113.

(39) Konermann, L.; Ahadi, E.; Rodriguez, A. D.; Vahidi, S. Anal. Chem. 2013, 85 (1), 2–9.

(40) Teunissen, S. F.; Eberlin, M. N. J. Am. Soc. Mass Spectrom. 2017, 28 (11), 2255–2261.

(41) Paul, W.; Steinwedel, H. Verfahren zur Trennung bzw. zum getrennten Nachweis von Ionen verschiedener spezifischer Ladung. DE944900C, 1956.

(42) March, R. E. J. Mass Spectrom. 1997, 32 (4), 351–369.

(43) Hager, J. W. Rapid Commun. Mass Spectrom. 2002, 16 (6), 512–526.

(44) Mayya, V. Mol. Cell. Proteomics 2004, 4 (2), 214–223.

(45) Douglas, D. J.; Frank, A. J.; Mao, D. Mass Spectrom. Rev. 2005, 24 (1), 1–29.

(46) Mayya, V.; Rezaul, K.; Cong, Y.-S.; Han, D. Mol. Cell. Proteomics 2005, 4 (2), 214–223.

(47) Watson, J. T.; Sparkman, O. D. Introduction to Mass Spectrometry: Instrumentation, Applications, and Strategies for Data Interpretation, 4th ed.; John Wiley & Sons Ltd, 2007.

113

(48) Costanzo, M. Investigation and Development of Novel Methodologies for Disease Diagnosis and Drug Monitoring Utilizing Mass Spectrometry and Field Asymmetric Ion Mobility Spectrometry. Ph.D. Dissertation, University of Florida, 2015.

(49) March, R. E. Mass Spectrom. Rev. 2009, 28 (6), 961–989.

(50) Thermo Fisher Scientific. LTQ XL Hardware Manual, Rev A.; 2006.

(51) Hedegaard, H.; Warner, M.; Minino, A. M. NCHS Data Brief 2017, No. 294, 1–8.

(52) Kolodny, A.; Courtwright, D. T.; Hwang, C. S.; Kreiner, P.; Eadie, J. L.; Clark, T. W.; Alexander, G. C. Annu. Rev. Public Health 2015, 36 (1), 559–574.

(53) Vowles, K. E.; McEntee, M. L.; Julnes, P. S.; Frohe, T.; Ney, J. P.; van der Goes, D. N. Pain 2015, 156 (4), 569–576.

(54) Gergov, M.; Nokua, P.; Vuori, E.; Ojanperä, I. Forensic Sci. Int. 2009, 186 (1–3), 36–43.

(55) Pergolizzi, J.; Pappagallo, M.; Stauffer, J.; Gharibo, C.; Fortner, N.; De Jesus, M. N.; Brennan, M. J.; Richmond, C.; Hussey, D. Pain Pract. 2010, 10 (6), 497–507.

(56) Manicke, N. E.; Belford, M. J. Am. Soc. Mass Spectrom. 2015, 26 (5), 701–705.

(57) Grissinger, M. P&T 2008, 33 (4), 194–247.

(58) Chouinard, C. D.; Cruzeiro, V. W. D.; Roitberg, A. E.; Yost, R. A. J. Am. Soc. Mass Spectrom. 2017, 28 (2), 323–331.

(59) Graham, M. R.; Davies, B.; Grace, F. M.; Kicman, A.; Baker, J. S. Sport. Med. 2008, 38 (6), 505–525.

(60) Pope, H. G.; Wood, R. I.; Rogol, A.; Nyberg, F.; Bowers, L.; Bhasin, S. Endocr. Rev. 2014, 35 (3), 341–375.

(61) El Osta, R.; Almont, T.; Diligent, C.; Hubert, N.; Eschwège, P.; Hubert, J. Basic Clin. Androl. 2016, 26 (1), 1–8.

(62) Franke, W. W.; Berendonk, B. Clin. Chem. 1997, 43 (7), 1262–1279.

(63) Rahnema, C. D.; Crosnoe, L. E.; Kim, E. D. Andrology 2015, 3 (2), 150–155.

(64) Rahnema, C. D.; Lipshultz, L. I.; Crosnoe, L. E.; Kovac, J. R.; Kim, E. D. Fertil. Steril. 2014, 101 (5), 1271–1279.

114

(65) Perry, P. J.; Lund, B. C.; Deninger, M. J.; Kutscher, E. C.; Schneider, J. Clin. J. Sport Med. 2005, 15 (5), 326–330.

(66) Maravelias, C.; Dona, A.; Stefanidou, M.; Spiliopoulou, C. Toxicol. Lett. 2005, 158 (3), 167–175.

(67) Ahonen, L.; Fasciotti, M.; Gennäs, G. B. af; Kotiaho, T.; Daroda, R. J.; Eberlin, M.; Kostiainen, R. J. Chromatogr. A 2013, 1310, 133–137.

(68) Chouinard, C. D.; Beekman, C. R.; Kemperman, R. H. J.; King, H. M.; Yost, R. A. Int. J. Ion Mobil. Spectrom. 2016, 1–9.

(69) Arthur, K. L.; Turner, M. A.; Brailsford, A. D.; Kicman, A. T.; Cowan, D. A.; Reynolds, J. C.; Creaser, C. S. Anal. Chem. 2017, 89 (14), 7431–7437.

(70) Sjöqvist, F.; Garle, M.; Rane, A. Lancet 2008, 371 (9627), 1872–1882.

(71) Whiteside, A.; Xantheas, S. S.; Gutowski, M. Chem. - A Eur. J. 2011, 17 (47), 13197–13205.

(72) ChemSpider; Royal Society of Chemistry, 2015.

115

BIOGRAPHICAL SKETCH

Michael Shenming Wei was born in Salt Lake City, Utah, in the year of 1990 to his mother, Shuchun Liang, and his father, Xing Wei, who was pursuing a Ph.D. in physics at the University of Utah. Michael and his family moved from Salt Lake City to

Alburquerque, New Mexico (notable for having four seasons) and then finally to in 1997 to Tallahassee, Florida (notable for not). Michael’s sister, Katherine, was born in 1998 in

Tallahassee. As child, Michael’s fascination with science began through his father’s work at the National High Magnetic Field Laboratory and particularly with chemistry in his high school Advanced Placement chemistry class. After graduating from high school, Michael attended Duke University for his undergraduate education, pursuing a

Bachelors of Science in chemistry.

At Duke University, Michael performed his financial aid work study in the Mass

Spectrometry Services lab headed by Dr. George Dubay, where he first learned about the analytical technique. He enjoyed his experience with mass spectrometry so much that he later participated in undergraduate research in the same lab under Dr. Dubay.

His project involved analyzing lipid residues on ancient Greek pottery for the characterization of food stuffs and trade goods. In his spare time, Michael indulged in board and video games with friends that he continues to play with today.

After graduating from Duke University, Michael began his graduate career at the

University of Florida under the direction of Dr. Richard Yost. After completing his degree, Michael would like to continue instrumentation research in either an industry or academic setting.

116