The Use of In Vitro and In Silico Technologies for Predicting Human Pharmacology and Toxicology of

Item Type dissertation

Authors Feasel, Michael

Publication Date 2017

Abstract Carfentanil is an ultra-potent of public health and chemical weapons defense concern. Exposure to carfentanil has been seen in both illicit drug use and in the resolution of a hostage situation in Moscow, Russia in 2002. While opioid pharmacol...

Keywords carfentanil; PBPK; ; ; Pharmacology; Toxicology

Download date 30/09/2021 18:28:51

Link to Item http://hdl.handle.net/10713/6764 Curriculum Vitae

Name: Michael G. Feasel

Contact: EMAIL: [email protected]

Degree and Date to be Conferred: Ph.D., 2017

Education: University of Maryland Baltimore, Graduate School (2010 - 2017) Ph.D. in Toxicology (2017) The University of Texas at Austin (2004 - 2008) B.S. in Biochemistry (2008)

Positions and Employment: 2012 – Present: Chemist, U.S. Army - Edgewood Chemical Biological Center, Operational Toxicology, Aberdeen Proving Ground, MD 21010. Principal investigator/researcher of various in vivo toxicity studies of riot control, non-lethal, simulant, and lethal chemical agents. Principal investigator of in vitro chemical compound metabolism, receptor efficacy, pharmacokinetics, and in silico modeling of physiochemical and pharmacological modeling.

2008 - 2012: Chemist, U.S. Army - Edgewood Chemical Biological Center, Smoke and Target Defeat, Aberdeen Proving Ground, MD 21010. Chemical non-lethal anti-personnel weapons/technology researcher. Studied operational use of riot control agents, malodorants, tagging, tracking and locating (TTL) technologies, and novel riot control agents. Interface with DoD and non-DoD community regarding chemical technology research and development and provide expertise regarding threat analysis, and threat assessment.

Professional Memberships: Society of Toxicology, Student Member, 2014 - Present International Society on Toxinology, Full Member, 2016 – Present National Science Foundation Graduate Research Fellowship Program Panelist, 2009- 2010

Technical Skills: Mouse, rat, guinea pig, rabbit handling, injection, bio-sampling, euthanasia, and dissection Powder and liquid aerosol generation Nose only inhalation techniques Head-out plethysmography techniques Physiological telemetry data acquisition and analysis Standard molecular biology and cell culturing techniques Cell-line based assay development

Pharmacological plate-reader based assay development Physiologically based pharmacokinetic (PBPK) modeling Small molecule structure elucidation via high resolution mass spectrometry Microscope use: bright field and fluorescent Technical Writing Animal use in research

Selected Publications: Feasel MG, Wohlfarth A, Nilles JM, Pang S, Kristovich RL, Huestis MA. Metabolism of Carfentanil, an Ultra-Potent Opioid, in Human Liver Microsomes and Human Hepatocytes by High-Resolution Mass Spectrometry. The AAPS Journal. 2016 Nov;18(6):1489-1499.

Salem, H, Feasel, MG, Ballantyne, B. Chapter 11: Inhalation Toxicology of Riot Control Agents. Inhalation Toxicology, 3rd Edition. 2014. CRC Press. PP. 211-244.

Corriveau, JL, Feasel, MG. Chapter 12: Incapacitating Agents. Inhalation Toxicology, 3rd Edition. 2014. CRC Press. PP. 245-256.

Selected Scientific Abstracts: Feasel, MG Methods for Non-Lethal Chemical Agent Screening. May 2017. Non-Lethal Weapons Symposium. European Working Group on Non-Lethal Weapons. Ettlingen, Germany.

ABSTRACT Title of Dissertation: The Use of In Vitro and In Silico Technologies for Predicting Human Pharmacology and Toxicology of Carfentanil

Author: Michael G. Feasel, Doctor of Philosophy, 2017

Dissertation Directed by: Robert L. Kristovich, Ph.D. Branch Chief, Molecular Toxicology Edgewood Chemical Biological Center

Abstract:

Carfentanil is an ultra-potent opioid of public health and chemical weapons

defense concern. Exposure to carfentanil has been seen in both illicit drug use and in the

resolution of a hostage situation in Moscow, Russia in 2002. While opioid pharmacology

and toxicology are well researched topics, carfentanil remains unstudied relative to its clinically used counterparts like fentanyl. Similarly to fentanyl, carfentanil elicits is

toxicity from central nervous system depression, largely in regions of the brain involved

in spontaneous respiratory rhythmogenesis. By depressing the respiratory centers of the

brain, respiratory failure can occur, and can lead to death. Carfentanil is of concern

because it has demonstrated 100 times higher potency than its prototype, fentanyl, in

assays in rodents. Carfentanil has very little human relevant data to indicate its

toxicity in man for use in public health or chemical defense risk assessments.

The present study was designed to test the hypothesis that carfentanil

pharmacokinetic modeling could accurately reflect observed pharmacokinetics in vivo in

a surrogate animal model, and be translated to a human equivalent predication of toxicity.

Studies were carried out to assess subtype specificity, potency, and efficacy. Carfentanil metabolism was studied in both rabbit and human liver microsomes to assess its intrinsic clearance. A metabolite identification study was undertaken to identify metabolites that could contribute to prolonged exposure or toxicity, and to generate a library of metabolites to be used in a forensic setting to identify carfentanil as a culprit agent in overdose or mass casualty exposures. Additionally, two key physicochemical properties of carfentanil were quantified in both rabbit and human blood: plasma protein binding and red blood cell – plasma partitioning. These properties have important roles in pharmacokinetic modeling, and can be used in a forensic setting to indicate where carfentanil can be found. Finally, these in vitro data were incorporated into an in silico physiologically based pharmacokinetic model to accurately predict in vivo pharmacokinetics seen in a surrogate species. This model was then used to translate to a human equivalent toxic dose.

The Use of In Vitro and In Silico Technologies for Predicting Human Pharmacology and Toxicology of Carfentanil

by Michael G. Feasel

Dissertation submitted to the faculty of the Graduate School of the University of Maryland, Baltimore in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2017

DEDICATION

To my mother and father, for teaching me the value of education and to never stop learning. To my son, you are the best in vitro experiment that I have ever performed. And to my wife, for going on this journey with me. I learn from you every day.

iii

ACKNOWLEDGEMENTS

First and foremost I would like to acknowledge the support of my research advisor Dr. Robert Kristovich. You have not only trained me as a scientist, but you have helped me navigate the bureaucracy of earning this degree while maintaining my full time duties at ECBC. Without your mentorship, this dissertation would not have been possible. I would also like to acknowledge the contributions of Dr. Joseph Corriveau, Mr.

Larry Bickford and Mr. Horace Pearce. Thank you for taking the time to find ways for

me to continue my graduate career in an ever changing political environment. Your

knowledge, experience, generosity, and guidance were invaluable assets during my

graduate education. I humbly thank you for this opportunity.

I would like to thank all of the members of the Molecular, Operational, and

Analytical Toxicology branches at ECBC who have helped me on my path to this

dissertation.

To my co-workers in Smoke and Target Defeat Branch, Ken, Terry, and Larry:

thank you for always finding ways to guide me to where I wanted to be in my career.

To U.S. Army Colonel (Ret.) William McManaway, thank you for helping me

achieve a career path I didn’t know existed but in movies.

I am a turtle on a fence post.

iv

TABLE OF CONTENTS

Chapter 1. Introduction ...... 1

1.1 Ripped from the Headlines: The 2002 Moscow Theatre Siege ...... 1 1.2 Historical Opioid Research ...... 3 1.3 Modern Opioid Research ...... 4 1.4 Mechanisms of Opioid Toxicity ...... 6 1.5 Carfentanil and other Ultra Potent Opioids and Their Use ...... 8 1.6 Carfentanil as a Public Health Risk ...... 11 1.7 Objectives of this Thesis ...... 13 Chapter 2. Receptor Pharmacology of Carfentanil and Other Fentanyl Derivatives ...... 15

2.1 Introduction ...... 15 2.2 Receptor Screening Methods ...... 19 2.2.1. Chemicals ...... 19 2.2.2. Cell Line ...... 20 2.2.3. Incubation and Standard Solutions ...... 20 2.2.4. Assay Protocol ...... 21 2.3 Results ...... 22 2.3.1. Cell Density and Forskolin Optimization ...... 22 2.3.2. Carfentanil and W-18 Receptor Specificity and Potency ...... 24 2.3.3. Fentanyl, , Norcarfentanil Receptor Specificity ...... 27 2.4 Discussion ...... 29 2.5 Conclusions ...... 32 Chapter 3. Metabolism of Carfentanil, an Ultra Potent Opioid, in Human Liver Microsomes and Human Hepatocytes by High Resolution Mass Spectrometry ...... 34

3.1 Introduction ...... 34 3.2 Materials and Methods ...... 37 3.2.1. Chemicals and Reagents ...... 37 3.2.2. Methods ...... 37 3.3 Results ...... 42 3.3.1. Metabolic Stability ...... 42 3.3.2. Carfentanil MS/MS Spectrum ...... 44 3.3.3. Metabolite Profiling in Hepatocytes ...... 46 3.4 Discussion ...... 58

v

3.4.1. Metabolite Intensities ...... 58 3.4.2. Support for Riches et al? ...... 58 3.4.3. Implications for Carfentanil Toxicity ...... 59 3.5 Conclusions ...... 60 Chapter 4. Physicochemical Properties of Carfentanil ...... 62

4.1 Introduction ...... 62 4.2 Methods ...... 63 4.2.1. Reagents ...... 63 4.2.2. 96-Well Equilibrium Dialysis ...... 64 4.2.3. Red Blood Cell Partitioning by Depletion ...... 65 4.2.4. Sample Preparation for LC/MS/MS Analysis ...... 66 4.2.5. LC/MS/MS Analysis ...... 66 4.2.6. Comparison with in silico Predictions ...... 67 4.3 Results ...... 68 4.3.1. Physicochemical Properties ...... 68 4.3.2. In silico Predictions of Physicochemical Properties ...... 69 4.4 Discussion ...... 69 4.5 Conclusions ...... 70 Chapter 5. PBPK Modeling of In Vivo Carfentanil Exposures in the Rabbit and Comparison with Human Predictions ...... 72

5.1. Introduction ...... 72 5.2. Methods ...... 74 5.2.1. RLM Stability Study ...... 74 5.2.2. In Vivo Exposure of Rabbits to Carfentanil ...... 76 5.2.3. Human and Rabbit PBPK profile simulations using GastroPlus ...... 77 5.3 Results ...... 78 5.3.1. Carfentanil Stability Study in RLMs ...... 78 5.3.2. In Vivo PK Study of Carfentanil in Rabbits ...... 79 5.3.3. In silico Modeling of Rabbit PK...... 80 5.3.4. In Silico Prediction of Bioequivalent Dose of Carfentanil ...... 84 5.4 Discussion ...... 85 5.5. Conclusions ...... 85 6. Conclusions and Future Directions ...... 87

References ...... 89

vi

LIST OF FIGURES

Figure 1. Chemical structure of ...... 3 Figure 2. Chemical structure of fentanyl ...... 4 Figure 3. Chemical structure of carfentanil ...... 8

Figure 4. Reported Ki value ranges of reference drug and fentanyl ...... 10 Figure 5. Chemical structure of W-18 ...... 12 Figure 6. Flow chart of planned program scheme ...... 14 Figure 7. Chemical structures for carfentanil and W-18 ...... 17 Figure 8. DOR cell density and forskolin optimization experiments...... 23 Figure 9. KOR cell density and forskolin optimization experiments...... 23 Figure 10. NOP cell density and forskolin optimization experiments...... 24 Figure 11. Dose-response of MOR expressing CHO cells...... 25 Figure 12. Dose-response of DOR expressing CHO cells...... 25 Figure 13. Dose-response of KOR expressing CHO cells...... 26 Figure 14. Dose-response of NOP expressing CHO cells...... 26 Figure 15. Dose-response curve for MOR expressing CHO cells...... 27 Figure 16. Dose-response curve for DOR expressing CHO cells...... 28 Figure 17. Dose-response curve for KOR expressing CHO cells...... 28 Figure 18. Dose-response curve for NOP expressing CHO cells...... 29

Figure 19. Major metabolic pathways for fentanyl, and , known metabolic pathways for remifentanil and proposed pathway for carfentanil. Chemical structure of ...... 36

Figure 20. Parent compound depletion of carfentanil in human liver microsome (HLM) incubation ...... 43 Figure 21. MS/MS of carfentanil and identified fragment structures...... 45

Figure 22. Proposed metabolic pathway for carfentanil in human hepatocytes. Metabolite structures derived from MS/MS and chromatographic analysis...... 47

Figure 23. MS/MS for identified metabolites M1, M2, and M3. Fragment structures used for metabolite identification are indicated in the figure...... 50

Figure 24. MS/MS for Identified Metabolites M5, M7, and M8. Fragment structures used for metabolite identification are indicated in figure...... 52

vii

Figure 25. MS/MS for Identified Metabolites M10, M11, and M12. Fragment structures used for metabolite identification are indicated in figure...... 55

Figure 26. MS/MS for Identified Metabolites M4, M6, and M9. Fragment structures used for metabolite identification are indicated in figure...... 57

Figure 27. Parent compound depletion of carfentanil in rabbit liver microsomes (RLM) incubation...... 79

Figure 28. Pharmacokinetic data for rabbits dosed at 1 ug/kg. Plotted as mean plasma concentrations with error bars as SD ...... 79

Figure 29. Naïve PBPK model using only predicted ADMET properties based on the 2D structure of carfentanil...... 80

Figure 30. PBPK model incorporating only CLint...... 81 Figure 31. PBPK prediction incorporating only PPB experimental data...... 82 Figure 32. PBPK model incorporating only RBP...... 83

Figure 33. PBPK model incorporating CLint, RBP, and PPB...... 83

Figure 34. Human-equivalent dosing of carfentanil extrapolated from in silico model of rabbit PK data...... 84

viii

LIST OF TABLES Table 1. Various species' therapeutic and toxic responses to carfentanil ...... 10

Table 2. EC50 and efficacy for carfentanil and W-18 at each human opioid receptor subtype .... 26

Table 3. EC50 and efficacy for fentanyl, remifentanil, and norcarfentanil ...... 29

Table 4. Relative Potencies of All Compounds ...... 30

Table 5. Comparison of Available Reported EC50 Values to Experimental Values ...... 31

Table 6. Detailed Information on All Identified Metabolic Products and Parent Compound ...... 48

Table 7. Preparation Volumes for Matrix-Matched RED Samples ...... 64

Table 8. Protein Binding of Fentanyl, Carfentanil, and Norcarfentanil in Human and Rabbit Plasma ...... 68

Table 9. Blood/Plasma Partitioning of Fentanyl, Carfentanil, and Norarfentanil in Human and Rabbit Blood...... 68

Table 10. In silico Predicted Values for logP, Plasma Protein Binding, and Blood/Plasma Partitioning ...... 69

Table 11. Predicted Therapeutic and Toxic Doses for Carfentanil in Monkeys and Humans ...... 86

ix

Chapter 1. Introduction

1.1 Ripped from the Headlines: The 2002 Moscow Theatre Siege

On an autumn night in October of 2002, during a production of Nord Ost at the

Dubrovka Theatre in Moscow, Russia, 50 Chechen militants stormed into the theatre,

taking 800+ theatre-goers hostage.1 With automatic weapons and wearing explosive

vests, the Chechen rebels herded the theatre-goers into the orchestra pit. For nearly four

days they were held as failed negotiations were conducted with Russian government personnel and police forces outside. Then, something changed. From the air vents came a mysterious smoky haze; but it was not a smoke from a feared fire; smoke from fire rises.

It sank from the air vents to the floor.2 Those closest to the air vents were presumably

affected first, subtly, experiencing at first disorientation, euphoria, and soon appeared to

lose consciousness. Others that were less affected were alarmed to hear the hostage takers

screaming “Gas! Gas!” while donning respirators. Fifteen minutes later, once most of the

occupants of the theatre were physically incapacitated, unconscious and sprawled on the

floor or over the theatre’s seats; the Russian Spetznas special police forces entered the

theatre, neutralized the Chechen rebels and removed casualties from the body-filled

theatre to the fresh air outside and much needed medical treatment.

Unfortunately, the medical treatment that some did receive was ill-informed. The

on scene first responders and health care professionals at local area hospitals were not

effectively communicated to regarding what these 800+ people had been exposed to.

Some had heard rumors that they should expect organophosphorus nerve agent casualties,

which was supported by observation of miosis, or pinpoint pupils and clinicians

accordingly gave atropine to reverse the cholinergic shock they suspected, but to no

1

effect. Other healthcare professionals identified the symptoms they saw as toxicity

and treated with , but this was inconsistent across the field of casualties and only

some responded positively while others did not at all. Over the next two weeks, the death

toll would rise to 127 excluding the hostage takers. Nine rescuers were also hospitalized for signs of opiate toxicity. Outside medical professionals concluded in the days and weeks after the siege, that most if not all of the deaths during this event stemmed from logistical errors, poor training, and poor patient care and transport.3 Heroically, the police

and first responders present during the siege hurried to bring people out of the theatre into

the fresh air outside. In this hurry, however, these people were left on their backs, in poor

recovery positions for upwards of an hour and half before being transported on buses, in

large groups, to area hospitals. Had these people been left in a proper recovery position,

had they had higher ratios of care-takers to casualties, had they been transported more

efficiently to hospitals, had naloxone been administered within the first few minutes of

the casualty recovery operation, had some or all of these things happened, medical professionals believe far fewer would have died.3

In the days and weeks following this siege, scrutiny would come from all over the

world on Russia’s admitted use of a still unidentified fentanyl derivative to quell a large scale hostage situation. World courts did decide, however, that Russia did not violate the

Chemical Weapons Convention (CWC) since they did not use this weapon as a method of warfare; but on their own people during an internal police action. The specific chemical

agent(s) used wouldn’t be known to the public until a decade later, when, in 2012, the

British Ministry of Defense’s Defense Science and Technology Laboratory (DSTL) would publish the results of analyses of clothing and urine samples from British citizens

2

present during the siege; the culprit agents: a cocktail of carfentanil and remifentanil.4

This use of pharmaceuticals to achieve military or police operational goals brought

pharmaceutical agents, specifically opioids, into the world spotlight and scope of the U.S.

chemical and biological (CB) defense mission.

1.2 Historical Opioid Research

Opioid have had a major role in modern medicine since 1932 with the

synthesis of ethyl 1-methyl-4-phenylpiperidine-4-carboxylate, or pethidine (Figure 1), by

Otto Eisleb.5 Pethidine was the first synthetic opioid derivative of the piperidine class.

Shortly after its synthesis, a newly graduated medical doctor, Dr. Paul Janssen, began his

career expanding upon this piperidine chemistry in an attempt to separate morphine-like

analgesia and atropine-like antispasmodic properties, both observed with pethidine.6 In

1960, Janssen created N-phenyl-N-[1-(2-phenylethyl)piperidin-4-yl]propanamide, or fentanyl (Figure 2), a potent analgesic with fewer side effects than morphine.7 Fentanyl’s

immediate role as a safer and more potent analgesic and anesthetic led to it being the

prototype of the 4-anilidopiperidine class of drugs.8 After the creation of fentanyl,

medicinal chemists, including Janssen, modified nearly every position of the fentanyl

piperidine backbone in an effort to improve further on this prototypical drug, despite not

knowing its molecular target until 1973, what they called the “opiate receptor.”9

Figure 1. Chemical structure of pethidine

3

Figure 2. Chemical structure of fentanyl

1.3 Modern Opioid Research

Following the discovery of the “opiate receptor”, researchers determined that this

so-thought single target was actually three distinct opiate receptors: identified as the mu,

delta, and kappa, and were each found to be responsible for different aspects of opioid’s

effects.10, 11 The delta opioid receptor’s (DOR, DOP) role in physiology and medicine is

more complicated than its relatives. Although it is unknown what drug interaction with

the pure DOR effects are, studies have shown that DOR agonism has a broad role in

pathways and outcomes surrounding analgesia, mood, chronic pain, immunity, and

cardiac control.12-15 It appears that the DOR’s major role is perhaps modulation of other

opioid receptors. Research into the protein structure of the MOR has shown structural

propensity for dimerization based on thermodynamic influences, and further there is

evidence for MOR/DOR heterodimerization with functionally different outcomes than

either individual receptor that supports the DOR’s role in analgesia.16-18

The kappa opioid receptor (KOR, KOP), like its family members, is involved in pain pathways, but has additional effects related to consciousness such as dissociation,

4

hallucination, and dysphoria.19 Interestingly, KOR agonism has shown to contradict

MOR agonist pathways and outcomes, including analgesia, tolerance/addiction, and

mood, but not respiratory depression.20 is one example of an opioid toxicity

reversal agent that is both a MOR antagonist and KOR agonist, reversing MOR activated

analgesia and euphoria with dysphoria.

The opioid receptor-like protein or receptor (ORL1 or NOP), is the

most recently discovered member of the opioid family of proteins. While it is largely

homologous with the MOR, DOR, and KOR, it is not agonized by any of the traditional

opiate agonists. 21, 22 The NOP has demonstrated roles in a variety of pathways related to immune system function, sensory perception, analgesia, anxiety, learning and memory, and tolerance/reward.22, 23

The mu opioid receptor (MOR) is the most well studied of the four major opioid

subtypes. This is largely because it is recognized as the target receptor for the most

desirable effects of : antinociception. Researchers have identified the major sites

responsible for the transmission of nociceptive signaling: the mid-brain periaqueductal

region (PAG), the rostroventral medulla (RVM), and the dorsal horn (DH) of the spinal

cord.24-26 These studies have demonstrated that inhibition of neurons in these regions by

MOR agonists is responsible for the antinociceptive properties of opiates. Additional effects of opiates include sedation and anesthesia. The mechanisms behind these effects are modulated by MOR expressing noradrenergic neurons found in the locus ceruleus

(LC).27 Silencing these neurons with MOR agonists leads to sedation. A second set of

neurons that express hypocretin/orexin located in the lateral hypothalamus, are also

known to express MORs in high concentration. Loss of these neurons is associated with

5

narcolepsy, a sleep [promoting] disorder.28, 29 These neurons have also been shown to

innervate the LC and multiplicatively elicit sedative/somnolent effects through the LC

pathways. Finally, a known arousal pathway that acts through hypocretin neurons which

is activated by histamine causing alertness, can also be silenced by MOR agonists,

inhibiting the histaminergic arousal pathways, further promoting sedation.30

1.4 Mechanisms of Opioid Toxicity

Through this search to find better opioid drug compounds, more potent fentanyl derivatives were created aimed at higher specificity for the desirable target receptor in the thought that this would decrease off-target and adverse effects. Paradoxically, both the therapeutic and adverse effects of opioids are largely mediated through the same receptor subtype, the MOR.31, 32 Opioids are desirable clinically and recreationally for their

euphoric and analgesic properties. The toxic effects of these drugs, however, are

numerous and can manifest at therapeutic doses. The most dangerous adverse effect of

opioid administration or exposure is respiratory depression, which if prolonged and

untreated can develop into hypoxic injury, respiratory failure, apnea, and subsequently

death.

Perioperatively, incidence of severe respiratory depression is low, at 0.2% of

patients receiving morphine, but when looking at this effect in a postoperative setting, the

incidence increases to 29% in the case of patient controlled analgesia (PCA) administered

morphine.33 The mechanism of action of opioid induced respiratory depression (OIRD)

has been observed to be the preferential inhibition of MORs on neurokinin-1 receptor-

expressing cells of the PreBötzinger complex,34, 35 the region of the medulla responsible

for respiratory rhythmogenesis in mammals.36 Not only has this mechanism been shown

6

both in vitro and in vivo, but successful reversal of OIRD by ampakines, e.g. CX717,

CX1942 has been demonstrated in both preclinical and clinical trials.37, 38 These trials are

of course aimed at addressing clinically relevant doses of opioids and not necessarily

overdoses seen in recreational use or in situations like the Moscow Theatre Siege. It is

unknown if non-opioid spontaneous respiration promoters such as ampakines would be

useful in those scenarios or with more potent congeners of the fentanyl class.

An entirely different mechanism of toxicity is known to occur with opioids if a

high enough dose is administered: a potentially fatal arrhythmia known as torsades de

pointes (TdP), a severe Q-T prolongation, that can cause a rapid, nearly instantaneous

death of the user or victim. This mechanism of action may work through pathways in the

opioid receptor system, but is more likely caused by an interaction with cardiac human

ether-a-go-go (hERG) channels and/or voltage gated sodium channels.39-43 While this is

well studied for the natural and semi-synthetic opiates morphine, , , and , the molecular cardiac effects of fentanyl and its congeners are largely unstudied.

While not a toxic effect of opioids, a complicating effect that could result in an adverse-outcome is vomiting/emesis. Once again, this is a MOR mediated effect centralized in the chemoreceptive emetic trigger zone, or area postrema.44 This area is

outside of the blood brain barrier (BBB) and thus will more likely be exposed to higher

levels of MOR agonist than more central neuronal systems. Activation of neurons in the

area postrema by MOR agonists induces vomiting. This is also something that cannot be

demonstrated in low-order species such as rats or mice since they do no vomit, but

requires a mammal with an emetic reflex such as a ferret or non-human primate.44, 45 This

7 contributes to adverse outcome pathways of opioids since vomiting while sedated or unconscious can lead to fatal aspiration. This risk therefore requires proper first aid and aspiration mitigation in the case of accidental overdose, or the incorporation of an antiemetic in an analgesic/anesthetic cocktail; a common clinical practice.

1.5 Carfentanil and other Ultra Potent Opioids and Their Use

While much is known about fentanyl and its clinically used congeners in regards to pharmacology, i.e. absorption (via various routes), distribution, metabolism, and excretion (ADME), little is known about its younger relatives that were developed to increase potency and decrease adverse effects.

Current interest in the ultra-potent synthetic opioid derivatives such as methyl 1-

(2-phenylethyl)-4-(N-propanoylanilino)piperidine-4-carboxylate, or commonly carfentanil (Figure 3), stems less from its candidacy in clinical medicine as a therapeutic, and more from its role as an illicit drug of abuse and, more esoterically, as a tactical pharmaceutical in the Moscow Theatre Siege.

Figure 3. Chemical structure of carfentanil Carfentanil is an opioid analgesic belonging to the fentanyl class of drugs. While not used in human clinical medicine, carfentanil does have historical use in veterinary medicine

8

for large animal sedation, take-down, and anesthesia.46 Carfentanil, is reported to be

10,000 times more potent than morphine in regard to analgesia which has traditionally

been measured by the rat tail withdrawal assay,47 and has the lowest effective half-

maximal concentration (EC50) of the fentanyl analogues in the guinea pig ilium assay.48

Carfentanil also has reportedly a very high selectivity/specificity for the MOR over the

DOR or KOR.48 Traditionally, two bioassays were employed to measure opioid potency: rodent nociceptive assays and guinea pig ilium electrophysiology. The rat-tail withdrawal

(RTW), tail flick, and hot plate tests are commonly used to measure analgesic potency.49-

51 The RTW provides a metric of the point at which the analgesic sufficiently masked the nociception/pain caused by the rat’s tail being dipped in a hot water bath, measured by latency to remove the rodent’s tail from the stimulus. Tail flick and hot plate tests measure the same response, but the stimulus is a noxious beam of light or heat respectively. These measured therapeutic values are commonly compared to an ex-vivo model: the guinea pig ileum (GPI) assay. The GPI is known to be rich in MORs and has historically been the in vitro gold-standard for µ-opioid receptor pharmacology.52-54 Other

in vitro/ex vivo based methods have been employed to measure opioid affinity, potency,

and efficacy, mostly different types of receptor inhibition assays (Ki). One such method

uses rat brain homogenate to analyze binding of agonists and inhibition of binding using

antagonists.55 Other methods exist and are too numerous to list here, but all do generate

some sort of relative affinity ranking for how well the drug binds the MOR compared to

another drug or test compound, but do not necessarily indicate how well it elicits a

response once bound, i.e. potency and/or efficacy.56 The varying methods of establishing

Ki’s of opioid drugs has led to wide ranges of Ki-correlated potencies being reported in

9

the literature (Figure 4), none of which are wrong, but all of which need expressed

caveats to their meaning.57

Morphine

Fentanyl

0.001 0.01 0.1 1 10 100 1000 Ki (nM)

52, 53 52, 54 Figure 4. Reported Ki value ranges of reference drug morphine and fentanyl. Ki values correlate with potency and efficacy.57

While rodent models can be useful tools for quantifying the therapeutic effects of opioid

analgesics, one problem with employing them for toxicity experiments targeting human

risk assessment is that there are great differences in how rodent species respond,

compared to the human, to opioid challenges which can be related directly to the

mechanism of action and symptomology.58, 59 Species differences in the therapeutic index

of opioids can be seen in Table 1.

Table 1. Various species' therapeutic and toxic responses to carfentanil.58, 59

Therapeutic Effective Dose Lethal Dose Species Effect Index (mg/kg) (mg/kg) (LD50/ED50) Rat 0.00032 Analgesia 3.39 10,594 Dog 0.0047 Analgesia 5.0 1,064 Monkey 0.0001 Immobilization 0.001 10

Opioid agonists act as central nervous system (CNS) depressants, slowing or stopping transmission of signals in nerves involved with pain and nociception, i.e. their therapeutic effects, but they also attenuate respiratory drive. Peripheral effects include muscular rigidity, gastrointestinal slowing, and in some cases cardiac arrhythmias.40, 60 Prolonged

10

or profound depression of respiratory drive leads to apnea and death if not treated with an

or airway management. Lower order species, such as rodents, do not

succumb to the apneic/hypercapnic effects of opioids as readily as higher order species,

such as the human, do. This species difference in central response and/or tolerance to

hypercapnia lead to an artificially large therapeutic index for these compounds in lower

order species (Table 1). Therefore, when it is claimed that carfentanil is 10,000 times

more potent than morphine or 100 times more potent than fentanyl, that doesn’t

necessarily correlate to toxicity, but only the therapeutic response in a less-than-ideal animal model.

1.6 Carfentanil as a Public Health Risk

From the public health standpoint, like carfentanil and W-18 (Figure 5)

are increasingly being reported as being found in illicit drug seizures or as part of drug

overdose investigations,61-68 but little is known about these compounds in most cases

regarding their pharmacology or toxicology. This is due in large part to a lack of clinical

data, because these compounds were never meant for human use and were therefore not

put through any preclinical or clinical trials. Generally speaking, these compounds were

synthesized in numerous variations on specific chemical themes aimed at down-selecting a small number of candidate drug materials from chemical libraries of many. Illicit drug manufacturers and distributors are delving into the published literature and finding

chemicals that are not covered by any legal bans, synthesizing them and selling them for

profit, to the demise of the end users. When one compound becomes listed or scheduled,

they make the next compound in the series to evade legal restrictions. Despite many of

these compounds being decades old, these drugs are termed New Psychoactive

11

Substances (NPS). For instance, the NPS was synthesized in 1968 and has

only in 2014 been recognized as a drug of abuse.69 W-18 was patented in 1986 and in

2015 was reported as being found in fentanyl tablets.68 Carfentanil was first synthesized

in 1974 by Paul Janssen, but was pushed into a veterinary niche because of its extreme

potency.46 The U.S. Drug Enforcement Agency (DEA) reports that there have been over

400 seizures of carfentanil in eight states over a period of four months in 2016 alone.70 In

2015, deaths due to overdose by illicit opioids breeched the 20,000 mark,71 indicating the

widespread use and deadliness of these types of compounds being abused in recreational

environments. Production of carfentanil was almost exclusively done by chemical

synthesis companies in China. Production by these Chinese companies was legal, since

carfentanil was not a controlled substance in its nation of production. However,

carfentanil is now on China’s controlled substances list as of March 1, 2017.70 U.S.

authorities are hoping that the Chinese ban of carfentanil will decrease the amount seen in

North American markets. But, questions still surround its use in the United States and

North America, and what we do and do not know about this reportedly very toxic

.

Figure 5. Chemical structure of W-18

12

1.7 Objectives of this Thesis

Generally speaking, very little is known about carfentanil regarding its behavior

in humans. Guinea pig brain homogenate studies have shown that carfentanil is more

selective for the MOR than the DOR or KOR, but this hasn’t been tested before using a

cloned human receptor in a standardized method.72 Metabolism of fentanyl, sufentanil, alfentanil, remifentanil, and other compounds has been show both in vitro and in vivo, yet

Riches et al. reports that no carfentanil metabolites have ever been generated

experimentally; only presumed based on the other congeners’ metabolic pathways.4, 73-75

Physicochemical properties for the clinically used derivatives have also been derived

through both in vitro and in vivo experimentation, yet carfentanil has been ignored by

these studies.76

The overarching objectives of the work done as part of this thesis were to provide

new and novel experimental data on the pharmacology and toxicology of carfentanil in an

environment lacking any human relevant data. These studies will be performed in an effort to support the generation of a human estimate of toxicity for carfentanil. The main thrust of this research was to develop and validate an in silico model that accurately predicted in vivo exposures to carfentanil in the rabbit animal model (Figure 6); later chapters discuss why the rabbit was chosen as the surrogate animal of choice. Feeding into this in silico model are four major aspects: 1) metabolism of carfentanil, which itself consisted of two parts, a) microsomal stability and intrinsic clearance investigation and b) human hepatocyte metabolite identification; 2) physicochemical properties of carfentanil and any available metabolites; 3) receptor activity assays to assess relative potency and

13 receptor subtype profile; 4) in vivo pharmacokinetics with which will be used to validate the in silico model.

Figure 6. Flow chart of planned program scheme

14

Chapter 2. Receptor Pharmacology of Carfentanil and Other Fentanyl Derivatives

2.1 Introduction

Opioid and opiate analgesics are popular clinical treatments for both acute and

chronic pain, but they are also drugs of abuse and have even been reported as being used

in a police-type action to quell a hostage situation.4 This wide use and variety of

applications, both on- and off-label have led to much interest in the toxicology and

pharmacology of these compounds, specifically the congeners for which no pre-clinical or clinical data exists.

Fentanyl is the prototype for the synthetic class of opioid analgesics. However, since its inception, more potent derivatives have been synthesized in an effort to increase therapeutic effect while minimizing adverse effect potential. This has come in the form of creating highly specific and incredibly potent agonists to the µ-opioid receptor (MOR); the receptor sub-type responsible for the profound analgesia and euphoria associated with acute pain relief, but also muscular rigidity, respiratory depression, and apnea.31, 32

Carfentanil was synthesized not long after fentanyl and was immediately noted

for its incredible potency in regard to its ability to relieve pain as measured by the rat-tail

withdrawal (RTW) assay.48 Since its synthesis, carfentanil has had very little pharmacological or toxicological data collected on it since it was never marketed as a human clinical drug. Carfentanil had a successful, albeit, short-lived role as a veterinary sedative used in large animal sedation and take-down,46 but has since been replaced by

other ultra-potent opioids such as buprenorphine and . Additionally, it is

reported that carfentanil is an incredibly specific agonist, or activator, of the MOR.77

15

Toxicity of carfentanil (Figure 7A) has been of interest to the chemical defense community since its reported use to quell a hostage situation in 2002 known as the

Moscow Theatre Siege,4 however, more recent public health concerns revolve around its incorporation into illicit drug sales and distribution via heroin and fentanyl markets, potentially contributing to the highest levels of opioid-related overdoses seen in history.78

News reports and county officials in Ohio report that eight overdose deaths have involved carfentanil being present in heroin and/or fentanyl tablets.79 This has coincided with a dramatic spike in opioid related emergency room visits and deaths over the past year.71

Previous studies have been conducted to rank the potency of various opioid compounds and have done so very successfully.59 These cover a very large swath of the clinically relevant opiate and opioid therapeutics. However, in all instances, carfentanil has been ignored due to its lack of use and/or utility in clinical medicine, be it human or veterinary.

4-chloro-N-[(2Z)-1-[2-(4-nitrophenyl)ethyl]piperidin-2-ylidene]benzene-1- sulfonamide, or W-18 (Figure 7B), is a compound that has recently been in the spotlight of public health concern since its reported seizures of neat material in kilogram quantity from Florida to Canada.68 This drug compound was synthesized in the 1980’s by the

University of Alberta in an effort to find a more potent opioid analgesic. It is reported to be 10,000 times more potent in regard to relieve pain in the mouse phenylquinone writhing assay (IC50 = 3.7 ng/kg, compared to 38 µg/kg for morphine).80 However, this assay is non-specific to opioids and does not characterize the drug’s receptor target.

Additionally, this assay was performed in the mouse animal model that, while acceptable for therapeutic/analgesic screening, is a less than ideal model when assessing opioid

16 toxicity. Rodents have an incredibly high tolerance of the drug, and while suitable for therapeutic assessments, toxicity/lethality assessments should be performed in higher order species. W-18 was therefore included in this study to assess the opioid activity, or lack thereof of W-18, and how potent or efficacious it is compared to control compounds and to the known ultra-potent opioid carfentanil.

Figure 7. Chemical structures for A) carfentanil and B) W-18 A standardized method has been employed in this study to compare opioid receptor activity of carfentanil to express its relative potency independent of species and high order physiological response. This method utilizes four separate Chinese Hamster

Ovary (CHO) cell lines transfected individually with the human µ-opioid receptor

(MOR), ƙ-opioid (KOR), ƍ-opioid (DOR) and (NOP) genes. These cells express Gαi-associated proteins that have been linked to the inhibitory responses seen in opioid receptor activation.81, 82 Upon agonism, GPCR transmembrane conformational changes and G-protein subunit translocation lead to the inhibition of adenylyl cyclase (AC), the enzyme responsible for cyclic adenosine-monophosphate

(cAMP) production. As a result, cAMP levels are suppressed. This dose-dependent change in cAMP production can be measured using a commercial off the shelf (COTS)

17

monoclonal antibody based assay kit. The kit functions based on competition between a

Europium (Eu) labeled cAMP molecule and endogenous cAMP in the cell for binding to

cAMP-specific monoclonal antibodies (mAb) labeled with a dye (ULight™).

The assay functions by inducing cAMP production with a compound called

Forskolin, and then inhibiting that induction with the opioid of interest. Forskolin acts directly at AC, inducing its activity and subsequent production of cAMP. In cells where there is an abundance of cAMP, i.e. those induced with Forskolin, the mAb-dye complex

binds to that free cAMP, dissociating it from the Eu-cAMP complex, resulting in a drop in fluorescence resonance energy transfer (FRET). When the cells are incubated with an opioid and AC is inhibited, cAMP levels in the cells drop. This increases the binding of

Eu-cAMP to the ULight™-mAb complex, resulting in an increased FRET signal. This signal is dose-dependent and increases with opioid concentration.

By investigating only the functioning of an opioid at the receptor itself, and being able to do so in a human receptor model in an in vitro system, relative potency of opioid compounds can be compared using the same metric. This eliminates variation from inter-

species differences, tissue culturing methods, inter-animal tissue differences, etc. and truly compares on an identical level the activity and efficacy of these compounds. This study aimed to assess the relative potency and receptor sub-type specificity of several compounds: fentanyl, remifentanil, carfentanil, norcarfentanil (the only commercially available carfentanil metabolite), and the reported ultra-potent opioid W-18.

In the current study, we used a very well characterized cell system expressing human opioid receptors specifically coupled to G-proteins associated with the cAMP

second messenger system. By utilizing a system that links receptor activation to a single

18

response, we were able to compare directly the functional activity of carfentanil, its

metabolite norcarfentanil, and W-18 to better studied compounds like fentanyl and

remifentanil. This relative potency value of carfentanil, combined with its metabolic and

physicochemical properties, will be incorporated into future human risk assessments

performed for this compound.

2.2 Receptor Screening Methods

2.2.1. Chemicals

The LANCE cAMP 10,000 assay point kit and 384-Proxiplates were purchased

from Perkin Elmer (Shelton, CT, USA). The LANCE kit consisted of cAMP Standard, 50

µM; Eu-cAMP tracer, ULight™-anti-cAMP; cAMP Detection Buffer; and BSA

Stabilizer. Carfentanil, was synthesized at Edgewood Chemical Biological Center

(ECBC) (APG, MD, USA) and purity verified by 13C and 1H NMR. W-18 was procured

from Cayman Chemical (Ann Arbor, MI, USA). Fentanyl citrate was procured from

Mallinckodt Pharmaceuticals (St. Louis, MO, USA). Remifentanil HCl was procured

from Pharm Agra Labs, Inc. (Brevard, NC, USA). Norcarfentanil was procured from

Toronto Research Chemicals, Inc. (Toronto, ON, CAN). DOR, KOR, MOR, and NOP

selective agonists [D-Pen2,D-Pen5] (DPDPE), (±)-U-50488 hydrochloride,

[D-Ala2, NMe-Phe4, Gly-ol5]-enkephalin (DAMGO), and Nociceptin, respectively were

purchased from Tocris Bioscience (Park Ellisville, MO, USA). HBSS 1X, HEPES 1M,

Versene Solution, and Geneticin were procured from Life Technologies (Grand Island,

NY, USA). DMSO, IBMX (3-isobutyl-1-methylxanthine), and Forskolin were procured

from Sigma-Aldrich (St. Louis, MO, USA). DPBS/Modified Buffer and Ham’s F-12

19

Media were procured from HyClone Laboratories, Inc. (Logan, UT, USA). Fetal Bovine

Serum (FBS) was procured from Mediatech, Inc. (Manassas, VA, USA).

2.2.2. Cell Line

ValiScreen® CHO-K1 cells expressing human µ-opioid receptors (ES-542-C) were purchased from Perkin Elmer (Waltham, MA, USA). ChanTest™ CHO-K1 cells expressing human ƍ-opioid (CT6607), ƙ-opioid (CT6606), and NOP (CT6604) receptors were purchased from Charles River Discovery (Cleveland, OH, USA). Cells were kept frozen in liquid nitrogen storage (vapor phase) until they were cultured. Cells were grown per Perkin Elmer and Charles River product literature. Cell cultures were passaged when they reached ~60-80% confluency and no cells were used past passage 10. Cells were used for opioid assay when they met requirements described in the product literature, i.e.

60-80% confluency. Cellular solutions used in plating were counted on a Countess II hemocytometer in duplicate prior to use.

2.2.3. Incubation and Standard Solutions

10 mM standard solutions of fentanyl, remifentanil, carfentanil, norcarfentanil, and W-18 were made in DMSO and stored until use in a freezer at 4°F. Standard solutions of specific agonists were made in sterile water in the following concentrations:

DAMGO, 1.95 mM; DPDPE, 1.55 mM; (±)-U-50488, 11.55 mM; nociception, 1.1 mM.

500 µM working solutions of carfentanil and the four specific agonists were prepared immediately prior to performing the assay. Stimulation Buffer, forskolin dilutions, and cAMP standards were all made as needed as described by the LANCE

Ultra cAMP Assay protocol immediately before performing the assay.

20

2.2.4. Assay Protocol

Assay development followed the protocols set out in PerkinElmer® LANCE Ultra cAMP Assay Development Guidelines. PerkinElmer 384-Proxiplates were used for all assays with the following dimensions: Plate Height 14.4mm, Well Diameter 3.15 mm,

Well Volume 25 µL. All plating was performed in triplicate: 5 µL of 2.00 × 105 cells/mL

cell solution were plated into full columns to achieve 1,000 cells/well; optimal density for

cAMP dose-response (data not shown). 2.5 µL of 3µM forskolin solution [optimized

concentration for cAMP dose-response] was then added into all wells containing cells.

2.5 µL of drug dilutions were then plated in descending order down the rows of cells.

Concentration of drug compound was dosed logarithmically from 1x10-4 M to 1x10-14 M.

The plate was then allowed to sit at room temperature for 30 minutes. All wells were then

administered 5 µL each of the kit tracer solution and antibody solution. The plate was

then allowed to sit for an additional 60 minutes covered with TopSeal-A sealing film

(PerkinElmer, Waltham, MA, USA) before being read on a Molecular Devices

SpectraMax Multi-Mode Microplate Detection platform (i3x) with SpectraMax Cisbio

HTRF detection cartridge (0200-7011POS) and data analyzed with SoftMax Pro v.6.5.1

(all Molecular Devices, Sunnyvale, CA, USA). Plate reader parameters included an

excitation wavelength of 340 nm and emission wavelengths of 615 and 665 nm and 20

flashes/read. Read height was optimized before each read to 7.22 mm. Mean value was

calculated for triplicate wells for each dose point with error bars of standard deviation.

21

2.3 Results

2.3.1. Cell Density and Forskolin Optimization

The cAMP standard curve was generated parallel to the cell density dilutions on the same 384-well plate. Cell density for all future experiments was determined based on the highest signal to noise ratio while staying within the IC10-IC90 dynamic range of the cAMP standard curve. Cells were plated in 5,000 cells/well (when possible), 3,000 cells/well, 1,000 cells/well, and 250 cells/well densities accomplished by serial dilution.

All conditions were performed in triplicate. It was measured that 1,000 cells per well met the criteria for cell density selection in all cell subtypes, being that this density had comparable signal/noise and was within the dynamic range of the linear phase of the cAMP standard curve having its top asymptote near the IC90 of the cAMP standard curve

(Figure 8, Figure 9, Figure 10).

Based on the 1,000 cells/well density, the experimental forskolin concentration was calculated to be the IC90 of that density, or 3 µM, again for all cell subtypes. The IC90 is used in these experiments to achieve the highest signal difference between forskolin- activated cells and cells co-stimulated with forskolin and the individual agonists of interest. Therefore, all further experiments were performed with 3 µM Forskolin concentration as the co-administered compound.

22

Figure 8. DOR cell density and forskolin optimization experiments. 1,000 cells/well is within the linear dynamic range of the instrument.

Figure 9. KOR cell density and forskolin optimization experiments. 1,000 cells/well is within the linear dynamic range of the instrument.

23

Figure 10. NOP cell density and forskolin optimization experiments. 1,000 cells/well is within the linear dynamic range of the instrument.

2.3.2. Carfentanil and W-18 Receptor Specificity and Potency

Efficacy and potency were assessed for each compound at each of the individual receptor subtypes (Figure 11, Figure 12, Figure 13, Figure 14). Carfentanil was in fact found to have very high selectivity for the MOR over the other three subtypes.

Additionally, W-18 was found to have no activity across the four human receptor subtypes tested in this study at any physiologically relevant concentration. A comparison of the EC50 values for both compounds and all four receptor subtypes is below (Table 2).

24

Figure 11. Dose-response curve for MOR expressing CHO cells.

Figure 12. Dose-response curve for DOR expressing CHO cells.

25

Figure 13. Dose-response of KOR expressing CHO cells.

Figure 14. Dose-response of NOP expressing CHO cells.

Table 2. EC50 and efficacy for carfentanil and W-18 at each human opioid receptor subtype reported as molar concentration (M).

Half-Maximal Concentration (M) Efficacy (% of Control) Drug µ ƍ ƙ NOP µ ƍ ƙ NOP Carfentanil 6.15×10-12 8.55×10-9 6.61×10-8 Inactive 97.1 94.4 67.9 13.7 W-18 Inactive Inactive Inactive Inactive 0 0 0 0

26

2.3.3. Fentanyl, Remifentanil, Norcarfentanil Receptor Specificity

Efficacy and potency were assessed for each of fentanyl, remifentanil, and norcarfentanil at the individual receptor subtypes (Figure 15, Figure 16, Figure 17, Figure

18). Fentanyl and remifentanil were observed to be nearly equipotent, both with very high selectivity for the MOR over the other three subtypes. Additionally, none of these three classical MOR agonists were found to have any activity at the NOP receptor subtype tested in this study at any physiologically relevant concentration. A comparison of the EC50 values for both compounds and all four receptor subtypes is below (Table 3).

Figure 15. Dose-response curve for MOR expressing CHO cells. Control compound was DAMGO.

27

Figure 16. Dose-response curve for DOR expressing CHO cells. Control compound was DPDPE.

Figure 17. Dose-response curve for KOR expressing CHO cells. Control compound was U-50488.

28

Figure 18. Dose-response curve for NOP expressing CHO cells. Control compound was Nociceptin.

Table 3. EC50 and efficacy for fentanyl, remifentanil, and norcarfentanil at each human opioid receptor subtype reported as molar concentration (M).

Half-Maximal Concentration (M) Efficacy (% of Control) Drug µ ƍ ƙ NOP µ ƍ ƙ NOP Fentanyl 5.11×10-10 3.77×10-8 9.73×10-7 Inactive 99.9 105.6 18.9 0 Remifentanil 5.99×10-10 1.03×10-7 ~1.08×10-6 Inactive 92.3 101.3 10.9 0 Norcarfentanil 1.13×10-7 9.01×10-7 ~9.73×10-6 Inactive 81.4 95.5 6.2 0

~ indicates an ambiguously fit non-linear regression of the dose-response

2.4 Discussion

Historical animal model work with opioids has shown that effectiveness and toxicity are species dependent, and that rodents have an artificially large therapeutic index compared to higher order animals. This has to do with both the receptor homology and physiology of the various species. By studying human receptors in vitro removes all interspecies variability while making the data more human-relevant when attempting to compare relative potencies (Table 4).

29

Table 4. Relative Potencies of all compounds tested to benchmark compound fentanyl

Potency Relative to Fentanyl Drug µ ƍ ƙ NOP Fentanyl 1 1 1 Inactive Remifentanil 0.85 0.37 0.90 Inactive Norcarfentanil 0.004 0.04 0.10 Inactive Carfentanil 83.1 4.41 14.7 Inactive W-18 Inactive Inactive Inactive Inactive

In addition to generating EC50’s that can relatively compare the potency of suspect or

known opioid compounds at all four receptor subtypes, this study also helped shed some

light on some of the interspecies differences in receptor homology.

The method utilized in this study which uses CHO cells transfected with human

receptor genes has clear advantages when compared to methods that utilized guinea pig

whole brain77 or guinea pig ilium and mouse vas deferens (Table 5).83 There is an order of magnitude more sensitivity difference in the activity of carfentanil at the human MOR when compared to the guinea pig brain method, while the DOR and KOR are relatively close to the reported values. Additionally, when using one tissue, as in the James et al. study to screen for three different receptor subtypes, this method is acknowledged to be more qualitative than quantitative. This method assesses receptor specificity based on dose-response curve shifts of the agonist of interest when using specific antagonists such as the MOR antagonist naloxone, KOR antagonist nor-, and DOR antagonist ICI 174,864.83 Receptor homology could be a contributing factor to the

sensitivity that higher order species to these compounds when compared with smaller

species and even more so with various tissues within those species. Other factors that can

contribute besides receptor homology are tissue-dependent densities and efficiencies.

These are yet more reasons why a standardized, single endpoint, human receptor based

30

system is preferable when assessing receptor specificity, whereas tissue based assays are

not optimized for selective target-receptor profiling.83

Table 5. Comparison of available reported EC50 values (nM) to experimental values from this study

Source µ ƍ ƙ NOP Fentanyl – 1.15 178 293 Not Reported Cometta-Morini77 Fentanyl 0.511 37.68 973 Inactive experimental Remifentanil – 2.4 Not Reported Not Reported Not Reported James et al.83 Remifentanil 0.599 103 1080 Inactive experimental Carfentanil – 0.024 3.28 43.1 Not Reported Cometta-Morini77 Carfentanil 0.00615 8.55 66.1 Inactive experimental

Based on the EC50 values of carfentanil at each of the receptor subtypes,

carfentanil does show overwhelming selectivity for the MOR. Carfentanil is 1,390 times

more specific for the MOR when compared to the DOR, and 10,747 times more MOR

specific when compared to the KOR. Even more overwhelmingly, carfentanil is at least

six orders of magnitude (~4,000,000) times more specific to the MOR when compared to

the NOP. It should be noted, however, that because of carfentanil’s low EC50 at its

intended target receptor, MOR, there is potential for DOR and KOR activity at

physiologically achievable levels of the drug, still falling in the 1-100 nM range. Of

course, in a controlled setting, carfentanil would ideally be dosed to a desired endpoint,

likely MOR-mediated, and thus off-target agonism of DOR and KOR would be minimal to non-existent, but in a mass-administration of carfentanil as reported during the

Moscow Theatre Siege, only the laws of physics control the dosage administered to each casualty. At these higher doses, over-dosing of the MOR toxic mechanisms, combined with an increased likelihood of off-target DOR and KOR mediated pathways, can further complicate symptomology and could lead to toxicity.

31

Additionally, while this receptor activity study gives insight into the observed potency of carfentanil, metabolic and physicochemical properties are what influence if, when, and how much of the drug makes it to its intended target, and if so, how quickly and for what kind of duration. The receptor activity study conducted here is a protein-free system, and therefore protein binding does not affect the receptor activation properties as it would in vivo. Protein binding alone can account for up to two orders of magnitude decrease in potency due to this sequestration. It is therefore important to measure the contribution that protein binding has on compound availability in blood and tissues when attempting to translate in vitro potency to in vivo plasma or tissue concentrations.

2.5 Conclusions

The effort proposed by this study was to assess opioid compounds in regard to their potency for each receptor subtype in a human model system. This assay utilizes

CHO cells expressing human opioid receptors of all four known subtypes. This is advantageous over other methods for assessing opioid potency/efficacy in that it is the only system based on the human receptor and most importantly, is not influenced by species or tissue specific characteristics that may or may not be representative of human receptors, tissues, or responses. This capability can be used to screen suspect opioid compounds for activity and specificity in a safe, rapid, and inexpensive manner. This capability is a useful screening tool for DoD and other public health-minded organizations concerned with opioid compounds and their potential pharmacology and toxicology.

This project confirmed carfentanil’s incredibly high potency and supports reports of carfentanil being roughly 100 times more potent than fentanyl.47 This study measured

32

carfentanil as having 83 times the potency of fentanyl at the MOR. We also further

reinforced the specificity that carfentanil has for the MOR. Despite this selectivity,

carfentanil could elicit off target effects at the other receptor subtypes at physiological

concentrations.

This study, for the first time, also refutes reports of W-18’s activity in the human opioid system. No other study to date has tested this compound at the human receptor. Its lack of agonist activity at the receptor leads to many more questions about its toxicity and mechanism of action if it is in fact bioactive and toxic. Further studies need to be done to elucidate its receptor targets. However, we can conclude that W-18 does not act either therapeutically or toxicologically through agonism of the opioid receptors.

Finally, norcarfentanil, the only commercially available metabolite of carfentanil was screened for receptor activity and showed five orders of magnitude less activity than its parent, carfentanil, confirming reports of its inactivity.84 As other metabolites

identified become available,85 future studies will screen them in an identical manner, to

assess what contribution metabolites make to the therapeutic and/or toxic effect of

carfentanil.

In order to get a better picture of why carfentanil is so toxic, metabolism will be

first studied, followed by physicochemical properties, to investigate what impact they

have on overall pharmacokinetics and toxicokinetics.

33

Chapter 3. Metabolism of Carfentanil, an Ultra Potent Opioid, in Human Liver Microsomes and Human Hepatocytes by High Resolution Mass Spectrometry1

3.1 Introduction

Carfentanil is an ultra-potent synthetic opioid analgesic belonging to the same

class of drugs as its prototype fentanyl. Little is known about human pharmacology and

toxicology, or analysis of carfentanil in biological samples. While not used in human

clinical medicine, carfentanil has limited use in veterinary medicine for large animal

sedation, take-down, and anesthesia.46 Carfentanil, seen in Figure 6, is 10,000 times more

potent than morphine for analgesia, as measured by the rat tail withdrawal assay.47

Carfentanil has the lowest EC50 of the fentanyl analogues.48 The reasons for its potency

are largely unknown, but can be attributed to its high lipophilicity, ease of crossing the

blood-brain barrier (BBB), high receptor efficacy, and high selectivity/specificity for the

µ-opioid receptor (MOR) over other opioid receptor types, such as ƙ or ƍ.48 The µ-opioid receptor is the receptor sub-type responsible for the desirable opioids effects such as analgesia and euphoria, but also the adverse and potentially toxic effects such as muscular rigidity, respiratory depression and apnea.31, 32

Carfentanil was administered to humans to map brain µ-opioid receptors, to study addiction, and to measure pleasure responses at the receptor level.86-88 Until recently,

these studies and veterinary use were the only applications for carfentanil. However, in

2002, carfentanil was reportedly administered to humans when Chechen rebels took more

1 Feasel, M. G.; Wohlfarth, A.; Nilles, J. M.; Pang, S. K.; Kristovich, R. L.; Huestis, M. A., Metabolism of Carfentanil, an Ultra-Potent Opioid, in Human Liver Microsomes and Human Hepatocytes by High-Resolution Mass Spectrometry. Aaps Journal 2016, 18 (6), 1489-1499.

34

than 800 theatre-goers hostage at the Dubrovka Theatre in Moscow, Russia.89 After days

of failed negotiations, an unknown aerosol was released into the theatre’s air conditioning

system. Fifteen minutes after release, police gained access to the theatre, neutralized the

hostage takers, and removed hostages from the theatre. 129 hostages died.90 It was later

reported that an opioid derivative was used.90 In 2012, British researchers analyzed urine

and clothing samples from British citizens who survived the incident, concluding that

carfentanil and remifentanil were present in the incapacitating cocktail.4 Additionally,

methyl-4-((propionyl)phenylamino)piperidine-4-carboxylate, the N-dealkylated product of carfentanil and remifentanil, was found in the urine sample. This report conceded that no study to date has definitively identified human carfentanil metabolites, but hypothesized that norcarfentanil is a likely candidate based on known metabolites of fentanyl and other widely used congeners.

Fentanyl, sufentanil and alfentanil are primarily metabolized via CYP3A4 in the liver, generating metabolites N-dealkylated at the piperidine ring (Figure 19A).73, 74

Whereas fentanyl is metabolized via N-dealkylation into the unique metabolite

norfentanyl, alfentanil and sufentanil are metabolized to the same N-dealkylated product, making forensic distinction impossible when only this metabolite is identified (Figure

19B). Morphine has active metabolites contributing to the therapeutic effect;91 fentanyl,

sufentanil, and alfentanil’s metabolites are inactive in the opioid system.92 Remifentanil

is the only family member of this class found to be ~95% metabolized in the blood and

tissues by non-CYP enzymes.93 This was attributed to an easily accessible ester group in

remifentanil allowing rapid hydrolysis by circulating blood esterases, yielding

remifentanil acid with 1/4000th of the parent’s analgesic potency.94 Methyl-4-

35

((propionyl)phenylamino)piperidine-4-carboxylate is a minor metabolite from remifentanil N-dealkylation in the liver, and is identical to N-dealkylated carfentanil.95

Therefore, there is no way to distinguish intake of remifentanil and carfentanil by

targeting the N-dealkylated metabolite only (Figure 19C).

Figure 19. Major metabolic pathways for fentanyl (a), sufentanil and alfentanil (b), known metabolic pathways for remifentanil and proposed pathway for carfentanil (c). Chemical structure of lofentanil (d)

To date, there are no controlled carfentanil administration studies to know how

carfentanil is distributed and metabolized in the human body. We performed a

comprehensive metabolism study for this fentanyl derivative with two major goals. First,

as carfentanil produces its effects over hours, we assessed metabolic stability in human

liver microsomes (HLM) to determine if carfentanil is toxic because of slow metabolism.

Second, we aimed to identify, for the first time, carfentanil metabolites after in vitro

human hepatocytes incubation. The results of our study might provide a basis for

interpretation of analytical results in future opioid exposures and might confirm Riches et

36

al.’s findings. In addition, it will enable future research on the pharmacological activity of

carfentanil metabolites, which may contribute to its longer duration of action.

3.2 Materials and Methods

3.2.1. Chemicals and Reagents

Caution: Carfentanil is hazardous and should be handled carefully: Narcan (Naloxone

HCl) intranasal injectors or alternative injectors should be readily available before working with this compound. It is highly recommended that multiple injectors be available per person and a buddy system be used..

Carfentanil citrate (95.8% pure determined by 1H and 13C NMR) was synthesized

at Edgewood Chemical Biological Center (ECBC), as it was not commercially available.

A 1 mg/mL solution, adjusted for the citrate salt formulation and purity, was prepared

gravimetrically in LC-MS grade methanol purchased from Sigma Aldrich (St. Louis,

Missouri, USA). LC-MS grade water and formic acid were acquired from Fisher

Scientific (Fair Lawn, New Jersey, USA). LC-MS grade acetonitrile was obtained from

Sigma Aldrich (St. Louis, Missouri, USA). Trypan Blue (0.4%) was used for measuring

cell viability. NADPH regenerating system solutions A (part #451220) and B (part

#451200) were purchased from Corning, Inc. (Corning, New York, USA).

3.2.2. Methods

The majority of experiments were performed according to our previous studies.96

In silico Metabolite Predictions

MetaSite software (v.5.0.3; Molecular Discovery, Pinner, UK) predicted in silico carfentanil metabolites. Predictions were generated using the CYP P450 Liver Model with the following parameters: reactivity correction, 39 common biotransformations, and

37 a minimum mass threshold of 100 Da for predicted metabolites. The algorithm considers both thermodynamic and kinetic factors. Predicted metabolites are assigned a probability score with 100% being the maximum likelihood of generation.

For comparison, ADMET Predictor (v.7.2.0001; Simulations Plus, Inc.,

California, USA) also predicted in silico carfentanil metabolites. This program uses a decision-learning based algorithm, which differs from MetaSite’s quantitative structure- activity relationship (QSAR) algorithm. ADMET Predictor generated predictions using the modeled CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4 enzymes.

Physicochemical properties, e.g. logD, also were predicted using this software simulating conditions at pH = 4.0. This software does not calculate probability scores.

Incubations

For the human liver microsome (HLM) incubations, carfentanil (5 µmol/L) was incubated for 1 h under constant shaking at 37°C with 1 mL of a solution containing

HLM pooled from 50 different donors (1 mg protein/mL; Bioreclamation IVT, Columbia,

MD, USA). HLM suspensions were prepared in duplicate by adding 100 µL 100 mmol/L potassium phosphate buffer, pH 7.4, 100 µL of 100 µg/mL carfentanil solution, 50 µL

NADPH regenerating solution A, 10 µL NADPH regenerating solution B and 50 µL

HLM (20 mg/mL) to 690 µL purified water. 100 µL samples were collected at 0, 3, 8, 13,

20, 30, 45, and 60 min and immediately mixed with an equal volume of ice-cold acetonitrile. Samples were stored at -80°C prior to analysis.

For the hepatocyte incubations, pooled cryopreserved human hepatocytes from ten different donors (Bioreclamation IVT, Columbia, MD, USA) were thawed at 37°C, transferred into pre-warmed InVitro Gro HT medium, centrifuged for 5 min at 50G and

38

the supernatant aspirated. As an additional washing step, the pellet was re-suspended in

InVitro Gro KHB buffer, 50 mL, centrifuged at 50G for 5 min again, and the supernatant aspirated. The pellet was finally re-suspended in 2 mL InVitro Gro KHB buffer. For cell viability measurements, 10 μL suspended cells were dyed with Trypan blue. Viability was >77%. The final drug/cell incubation suspensions contained carfentanil drug solution

(10 µmol/L), and cell suspension (500,000 cells/well). Incubation temperature was 37°C.

Samples were collected at 0, 0.5, 1, 2, 4 and 6 h incubation and were immediately mixed with 0.5 mL ice-cold acetonitrile. Positive diclofenac control samples were incubated and formation of 4’-hydroxydiclofenac and diclofenac acyl glucuronide were evaluated to verify metabolic activity. Samples were stored at -80°C prior to analysis.

Sample Preparations

HLM samples were centrifuged at 4°C for 5 min at 15,000G to remove debris, and supernatants diluted 1:5 with mobile phase A, 0.1% formic acid in water. Ten µL was injected into the MS system. Mobile phases A and B, 0.1% formic acid in acetonitrile (90:10, v/v), and 10 µg/L carfentanil in mobile phase (90:10, v/v) were also analyzed.

Human hepatocyte samples were thawed, vortexed and centrifuged at 4°C for 10 min at 15,000G to remove cell debris, and supernatants diluted 1:5 with mobile phase A

(0.1% formic acid in water) before injecting 10 µL into the MS system. Mobile phase

(50:50, v/v) mixture and 50 µg/L carfentanil in mobile phase solution (90:10, v/v) were analyzed with hepatocyte samples.

Instrumentation

39

HLM samples were analyzed on a 3200 QTRAP mass spectrometer (Sciex,

Redwood City, CA, USA) and data acquired with Analyst v1.6. Chromatographic

separation by a Shimadzu Prominence HPLC system consisting of two LC-20 AD XR

pumps, a DGU-20A5R degasser, SIL-20 AC XR auto-sampler, and CTO-20 AC column

oven (Shimadzu Corp., Columbia, MD, USA).

Hepatocyte samples were analyzed by liquid chromatography with a Shimadzu

Prominence HPLC system consisting of two LC-20 AD XR pumps, a DGU-20A5R degasser, SIL-20 AC XR auto-sampler, and CTO-20 AC column oven (Shimadzu Corp.,

Columbia, MD, USA) and a 5600+ TripleTOF mass spectrometer (Sciex, Redwood City,

CA, USA). Data were acquired with Analyst v1.6.

LC-MS Analysis

For the HLM samples, chromatographic separation was performed on a Kinetex™

C18 column (100 × 2.1 mm, 2.6 µm, Phenomenex, Torrance, CA, USA). The HPLC mobile phases consisted of 0.1% formic acid in water (A) and acetonitrile (B), respectively. Gradient elution was as follows: 10% B from 0 – 0.5 min, ramping up to

95% B from 0.5 – 10 min, holding 95% B from 10 – 12.5 min, followed by re- equilibration of 10% B at 12.5 min until completion at 15 min. LC flow was 0.5 mL/min.

MS parameters were as follows: interface, positive electrospray ionization (ESI) mode; gas 1 and 2, nitrogen, 50 psi; curtain gas, nitrogen, 40 psi; source temperature, 500°C; ion spray voltage, 5500 V. The transitions monitored were m/z 395.2/335.3 (collision energy

CE 25 eV), m/z 395.2 /113.1 (CE 39 eV), and m/z 395.2/246.2 (CE 27 eV), all of which agree with previously published transitions for carfentanil.84 Declustering potential (DP),

40

entrance potential (EP), and collision cell exit potential (CXP) were optimized to the

following values: DP 51 V, EP 4.5 V, CPX 4 V.

For the hepatocyte samples, chromatographic separation was performed similarly

to the HLM protocol, using the same column and mobile phases. Gradient elution

followed the following regimen: 10% B from 0 - 2 min, to 50% B from 2 - 20 min, then increasing to and holding at 95% B from 20 - 22 min, followed by re-equilibration to

10% B at 22 min, until completion at 25 min. MS and MS/MS data were acquired via information-dependent acquisition (IDA) with dynamic background subtraction in positive ESI. For IDA, signals in the TOF-MS survey scan (range, m/z 100-950; accumulation time, 0.1 s) exceeding 500 cps were selected for dependent product ion scans (range, m/z 60-950; accumulation time, 0.075 s) excluding isotopes within 3 Da.

Mass tolerance was 50 mDa. 10 candidates per cycle were allowed. MS source parameters were: gas 1 and 2, nitrogen, 50 psi; curtain gas, nitrogen, 45 psi; source temperature, 500°C, ion spray voltage, 5500 V; DP 80 V. Collision energy for the product ion scans was 35±15 eV. An automated calibration was performed after every third injection to maintain mass calibration.

Human Hepatocyte Data Processing

Metabolite evaluation was performed using MetabolitePilot v1.5 (Sciex, Redwood

City, CA, USA). A list of theoretically possible biotransformations was created based on structures and analogous metabolism of other fentanyl-family compounds. Four different peak finding strategies were utilized: 1) search for predicted metabolites, 2) mass defect filtering with a mass defect of ±50 mDa of the parent mass and several predicted metabolites, 3) search for characteristic product ions and neutral losses (≥2) and 4)

41

generic peak finding, which searches for intense signals not identified by the other three

algorithms and suggests a number of “unexpected metabolites”, which we limited to 10.

MS threshold intensity was 400 cps, MS/MS threshold intensity was 60 cps, and mass

tolerance was set to 50 ppm. Minimum peak width was 2.5 s with a minimum

chromatographic intensity of 1500 cps. Potential metabolites were evaluated based on

three factors: mass measurement error <5ppm; intense product ions had to either match

those in parent compound fragmentation or have a mass shift corresponding to the

metabolite biotransformation(s) with a mass error <10 ppm; retention time had to be

consistent with expected hydrophilic/lipophilic properties in the context of the entire metabolic profile. The in silico predicted logD was instrumental in helping explain retention times. Three controls were used in the hepatocyte experiment: 1) a mobile phase sample accounting for impurities in the solvents and the instrument, 2) a sample of carfentanil in mobile phase checking for impurities in the standard and 3) a hepatocyte sample that was fortified with carfentanil, with hepatocytes precipitated before adding the standard accounting for matrix compounds and consumables.

3.3 Results

3.3.1. Metabolic Stability

HLM half-life 7.8 ± 0.4 min was calculated from log-linear transformation of carfentanil depletion over 1 h (Figure 20), which was translated to microsomal intrinsic

97 clearance (CLint=89.35 µL/min /mg protein) and finally into predicted human hepatic clearance (CL=16.2 mL/min/kg).97

42

Figure 20. Parent compound depletion of carfentanil in human liver microsome (HLM) incubation. Performed in duplicate. However, in hepatocytes, carfentanil peak intensity did not decrease significantly over 6 h. The only slight changes in peak area probably need to be attributed to detector saturation. In contrast to HLM, hepatocytes contain intact cell membranes that must be passed prior to metabolism, which could be a possible reason for the observed differences in depletion.

Factors that may potentially contribute to the longer half-life in vivo could be carfentanil’s lipophilicity or volume of distribution. If this drug is either highly lipophilic and distributes into cellular membranes easily with slow liberation from this medium, or both, this could render less carfentanil available for hepatic extraction and metabolism.

Regarding these factors, evidence exists to suggest that carfentanil is potentially highly bound by plasma proteins, and has a larger volume of distribution than its less potent analogues. This is inferred from experimental data collected on more widely used fentanyl drugs; fentanyl, sufentanil, and lofentanil (Figure 19D) that are 84.4%, 92.5%, and 93.6% bound, respectively.98 Based on structural similarity, we hypothesize that

43

carfentanil should also be highly bound. In regard to volume of distribution, carfentanil

caused renarcotization,99 where an agonist has a longer duration than its reversal agent. It

can be inferred from carfentanil’s reputation for possessing a high risk of renarcotization

that it distributes more widely than other opioids, and is possibly largely sequestered by

plasma proteins. These factors potentially contribute to longer exposure and duration of

effects in vivo.

3.3.2. Carfentanil MS/MS Spectrum

Fragmentation of the parent compound is largely analogous to the collision-

induced dissociation (CID)-MS of remifentanil,100 as we verified in Figure 21. Fragment

m/z 363.2075 is generated by methanol loss from the methyl ester functional group, the

base peak at m/z 335.2126 by loss of the methyl ester group. Fragment m/z 279.1862 is

generated by loss of both the carbomethoxy group and propenone. Fragment m/z

246.1494 results from cleavage at the central bond between carboxamide nitrogen and

piperidine leading to a loss of N-phenyl-propionamide, a fragmentation common for fentanyl and derivatives.100

44

Figure 21. MS/MS of carfentanil and identified fragment structures. Protons are not indicated on metabolite fragment formulae. A 2.016 Da loss, characteristic of a carbomethoxy group loss, can be followed by

a double bond formation on the piperidine, likely through tertiary carbo-cation formation

and a hydrogen shift mechanism. For instance, m/z 202.1226 generated from an

inhomogeneous piperidine ring cleavage results in a protonated N-phenyl-butadienal-

propionamide fragment. This fragment can undergo further propenone loss yielding m/z

146.0966. Instead of tertiary carbo-cation formation and hydrogen shifting, both C-N

bonds are broken during the inhomogeneous piperidine ring cleavage and an additional

double bond is formed yielding m/z 158.0964. Additional evidence for this characteristic

2.016 Da loss is from the m/z 186.1277 fragment, where the central N-piperidyl bond is

cleaved and the charge resides on the phenyl-ethyl-piperidine side. Conversely to the fragment at m/z 202.1226, the charge remaining on the phenethyl side results in the methyl phenthylamine fragment at m/z 134.0969. Likewise, the m/z 113.0606 fragment has two double bonds initiated by the C‐N bond breakages, and m/z 105.0708

45 corresponds to the phenyl ethyl fragment indicating double bond formation, but at this specific location. Later analysis using d5-carfentanil labeled at the N-phenyl ring helped confirm our fragment structures, as shifts of 5 Da were observed in all of the parent fragments containing the N-phenyl group: m/z 146.0966, m/z 158.0964, m/z 202.1226, m/z 279.1862, and m/z 335.2126. This helped correct for discrepancies in predicted versus elucidated fragments within MetabolitePilot.

3.3.3. Metabolite Profiling in Hepatocytes

Metabolite identification includes two steps: First, potential metabolites are detected by data mining the HRMS raw data with software help applying several search algorithms like predicted metabolites, mass defect filtering and search for common product ions and neutral losses. Second, metabolite structures are elucidated based on accurate masses for the protonated molecule and product ions and mass shifts between parent and metabolite. Chemical structures of the identified metabolites and a potential pathway are shown in Figure 22. Retention time, diagnostic fragment ions, precursor ion m/z, and MS areas for all time points for identified metabolites are shown in Table 6.

Metabolites were labeled M1 to M12 based on increasing elution time.

46

Figure 22. Proposed metabolic pathway for carfentanil in human hepatocytes. Metabolite structures derived from MS/MS and chromatographic analysis.

47

Table 6. Detailed Information on All Identified Metabolic Products and Parent Compound

Peak Biotransformation Elemental Precursor Ion Diagnostic Fragment Ions Mass error RT (min) Relative MS Relative MS Relative MS Relative MS Relative MS ID Composition (m/z) (m/z) (ppm) area, 0.5 h area, 1 h area, 2 h area, 4 h area, 6 h

M1 N-dealkylation + C15H20N2O3 277.1547 110, 113, 175, 217, 235, 245 0.1 4.05 0.35 0.63 0.74 0.89 1.00 ester hydrolysis

M2 N-Dealklylation of C16H22N2O3 291.1711 142, 150, 175, 231, 235, 259 2.6 6.89 0.24 0.51 0.57 0.86 1.00 Piperidine Ring

M3 N-dealkylation of C16H22N2O4 307.1655 126, 173, 247, 251, 275 0.7 8.20 0.24 0.64 0.43 0.80 1.00 Piperidine Ring + Hydroxylation of Propanoic Group

M4 Ester Hydrolysis + C23H28N2O4 397.2126 132, 162, 188, 232, 262, 337, 1.1 9.50 0.12 0.39 0.38 0.78 1.00 Hydroxylation of 365, 379 Piperidine Ring

M5 Hydroxylation of C24H30N2O4 411.2285 186, 218, 351 1.6 9.50 0.00 0.20 0.73 1.23 1.00 Propanoic Group

M6 Hydroxylation of C30H38N2O10 587.26 121, 326, 351, 393, 411, 438 1.3 9.70 0.17 0.55 0.42 0.81 1.00 Phenethyl Group + Glucuronidation

M7 Hydroxylation of C24H30N2O4 411.2279 121, 262, 295, 351, 379 0.3 10.42 0.23 0.68 0.46 0.85 1.00 Phenylethyl Structure

M8 Hydroxylation of C24H30N2O4 411.2292 162, 232, 262, 295, 351, 393 3.3 11.20 0.18 0.60 0.43 0.81 1.00 Piperidine Ring

M9 Ketone Formation C24H28N2O4 409.2135 120, 148, 260, 293, 353, 349, 3.3 11.80 0.21 0.49 0.53 0.68 1.00 on Phenethyl Linker 377

M10 N-oxidation of C24H30N2O5 427.2229 166, 172, 274, 278, 303 0.4 12.23 0.49 1.00 0.40 0.65 0.73 Piperidine N + Hydroxylation of Phenethyl Group

P Parent C24H30N2O3 395.2339 105, 113, 134, 146, 158, 186, 2.5 12.59 0.36 0.56 0.77 0.51 1.00 202, 214, 246, 279, 335, 363

M11 N-oxidation of 4' C24H30N2O4 411.2287 105, 154, 247, 262, 274, 290, 2.1 13.48 0.39 0.64 0.52 0.78 1.00 Nitrogen 303

M12 N-oxidation of C24H30N2O4 411.2287 105, 148, 230, 262, 274, 275, 2.0 13.68 0.32 0.62 0.67 0.87 1.00 Piperidine N 303, 379 N-Dealkylated Metabolites

N-dealkylated metabolites (Figure 23) were expected based on fentanyl and other analogue metabolism studies. Indeed, carfentanil demonstrated similar metabolic products - in total, three metabolites (M1, M2, and M3) were generated by N- dealkylation.

M2 (m/z 291.1711, retention time RT 6.89 min) is the N-dealkylated metabolite, or norcarfentanil, which has completely lost the phenethyl substructure indicated by absence of m/z 105.0704, m/z 134.0964 and m/z 186.1277. The remaining parts of the molecule are unchanged as fragments corresponding to portions of the piperidine ring and

4’ bound structures (m/z 146.0964, m/z 158.0964, and m/z 202.1226) remained in the

MS/MS spectrum. M2 was the second most abundant metabolite overall at each time point. However, it should be noted MS peak areas can be affected by ionization efficiencies and matrix effects. Based on the assumption that phase I metabolites show

48

similar peak areas if present in the same concentrations, we estimated and compared their

relative abundances. Carrying an additional acidic functional group, phase II metabolites

like glucuronides and sulfates usually show lower ionization efficiencies in positive ESI

mode.

M2 underwent further biotransformations yielding M1 (m/z 277.1547, RT 4.05

min) and M3 (m/z 307.1655, RT 8.20 min). Ester hydrolysis generated M1 as indicated

by m/z 175.1226 corresponding to an unmodified piperidine ring and 4’ phenyl ring and

m/z 217.1328 verifying that the N-propanoic group was unchanged. The fragment at m/z

113.0603 represents the hydrolyzed ester group and piperidine ring less its nitrogen.

Interestingly, ester hydrolysis often is the major metabolic pathway, as esters are “built in” to make “soft” drugs more labile and easily metabolized. Remifentanil, for example, contains an additional ester moiety that when hydrolyzed results in rapid hepatic-

independent clearance and subsequent offset of effects (t1/2= 3-10 min). However,

carfentanil’s ester moiety might be more sterically hindered than remifentanil’s as it is

protected by the N-phenyl ring.

M3 (m/z 307.1655, RT 8.20) was generated from M2 by monohydroxylation at

the N-propanoic group. Much like M1, m/z 146.0966, m/z 158.0813 and m/z 173.1075

indicates unmodified piperidine and phenyl rings, whereas fragment m/z 275.1395

indicates an intact carbomethoxy group, leaving the propanoic group the only available

site for hydroxylation. This hypothesis is further supported by m/z 247.1442, which

corresponds to M3 less its carbomethoxy group.

49

Figure 23. MS/MS for identified metabolites M1, M2, and M3. Fragment structures used for metabolite identification are indicated in the figure. Protons are not indicated on metabolite fragment formulae.

50

Monohydroxylated metabolites

In total, three metabolites (M5, M7, M8) (Figure 24) were generated by

monohydroxylation. Hydroxylation occurred at three carfentanil sites, most dominantly at

the piperidine ring, and to a lesser extent at the N-propanoic group and the phenethyl substructure.

M8 (m/z 411.2292, RT 11.20 min) is monohydroxylated at the piperidine ring, indicated by m/z 105.0706, which corresponds to an unmodified phenethyl group, and m/z 162.0920, corresponding to a hydroxylated parental fragment at m/z 134.0969 with an additional carbon atom from the piperidine ring. Finally, m/z 113.0607 indicates an unmodified carbomethoxy group, while m/z 202.1229 rules out hydroxylation of the N- propanoic group. Overall, M8 was the most abundant metabolite detected at each time point.

M5 (m/z 411.2285, RT 9.50 min) is hydroxylated at the N-propanoic group. This position was indicated by preservation of the phenethyl group (m/z 105.0704), piperidine ring (m/z 186.1275), N-phenyl ring (m/z 146.0963 and 158.0963), and carbomethoxy

(m/z 113.0602) group. Absence of m/z 202.1229 and presence of m/z 218.1171, a mass shift consistent with hydroxylation, indicated hydroxylation at the N-propanoic group.

M7 (m/z 411.2279, RT 10.42) is hydroxylated at the phenethyl group and a potential precursor to M6. The hydroxylation site is concluded based on fragment m/z 121.0650, which represents the phenethyl precursor fragment m/z 105.0708 plus the mass of an oxygen atom.

51

Figure 24. MS/MS for Identified Metabolites M5, M7, and M8. Fragment structures used for metabolite identification are indicated in figure. Protons are not indicated on metabolite fragment formulae.

52

Metabolites Generated by N-Oxidation

In total, three metabolites were generated through N-oxidation (M11, M12) or

combination of N-oxidation and hydroxylation (M10) (Figure 25). These metabolites all

eluted late, either slightly before or even after the parent, as is often seen for N-oxides.101,

102 M10 (m/z 427.2229, RT 12.23) results from N-oxidation of M7, the phenethyl

hydroxylated metabolite. The fragment at m/z 274.1434 indicates that the piperidine

carbons, the carbomethoxy and the N-propanoic groups remained unchanged. Fragment m/z 166.0854 indicates di-hydroxylation of the phenethyl region including the piperidine nitrogen. Di-hydroxylation of the aliphatic carbons would yield m/z 121.1559 and m/z

137.1553, indicating single and double hydroxylation of m/z 105.0708, respectively however, these fragments were not found. N-oxidation at the piperidine nitrogen explains the m/z 166.0854 fragment and is further supported by M10’s late elution time.

M11 (m/z 411.2287, RT 13.48 min) is generated by N-oxidation of the 4’-position nitrogen. As indicated by parent fragments m/z 105.0703, m/z 113.0598 and m/z

132.0805, the phenethyl group, the carbomethoxy group and the piperidine ring were unchanged. The fragment at m/z 154.0858 indicated N-oxidation of the 4’-position, further supported by m/z 274.1433 and m/z 290.1392; m/z 274.1433 represents the carbomethoxy, N-phenyl, and N-propanoic groups along with all of the piperidine carbons. Further N-oxidation of this group led to m/z 290.1392.

M12 (m/z 411.2287, RT 13.68 min) is the N-oxide metabolic product of the piperidine nitrogen, indicated by fragments at m/z 230.1174 and m/z 262.1436. These are both N-oxide products of parent fragments m/z 214.1225 and 246.1494, both containing

53 only the piperidine nitrogen. The phenethyl ring is determined to be unmodified (m/z

105.0703). Additionally, signature fragments at m/z 146.0966 and m/z 158.0962 are still prominent, indicating that oxidation did not occur at the piperidine carbons or any of the

4’-bound carbon positions.

54

Figure 25. MS/MS for Identified Metabolites M10, M11, and M12. Fragment structures used for metabolite identification are indicated in figure. Protons are not indicated on metabolite fragment formulae. Other Metabolites

Three metabolites were generated by carbonylation, or ketone formation, ester hydrolysis, and glucuronidation (Figure 26). M9 (m/z 409.2135, RT 11.80) was the only

55

ketone product identified. The fragment at m/z 120.0812 indicated a double-bound oxygen on the phenethyl group, and m/z 148.0759, a double-bound oxygen added to the parent fragment (m/z 134.0969) that contained phenethyl group, piperidine nitrogen, and one carbon atom of the piperidine ring. Parent fragments m/z 113.0604, m/z 146.0967, and m/z 158.0968 suggest an intact, unmodified carbomethoxy group, N-phenyl, and piperidine ring. The most likely place for the keto group would be the ethyl-linker region between the piperidine nitrogen and phenyl ring.

M4 (m/z 397.2126, RT 9.50 min) is produced by ester hydrolysis of M8, based on the mass shift of 14.0166, corresponding to methyl group loss. The fragment at m/z

162.0909 is identical to the M8 signature fragment indicating piperidine ring monohydroxylation. The fragment at m/z 379.2004 indicates ester hydrolysis and includes the piperidine ring hydroxyl group. M6 (m/z 587.2607, RT 9.74 min) is generated by hydroxylation of the phenethyl group followed by glucuronidation and is the only phase II metabolite identified. Hydroxylation position and subsequent glucuronidation on the phenethyl group was suggested by parental fragments m/z

113.0604, m/z 146.0960, m/z 158.0966 and m/z 202.1221 indicating an intact carbomethoxy group, N-phenyl ring, and N-propanoic group. Most importantly, m/z

105.0708 was absent and m/z 121.0651 present, indicating phenethyl group hydroxylation. Glucuronic acid is bound to the phenethyl structure (m/z 326.1237).

56

Figure 26. MS/MS for Identified Metabolites M4, M6, and M9. Fragment structures used for metabolite identification are indicated in figure. Protons are not indicated on metabolite fragment formulae.

57

3.4 Discussion

3.4.1. Metabolite Intensities

Signal intensities of the 12 metabolites were tracked at 1, 4 and 6 h. The most

intense peaks were M8 (piperidine ring hydroxylation) and M2 (N-dealkylation forming norcarfentanil), followed by M9 (ketone formation at phenethyl linker), M5 (N-propanoic

hydroxylation), M7 (phenethyl ring hydroxylation), M12 (piperidine N-oxidation) and

M11 (4-position nitrogen N-oxidation). The only phase II metabolite identified was a glucuronide conjugate of hydroxylated carfentanil (M6) with low intensity. As there was

a constant high amount of parent compound available over time, signal intensities of

second-generation metabolites increased, but without decreasing first-generation

metabolite signal intensities. Overall, the metabolic profiles at all-time points look

similar.

3.4.2. Support for Riches et al?

M2, or norcarfentanil, showed the second most abundant signal in our human

hepatocytes profile, ranking second after M8. These results confirm Riches et al.’s

hypothesis that norcarfentanil is a carfentanil metabolite; however, norcarfentanil is a

metabolite of both carfentanil and remifentanil. Riches et al. suggested use of both

compounds based on positive results of clothing samples and did not attribute the

presence of norcarfentanil to either. Future forensic analyses in which opioid exposure is suspected require identification of more specific metabolites. Our data suggest the piperidine ring-hydroxylated carfentanil metabolite (M8), containing the entire parent compound including the ester group and showing the most intense peak in our samples.

However, verification in post-exposure urine or blood is needed.

58

3.4.3. Implications for Carfentanil Toxicity

HLM half-life was 7.8 min, which indicates that toxicity would not be the result

of inadequate enzymatic capacity to metabolize carfentanil. However, substrate depletion

in the human hepatocyte samples was slow. According to common classifications,103

carfentanil is categorized as a “high clearance” compound. This is an interesting result as

all observations about carfentanil effects point towards a long in vivo half-life. The HLM experiment definitely excludes metabolic clearance as the time-limiting factor.

Based on its calculated logP value of 3.7 104, 105 to 4.01,106, 107 indicating high

lipophilicity, and the well-studied behavior of other fentanyl compounds,108 we assume

that carfentanil is rapidly distributed to and stored in fatty tissues. This could prolong low

level re-absorption into the systemic circulation and prolong toxicity when combined

with carfentanil’s high receptor affinity. Opioids have a central nervous system

depressant effect that manifests as decreased respiration, which can lead to respiratory

failure, apnea, and death.109 Except for the N-dealkylated metabolites that lost the entire phenethyl structure, many metabolites were only slight modifications of carfentanil.

Metabolites with a single hydroxylation could maintain activity at the µ-opioid receptors, potentially prolonging the effect. For example, β-hydroxyfentanyl, though not a fentanyl metabolite, is a potent fentanyl analogue distributed on the black market in the 1980’s.110

M7, hydroxylated at the phenethyl group (that includes the mentioned β-position), could be an active carfentanil metabolite. On the other hand, M2 and M3 are similar to fentanyl metabolites that showed little to no activity at µ-opioid receptors in vitro. Norfentanyl, analogous to carfentanil M2, and hydroxyl-norfentanyl, analogous to M3, are expected to be inactive,111 but additional research is needed for definitive conclusions.

59

3.5 Conclusions

HLM metabolic stability indicated that carfentanil is readily metabolized by CYP

enzymes. We propose that slower hepatocyte metabolism was likely due to poor

accessibility of the compound to these enzymes for reasons that need to be addressed in

future studies. Despite slow parent depletion in hepatocytes, we identified 12 carfentanil

metabolites. Phase I biotransformations dominated metabolite formation and included

monohydroxylation, N-dealkylation, N-oxidation, ester hydrolysis, and carbonylation.

The two most abundant metabolites were M8 and M2. M2 is the N-dealkylated carfentanil, more commonly known as norcarfentanil. Only one phase II metabolite (M6) was identified, the glucuronide of M7. M7 was monohydroxylated at an undetermined position on the phenethyl group.

M2, or norcarfentanil, is a good indicator of exposure to carfentanil being one of the two most intense peaks during hepatocyte incubation. However, M2 is not exclusive to carfentanil, and carfentanil administration must be confirmed by looking for other more unique metabolites such as M8. This metabolite had even higher abundance than

M2, and contains the entire carfentanil structure, lending itself to forensic identification if found. This study has opened the door to new questions about carfentanil and how it exhibits its toxicity. First and foremost, in vivo studies need to be conducted to confirm the metabolites identified in this study. Plasma protein binding assays can assess carfentanil binding, and in silico models might predict plasma drug concentrations, exposure duration after administration, and toxicity by physiologically based pharmacokinetic studies (PBPK). This might explain the slower than anticipated carfentanil depletion observed in hepatocytes. Finally, performing a similar study in liver

60 spheroids, more representative of a whole liver, could compare performance of this newer technology in assessing carfentanil clearance and metabolite formation.

This study demonstrated that carfentanil metabolism by CYP enzymes is not a restrictive process, and thus likely doesn’t contribute to prolonged exposure to the drug, but the slow parent compound depletion observed during the hepatocyte incubations does indicate some sort of sequestration process that is likely based on a physicochemical property. Therefore, properties such as plasma protein binding and RBC/plasma partitioning need to be assessed in an attempt to explain the metabolic differences seen between the microsomal and hepatocyte incubations.

61

Chapter 4. Physicochemical Properties of Carfentanil

4.1 Introduction

In addition to metabolism, other influences on pharmacological activity, pharmacological profile, and subsequent toxicity, are the physicochemical properties of the compound; none of which are known for carfentanil. This study aimed to establish, in human in vitro models, both the plasma protein binding (PPB) and red blood cell

(RBC)/plasma partitioning (RBP) of carfentanil. PPB influences how quickly the drug is cleared from systemic circulation and how much is available to act on its receptor targets.

RBP indicates if the RBCs act as a reservoir for the compound, directing what type of biosample is best suited to detect the minute quantities of drug able to illicit a therapeutic or toxic response in humans.112

Physicochemical properties are additionally vital to any pharmacokinetic modeling effort. Next to metabolism, PPB and RBP are key modeling parameters because they directly influence how much unbound drug is free in systemic circulation, and where that free drug goes to within the central blood compartment.113 Because of carfentanil’s high calculated lipophilicity, it is a candidate to be sequestered in lipid membranes, to include those of the RBCs themselves.104, 105 Therefore, if carfentanil migrates heavily into the RBC partition of the blood, the RBCs can act as a sink, prolonging exposure.114

Additionally, knowing the RBP will guide forensic specialists and medical personnel to the proper sampling protocol, be it whole blood, plasma, or serum.115, 116

62

4.2 Methods

4.2.1. Reagents

Carfentanil was synthesized at Edgewood Chemical Biological Center (ECBC)

(APG, MD, USA) and purity verified by 13C and 1H NMR. Norcarfentanil was purchased

from Toronto Research Chemicals, Inc. (Toronto, CA). Solutions for protein binding and

red blood cell (RBC) partitioning experiments were prepared at 10 mM (free base) in

dimethyl sulphoxide (DMSO) EP/USP grade (Sigma Aldrich, St. Louis, MO, USA ).

Stock solutions for calibration standards and quality control samples were prepared

gravimetrically at 1.0 mg/mL (free base equivalents) in methanol (LC-MS Ultra

Chromasolv, Sigma Aldrich, St. Louis, MO, USA). Isotopically labelled internal standard (ISTD) compounds were obtained from Toronto Research Chemicals (Toronto,

CA) and prepared at 4.9 mg/mL (carfentanil-d5, free base) and 1.0 mg/mL

(norcarfentanil-d5, free base) in methanol. All solutions were stored at -20°C until use.

Formic acid (FA, reagent grade >95%) and acetonitrile (ACN, Chromasolv Plus for HPLC, ≥99.9%) were obtained from Sigma-Aldrich (St. Louis, MO). Isopropanol

(IPA, ACS reagent grade 99.6%) was purchased from Acros Organics (VWR, West

Chester, PA).

Phosphate-buffered saline (PBS) 10X was obtained from Electron Microscopy

Sciences EMS (Hatfield, PA) and diluted 1:10 with deionized water (dH2O) from a

Millipore (Darmstadt, Germany) Academic water purification system. The resultant salt concentrations were 2.7 mM potassium chloride, 10 mM disodium phosphate, and 137 mM sodium chloride. The pH was verified as 7.4 and the solution was stored at 4°C until use.

63

Pooled gender human and rabbit heparinized whole blood and plasma were

obtained from Bioreclamation IVT (Westbury, NY) and stored at 4°C or -80°C,

respectively, until required.

4.2.2. 96-Well Equilibrium Dialysis

Thermo Scientific™ single-use rapid equilibrium dialysis (RED), 8K molecular weight cut-off, 96-well plates were employed to measure plasma protein binding properties for carfentanil and norcarfentanil. Human and rabbit plasma were each thawed and centrifuged (2,300 g for 15 minutes) to remove any particulate. 1.0 µM and 10 µM test concentrations were prepared similarly in human and rabbit plasma (0.01 and 0.1%

DMSO, v/v). Nine aliquots of 300 µL test plasma/500 µL PBS buffer for each test concentration were added to adjoining wells of a RED plate to allow three experiments for each concentration and time point. Unadulterated plasma experiments were included for analytical quality control (QC) blanks. The plate was covered with adhesive sealing tape and incubated at 37°C with platform mixing. Aliquots were transferred to micro centrifuge tubes at 2 Hr intervals (2, 4, and 6 Hrs) and matrix-matched before LC/MS/MS sample preparation/analysis as indicated in Table 7.

Table 7. Preparation volumes for matrix-matched RED samples

Test Concentration Transfer Volume (µL) Matrix Match Additional 1 µM RED Plasma 100 100 µL PBS 1 µM RED PBS 100 100 µL Hu Plasma 10 µM RED Plasma 10 100 µL PBS 90 µL Hu Plasma 10 µM RED PBS 100 100 µL Hu Plasma 1 µM RED Plasma 100 100 µL PBS 1 µM RED PBS 100 100 µL Rb Plasma 10 µM RED Plasma 10 100 µL PBS 90 µL Rb Plasma 10 µM RED PBS 100 100 µL Rb Plasma

64

4.2.3. Red Blood Cell Partitioning by Depletion

Aliquots of whole blood were centrifuged (1,200 g for 10 minutes) to generate common-origin plasma used to prepare spiking and reference solutions. 10 µM fentanyl, carfentanil, and norcarfentanil spiking solutions were prepared in plasma from 10 mM

DMSO stocks. RBC partitioning experiments were conducted at 30 and 500 nM concentrations. Each 30 nM whole blood test solution was prepared by adding 15 µL of the 10 µM fentanyl, carfentanil, or norcarfentanil spiking solution to 4,985 µL of whole blood. Each 500 nM whole blood test solution was prepared similarly (250 µL to 4,750

µL). The blood test solutions were gently mixed with inversion. 30 and 500 nM reference plasma solutions were also prepared, substituting common-origin plasma for whole blood.

Each whole blood test and reference plasma solution was evenly distributed into five micro centrifuge tubes. The tubes were labelled for time points of 0, 10, 30, 60, and

120 minutes. The 10-120 minute tubes were incubated at 37°C, with rotational mixing.

Whole blood test and reference plasma tubes were removed from incubation at each time point. The whole blood test solution tubes were centrifuged; the resulting plasma was transferred to a clean micro centrifuge tube and mixed. Three 100 µL aliquots were transferred to clean micro centrifuge tubes to provide triplicate preparation and analysis.

The reference plasma tubes were vortex mixed and directly aliquoted.

The hematocrit was determined for each lot of whole blood evaluated and was measured on a HESKA® HemaTrue® Veterinary Hematology Analyzer (Loveland, CO,

USA).

65

4.2.4. Sample Preparation for LC/MS/MS Analysis

Plasma samples were spiked with ISTD and mixed. Proteins were precipitated

with ACN containing 0.1% FA (v/v). Samples were mixed, batched, and stored at -80°C

overnight.

Upon analysis, samples were thawed, mixed, and centrifuged. For filtration and

the removal of phospholipids, the supernatants were transferred to a HybridSPE-

Phospholipid Technology 96-well protein precipitation plate (Sigma-Aldrich, St. Louis,

MO) pretreated with ACN/0.1% FA (v/v). Vacuum was applied to transfer the samples

into a 96-well collection plate.

The collection plate was dried using a Biotage (Charlotte, NC) TurboVap 96

Automated Evaporation System. The samples were reconstituted with 100 µL of 90:10 dH2O:ACN, covered with a cap mat, mixed for 5 minutes, and transferred to the instrument auto sampler for LC/MS/MS analysis.

4.2.5. LC/MS/MS Analysis

The analytical instrumentation used for the quantitative analyses of carfentanil

and norcarfentanil consisted of an Agilent (Santa Clara, CA) 6490 Triple Quad

LC/MS/MS equipped with a 1290 Infinity LC stack with binary pump, auto sampler, and

column heater. Agilent MassHunter Workstation software was used for data acquisition

and analysis. Reverse phase separation was conducted using a Waters (Beverly, MA)

Acquity UPLC BEH 1.7 µm, 2.1x50mm column, with a phase matched VanGuard guard

column, isothermal at 40°C. Mobile phase consisted of (A) dH2O with 0.1% FA and (B)

ACN w/ 0.1% FA. For carfentanil, a 2 minute gradient elution was employed starting

with 10% B for 0.5 min, to 99% B at 1 min, hold for 0.5 min, then return to initial

66

conditions/equilibrate for 0.5 min. For norcarfentanil, a 3 minute gradient elution was

used starting with 15% B for 1 min, to 99% B at 1.5 min, hold for 1 min, then return to

initial conditions/equilibrate for 0.5 min. Flow rate was 0.5 mL/min.

The mobile phase flow was delivered to an Agilent Jet Stream electrospray

ionization (ESI) source maintained in positive ion mode. MS/MS discrimination was

performed via the multiple reaction monitoring (MRM) technique, incorporating isotope

dilution. Three mass transitions were monitored for each analyte: quantitation,

confirmation, and internal standard. Linear calibration curves (1/x weighting,

analyte:ISTD peak area ratio) were generated for each analyte over the range of 1-1,000

ng/mL.

For the RED experiments, the unbound fraction in plasma was calculated from the ratio of the average buffer side concentration to the average plasma side concentration.

For the RBC:plasma partitioning experiments, the partition coefficient (KRBC/PL) was

calculated using Equation 1.117

Equation 1: = × 𝑅𝑅𝑅𝑅𝑅𝑅 1 + 1 𝐼𝐼𝑃𝑃𝑃𝑃 𝑅𝑅𝑅𝑅𝑅𝑅 1 𝐾𝐾 𝑃𝑃𝑃𝑃 𝐻𝐻 � 𝐼𝐼𝑃𝑃𝑃𝑃 − � where is the partition coefficient of the analyte in red blood cells, H is the 𝑅𝑅𝑅𝑅𝑅𝑅 𝐾𝐾 𝑃𝑃𝑃𝑃 hematocrit value, is the average analyte concentration in the reference plasma, and 𝑅𝑅𝑅𝑅𝑅𝑅 𝑃𝑃𝑃𝑃 is the average 𝐼𝐼analyte concentration in the whole blood test plasma.

𝑃𝑃𝑃𝑃 𝐼𝐼 4.2.6. Comparison with in silico Predictions

Results of the protein binding and plasma partitioning were compared to predicted

values generated in ADMET Predictor™ version 8.0.4.6 (Simulation Plus, Inc. ©2016).

Model parameters were used unmodified and based on structure alone. Properties

67 compared were “PrUnbnd” or percent unbound to blood plasma proteins and “RBP” or blood-to-plasma concentration ratio. Calculated logP (octanol:water partition coefficient) is also reported in ADMET Predictor™, and is reported in this manuscript to give context to the overall pharmacological differences between compounds.

4.3 Results

4.3.1. Physicochemical Properties

While drug potency is an important aspect in regards to its pharmacological action, it should be understood that the effect of any drug is a composite of various properties that include all of the ADME paradigm. Specifically, as logP, PPB, and RBP all contribute to absorption, distribution, and metabolism of the drug. These properties were previously unknown for carfentanil and norcarfentanil. This study set out to determine these values for the purposes explaining the differences in metabolism seen when human liver microsomes and human hepatocytes were incubated with carfentanil.85

Both physicochemical properties were time and concentration dependent; 30 minute values for protein binding and 60 minute values for blood/plasma partitioning are reported here (Table 8 and Table 9 respectively).

Table 8. Protein binding of fentanyl, carfentanil, and norcarfentanil in human plasma

Human Total % protein binding Rabbit Total % protein binding Conc. Fentanyl Carfentanil Norcarfentanil Fentanyl Carfentanil Norcarfentanil 1 µM 88.3 87.6 11.3 80.9 83.3 27.8 10 µM 86.7 83.8 10.9 75.4* 82.8 30.7 * value was outside of calibration curve

Table 9. Blood/plasma partitioning of fentanyl, carfentanil, and norarfentanil.

Human RBP Rabbit RBP Conc. Fentanyl Carfentanil Norcarfentanil Fentanyl Carfentanil Norcarfentanil 30 nM 1.0656 0.2746 1.2063 0.8643 0.9287 1.5594 500 nM 0.8295 0.7685 0.8499 1.3041 0.9681 1.1860

68

4.3.2. In silico Predictions of Physicochemical Properties

The in silico predictions, utilizing the Simulations Plus ADMET Predictor, which

is based on the 2-dimensional structure of carfentanil and norcarfentanil, were generally

found to not correspond well with experimentally derived values (Table 10). Plasma protein binding was over-predicted in both cases, and blood/plasma partitioning coefficients were under-predicted, but could also be time dependent, an undefined parameter in this model.

Table 10. In silico predicted values for octanol:water partitioning coefficient (logP), plasma protein binding expressed as percent bound, and blood/plasma partitioning ratio from Simulations Plus, Inc.’s ADMET Predictor.

Carfentanil Norcarfentanil LogP (calculated) 3.786 1.289 Protein Binding 92.747 42.778 Blood/Plasma Partitioning 0.702 0.755

4.4 Discussion

This study helped demonstrate that carfentanil’s potency and subsequent toxicity

are not influenced greatly by its PPB compared to fentanyl, but carfentanil may be influenced by its RBP. The observed human PPB of carfentanil in this study was

determined to be 87.6%, compared to fentanyl’s 88.3% (previously reported values for

fentanyl human PPB of 84.4%98 and 86-89%118). This is a rather unremarkable difference

and doesn’t account for the vastly different drug behavior compared to receptor potency

and efficacy see here. Norcarfentanil is, as expected, much less bound by plasma

proteins. This indicates its ability to be cleared and excreted more readily.

The biggest contribution regarding carfentanil’s distribution in blood and other

tissues comes from the RBP. Carfentanil, surprisingly, and unlike fentanyl, has a strong

proclivity to migrate into the plasma of the human samples (RBP of 0.2746), and is in

69

fact not sequestered in the membranes of RBC’s, which would be assumed to happen

based on its logP alone. Highly lipophilic drugs tend to be sequestered in cell membranes,

including those of RBCs, and RBCs provide very good sink conditions for these types of

compounds. However, in the case of carfentanil, it tends to exist in the plasma despite its

higher logP than fentanyl, which is more equally split between the RBC and plasma

compartments with a ratio of 1.0656. This observation was not made in the rabbit plasma

with an RBP value of 0.9287, indicating a slightly higher probability to migrate to the

plasma over RBCs. Norcarfentanil’s partitioning puts it slightly more in the RBC

compartment than plasma, but this information combined with its relatively low potency

at the receptor make it a low-risk toxicant, especially considering the low concentration

of norcarfentanil that would be present, even in a carfentanil overdose scenario.

This partitioning property has dual use, both in pharmacokinetics and forensic

analysis or analytical chemistry. Detection of these compounds is of interest, be it for

point-of-care diagnostics, sports medicine or anti-doping requirements, or post-mortem

toxicological screenings. Knowing which bio-compartment to best sample from helps to increase the chance of detection. This study indicates that for carfentanil in humans, the plasma is the most advantageous compartment to target for detection, as it will have roughly 3-4 times the amount of carfentanil compared to whole blood or RBC fractions.

Norcarfentanil, comparably, would be detectable slightly more in the RBC/whole blood fraction than plasma in both human and rabbit exposures.

4.5 Conclusions

This study provides important data on the physicochemical properties of both carfentanil and norcarfentanil in rabbit and human blood; data that were previously

70 unreported. Available experimental data for these properties for carfentanil is vital to the pharmacokinetic (PK) modeling portion of this research (see Chapter 5). PPB and RBP contribute heavily to in vivo PK, and the in silico modeling of kinetics. This is because they have a large role in determining the absorptive, distributive, and metabolic properties. These properties in turn impact how much carfentanil is in the various tissue compartments (and volume of distribution), how much of that is in systemic circulation and available to cross the BBB, act on its target receptor(s), or be cleared by metabolizing organs/enzymes; all of which determine overall exposure intensity, duration, and subsequently, toxicity. These data will also provide useful tools when performing forensic or analytical chemical analysis on carfentanil containing samples. This information, combined with that of the microsomal stability and metabolism study, gives a better understanding of the pharmacokinetics of this poorly studied drug compound, carfentanil.

71

Chapter 5. PBPK Modeling of In Vivo Carfentanil Exposures in the Rabbit and Comparison with Human Predictions

5.1. Introduction

The pinnacle purpose of the work performed as part of this thesis is to generate an

in silico model of exposure to carfentanil for use in human toxicity prediction. In order to

validate the in silico model, it is necessary to have either a) human data or b) a surrogate

animal model through which to build the computer model. Since that former is outside of

the scope of this thesis and institution, a surrogate animal model was chosen and the in

silico model was conducted incorporating in vitro experimental data to investigate the

effectiveness of providing experimental data to the model. It is in the effort to reduce,

refine, and replace animal models that in silico and in vitro technologies are taking on a

greater role.

The choice of animal model is vital to the successful translation of any

xenobiotic’s effects to human relevance, regardless of the endpoint. Of course, having an

appropriate animal for both the mechanism of the drug and potential off-target or adverse

outcome pathways (AOP) would be ideal. Take, for example, fentanyl. It was first

screened in mice and rats, animals for which there is a very wide therapeutic index (TI),

or difference between therapeutic and lethal doses. These animals are very good

therapeutic models in an appropriate assay that measures the endpoint, i.e. the RTW

assay.59 However, this is not representative of higher order species’ response, since they

are more susceptible to the toxic effects of opioids, and therefore have a TI orders of magnitude less. The animal model chosen for modeling the effects of carfentanil was the

New Zealand White Rabbit (Oryctolagus cuniculus).The rabbit was chosen for this study because it is a higher order species than both rats and mice, and has been shown to be a

72

good model of opioid therapy as well as experiencing OIRD at a lower dose than that of

rodents, i.e. a smaller TI.119-121 Additionally, rabbits are logistically and economically less

cumbersome than non-human primate species that may be more representative of

humans.

In order to better model in vivo pharmacokinetic data and translate the in silico

modeling from rabbit to human, experimentally derived values for drug clearance and

physicochemical properties will be used to populate the model. In Chapter 3 intrinsic

clearance values were derived in HLMs, but not rabbit liver microsomes (RLMs).

Physicochemical properties in both rabbit and human blood were measured in Chapter 4.

Drug clearance is the single most important parameter in modeling.113 Clearance,

while influenced by the appropriate prediction or experimental determination of

physicochemical properties, affects predicted exposure in the central compartment,

circulation, and thus bioavailability over time. It is therefore vital that when building a model to predict human pharmacokinetics (PK), clearance is scaled properly, in the case

of human prediction, or known in the case of modeling surrogate animal kinetics. The

same study performed with HLMs was therefore repeated here using RLMs.85 From this

study, rabbit microsomal half-life was calculated and used to derive intrinsic clearance

values used in the in silico model.

The in silico pharmacokinetic model used in this study is a physiologically based

pharmacokinetic (PBPK) model called GastroPlus. PBPK models can be advantageous

over traditional compartmental models because they take into account physiological

properties of each organ system in a body and treat each as its own compartment. Some

of these properties include perfusion rates, partitioning coefficients, and CYP enzyme

73 content and activity.122 These values are based not only on the organ/tissue in question, but are species specific as well. PBPK models, therefore, are also more useful, typically, when attempting to translate from non-human species to human.

GastroPlus contains PBPK models of humans of various ethnicities, age, weight, and gender, well as model species such as mouse, rat, rabbit, dog, minipig, and monkey.

The organ systems for these different species/humans include the gut, liver, lungs, spleen, heart, brain, kidney, skin, adipose, venous blood, arterial blood, as well as yellow and red marrow.

This study compared a naïve rabbit model, i.e. no experimental data provided, to pharmacokinetics seen in in vivo intravenous (IV) exposures of rabbits to a single dose of carfentanil. Experimentally derived data was then substituted in place of the predicted intrinsic clearance and physicochemical properties values in a stepwise fashion, running the model between each data point replacement. This way, we assessed which properties, or combination of properties, influenced the model to be more representative of the in vivo exposures. Finally, equivalent dosing was determined in the human by switching to the human PBPK model and substituting experimental HLM intrinsic clearance and each of the human-based physicochemical properties.

5.2. Methods

5.2.1. RLM Stability Study

Caution: Carfentanil is hazardous and should be handled carefully: Narcan (Naloxone

HCl) intranasal injectors or alternative injectors should be readily available before working with this compound. It is highly recommended that multiple injectors be available per person.

74

Carfentanil citrate (95.8% pure determined by 1H and 13C NMR) was synthesized at Edgewood Chemical Biological Center (ECBC), as it was not commercially available.

A 1 mg/mL solution, adjusted for the citrate salt formulation and purity, was prepared gravimetrically in LC-MS grade methanol purchased from Sigma Aldrich (St. Louis,

Missouri, USA). LC-MS grade water and formic acid were acquired from Fisher

Scientific (Fair Lawn, New Jersey, USA). LC-MS grade acetonitrile was obtained from

Sigma Aldrich (St. Louis, Missouri, USA). Trypan Blue (0.4%) was used for measuring cell viability. NADPH regenerating system solutions A (part #451220) and B (part

#451200) were purchased from Corning, Inc. (Corning, New York, USA).

For the rabbit liver microsome (RLM) incubations, carfentanil (5 µmol/L) was incubated for 1 h under constant shaking at 37°C with 1 mL of a solution containing

RLM (1 mg protein/mL; Bioreclamation IVT, Columbia, MD, USA). RLM suspensions were prepared in duplicate by adding 100 µL 100 mmol/L potassium phosphate buffer, pH 7.4, 100 µL of 100 µg/mL carfentanil solution, 50 µL NADPH regenerating solution

A, 10 µL NADPH regenerating solution B and 50 µL RLM (20 mg/mL) to 690 µL purified water. 100 µL samples were collected at 0, 3, 8, 13, 20, 30, 45, and 60 min and immediately mixed with an equal volume of ice-cold acetonitrile. Samples were stored at

-80°C prior to analysis.

RLM samples were centrifuged at 4°C for 5 min at 15,000G to remove debris, and supernatants diluted 1:5 with mobile phase A, 0.1% formic acid in water. Ten µL was injected into the MS system. Mobile phases A and B, 0.1% formic acid in acetonitrile (90:10, v/v), and 10 µg/L carfentanil in mobile phase (90:10, v/v) were also analyzed.

75

RLM samples were analyzed on a 3200 QTRAP mass spectrometer (Sciex,

Redwood City, CA, USA) and data acquired with Analyst v1.6. Chromatographic separation by a Shimadzu Prominence HPLC system consisting of two LC-20 AD XR pumps, a DGU-20A5R degasser, SIL-20 AC XR auto-sampler, and CTO-20 AC column oven (Shimadzu Corp., Columbia, MD, USA).

Chromatographic separation was performed on a Kinetex™ C18 column (100 ×

2.1 mm, 2.6 µm, Phenomenex, Torrance, CA, USA). The HPLC mobile phases consisted of 0.1% formic acid in water (A) and acetonitrile (B), respectively. Gradient elution was as follows: 10% B from 0 – 0.5 min, ramping up to 95% B from 0.5 – 10 min, holding

95% B from 10 – 12.5 min, followed by re-equilibration of 10% B at 12.5 min until completion at 15 min. LC flow was 0.5 mL/min. MS parameters were as follows: interface, positive electrospray ionization (ESI) mode; gas 1 and 2, nitrogen, 50 psi; curtain gas, nitrogen, 40 psi; source temperature, 500°C; ion spray voltage, 5500 V. The transitions monitored were m/z 395.2/335.3 (collision energy CE 25 eV), m/z 395.2

/113.1 (CE 39 eV), and m/z 395.2/246.2 (CE 27 eV), all of which agree with previously published transitions for carfentanil.84 Declustering potential (DP), entrance potential

(EP), and collision cell exit potential (CXP) were optimized to the following values: DP

51 V, EP 4.5 V, CPX 4 V.

5.2.2. In Vivo Exposure of Rabbits to Carfentanil

All animal exposures were conducted in accordance with the U.S. Army –

Edgewood Chemical Biological Center Institutional Animal Care and Use Committee

(IACUC) under animal use protocol 16-473. Four male New Zealand White Rabbits

(Covance, Inc., Denver, PA, USA) were given a 1 µg/kg intravenous bolus dose of

76

carfentanil in sterile saline in the marginal ear vein. At the initiation of the study, the

rabbits weighed from 2.49 to 2.63 kg. Blood samples were collected from the non-

injection ear’s marginal ear vein in tubes containing Li-EDTA as the anticoagulant, pre- exposure, and at 0.017 (1 min.), 0.083 (5 min.), 0.17 (10 min.), 0.5 (30 min.), 1, and 24 hours after the intravenous administration. 100% of the animals dosed were collapsed and

75% were prostrate. All animals recovered to normal apparent behavior within 24 hours.

Plasma was collected and stored at -80°C until analysis. The concentration of carfentanil in each plasma sample was determined by LC-MS/MS analysis. Time zero (immediately after injection) values were calculated based on the 1 µg/kg dose, actual body weight of each rabbit, and a blood volume of 60 mL/kg.123 Two phase decay analysis was

performed using GraphPad Prism v.7.02 (Graphpad Software, Inc., La Jolla, CA, USA).

5.2.3. Human and Rabbit PBPK profile simulations using GastroPlus

GastroPlus v.9.0.0007 simulation software (Simulations Plus, Lancaster, CA) was

used to predict the PK profile of carfentanil in human and rabbit based on the species

specific physicochemical properties of the compound (Chapter 4) and the predicted clearance (Chapters 3 and 5). Specifically, the rabbit PPB = 83.3%, RBP = 0.9287, and

RLM derived CLint values were used to populate the rabbit PBPK simulation. For all of

the other parameters, the default settings of GastroPlus were used. The PK profile

simulated with a dose of 1 µg/kg body weight administered IV were compared with those obtained in live rabbits through the same dose and route of administration (See 5.2.2).

Based on the in vivo findings, the rabbit PBPK model is associated with 100% collapse of the animals at this dose.

77

Human PBPK simulations were conducted similarly, substituting the human values from Chapter 3 and Chapter 4 for predicted values. Specifically, the human PPB =

87.6%, RBP = 0.2746, and both CLint values were used (16.2 mL/min/kg-weight and

89.35 µL/min/mg-protein) to assess the extrapolation from HLM to whole body clearance presented in Chapter 3.103 The dose was adjusted manually from equal to that of the rabbit (1 µg/kg) lower until it achieved the same peak plasma concentration and PK profile as the model-validated rabbit data. The human PBPK parameters used were the default 70 kg, 30 year old male human standard present in the GastroPlus software.

5.3 Results

5.3.1. Carfentanil Stability Study in RLMs

RLM half-life 2.1 ± 0.1 min was calculated from log-linear transformation of carfentanil depletion over 1 h (Figure 27), which was translated to microsomal intrinsic

97 clearance (CLint=336.1 µL/min x mg protein) and finally into predicted human hepatic clearance (CL=19.8 mL/min/kg) based on a liver weight of 45.1 grams per kilogram of body weight and protein concentration of 109 mg protein per g liver for the New Zealand

White Rabbit.97, 124, 125 This clearance rate is 3.7 times faster than the clearance in HLMs

(7.8 min.).

78

Figure 27. Parent compound depletion of carfentanil in rabbit liver microsomes (RLM) incubation. Performed in duplicate.

5.3.2. In Vivo PK Study of Carfentanil in Rabbits

A single bolus dose of 1.0 µg carfentanil per kg of body weight in four male New Zealand White Rabbits yielded a pharmacokinetic curve indicating rapid redistribution from the central compartment and slower clearance from the blood. The two phase decay half-life of carfentanil in these animals was calculated to be 2.6E-8

minutes for the redistribution phase and 6.09 minutes for the clearance phase (R square =

0.9995) (Figure 28).

Figure 28. Pharmacokinetic data for rabbits dosed at 1 ug/kg. Plotted as mean plasma concentrations with error bars as SD. n=4

79

5.3.3. In silico Modeling of Rabbit PK

Modeling of the actual PK data in a naïve model was slightly under-predictive of the overall clearance of carfentanil (Figure 29). The predicted plasma concentrations

(solid line) were higher than those observed (black dots). As stated previously, metabolism, i.e. clearance is the most influential parameter in most PBPK models,113 and is obviously under-predicted in this naïve model. Additionally, it can be observed as well that the initial redistribution phase isn’t nearly rapid enough in the predicted model. This is something not due to clearance, but the physicochemical properties, likely the RBP.

Redistribution influenced by RBP is the most influential property that can act rapidly enough to influence the first few minutes of the model. With no experimental data, though, the predicted PK runs parallel to the actual PK, only slightly higher. If nothing was known about this compound, this prediction wouldn’t be an irresponsible place to start considering it overestimates the plasma levels, and subsequently the toxicity; a prudent and conservative place to begin with an unknown toxicant.

Figure 29. Naïve PBPK model using only predicted ADMET properties based on the 2D structure of carfentanil. Line = predicted PK. Dots represent actual in vivo data. Error bars are SD of actual plasma concentrations.

80

Because metabolism is reported to be the most influential parameter to change in

a PBPK model, this was added into the model first (Figure 30). Compared to the purely in silico predicted model above, this prediction incorporating the experimentally derived

CLint is closer to the observed in vivo PK. This model improves upon the pure prediction

in the later time points only, however. This appears to agree with analogous compounds

that are better studied. Fentanyl has been shown to undergo a three phase PK curve: rapid

redistribution, slower distribution, and finally an elimination phase.126 The first two

phases, therefore, would likely not be highly influenced by clearance, but by

physicochemical properties that dictate their redistribution.

Figure 30. PBPK model incorporating only CLint. Experimental Data. Line = predicted PK. Dots represent actual in vivo data. Error bars are SD of actual plasma concentrations.

Following a stepwise approach, the PPB property was added to the model,

without clearance, to see what effect it had on the PK prediction alone (Figure 31). This

property had the least influence over the model. It was anticipated that with other

81

experimentally derived properties being included with PPB into the model, it could work

synergistically with RBP or clearance values to improve the model.

Figure 31. PBPK prediction incorporating only PPB experimental data. Line = predicted PK. Dots represent actual in vivo data. Error bars are SD of actual plasma concentrations.

The next PBPK model conducted was to see how using the experimentally

derived RBP value improved the model (Figure 32). This property improved the

“shoulder” in the PBPK curve seen in the in vivo PK profile. RBP plays a larger role than

anticipated in influencing rapid (phase I) and slower (phase II) distribution from the

central compartment. However, without clearance and PPB contributing to the third

phase elimination phase, the later time points are still over predicted.

82

Figure 32. PBPK model incorporating only RBP. Experimental Data. Line = predicted PK. Dots represent actual in vivo data. Error bars are SD of actual plasma concentrations. Finally, all three in vitro properties were incorporated into a single model (Figure

33). It is apparent that RBP, as shown previously, affects the rapid distribution phase, increasing the apparent clearance within the first few minutes of the simulation, representative of observed values in vivo. PPB and CLint more so affect the second phase distribution and elimination. While the PBPK model doesn’t fall directly on the experimental values, it very closely approximates the overall actual PK profile.

Figure 33. PBPK model incorporating CLint, RBP, and PPB. Experimental Data. Line = predicted PK. Dots represent actual in vivo data. Error bars are SD of actual plasma concentrations.

83

5.3.4. In Silico Prediction of Bioequivalent Dose of Carfentanil

It has been demonstrated that the predictive power of the carfentanil-rabbit PBPK model can be improved simply by adding a few key properties into the simulation, namely CLint, PPB, and RBP. This model transitioned from being under-predictive for

carfentanil PK via IV administration in the rabbit in its naïve state, to more representative

of in vivo PK, by populating the model with experimental data. The next and final step in

this model building exercise is to attempt to equate this to a human-equivalent dose using plasma concentration as the metric of bioequivalence (Figure 34).

Figure 34. Human-equivalent dosing of carfentanil extrapolated from in silico model of rabbit PK data.

The direct equivalent dose by mass to the human achieves roughly three times the

maximum plasma concentration. When the dose is lowered to ~1/3 of that, the predicted

plasma concentrations align very closely between human and rabbit predicted plots. The

prediction lies nearly on top of the rabbit PK prediction upon changing the units of

intrinsic clearance from that directly measured by HLM clearance (89.35 µL/min/mg

84

protein), to the extrapolated organ clearance (16.2 mL/min/kg body weight)103

demonstrating higher accuracy of the extrapolated value when using in a PBPK model.

5.4 Discussion

The use of in silico PBPK modeling is extremely valuable when attempting to

translate from purely animal model and/or in vitro data to a human model. In the case of

carfentanil, there is very limited data on human administration of carfentanil and its

subsequent PK properties. In all studies where humans were given carfentanil, it was for

the purpose of mapping MORs in the brain using PET scanning, or in one case,

measuring the dosimetry of that same carfentanil radioisotope.86-88, 127

When translating from surrogate animal model to human PBPK simulation in

GastroPlus, the most important factor was clearance rates. As observed in the HLM and

RLM studies, the difference in CLint was calculated to be 3.7x faster in RLM than in

HLM. The human equivalent dose, therefore, was roughly 1/3rd of that of the rabbit,

likely attributable to having roughly 1/3rd the metabolic clearance potential. The

remainder of that difference is likely due to the physicochemical properties’ slight

differences from the in silico predicted values. Together, the in silico model provides an

accurate PBPK of carfentanil exposure when used in conjunction with key

experimentally derived property values, in this case PPB and RBP.

5.5. Conclusions

What is perhaps the keystone of this research, in addition to the better fitment of

the PBPK model to in vivo data, and the human-equivalence dosing to achieve similar

PBPK, is the plausibility of the human predicted dose for toxicity of carfentanil (Table

11). The model employed in this study provides a human estimate for a dose that

85 achieved a “toxic” endpoint in the rabbit surrogate (0.34 µg/kg), i.e. collapse. Previous studies in humans have shown successful administration of 11C-carfentanil at doses of

0.03 µg/kg with only reports of drowsiness in some patients.127

Table 11. Therapeutic, toxic doses for carfentanil in monkeys (experimental) and humans (predicted) and their respective therapeutic indices

Effective Dose (µg/kg) Toxic Dose (µg/kg) Therapeutic Index Monkey58 0.1 1 10 Human 0.03127 0.34 (predicted) 11

The administered human dose reported is almost exactly one order of magnitude less than the predicted toxic dose. This human predicted toxic dose, therefore agrees with the therapeutic index reported for carfentanil in non-human primates.58

It has been demonstrated in this study that accurate predictions can be achieved by populating the in silico model with experimental data and by using multiple-species in vitro data, human extrapolation can achieve realistic predictive power. By utilizing in silico and in vitro technologies, as demonstrated here, in vivo experimentation can be refined, reduced, and potentially replaced while still working effectively and efficiently towards human relevant toxicity estimates.

86

6. Conclusions and Future Directions

The results of the current study demonstrate, for the first time, several different

aspects of carfentanil pharmacology and toxicology. Carfentanil has now been

experimentally observed to be 83 times more potent than its prototype fentanyl in a

human relevant cell-based screening assay; an assay that can be used to rapidly screen de

novo compounds for opioid activity in the future. Carfentanil has also been observed to

have rapid clearance in HLMs, but not in human hepatocytes. Twelve carfentanil

metabolites were identified, confirming the reported existence of one presumed

metabolite, norcarfentanil. Norcarfentanil was, in turn, observed to have decreased opioid

activity compared to its parent. Previously unknown physicochemical properties, PPB

and RBP, were measured for carfentanil for the first time, in both human and rabbit blood

and plasma samples, demonstrating species differences in these properties. These

metabolic and physicochemical properties were then incorporated into an in silico PBPK

model to refine its prediction of carfentanil disposition in vivo in rabbits administered 1

µg/kg of carfentanil, a dose high enough to elicit toxic responses like collapse and apnea.

This rabbit-predictive model was then translated to a human prediction of an equivalent toxic dose, a dose roughly 1/3rd of that required for the rabbit.

The in vitro results are impactful because they demonstrate the ability to compare

ADME properties of compounds for which very limited or no data exists, to benchmark compounds and make rapid assessments of toxicity or pharmacology. Additionally, the in silico studies demonstrate that animal-alternative studies can be conducted using purely in silico methods, but the incorporation of important in vitro data can bolster the predictive capabilities of these systems. Additionally, this study also demonstrated the

87 relevance of an appropriate animal model in making human health risk assessments. The rabbit, while not completely representative of the human response to carfentanil, served as a surrogate to predict a human toxic dose with the inclusion of the species specific in vitro data collected.

Future work using these preclinical models includes validating the metabolite identification study with animal urine and blood. This type of validation can be incorporated into point-of-care diagnostic tools used to identify opioid, and specifically, carfentanil toxicity. Additionally, the opioid receptor screening capability will be used to assess the opioid receptor profile of new compounds as they are identified by both public health and defense communities. Metabolic clearance in hepatocytes will be reassessed to attempt to elucidate the reason for its slow hepatocytic, yet rapid microsomal clearance.

Finally, the in silico model will be further refined by performing additional studies to acquire more experimental data with which to replace the purely predicted values. LogP and solubility properties are inexpensive and rapidly conducted studies that can add much value to the model. Additionally, as compounds of opioid derivation are identified, the current in silico model adds confidence to initial threat assessments and can guide both in vitro and in vivo research of these compounds in an attempt to provide a more accurate and representative depiction of opioid toxicity and pharmacology in humans.

88

References

1. Wax, P. M.; Becker, C. E.; Curry, S. C., Unexpected "gas" casualties in Moscow: A medical toxicology perspective. Annals of Emergency Medicine 2003, 41 (5), 700-705. 2. Tavernise, S.; Kishkovsky, S. Hostage Drama in Moscow: The Scene;The Survivors Dribble Out, All With a Story to Tell The New York Times [Online], 2002. http://www.nytimes.com/2002/10/28/world/hostage-drama-moscow-scene-survivors- dribble-all-with-story-tell.html. 3. Kutukova, N. Zamoskvoretskiy Court shall continue the examination of complaints of victims in "Nord-Ost" on the Moscow prosecutor's office Zaprava [Online], 2003. http://feedbackgroup.info/proetcon/sudnordost.htm (accessed March 9, 2017). 4. Riches, J. R.; Read, R. W.; Black, R. M.; Cooper, N. J.; Timperley, C. M., Analysis of Clothing and Urine from Moscow Theatre Siege Casualties Reveals Carfentanil and Remifentanil Use. Journal of Analytical Toxicology 2012, 36 (9), 647-656. 5. Michaelis, M.; Scholkens, B.; Rudolphi, K., An anthology from Naunyn-Schmiedeberg's archives of pharmacology. Naunyn-Schmiedebergs Archives of Pharmacology 2007, 375 (2), 81-84. 6. Black, J., A personal perspective on Dr. Paul Janssen. Journal of Medicinal Chemistry 2005, 48 (6), 1687-1688. 7. Stanley, T. H., The History and Development of the Fentanyl Series. Journal of Pain and Symptom Management 1992, 7 (3), S3-S7. 8. Bagley, J. R.; Kudzma, L. V.; Lalinde, N. L.; Colapret, J. A.; Huang, B. S.; Lin, B. S.; Jerussi, T. P.; Benvenga, M. J.; Doorley, B. M.; Ossipov, M. H.; Spaulding, T. C.; Spencer, H. K.; Rudo, F. G.; Wynn, R. L., Evolution of the 4-Anilidopiperidine Class of Opioid Analgesics. Medicinal Research Reviews 1991, 11 (4), 403-436. 9. Pert, C. B.; Snyder, S. H., Opiate Receptor - Demonstration in Nervous-Tissue. Science 1973, 179 (4077), 1011-1014. 10. Martin Wr Fau - Eades, C. G.; Eades Cg Fau - Thompson, J. A.; Thompson Ja Fau - Huppler, R. E.; Huppler Re Fau - Gilbert, P. E.; Gilbert, P. E., The effects of morphine- and - like drugs in the nondependent and morphine-dependent chronic spinal dog. (0022-3565 (Print)). 11. Lord, J. A. H.; Waterfield, A. A.; Hughes, J.; Kosterlitz, H. W., ENDOGENOUS OPIOID PEPTIDES - MULTIPLE AGONISTS AND RECEPTORS. Nature 1977, 267 (5611), 495-499. 12. Broom, D. C.; Jutkiewicz, E. M.; Rice, K. C.; Traynor, J. R.; Woods, J. H., Behavioral effects of delta-opioid receptor agonists: Potential antidepressants? Japanese Journal of Pharmacology 2002, 90 (1), 1-6. 13. Gaveriaux-Ruff, C.; Karchewski, L. A.; Hever, X.; Matifas, A.; Kieffer, B. L., Inflammatory pain is enhanced in delta opioid receptor-knockout mice. European Journal of Neuroscience 2008, 27 (10), 2558-2567. 14. Filliol, D.; Ghozland, S.; Chluba, J.; Martin, M.; Matthes, H. W. D.; Simonin, F.; Befort, K.; Gaveriaux-Ruff, C.; Dierich, A.; LeMeur, M.; Valverde, O.; Maldonado, R.; Kieffer, B. L., Mice deficient for delta- and gamma-opioid receptors exhibit opposing alterations of emotional responses. Nature Genetics 2000, 25 (2), 195-200. 15. Deo, S. H.; Barlow, M. A.; Gonzalez, L.; Yoshishige, D.; Caffrey, J. L., Cholinergic location of delta-opioid receptors in canine atria and SA node. American Journal of Physiology-Heart and Circulatory Physiology 2008, 294 (2), H829-H838.

89

16. Manglik, A.; Kruse, A. C.; Kobilka, T. S.; Thian, F. S.; Mathiesen, J. M.; Sunahara, R. K.; Pardo, L.; Weis, W. I.; Kobilka, B. K.; Granier, S., Crystal structure of the mu-opioid receptor bound to a morphinan antagonist. Nature 2012, 485 (7398), 321-U170. 17. Johnston, J. M.; Aburi, M.; Provasi, D.; Bortolato, A.; Urizar, E.; Lambert, N. A.; Javitch, J. A.; Filizola, M., Making Structural Sense of Dimerization Interfaces of Delta Opioid Receptor Homodimers. Biochemistry 2011, 50 (10), 1682-1690. 18. Zhang, Z.; Pan, Z. Z., Synaptic Mechanism for Functional Synergism between delta- and mu-Opioid Receptors. Journal of Neuroscience 2010, 30 (13), 4735-4745. 19. Land, B. B.; Bruchas, M. R.; Lemos, J. C.; Xu, M.; Melief, E. J.; Chavkin, C., The dysphoric component of stress is encoded by activation of the kappa-opioid system. Journal of Neuroscience 2008, 28 (2), 407-414. 20. Pan, Z. Z., mu-opposing actions of the kappa-opioid receptor. Trends in Pharmacological Sciences 1998, 19 (3), 94-98. 21. Mollereau, C.; Parmentier, M.; Mailleux, P.; Butour, J. L.; Moisand, C.; Chalon, P.; Caput, D.; Vassart, G.; Meunier, J. C., ORL1, A NOVEL MEMBER OF THE OPIOID RECEPTOR FAMILY - CLONING, FUNCTIONAL EXPRESSION AND LOCALIZATION. Febs Letters 1994, 341 (1), 33-38. 22. Zaveri, N., Peptide and nonpeptide ligands for the nociceptin/orphanin FQ receptor ORL1: Research tools and potential therapeutic agents. Life Sciences 2003, 73 (6), 663- 678. 23. Mollereau, C.; Mouledous, L., Tissue distribution of the opioid receptor-like (ORL1) receptor. Peptides 2000, 21 (7), 907-917. 24. Herz, A.; Albus, K.; Metys, J.; Schubert, P.; Teschemacher, H., CENTRAL SITES FOR ANTINOCICEPTIVE ACTION OF MORPHINE AND FENTANYL. Neuropharmacology 1970, 9 (6), 539-+. 25. Osborne, P. B.; Vaughan, C. W.; Wilson, H. I.; Christie, M. J., Opioid inhibition of rat periaqueductal grey neurones with identified projections to rostral ventromedial medulla in vitro. Journal of Physiology-London 1996, 490 (2), 383-389. 26. Satoh, M.; Takagi, H., ENHANCEMENT BY MORPHINE OF CENTRAL DESCENDING INHIBITORY INFLUENCE ON SPINAL SENSORY TRANSMISSION. European Journal of Pharmacology 1971, 14 (1), 60-&. 27. Astonjones, G.; Chiang, C.; Alexinsky, T., DISCHARGE OF NORADRENERGIC LOCUS-CERULEUS NEURONS IN BEHAVING RATS AND MONKEYS SUGGESTS A ROLE IN VIGILANCE. Neurobiology of the Locus Coeruleus 1991, 88, 501-520. 28. Kilduff, T. S.; Peyron, C., The hypocretin/orexin ligand-receptor system: implications for sleep and sleep disorders. Trends in Neurosciences 2000, 23 (8), 359-365. 29. Georgescu, D.; Zachariou, V.; Barrot, M.; Mieda, M.; Willie, J. T.; Eisch, A. J.; Yanagisawa, M.; Nestler, E. J.; DiLeone, R. J., Involvement of the lateral hypothalamic peptide orexin in morphine dependence and withdrawal. Journal of Neuroscience 2003, 23 (8), 3106-3111. 30. Huang, Z. L.; Qu, W. M.; Li, W. D.; Mochizuki, T.; Eguchi, N.; Watanabe, T.; Urade, Y.; Hayaishi, O., Arousal effect of orexin A depends on activation of the histaminergic system (vol 98, pg 9965, 2001). Proceedings of the National Academy of Sciences of the United States of America 2002, 99 (2), 1098-1098. 31. Ling, G. S. F.; Spiegel, K.; Nishimura, S. L.; Pasternak, G. W., Dissociation of Morphine's Analgesic and Respiratory Depressant Actions. European Journal of Pharmacology 1983, 86 (3-4), 487-488. 32. Shook, J. E.; Watkins, W. D.; Camporesi, E. M., Differential Roles of Opioid Receptors in Respiration, Respiratory-Disease, and Opiate-Induced Respiratory Depression. American Review of Respiratory Disease 1990, 142 (4), 895-909.

90

33. Dahan, A.; Aarts, L.; Smith, T. W., Incidence, Reversal, and Prevention of Opioid- induced Respiratory Depression. Anesthesiology 2010, 112 (1), 226-238. 34. Montandon, G.; Qin, W. X.; Liu, H.; Ren, J.; Greer, J. J.; Horner, R. L., PreBotzinger Complex Neurokinin-1 Receptor-Expressing Neurons Mediate Opioid-Induced Respiratory Depression. Journal of Neuroscience 2011, 31 (4), 1292-1301. 35. Greer, J. J.; Ren, J., Ampakine therapy to counter fentanyl-induced respiratory depression. Respiratory Physiology & Neurobiology 2009, 168 (1-2), 153-157. 36. Smith, J. C.; Ellenberger Hh Fau - Ballanyi, K.; Ballanyi K Fau - Richter, D. W.; Richter Dw Fau - Feldman, J. L.; Feldman, J. L., Pre-Botzinger complex: a brainstem region that may generate respiratory rhythm in mammals. (0036-8075 (Print)). 37. Haw, A. J.; Meyer, L. C. R.; Greer, J. J.; Fuller, A., Ampakine CX1942 attenuates opioid- induced respiratory depression and corrects the hypoxaemic effects of etorphine in immobilized goats (Capra hircus). Veterinary Anaesthesia and Analgesia 2016, 43 (5), 528-538. 38. Ren, J.; Lenal, F.; Yang, M.; Ding, X. Q.; Greer, J. J., Coadministration of the AMPAKINE CX717 with Propofol Reduces Respiratory Depression and Fatal Apneas. Anesthesiology 2013, 118 (6), 1437-1445. 39. Katchman, A. N.; McGroary, K. A.; Kilborn, M. J.; Kornick, C. A.; Manfredi, P. L.; Woosley, R. L.; Ebert, S. N., Influence of opioid agonists on cardiac human ether-a-go- go related gene K+ currents. Journal of Pharmacology and Experimental Therapeutics 2002, 303 (2), 688-694. 40. Stringer, J.; Welsh, C.; Tommasello, A., Methadone-associated Q-T interval prolongation and torsades de pointes. American Journal of Health-System Pharmacy 2009, 66 (9), 825-833. 41. Ramakrishna, H.; O'Hare, M.; Mookadam, F.; Gutsche, J. T.; Shah, R.; Augoustides, J. G. T., Sudden Cardiac Death and Disorders of the QT Interval: Anesthetic Implications and Focus on Perioperative Management. Journal of Cardiothoracic and Vascular Anesthesia 2015, 29 (6), 1723-1733. 42. Zuenkler, B. J.; Wos-Maganga, M., Comparison of the Effects of Methadone and Heroin on Human ether-A -go-go-Related Gene Channels. Cardiovascular Toxicology 2010, 10 (3), 161-165. 43. Haeseler, G.; Foadi N Fau - Ahrens, J.; Ahrens J Fau - Dengler, R.; Dengler R Fau - Hecker, H.; Hecker H Fau - Leuwer, M.; Leuwer, M., , fentanyl and sufentanil but not morphine block voltage-operated sodium channels. (1872-6623 (Electronic)). 44. Borison, H. L., AREA POSTREMA - CHEMORECEPTOR CIRCUMVENTRICULAR ORGAN OF THE MEDULLA-OBLONGATA. Progress in Neurobiology 1989, 32 (5), 351-390. 45. Wynn, R. L.; Essien, E.; Thut, P. D., THE EFFECTS OF DIFFERENT ANTIEMETIC AGENTS ON MORPHINE-INDUCED EMESIS IN FERRETS. European Journal of Pharmacology 1993, 241 (1), 47-54. 46. Cole, A.; Mutlow, A.; Isaza, R.; Carpenter, J. W.; Koch, D. E.; Hunter, R. P.; Dresser, B. L., Pharmacokinetics and pharmacodynamics of carfentanil and in female common eland (Taurotragus oryx). Journal of Zoo and Wildlife Medicine 2006, 37 (3), 318-326. 47. Vanbever, W. F. M.; Niemegeers, C. J. E.; Schellekens, K. H. L.; Janssen, P. A. J., N-4- Substituted 1-(2-arylethyl)-4-piperidinyl-N-phenylpropanamides, a Novel Series of Extremely Potent Analgesics with Unusually High Safety Margin. Arzneimittel- Forschung/Drug Research 1976, 26 (8), 1548-1551. 48. Subramanian, G.; Paterlini, M. G.; Portoghese, P. S.; Ferguson, D. M., Molecular docking reveals a novel binding site model for fentanyl at the mu-opioid receptor. Journal of Medicinal Chemistry 2000, 43 (3), 381-391.

91

49. Kataoka, K.; Hara, K.; Haranishi, Y.; Terada, T.; Sata, T., The Antinociceptive Effect of SNAP5114, a Gamma-Aminobutyric Acid Transporter-3 Inhibitor, in Rat Experimental Pain Models. Anesthesia and Analgesia 2013, 116 (5), 1162-1169. 50. Mogil, J.; Wilson, S.; Wan, Y., Assessing Nociception in Murine Subjects. In Methods in Pain Research, CRC Press: 2001. 51. Carter, M.; Shieh, J. C., Chapter 2 - Animal Behavior. In Guide to Research Techniques in Neuroscience, Academic Press: New York, 2010; pp 39-71. 52. Brasel, C. M.; Sawyer, G. W.; Stevens, C. W., A pharmacological comparison of the cloned frog and human mu opioid receptors reveals differences in opioid affinity and function. European Journal of Pharmacology 2008, 599 (1-3), 36-43. 53. Chen, J. C.; Smith, E. R.; Cahill, M.; Cohen, R.; Fishman, J. B., The Opioid Receptor- Binding of , Morphine, Fentanyl, , and . Life Sciences 1993, 52 (4), 389-396. 54. Traynor, J. R.; Nahorski, S. R., Modulation by Mu-Opioid Agonists of Guanosine-5'-O- (3-S-35 Thio)Triphosphate Binding to Membranes from Human Neuroblastoma SH- SY5Y Cells. Molecular Pharmacology 1995, 47 (4), 848-854. 55. Grinnell, S. G.; Majumdar, S.; Narayan, A.; Le Rouzic, V.; Ansonoff, M.; Pintar, J. E.; Pasternak, G. W., Pharmacologic Characterization in the Rat of a Potent Analgesic Lacking Respiratory Depression, IBNtxA. Journal of Pharmacology and Experimental Therapeutics 2014, 350 (3), 710-718. 56. Taylor, D. A., In vitro opioid receptor assays. (1934-8290 (Electronic)). 57. Volpe, D. A.; Tobin, G. A. M.; Mellon, R. D.; Katki, A. G.; Parker, R. J.; Colatslcy, T.; Kropp, T. J.; Verbois, S. L., Uniform assessment and ranking of opioid Mu receptor binding constants for selected opioid drugs. Regulatory Toxicology and Pharmacology 2011, 59 (3), 385-390. 58. Port, J.; Stanley, T.; Steffey, E.; Pace, N.; Henrickson, R.; McJames, S., Intravenous Carfentanil in the Dog and Rhesus Monkey. Anesthesiology 1984, 61 (3), A378-A378. 59. Feldman, P. L.; James, M. K.; Brackeen, M. F.; Bilotta, J. M.; Schuster, S. V.; Lahey, A. P.; Lutz, M. W.; Johnson, M. R.; Leighton, H. J., Design, Synthesis, and Pharmacological Evaluation of Ultrashort-Acting to Long-Acting Opioid Analgesics. Journal of Medicinal Chemistry 1991, 34 (7), 2202-2208. 60. Pan, J.-B.; Ji, N.; Pan, W.; Hong, R.; Wang, H.; Ji, Z.-L., High-throughput identification of off-targets for the mechanistic study of severe adverse drug reactions induced by analgesics. Toxicology and Applied Pharmacology 2014, 274 (1), 24-34. 61. DeMio, T. Drug overdoses killed 192 in Butler County in 2016 Cincinnati [Online], 2017. http://www.cincinnati.com/story/news/2017/02/24/drug-overdoses-killed-192- butler-county-2016/98349728/ (accessed March 8, 2017). 62. Kinetz, E.; Butler, D. Chinese companies are selling a chemical weapon turned narcotic that's killing unsuspecting users Business Insider [Online], 2016. http://www.businessinsider.com/chinese-companies-are-selling-a-chemical-weapon- turned-narcotic-thats-killing-unsuspecting-users-2016-10 (accessed March 4, 2017). 63. Kounang, N.; Marco, T. Heroin laced with elephant tranquilizer hits the streets CNN [Online], 2016. http://www.cnn.com/2016/08/24/health/elephant-tranquilizer-carfentanil- heroin/ (accessed March 4, 2017). 64. Influx of Fentanyl-laced Counterfeit Pills and Toxic Fentanyl-related Compounds Further Increases Risk of Fentanyl-related Overdose and Fatalities. CDC: CDC Health Alert Network, 2016; Vol. 00395. 65. . Deadly Carfentanil Has Been Detected in Cuyahoga County Cuyahoga County Office of the Executive [Online], 2016. http://executive.cuyahogacounty.us/en-US/ME-Public- Health-Warning.aspx.

92

66. Helander, A.; Backberg, M.; Beck, O., Intoxications involving the fentanyl analogs acetylfentanyl, 4-methoxybutyrfentanyl and : results from the Swedish STRIDA project. Clinical Toxicology 2016, 54 (4), 324-332. 67. Rogers, J. S.; Rehrer, S. J.; Hoot, N. R., ACETYLFENTANYL: AN EMERGING DRUG OF ABUSE. Journal of Emergency Medicine 2016, 50 (3), 433-436. 68. Kroll, D. W-18, The High-Potency Research Chemical Making News: What It Is And What It Isn't Forbes [Online], 2016. https://www.forbes.com/sites/davidkroll/2016/04/30/w-18-the-high-potency-research- chemical-making-news-what-it-is-and-what-it-isnt/#1492543f4757. 69. Katselou, M.; Papoutsis, I.; Nikolaou, P.; Spiliopoulou, C.; Athanaselis, S., Old opioids, new concerns: the case of acetylfentanyl. Forensic Toxicology 2016, 34 (2), 201-212. 70. Chappell, B. China Blacklists Carfentanil, Closing A Deadly Loophole NPR.com [Online], 2017. http://www.npr.org/sections/thetwo-way/2017/02/16/515544874/china- blacklists-carfentanil-closing-a-deadly-loophole. 71. National Overdose Deaths - 2015. National Institute on Drug Abuse, 2015. 72. Freye, E., Opioids in Medicine: A Comprehensive Review on the Mode of Action and the Use of Analgesics in Different Clinical Pain States. Springer: Netherlands, 2008. 73. Feierman, D. E.; Lasker, J. M., Metabolism of fentanyl, a synthetic opioid analgesic, by human liver microsomes - Role of CYP3A4. Drug Metabolism and Disposition 1996, 24 (9), 932-939. 74. Guitton, J.; Buronfosse, T.; Desage, M.; Lepape, A.; Brazier, J. L.; Beaune, P., Possible involvement of multiple cytochrome P450S in fentanyl and sufentanil metabolism as opposed to alfentanil. Biochemical Pharmacology 1997, 53 (11), 1613-1619. 75. Peacock, J. E.; Francis, G., Remifentanil pharmacokinetics and metabolism. Drugs of Today 1997, 33 (9), 611-618. 76. Mather, L. E., Clinical Pharmacokinetics of Fentanyl and Its Newer Derivatives. Clinical Pharmacokinetics 1983, 8 (5), 422-446. 77. Cometta-Morini, C.; Maguire, P. A.; Loew, G. H., Molecular Determinants of Mu- Receptor Recognition for the Fentanyl Class of Compounds. Molecular Pharmacology 1992, 41 (1), 185-196. 78. Gladden, R. M., Fentanyl Law Enforcement Submissions and Increases in Synthetic Opioid-Involved Overdose Deaths - 27 States, 2013-2014. Prevention, C. f. D. C. a., Ed. Centers for Disease Control and Prevention, 2016; Vol. 65, pp 844-849. 79. Kounang, N. Elephant tranquilizer to blame for at least 8 Ohio deaths CNN [Online], 2016. http://www.cnn.com/2016/09/06/health/carfentanil-deaths-ohio/. 80. Knaus, E. E.; Warren, B. K.; Ondrus, T. A., Analgesic substituted piperidylidene-2- sulfon(cyan)amide derivatives. Google Patents: 1984. 81. Al-Hasani, R.; Bruchas, M. R., Molecular Mechanisms of Opioid Receptor-dependent Signaling and Behavior. Anesthesiology 2011, 115 (6), 1363-1381. 82. Minneman, K. P.; Iversen, L. L., Enkephalin and Opiate Narcotics Increase Cyclic-GMP Accumulation in Slices of Rat Neostriatum. Nature 1976, 262 (5566), 313-314. 83. James, M. K.; Feldman, P. L.; Schuster, S. V.; Bilotta, J. M.; Brackeen, M. F.; Leighton, H. J., OPIOID RECEPTOR ACTIVITY OF GI-87084B, A NOVEL ULTRA-SHORT ACTING ANALGESIC, IN ISOLATED-TISSUES. Journal of Pharmacology and Experimental Therapeutics 1991, 259 (2), 712-718. 84. Wang, L. Q.; Bernert, J. T., Analysis of 13 fentanils, including sufentanil and carfentanil, in human urine by liquid chromatography-atmospheric-pressure ionization-tandem mass spectrometry. Journal of Analytical Toxicology 2006, 30 (5), 335-341. 85. Feasel, M. G.; Wohlfarth, A.; Nilles, J. M.; Pang, S. K.; Kristovich, R. L.; Huestis, M. A., Metabolism of Carfentanil, an Ultra-Potent Opioid, in Human Liver Microsomes and

93

Human Hepatocytes by High-Resolution Mass Spectrometry. Aaps Journal 2016, 18 (6), 1489-1499. 86. Newman, A.; Channing, M.; Finn, R.; Dunn, B.; Simpson, N.; Carson, R.; Ostrowski, N.; Cohen, R.; Burke, T.; Larson, S.; et al., Ligands for imaging opioid receptors in conscious humans by positron emission tomography (PET). NIDA Res Monogr 1988, 90, 117-21. 87. Endres, C. J.; Bencherif, B.; Hilton, J.; Madar, I.; Frost, J. J., Quantification of brain mu- opioid receptors with C-11 carfentanil: reference-tissue methods. Nuclear Medicine and Biology 2003, 30 (2), 177-186. 88. Villemagne, P. S. R.; Dannals, R. F.; Ravert, H. T.; Frost, J. J., PET imaging of human cardiac opioid receptors. European Journal of Nuclear Medicine and Molecular Imaging 2002, 29 (10), 1385-1388. 89. Krechetnikov, A. Moscow theatre siege: Questions remain unanswered BBC News [Online], 2012. http://www.bbc.com/news/world-europe-20067384 (accessed May 13, 2015). 90. Glasser, S.; Baker, P., Russia Confirms Suspicions About Gas Used in Raid; Potent Anesthetic Pumped Into Theater; 2 More Hostages Die From Drug's Effects. The Washington Post 2002, p A.15. 91. Hanks, G. W.; Aherne, G. W.; Hoskin, P. J.; Turner, P.; Poulain, P., EXPLANATION FOR POTENCY OF REPEATED ORAL DOSES OF MORPHINE. Lancet 1987, 2 (8561), 723-725. 92. Schneider, E.; Brune, K., OPIOID ACTIVITY AND DISTRIBUTION OF FENTANYL METABOLITES. Naunyn-Schmiedebergs Archives of Pharmacology 1986, 334 (3), 267- 274. 93. Ultiva, Remifentanil Hydrochloride. In Package Insert, Mylan Institutional, L., Ed. 2015. 94. Burkle, H.; Dunbar, S.; VanAken, H., Remifentanil: A novel, short-acting, mu-opioid. Anesthesia and Analgesia 1996, 83 (3), 646-651. 95. Van Nimmen, N. F. J.; Poels, K. L. C.; Veulemans, H. A. F., Highly sensitive gas chromatographic-mass spectrometric screening method for the determination of picogram levels of fentanyl, sufentanil and alfentanil and their major metabolites in urine of opioid exposed workers. Journal of Chromatography B-Analytical Technologies in the Biomedical and Life Sciences 2004, 804 (2), 375-387. 96. Wohlfarth, A.; Castaneto, M. S.; Zhu, M. S.; Pang, S. K.; Scheidweiler, K. B.; Kronstrand, R.; Huestis, M. A., Pentylindole/Pentylindazole Synthetic Cannabinoids and Their 5-Fluoro Analogs Produce Different Primary Metabolites: Metabolite Profiling for AB-PINACA and 5F-AB-PINACA. Aaps Journal 2015, 17 (3), 660-677. 97. Lave, T.; Dupin, S.; Schmitt, C.; Valles, B.; Ubeaud, G.; Chou, R. C.; Jaeck, D.; Coassolo, P., The use of human hepatocytes to select compounds based on their expected hepatic extraction ratios in humans. Pharmaceutical Research 1997, 14 (2), 152-155. 98. Meuldermans We Fau - Hurkmans, R. M.; Hurkmans Rm Fau - Heykants, J. J.; Heykants, J. J., Plasma protein binding and distribution of fentanyl, sufentanil, alfentanil and lofentanil in blood. (0003-9780 (Print)). 99. Allen, J. L., RENARCOTIZATION FOLLOWING CARFENTANIL IMMOBILIZATION OF NONDOMESTIC UNGULATES. Journal of Zoo and Wildlife Medicine 1989, 20 (4), 423-426. 100. Thevis, M.; Geyer, H.; Bahr, D.; Schanzer, W., Identification of fentanyl, alfentanil, sufentanil, remifentanil and their major metabolites in human urine by liquid chromatography/tandem mass spectrometry for doping control purposes. European Journal of Mass Spectrometry 2005, 11 (4), 419-427. 101. Cashman, J. R.; Park, S. B.; Yang, Z. C.; Wrighton, S. A.; Jacob, P.; Benowitz, N. L., METABOLISM OF NICOTINE BY HUMAN LIVER-MICROSOMES -

94

STEREOSELECTIVE FORMATION OF TRANS-NICOTINE N'-OXIDE. Chemical Research in Toxicology 1992, 5 (5), 639-646. 102. Pirmohamed, M.; Williams, D.; Madden, S.; Templeton, E.; Park, B. K., METABOLISM AND BIOACTIVATION OF CLOZAPINE BY HUMAN LIVER IN-VITRO. Journal of Pharmacology and Experimental Therapeutics 1995, 272 (3), 984-990. 103. McNaney, C. A.; Drexler, D. M.; Hnatyshyn, S. Y.; Zvyaga, T. A.; Knipe, J. O.; Belcastro, J. V.; Sanders, M., An automated liquid chromatography-mass spectrometry process to determine metabolic stability half-life and intrinsic clearance of drug candidates by substrate depletion. Assay and Drug Development Technologies 2008, 6 (1), 121-129. 104. Tetko, I. V.; Gasteiger, J.; Todeschini, R.; Mauri, A.; Livingstone, D.; Ertl, P.; Palyulin, V.; Radchenko, E.; Zefirov, N. S.; Makarenko, A. S.; Tanchuk, V. Y.; Prokopenko, V. V., Virtual computational chemistry laboratory - design and description. Journal of Computer-Aided Molecular Design 2005, 19 (6), 453-463. 105. Virtual Computational Chemistry Laboratory. 2005 ed.; www.vcclab.org, 2005. 106. Hansch, C.; Leo, A., Exploring QSAR: Hydrophobic, Electronic, and Steric Constants. ACS Publications 1995, 2. 107. ADMET Predictor, 7.2.0001; Simulations Plus: 2015. 108. Hug, C. C.; Murphy, M. R., TISSUE REDISTRIBUTION OF FENTANYL AND TERMINATION OF ITS EFFECTS IN RATS. Anesthesiology 1981, 55 (4), 369-375. 109. Dart, R., Medical Toxicology. 3rd ed.; Lippincott Williams and Wilkins: Philadelphia, PA, 2004. 110. Henderson, G. L., DESIGNER DRUGS - PAST HISTORY AND FUTURE- PROSPECTS. Journal of Forensic Sciences 1988, 33 (2), 569-575. 111. Mahlke, N. S.; Ziesenitz, V.; Mikus, G.; Skopp, G., Quantitative low-volume assay for simultaneous determination of fentanyl, norfentanyl, and minor metabolites in human plasma and urine by liquid chromatography-tandem mass spectrometry (LC-MS/MS). International Journal of Legal Medicine 2014, 128 (5), 771-778. 112. Deshmukh, P. V.; Badgujar, P. C.; Gatne, M. M., In-vitro red blood cell partitioning of doxycycline. Indian Journal of Pharmacology 2009, 41 (4), 173-175. 113. Espie, P.; Tytgat, D.; Sargentini-Maier, M. L.; Poggesi, I.; Watelet, J. B., Physiologically based pharmacokinetics (PBPK). Drug Metab Rev 2009, 41 (3), 391-407. 114. Hinderling, P. H., Red blood cells: A neglected compartment in pharmacokinetics and pharmacodynamics. Pharmacological Reviews 1997, 49 (3), 279-295. 115. Bista, S. R.; Haywood, A.; Hardy, J.; Lobb, M.; Tapuni, A.; Norris, R., Protein binding of fentanyl and its metabolite nor-fentanyl in human plasma, albumin and alpha-1 acid glycoprotein. Xenobiotica 2015, 45 (3), 207-212. 116. Ishizaki, J.; Yokogawa, K.; Nakashima, E.; Ichimura, F., Relationships between the hepatic intrinsic clearance or blood cell-plasma partition coefficient in the rabbit and the lipophilicity of basic drugs. Journal of Pharmacy and Pharmacology 1997, 49 (8), 768- 772. 117. Yu, S.; Li, S.; Yang, H.; Lee, F.; Wu, J. T.; Qian, M. G., A novel liquid chromatography/tandem mass spectrometry based depletion method for measuring red blood cell partitioning of pharmaceutical compounds in drug discovery. Rapid Commun. Mass Spectrom. 2005, 19 (2), 250-4. 118. Bista, S. R.; Haywood A Fau - Hardy, J.; Hardy J Fau - Lobb, M.; Lobb M Fau - Tapuni, A.; Tapuni A Fau - Norris, R.; Norris, R., Protein binding of fentanyl and its metabolite nor-fentanyl in human plasma, albumin and alpha-1 acid glycoprotein. (1366-5928 (Electronic)).

95

119. Flecknell, P. A.; Liles, J. H.; Wootton, R., Reversal of fentanyl/fluanisone neuroleptanalgesia in the rabbit using mixed agonist/antagonist opioids. Lab Anim 1989, 23 (2), 147-55. 120. Hyatt, J.; Coro, C.; Bergman, S. A.; Wynn, R. L., Effects of midazolam on fentanyl antinociception and respiration in a rabbit model. J Oral Maxillofac Surg 1989, 47 (12), 1298-302. 121. Stephen, G. W.; Cooper, L. V., The role of analgesics in respiratory depression: a rabbit model. Anaesthesia 1977, 32 (4), 324-7. 122. Campbell, J. L., Jr.; Clewell, R. A.; Gentry, P. R.; Andersen, M. E.; Clewell, H. J., 3rd, Physiologically based pharmacokinetic/toxicokinetic modeling. Methods Mol Biol 2012, 929, 439-99. 123. Suckow, M. A.; Schroeder, V., The Laboratory Rabbit. 2 ed.; 2010. 124. Jelenko, C., 3rd; Anderson, A. P.; Scott, T. H., Jr.; Wheeler, M. L., Organ weights and water composition of the New Zealand albino rabbit (Oryctalagus cuniculus). Am J Vet Res 1971, 32 (10), 1637-9. 125. Sohlenius-Sternbeck, A. K., Determination of the hepatocellularity number for human, dog, rabbit, rat and mouse livers from protein concentration measurements. Toxicology in Vitro 2006, 20 (8), 1582-1586. 126. Halliburton, J. R., The pharmacokinetics of fentanyl, sufentanil and alfentanil: a comparative review. Aana j 1988, 56 (3), 229-33. 127. Newberg, A. B.; Ray, R.; Scheuermann, J.; Wintering, N.; Saffer, J.; Schmitz, A.; Freifelder, R.; Karp, J.; Lerman, C.; Divgi, C., Dosimetry of C-11-carfentanil, a mu- opioid receptor imaging agent. Nuclear Medicine Communications 2009, 30 (4), 314-318.

96