Bifunctional Peptidomimetics: Studies on Affinity and Efficacy

by

Nicholas W. Griggs

A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Pharmacology) in the University of Michigan 2019

Doctoral Committee:

Professor John R. Traynor, Chair Assistant Professor Emily M. Jutkiewicz Professor Henry I. Mosberg Associate Professor Manojkumar A. Puthenveedu

Nicholas W. Griggs

[email protected]

ORCID iD: 0000-0003-0714-4712

© Nicholas W. Griggs 2019

Dedication

I dedicate this dissertation to my wife, Angela. I am forever thankful for all that she does for me. This dissertation would not have been possible without her.

I also dedicate this dissertation to my family. My parents, Lynn and Mark, and my sister Ellen. I am forever grateful for all that they have done for me.

Finally, I dedicate this dissertation to everyone in the world struggling with either chronic or opioid addiction.

ii Acknowledgements

I would like to express my sincere gratitude to my academic advisor, Dr. John Traynor. He has been a generous mentor who has always taken an interest in my development as a scientist and as a person. I am very appreciative of his patience, wisdom, and guidance, and I very much enjoyed having him as my mentor during my graduate school training.

I would also like express my gratitude to the members of my thesis committee, Drs. Emily Jutkiewicz, Henry Mosberg, and Manoj Puthenveedu, for their support and valuable advice.

I would like to thank all of the members of the Traynor lab, both past and present, and everyone in the Department of Pharmacology. I have made many friends during my years as a wolverine at the University of Michigan (2007-2019), but I will remember the years in graduate school fondly. I considered the Traynor lab my home away from home, and the people in Pharmacology as my family away from family.

Finally, I am grateful for the sources of funding that have supported my graduate studies: the Program in Biomedical Sciences (PIBS, University of Michigan), the Interdepartmental Training in Pharmacological Sciences (NIGMS T32 GM007767), the NIDA Training Program in Neuroscience (NIDA T32 DA007281), the Endowment for the Basic Sciences (University of Michigan), and several NIH grants held by Drs. John Traynor and Henry Mosberg. .

iii

Table of Contents

Dedication ii

Acknowledgements iii

List of Tables viii

List of Figures ix

List of Abbreviations xi

Abstract xii

Chapter I: General Introduction 1

Opioids and Opioid Receptors 1

Opioid Use in Pain 5

The Opioid Crisis 7

Targeting Multiple Opioid Receptors 9

Bifunctional Opioid Compounds 12

Bifunctional Opioid Peptides 14

Bifunctional Opioid Peptidomimetics 16

Aims of the Dissertation 20

Chapter II: Characterization of Bifunctional Opioid Peptidomimetics 21

iv

Introduction 21

Results 24

μOR SAR: C6 position Substitutions (R1) 24

δOR SAR: C6 position Substitutions (R1) 27

δOR SAR: N1 Substitutions (R2) 28

μOR SAR: N1 Substitutions (R2) 29

C8 position Substitutions (R3) 33

In Vitro δOR Antagonism 35

Discussion 37

In Vitro SAR Exploration 37

Relationship Between In Vitro SAR and In Vivo Behavioral Effects 39

Methods 43

Materials and Reagents 43

Cell Lines and Membrane Preparation 43

Radioligand Binding Assays 44

Stimulation of GTPγ35S Binding Assays 45

In Vivo Antinociception Assays 46

Data Analysis 46

Appendix to Chapter II 48

Introduction 48

v Methods 49

Results 50

Chapter III: Agonism of Bifunctional Opioid Peptidomimetics 53

Introduction 53

Results 61

Stimulation of GTPγ35S Binding in C6 μOR Cell Membranes 61

Affinity Shifts at the μOR in C6 μOR Cell Membranes 64

Agonism Correlation Plots 69

Kinetics of Binding at μOR in C6 μOR Cell Membranes 71

Stimulation of GTPγ35S Binding and Affinity Shifts in C6 δOR Cell Membranes 73

Stimulation of GTPγ35S Binding in CHO δOR Cell Membranes 76

Stimulation of GTPγ35S Binding in SH-SY5Y Cell Membranes 78

Discussion 87

Methods 92

Materials and Reagents 92

Cell Lines and Membrane Preparation 92

Radioligand Binding Assays 94

Stimulation of GTPγ35S Binding Assays 95

PathHunter β-Arrestin Assays 95

Data Analysis 96

vi Chapter IV: Conclusions and Future Directions 97

Bibliography 103

vii List of Tables

Table 1. Effects of Benzyl R1 Substitutions and Carbonyl Groups at R2. 26

Table 2. Effects of the 2-naphthyl R1 Substitution and Carbonyl Groups at R2. 30

Table 3. R1 2-methyl-THIQ Substitutions Increase μOR Agonism and R2 Substitutions Improve μOR /

δOR Affinity Balance. 31

Table 4. R2 Mesyl Substitutions Increase μOR Agonism and Improve μOR / δOR Affinity Balance. 32

Table 5. Bulky C8 Carbonyl Substitutions Maintain δOR Antagonist Profile. 34

Table 6. In Vitro δOR Antagonism of Lead Peptidomimetics. 36

Table 7. Relative Efficacy, , and an Index of Agonism of at μOR. 63

+ Table 8. Binding Affinity (Ki) of Opioid Ligands and the Affinity Shift by Na ions and Guanine

Nucleotide at μOR. 66

Table 9. Binding Affinity of Opioid Ligands and the Affinity Shift by Na+ ions and Guanine Nucleotide at

δOR. 75

Table 10. Stimulation of GTPγ35S Binding by Opioid in SH-SY5Y Membranes. 80

Table 11. Bias Factors for Opioid Compounds at the μOR. 85

Table 12. Bias Factors for Opioid Compounds at the δOR. 86

viii List of Figures

Figure 1. Chemical Structures of μOR / δOR Antagonist Bifunctional Opioid Compounds. 13

Figure 2. Chemical Structures of Analogues and the Design of a Bifunctional Opioid

Peptidomimetic. 17

Figure 3. In Vivo Antinociceptive Activity of the Bifunctional Opioid Peptidomimetic, compound 1. 19

Figure 4. Positions for Substitution of the THQ core of compound 1. 23

Figure 5. In Vitro SAR Summary. 35

Figure 6. In Vivo Behavioral Effects of Bifunctional Opioid Peptidomimetics. 41

Figure A1. Inhibition of δOR Does Not Prevent DAMGO-mediated µOR Tolerance in SH-SY5Y cells. 52

Figure 7. Chemical Structures of G protein-biased Agonists. 57

Figure 8.Chemical Structures and Physiochemical Properties of a Small Series of Bifunctional Opioid

Peptidomimetics. 59

Figure 9. Chemical Structures and Physiochemical Properties of Several Classical and Highly Efficacious

Opioid Agonists. 60

Figure 10. Relative Efficacy of Opioid Agonists and Peptidomimetics at μOR. 62

Figure 11. Inhibition of [3H]- Binding by Opioid Ligands at μOR. 67

Figure 12. Inhibition of [3H]-diprenorphine Binding by Opioid Ligands at μOR (cont’d). 68

Figure 13. Correlation of Agonism (∆Log (Max/EC50)) with Affinity Shift by Na+ and Guanine

Nucleotide at μOR. 70

Figure 14. Correlation of Ehlert’s Efficacy with Affinity Shift by Na+ ions and Guanine Nucleotide at

μOR. 70

Figure 15. Kinetics of Competitive Binding at the μOR. 72 ix Figure 16. AFN42 is a δOR Antagonist and is Insensitive to the Na+ Effect on Binding Affinity. 74

Figure 17. Bifunctional Peptidomimetics are Partial Agonists in CHO δOR Membranes. 77

Figure 18. Stimulation of GTPγ35S Binding by Opioid Agonists in SH-SY5Y Membranes. 79

Figure 19. β-Arrestin Recruitment of Opioid Ligands Following Pertussis Toxin Treatment at the μOR. 83

Figure 20. β-Arrestin Recruitment of Opioid Ligands at the μOR and δOR and Biased Agonism Plots. 84

x List of Abbreviations

DAMGO [D-Ala2, N-MePhe4, Gly-ol]- DMT 2’,6’-Dimethyl-L-tyrosine DPN Diprenorphine δOR Delta DPDPE [D-Pen2, D-Pen5]-enkephalin GDP Guanosine 5’-diphosphate GPCR G protein-coupled receptor GTPγ35S Guanosine‐5′‐O‐(3‐[35S]thio)triphosphate κOR Kappa opioid receptor NOP /Orphanin FQ (N/OFQ) Peptide NTI NTB PTX Pertussis Toxin SAR Structure-Activity Relationship THQ Tetrahydroquinoline THIQ Tetrahydroisoquinoline µOR Mu opioid receptor

xi Abstract

Opioids are effectively used for the treatment of acute and chronic pain, however, serious

problems such as fatal overdose due to respiratory and opioid addiction are associated

with the use of these medications. A strong body of evidence suggests that the simultaneous

targeting of mu opioid receptors and other opioid receptor types may be beneficial in improving

mu opioid receptor-mediated side effects. This dissertation reports in vitro experiments designed

to pharmacologically analyze opioid compounds that activate the mu opioid receptor and inhibit

the delta opioid receptor to provide information that allows for assessment of their potential as novel .

Such “bifunctional opioid peptidomimetics” were evaluated to define their structure- activity relationship (SAR) properties and improve our understanding of how such compounds bind to and activate the mu and delta opioid receptors and influence downstream signaling. Some key findings were that substitutions to the N1 position of the tetrahydroquinoline (THQ) core of

the peptidomimetic improved delta opioid receptor affinity while maintaining an already high mu

opioid receptor affinity. In addition, larger substituents at the C6 position of the THQ core, such

as a 2-napthyl or 2-methyl-THIQ, were instrumental in achieving the desired mu opioid receptor

agonist and delta opioid pharmacological profile. These insights provide

valuable information to advance the design and development of bifunctional opioid

peptidomimetics that exhibit high binding affinity for both mu and delta opioid receptors.

xii Several bifunctional opioid peptidomimetics demonstrate a greater degree of mu opioid receptor agonism than the prototypical opioid full agonist, DAMGO. The assessment of this

“superagonism” was achieved by using multiple cell lines, performing G protein activation assays, and competition binding assays with or without sodium ions and guanine nucleotide. Several of these peptidomimetics also demonstrated a difference in the way they signal downstream of the mu opioid receptor, with some compounds preferring G protein activation and others preferring the recruitment of arrestin proteins. In addition to characterizing superagonism and biased agonism, interactions between the mu and delta opioid receptors at the single cell level were

examined to further understand the mechanism of action of these bifunctional opioid peptidomimetics in an attempt to explain the beneficial pharmacology generated by such compounds. In conclusion, these compounds may serve as valuable pharmacological tools to further define biased agonism at the opioid receptors and help drive the development of bifunctional opioid compounds as safe and non-addictive analgesics.

.

xiii Chapter I

General Introduction

Opioids and Opioid Receptors

The term “opioids” refers to all organic compounds that bind to opioid receptors, and includes the endogenous opioid peptides, the , and an ever-expanding list of synthetic opioid compounds. The terms “opiates” represents the naturally occurring alkaloid compounds extracted from , which is the dried latex of the plant. These alkaloid compounds have been used by for medicinal, recreational or spiritual purposes for an estimated eight thousand years (Brook, Bennett and Desai, 2017). is the most commonly used , typically in prescription-grade cough suppressants, and it is on the Essential

Medicines List (EML) of the World Health Organization (WHO). Other common opiates include , which serves as the raw material for several semisynthetic opioids, and , which is the most abundant active alkaloid in opium and also on the EML of the WHO. Morphine was first extracted from opium by Friedrich Wilhelm Sertürner in the year 1804. For its tendency to induce sleep the compound was named morphium, after the Greek God of dreams, Morpheus. The elucidation of its chemical formula and derivation of its structure earned Sir Robert Robinson the

1947 Nobel Prize in Chemistry. Today, morphine is regarded as the gold standard treatment for acute and chronic severe pain. The opioids are an undeniably important class of medicines that have greatly improved our health and well-being throughout much of history.

1 The existence of molecular targets for opioid drugs was first proposed in 1954 (Beckett and Casy, 1954) and over the course of two decades multiple receptor proteins were identified by several teams of scientific researchers (Portoghese, 1965; Pert and Snyder, 1973; Gilbert and

Martin, 1976; Lord et al., 1977). Their discoveries were later confirmed when mRNAs for three opioid receptor types were cloned and characterized (Evans et al., 1992; Yasuda et al., 1993; Chen et al., 1994). The µ Opioid Receptor (µOR) is the target for the vast majority of opioids and is

responsible for most, if not all, of the observed acute and chronic effects of morphine and other

opioid analgesics. The other members of the opioid receptor family include the δ Opioid Receptor

(δOR), the κ Opioid Receptor (κOR) and the closely related Nociceptin/Orphanin FQ (N/OFQ)

Peptide (NOP) receptor, the latter of which was found to have high sequence homology with the

previously identified opioid receptors (Mollereau et al., 1994).

Opioid receptors are G protein-coupled receptor (GPCR) proteins, which are 7-

transmembrane domain proteins that interact with heterotrimeric G proteins and other intracellular

signaling proteins. Opioid receptors are abundantly expressed in the brain and spinal cord in the

central nervous system and throughout the peripheral nervous system. As these receptors are

essential for the inhibition of the sensation of pain, they are localized to areas along the pain

pathways. µORs are expressed on neurons in brain regions such as the amygdala, anterior cingulate

cortex, caudate putamen, hippocampus, locus coeruleus, nucleus accumbens, periaqueductal gray,

and thalamus. (Mansour et al., 1994). In the spinal cord, µORs are predominantly expressed on

dorsal root ganglion neurons located within the dorsal horn (Besse et al., 1990). The δORs are

more restricted in their distribution in the central nervous system and in the brain are expressed in

the amygdala, anterior cingulate cortex, caudate putamen, nucleus accumbens, striatum, and

temporal cortex (Blackburn et al., 1988; Mansour et al., 1994). The broad expression pattern of

2 opioid receptors throughout the body, compounded by an assortment of endogenous opioid peptides, which include the , endorphins, (Lord et al., 1977) and (Zadina et al., 1997), raises the possibility that each endogenous opioid peptide- receptor complex may exhibit a unique profile of signal transduction events. Together, endogenous opioid peptides, opiates, and synthetic opioids generate a broad spectrum of behavioral and physiological effects to regulate complex human behavior and physiology.

At the level of the cell, opioid agonist-receptor binding initiates the activation of heterotrimeric G proteins of the inhibitory Gαi/o class (Al-Hasani and Bruchas, 2011). The coupling of G proteins to membrane-bound GPCRs facilitates the exchange of guanosine 5’-diphosphate

(GDP) for guanosine 5’-triphosphate (GTP), leading to the functional dissociation of GTP-bound

Gαi/o and Gβγ. The activated heterotrimeric G protein subunits bind to and either activate or inhibit

effector enzymes, catalyzing various signaling events, which have been well studied and outlined

in an extensive review (Williams et al., 2013). Gαi/o-mediated inhibition of adenylyl cyclase

reduces the production of cyclic adenosine monophosphate (cAMP), a second messenger involved in many biological processes. Additionally, heterotrimeric G protein subunits differentially activate G protein-coupled inwardly rectifying potassium (GIRK) channels and various kinases

(e.g., extracellular signal-regulated kinase (ERK1/2), mitogen-activated protein kinase (MAPK), c-Jun N-terminal kinase (JNK) and AKT/protein kinase B).

Gαi/o protein signaling is terminated when the intrinsic GTPase activity of the Gαi/o subunit facilitates GTP-Gαi/o hydrolysis back to Gαi/o-GDP, ultimately inactivating Gαi/o. Accessory

proteins called regulators of G protein signaling (RGS proteins) accelerate the process of GTP-

Gαi/o hydrolysis and are fundamentally responsible for the rapid termination of Gαi/o protein

3 signaling (Ross and Wilkie, 2002). Overall, the activity of opioid receptors in functioning neurons

is determined by their recent history of activation and their association with other proteins.

Frequent and repeated exposure of an opioid receptor to an opioid agonist produces opioid

tolerance. In general terms, tolerance is defined as a reduction in effect following prolonged drug

administration that leads to a loss of drug potency. At the molecular level, a series of signal

transduction events reduce the total number of functional receptors expressed on the cell

membrane. Following receptor activation and the functional dissociation of heterotrimeric G

proteins, G protein receptor kinases (GRKs) bind to agonist-bound receptor and phosphorylate

several amino acid residues on the C-terminus and intracellular loops of the 7-transmembrane

GPCR. Receptor phosphorylation provides the site for β-arrestin recruitment and this initiates

further downstream signaling events. Such events include clathrin-mediated endocytosis of the receptor, its transportation through various intracellular compartments, and degradation or recycling of the receptor back to the plasma membrane. Loss of receptor expression on the plasma membrane over time is a central mechanism that produces opioid tolerance. Moreover, different opioid agonists variably induce receptor internalization and loss of receptor expression, which differentially influences the extent of opioid tolerance (Cahill et al., 2016). Pharmacokinetic factors such as the absorption, distribution, metabolism, and excretion (ADME) of the opioid, as well as frequency of administration and dosage, contribute to the degree of opioid tolerance.

In conclusion, thousands of opioid compounds have been developed as pharmacological tools (i.e., ligands) for the study of the opioid receptors. A network of intracellular processes regulates the activity of the opioid receptors, and the duration and sensitivity of G protein signaling ultimately produces adequate (or inadequate) analgesia for a patient suffering with pain, thus highlighting the extreme importance of opioid receptors in human biology and pathophysiology.

4

Opioid Use in Pain

It is estimated that 20% of adults in the United States have chronic pain and 8% have high-

impact chronic pain that frequently limits life or work activities (Dahlhamer et al., 2018). Pain is

classified as acute or chronic; acute pain, generally caused by tissue damage, resolves quickly and

typically persists less than three months, whereas chronic pain persists for greater than three

months. Opioids are the most effective medications to provide pain relieving effects (analgesia) to

an individual suffering from acute, moderate to severe pain. Studies in µOR knockout mice

(Kieffer and Gavériaux-Ruff, 2002) have confirmed that both analgesia and the side effects of morphine and other opioids are mediated by the µOR. In support of the key role of the activation of µOR for producing analgesia, compounds selective for the activation of the δOR and κOR have

not achieved success in clinical development (Chavkin and Martinez, 2015; Spahn and Stein,

2017). Although δOR-activating compounds have shown promise in preclinical in vivo models for

certain indications (i.e. producing -like effects), δOR-selective compounds that lack

µOR activity produce limited analgesia. κOR-selective compounds produce limited analgesia as well, and δOR and κOR agonists have their own unique set of side effects (i.e. producing convulsions and dysphoria, respectively).

Although µOR-selective opioid medications are the gold standard for the treatment of acute pain, they are not as effective for certain types of chronic pain, such as neuropathic pain, and they produce unpleasant side effects, which include constipation, drowsiness, nausea, and pruritus. For the management of chronic pain, an opioid is required for a long period of time, and in most circumstances, the prolonged administration of an opioid leads to the development of opioid tolerance (Trescot et al., 2008; Hayhurst and Durieux, 2016). Additionally, prolonged opioid

5 administration may also lead to opioid-induced hyperalgesia (Mao, 2002; Angst and Clark, 2006).

Although there are strategies for improving patient safety and satisfaction, opioid tolerance and

opioid-induced hyperalgesia reduce the long-term utility of opioids in the treatment of chronic pain

(Weber, Yeomans and Tzabazis, 2017). The clinical procedures to avoid opioid tolerance are to

either increase the dosage or switch to a more potent opioid agonist, but these two procedures

elevate the risk of the patient developing a physical dependence to the effects of the opioid

medication and once opioid treatment is stopped there will be withdrawal effects. Another

drawback to prolonged opioid treatment is that tolerance to side effects develops less rapidly

compared to tolerance to the desired therapeutic effect. For example, tolerance to constipation

develops at a slower rate compared to analgesia and thus constipation proportionally increases in

severity over time. As such, opioid-induced constipation is a very common issue in the clinic,

although this is now managed with peripherally-acting μOR antagonists (Holzer, 2008).

The management of acute and chronic pain places a tremendous burden on our economy

and society. Ineffective pain management has resulted in prolonged hospital visits, increased rates

of hospitalization, and a decreased ability for people to function in everyday life. Combining the

cost of associated health care, patient rehabilitation, and lost productivity in the workplace, the estimated annual cost to the United States economy (as of 2010) is between $560-$635 billion

(Gaskin and Richard, 2012). Patient care requires a considerable amount of money, time, and effort from physicians, nurses, pharmacists, and other healthcare professionals. Thus, the treatment of chronic pain is a formidable and growing public health issue. While opioids have been effective

in eliminating acute, moderate to severe pain and are recommended in circumstances of chronic,

neuropathic and inflammatory pain, the detrimental side effects that limit their clinical utility and

responsible long-term use have triggered a serious public health emergency. New strategies are

6 needed that harness the strong pain-relieving effects that the µOR provides, and these strategies must evade the adverse effects that are typically observed with opioid treatment.

The Opioid Crisis

The ongoing “opioid crisis” is arguably the most devastating drug epidemic in recent history (Jalal et al., 2018). According to the Centers for Disease Control and Prevention (CDC), every day 130 Americans die from an , and in the span of 18 years [1999-2017] nearly 400,000 people have died from an opioid overdose. The number of opioid overdose deaths has increased 6-fold since 1999, prompting the U.S. government to declare the opioid crisis a public health emergency (The Lancet, 2017). This state of emergency has arisen due to various contributing factors, such as the over-prescription of clinically prescribed opioid drugs (i.e. , , and ) and the greater prevalence of use, which is conceivably a consequence of the eventual inaccessibility and high cost of prescription opioids

(Kolodny et al., 2015). Another causative factor is the increase in availability of very potent synthetic opioids, such as fentanyl and its numerous analogues, which have greatly elevated the risk of fatality due to accidental overdose; is one such analogue that is a staggering

10,000 times more potent than morphine (Elzey, Barden and Edwards, 2016). Despite the good intentions to improve the therapeutic window with potent opioids, the physical handling of even tiny amounts of these compounds can be extremely hazardous. The opioid receptor antagonist , which has life-saving potential to reverse a heroin overdose by displacing it from receptors in the brain stem and medulla, may not be able to counteract extremely potent opioids.

Heroin, and now fentanyl, have been selected as the opioid drugs to be sensationalized by the mass

7 media, however, the seemingly innocuous prescription painkiller (i.e. OxyContin®, Vicodin®) is

the veiled menace that has fueled the opioid crisis.

One in five heroin addicts claim the first opioid drug they used was a prescription painkiller

(Ahrnsbrak et al., 2017). According to data from the 2017 National Survey on Drug Use and Health

(NSDUH), the number of people using heroin has more than doubled in the past 15 years to nearly one million people in 2016. From the analysis of the NSDUH data, the National Institute on Drug

Abuse (NIDA) concluded that excessive prescribing of and easy access to legal opioids prompted an increase in opioid diversion and illegal drug-seeking behavior. An over-reliance on opioid drugs for conditions of minor pain and an under-appreciated need for social support and better treatment options for people struggling with opioid addiction have exacerbated the opioid crisis. The opioid crisis represents a focal point at the center of two challenging and long-standing public health problems; chronic pain and opioid addiction.

In a “call to arms” press report by Dr. Nora Volkow and Dr. Francis Collins of the National

Institutes of Health (NIH), there is an urgency to simultaneously reduce suffering due to pain and to tackle the problems associated with long-term use of opioid medications (Volkow and Collins,

2017). Strategies to fight the opioid crisis they fall into three general categories: (i) overdose prevention and reversal, (ii) treatment of opioid use disorders (OUDs), and (iii) treatment of chronic pain. Moreover, within each category there are several short-term, intermediate, and long- term goals. Drs. Volkow and Collins suggest that making improvements in (i) and (ii) is undeniably obligatory, but the innate problems will continue if the treatment needs of those with chronic pain are not adequately addressed. Consequently, long-term strategies that address OUDs and chronic pain require deliberate consideration and execution.

8 For the improved treatment of chronic pain, there is a need for safer, non-addictive, and

more effective opioid analgesics. Currently, there are several opioid drug discovery campaigns

ongoing, with each campaign incorporating a unique biological concept to target opioid receptors.

One is allostery and the development of allosteric modulators for the opioid receptors and another is biased agonism and the development of G protein-biased agonists, both of which could lead to the development of safer and more effective opioids for the management of chronic pain. The unique biological concept addressed in this dissertation is to target multiple opioid receptors simultaneously with a “bifunctional” opioid compound. Such a bifunctional compound would bind to two receptors simultaneously and either activate both receptors (i.e. μOR agonist / δOR agonist) or exhibit agonism at one receptor and antagonism at the other (i.e. μOR agonist / δOR antagonist); for an extensive review on bifunctional opioid compounds, consult the following (Schiller, 2010;

Anand and Montgomery, 2018; Günther et al., 2018).

Targeting Multiple Opioid Receptors

The immense opioid catalog largely contains opioids that have been developed to selectively bind to and activate the μOR, because the activation of µOR is responsible for most, if not all, of the opioid drug effects. Opioids were developed to be selective for δOR, κOR, and the

NOP receptor because they were extremely useful as pharmacological tools in understanding the properties of each opioid receptor. The current consensus, however, is that opioids selective for

μOR, δOR, κOR, and the NOP receptor all have limitations in regard to their clinical utility, but opioids that target more than one type of receptor might have therapeutic potential as analgesics with reduced side effects.

9 It has been widely reported in the literature that the δOR modulates μOR-agonist-mediated effects such as the development of antinociceptive tolerance, the development of physical dependence, rewarding effects, and respiratory depression, providing support for δOR as an important drug target. The first evidence that δOR modulated μOR activity was from in vivo studies that reported that δOR blockade greatly reduced the development of morphine tolerance and dependence (Abdelhamid et al., 1991). In this study, the development of morphine tolerance was considerably reduced in mice that were treated with the δOR-selective antagonist naltrindole

(NTI) in conjunction with chronic morphine administration. Another study that used the δOR- selective peptide antagonist TIPP[ψ] added support to this preliminary observation (Fundytus et al., 1995). In studies using rats, the co-administration of NTI with chronic morphine treatment also decreased the development of morphine tolerance, and withdrawal effects following a naloxone challenge were significantly attenuated (Hepburn et al., 1997). In another study, following i.c.v. administration of the irreversible δOR antagonist, 5’-NTII, there was a reduction in the reinforcing effects of heroin, without affecting antinociception (Martin et al., 2000). Finally, there is evidence

that the δOR-selective antagonists naltrindole (NTI) and naltriben (NTB) prevented the

development of tolerance to μOR-mediated antinociception (Shippenberg, Chefer and Thompson,

2009). These in vivo pharmacological studies using various δOR-selective antagonists suggest that

the inhibition of δOR could be beneficial in reducing opioid tolerance, physical dependence, opioid

withdrawal, and abuse-related rewarding effects.

To strengthen the data reporting that a reduction of chronic morphine-induced side effects was an effect of pharmacological δOR inhibition, δOR expression levels were decreased via i.c.v.

administration of a 20mer antisense oligodeoxynucleotide targeting the mRNA of the cloned delta

opioid receptor (Kest et al., 1996); led to an inhibition of acute morphine tolerance and

10 dependence. In several other studies, the complete elimination of δOR function was achieved by

using δOR knockout mice. In one study, morphine analgesia was significantly reduced after 5 days of chronic administration of morphine, whereas morphine analgesia was preserved in δOR knockout mice (Zhu et al., 1999). In support of this investigation, our laboratory employed δOR knockout mice and observed a threefold rightward shift in the morphine dose-response curve in wild-type mice, but failed to see any shift in δOR knockout mice after repeated treatment with escalating doses of morphine (Anand et al., 2018). Together, these results suggest a novel opioid compound with a μOR agonist and δOR antagonist pharmacological profile to be an analgesic with a lesser propensity to produce opioid tolerance.

Other investigators considered how δOR inhibition could regulate respiratory depression.

In one study, -induced respiratory depression could be reversed by NTI (Freye, Latasch and Portoghese, 1992), and in another study, peripheral administration of NTI could reverse -induced respiratory depression but not the antinociceptive activity of alfentanil (Su,

McNutt and Chang, 1998). In summary, δOR knockout mice show resistance to morphine-induced respiratory depression, and reduced development of opioid tolerance, physical dependence, and withdrawal symptoms (Chefer and Shippenberg, 2009). These studies lend support for the hypothesis that δOR activity is a key feature in the various side effects mediated through the μOR.

In corroborate of the results of these studies, is a bifunctional opioid compound with a NOP and μOR partial agonist / δOR and κOR antagonist pharmacological profile, and it

demonstrates a remarkable safety margin as it does not suppress respiratory function to the degree

of other classical opioids (Lewis and Husbands, 2005). Taken together, a centrally-acting

bifunctional opioid that has a μOR agonist / δOR antagonist pharmacological profile may produce

less respiratory depression (Freye, Latasch and Portoghese, 1992; Su, McNutt and Chang, 1998;

11 Chefer and Shippenberg, 2009), which would be encouraging for the development of safer opioid

analgesics, that are comparable to Buprenorphine and that lessen respiratory suppressing effects.

In conclusion, simultaneous targeting of more than one type of opioid receptor may be a

beneficial strategy to mitigate opioid drug side effects. The combination therapy of buprenorphine and the naloxone (Suboxone®) for the management of opioid addiction (Chen,

Chen and Mao, 2014) was formulated to discourage intravenous administration and may have less abuse potential than buprenorphine alone. As mentioned earlier, combining a μOR agonist with a

peripherally-active non-selective opioid antagonist is a polypharmacy approach to reduce the

occurrence of opioid-induced constipation (Holzer, 2008). Although opioid polypharmacy is a reliable approach for mitigating certain side effects, combining two or more opioid drugs may bring unnecessary complication in regard to drug-drug interactions, unpredictable pharmacokinetics, and concerns regarding patient compliance (Morphy and Rankovic, 2006). To address these limitations, single compounds that have a distinct pharmacophore for each opioid receptor type have been designed, characterized, and termed bifunctional opioid compounds.

Bifunctional Opioid Compounds

Eluxadoline (VIBERZI®) is a bifunctional opioid compound that is an FDA approved drug and prescribed to patients for the treatment of irritable bowel syndrome with diarrhea (IBS-D)

(Figure 1) (Lembo et al., 2016; Levio and Cash, 2017). This compound activates μORs while simultaneously inhibiting δORs within the peripheral nervous system, and because it does not cross the blood-brain barrier it does not produce rewarding effects (Breslin et al., 2012; Fujita et al.,

2014). Accordingly, it does not produce a robust analgesic effect, as results from two phase III trials did not reach clinical significance (over placebo) for relieving abdominal pain (Barshop and

12 Staller, 2017). Despite the uncertainty that its δOR antagonist feature is responsible for stabilizing

gastrointestinal distress, does afford significant improvements in the symptoms of

IBS-D. Nevertheless, the success of eluxadoline in achieving a clinically significant outcome has bolstered confidence in developing single chemical entities that function in different ways at multiple opioid receptor types.

Figure 1. Chemical Structures of μOR Agonist / δOR Antagonist Bifunctional Opioid Compounds. (A) Eluxadoline (Breslin et al., 2012) (B) DIPP-NH2 (Dmt-Tic-[CH2NH]-Phe-Phe-NH2) (Schiller, Weltrowska, et al., 1999) (C) UFP-505 (Dmt-Tic-Gly-NH-Bzl) (Balboni et al., 2010) (D) UMB-425 (Healy et al., 2013) (E) UMB-246 (Healy et al., 2017) and (F) Pseudoindoxyl (Váradi et al., 2016).

Such compounds were first reported in the 1990’s and have since been discussed in several

literature reviews (Ananthan, 2008; Schiller, 2010; Anand and Montgomery, 2018). From the in

vivo studies already discussed, it is well documented that μOR agonist / δOR antagonist

bifunctional opioid compounds confer a greater therapeutic effect to side effect profile. Some

notable examples of μOR agonist / δOR antagonist bifunctional opioid compounds, which are

shown in Figure 1 include Eluxadoline (VIBERZI®), the tetrapeptide DIPP-NH2 (Dmt-Tic- 13 [CH2NH]-Phe-Phe-NH2) (Schiller, Weltrowska, et al., 1999), the peptide UFP-505 (Dmt-Tic-Gly-

NH-Bzl) (Balboni et al., 2010), the morphinan analogues UMB-425 (Healy et al., 2013) and UMB-

246 (Healy et al., 2017), and the corynanthe analogue mitragynine pseudoindoxyl (F) (Váradi et

al., 2016). Moreover, bifunctional peptide-like opioid compounds that target the μOR and δOR will be the focus of this dissertation.

Bifunctional Opioid Peptides

The collaborative efforts of the Mosberg lab and several labs in the Department of Pharmacology

led to the development of many bifunctional opioid peptides. The inspiration for the design of bifunctional peptide compounds comes from the endogenous peptide Leu-enkephalin, and the

synthetic peptide DPDPE ([D-Pen2, D-Pen5]-enkephalin) (Mosberg et al., 1983), which is shown in Figure 2A. DPDPE is a cyclic disulfide-containing pentapeptide analogue of the endogenous opioid, Leu-enkephalin. The carboxylic acid on the C-terminus of DPDPE is critical for its high

δOR affinity, selectivity, and agonism for δOR over the μOR. The key components of its δOR selectivity can be visualized by comparing the C termini of two synthetic analogues of DPDPE,

JOM-6 (Tyr-c-(S-CH2-CH2-S)-[D-Cys-Phe-D-Pen]-NH2) (Figure 2B) and JOM-13 (Tyr-c-(SS)-

[D-Cys-Phe-D-Pen]-OH (Figure 2C). The carboxylic acid on the C-terminus of DPDPE was modified to a carboxamide in JOM-6 (McFadyen et al., 2000), and due to this structural change there was a significant enhancement in its μOR affinity; the μOR binding affinity (Ki) of JOM-6

is 0.29 nM whereas the μOR Ki of JOM-13 is 52 nM (Mosberg and Fowler, 2002).

Further optimization of the disulfide-containing tetrapeptides led to discovery of more conformationally constrained cyclic tetrapeptides that demonstrated the desired μOR agonist /

δOR antagonist pharmacological profile. KSK-103 (DMT-c(SEtS)-[D-Cys-Aic-D-Pen]-OH),

14 (Figure 2D) incorporated an ethylene dithioether linkage and replaced the residue

(Phe3) with 2-aminoindane-2-carboxylic acid (Aic), a bulkier and more constrained aromatic

residue (Purington et al., 2011). Another important modification was the replacement of the Tyr1

residue with 2’,6’-Dimethyl-L-tyrosine (DMT), a commonly used chemical moiety in the synthesis

of opioid peptides and peptidomimetics (Shiotani et al., 2007; Balboni et al., 2010). Previous

studies had shown there were significant improvements in Ki, potency (EC50) and biological

activity for analogues that incorporated the DMT moiety, as opposed to L-Tyrosine (Hansen et al.,

1992).

Whereas JOM-6 lacked the DMT moiety and had a μOR Ki of 0.29 nM and a δOR Ki of 25 nM, KSK-103, which incorporated the DMT moiety, demonstrated an increase in its δOR Ki and

afforded an equivalent affinity for μOR (Ki = 2.4 nM) and δOR (Ki = 2.3 nM). In addition to

balanced μOR and δOR binding affinities, this compound was a µOR partial agonist and a δOR

antagonist. In the GTPγ35S (Guanosine‐5′‐O‐(3‐[35S]thio)triphosphate) binding assay using μOR

35 cell membranes, KSK-103 had an EC50 of 4.7 nM and a stimulation of GTPγ S binding of 59%

relative to the prototypical μOR full agonist, DAMGO ([D-Ala2, N-MePhe4, Gly-ol]-enkephalin).

In the GTPγ35S binding assay using C6 δOR membranes and as determined by competitive inhibition of δOR agonist-mediated stimulation of GTPγ35S binding, KSK-103 had an antagonist

affinity constant (Ke) of 4.4 nM, indicating that it is a high affinity δOR antagonist (Purington et

al., 2011). The innovative structural modifications and the pharmacological evaluation of KSK-

103 enabled the discovery of several other peptide analogues that exhibited the desired μOR

agonist / δOR antagonist profile and had high binding affinities for both the μOR and the δOR.

While improvements were made in rational ligand design and facile chemical synthesis,

and structure-activity relationship (SAR) efforts toward the goal of producing high affinity μOR

15 agonist / δOR antagonist bifunctional peptides were successful by our group and others, many of the peptide analogues, such as UFP-505 (Figure 1C), failed to demonstrate centrally-mediated biological activity following peripheral administration in mice (Schiller, Fundytus, et al., 1999;

Balboni et al., 2010). Peptides have poor bioavailability, poor BBB permeability, and are susceptible to rapid proteolysis in the body. To account for the poor pharmacokinetics of peptides,

KSK-103 was glycosylated via the addition of a C-terminal β-glucosylserine residue. The resulting peptide, VRP-26 (Figure 2E), demonstrated a significant improvement in its bioavailability and metabolic stability, as demonstrated by its robust in vivo antinociceptive activity (Mosberg et al.,

2014). In fact, this was arguably the first bifunctional μOR agonist / δOR antagonist compound to demonstrate central bioavailability following peripheral administration. Moreover, VRP-26 produced significant acute and chronic antinociceptive activity and reduced side effects after prolonged administration in vivo, and its activity was comparable to the activity of morphine

(Anand et al., 2016). Overall, steady progress has been made in the development of bioavailable bifunctional opioid peptides that show reductions in acute antinociceptive tolerance. Although the lessons learned along the way serve as valuable benchmarks, bifunctional opioid peptidomimetics come with their own pharmacodynamic and pharmacokinetic challenges.

Bifunctional Opioid Peptidomimetics

The structure-activity relationships (SAR) of the peptide analogues were analyzed to design more drug-like ‘peptidomimetics’. As the name intuitively suggests, a peptidomimetic is a compound that incorporates key moieties of a peptide into a more drug-like, small molecule scaffold, thereby conferring greater resistance to proteolytic degradation and allowing penetration through biological membranes. A tyrosine residue and an aromatic side chain are two critical elements of

16 endogenous opioid peptides required for receptor binding and activity. Therefore, the tyramine

group and benzyl side chain of the tetrapeptide analogues were incorporated into a small molecule

scaffold to engineer novel opioid peptidomimetics.

Figure 2. Chemical Structures of Opioid Peptide Analogues and the Design of a Bifunctional Opioid Peptidomimetic. (A) DPDPE (Clark et al., 1986) (B) JOM-6 (McFadyen et al., 2000) (C) JOM-13 (Mosberg and Fowler, 2002) (D) KSK-103 (Purington et al., 2011) (E) VRP-26, also known as compound 4 (Mosberg et al., 2014) and (F) compound 1 (Mosberg et al., 2013). The C-terminal β-glucosylserine residue of VRP-26 and the 2’,6’- Dimethyl-L-tyrosine (DMT) moiety and benzyl side chain of compound 1 are highlighted in blue.

The goal in the design of opioid peptidomimetics was to eliminate the bulky 11-membered

disulfide-containing moiety that was characteristic of the opioid tetrapeptides (Figure 2). This was accomplished by replacing the C-terminal tripeptide fragment with a tetrahydroquinoline (THQ)

moiety while retaining the critical benzyl side chain residue. The incorporation of a THQ structure,

which is an aromatic heterocyclic amine composed of a benzene ring fused to a piperidine, afforded

greater flexibility in making chemical substitutions thereby improving the process of synthesis as

compounds could be tailored to achieve various binding and efficacy profiles. The lead

17 bifunctional opioid peptidomimetic, (S)-2-amino-N-((R)-6-benzyl-1,2,3,4-tetrahydroquinolin-4- yl)-3-(4-hydroxy-2,6-dimethylphenyl)-propanamide, also known as KSKPP1E, and which has been published in the literature as compound 1 (4R) (Mosberg et al., 2013), shares several key features with its parent peptide, KSK-103 (Figure 2D). The superposition of the tyramine moiety and benzyl side chain, which is depicted in blue for compound 1 (4R) (Figure 2F), shows that the

THQ structure is a peptide backbone replacement that also mimics the phenylalanine (Phe3) side

chain of the parent tetrapeptide.

The pharmacological properties of compound 1 (4R) were first examined using C6 rat glioma cells that stably overexpress the rat µOR and C6 cells that stably overexpress the rat δOR

(Wang et al., 1998). In these studies, at μOR compound 1 (4R) was reported to have a Ki of 0.22

35 nM, EC50 of 1.6 nM, and stimulation of GTPγ S binding of 81% compared to DAMGO. At the

35 δOR it was reported to have a Ki of 9.4 nM, an EC50 of 110 nM, and stimulation of GTPγ S

binding of 16% compared to DPDPE. These data in the context of an in vitro model system were

used to characterize compound 1 (4R) as a potent μOR agonist and potent δOR partial agonist.

After initial in vitro evaluation, compound 1 (4R) was examined in a standard in vivo model of antinociception, the warm water tail withdrawal (WWTW) assay. Following its peripheral administration in mice, compound 1 (4R) produced dose-dependent antinociception in the WWTW assay. Its antinociceptive activity, with an ED50 of approximately 3 mg/kg, was

reversible demonstrating that the effect was mediated by opioid receptors (Figure 3A) (Mosberg

et al., 2013). Similar to the lead bifunctional peptides that demonstrated a prolonged

antinociceptive effect in vivo, compound 1 (4R) demonstrated a duration of action comparable to

but slightly shorter than that of morphine (Figure 3B).

18 In conclusion, the lead peptidomimetic compound 1 (4R) was determined to be a

bioavailable and centrally active opioid compound with a μOR agonist / δOR partial agonist pharmacological profile. Based on several improvements in its structural design (i.e. incorporation of a THQ) that allowed for efficient chemical synthesis (Bender, Griggs, Gao, et al., 2015) and extensive structural modification, it was chosen as the lead compound and starting point for the development of a series of peptidomimetic analogues that are explored in this dissertation.

Figure 3. In Vivo Antinociceptive Activity of the Bifunctional Opioid Peptidomimetic, compound 1. (A) Data represent a cumulative dose response of compound 1 following a 30 min pretreatment with saline [open circles] or the opioid antagonist naltrexone (NTX) [filled circles]. The antinociceptive activity of compound 1 increased in a dose- dependent manner and was opioid receptor mediated, because 1 mg/kg NTX produced a rightward shift in the dose response curve. (B) Data represent a time course [120 minutes] for the antinociceptive activity of morphine [open circles] and compound 1 [filled squares] following i.p. injection. Administration of 10 mg/kg of compound 1 produced maximal antinociceptive activity that was maintained for approximately 60 minutes. This figure is published in (Mosberg et al., 2013).

19 Aims of the Dissertation

This dissertation reports in vitro experiments designed to pharmacologically analyze

bifunctional opioid peptidomimetic compounds and provide information that allows for

assessment of their potential as novel analgesics. [3H]-diprenorphine competition binding assays

35 are used to evaluate the binding affinity (Ki) of the analogues, and stimulation of GTPγ S binding

assays are used to evaluate the potency (EC50) and relative efficacy of the analogues. The structure- activity relationship (SAR) data of the bifunctional peptidomimetic analogues are explored to clearly define their µOR and δOR agonist properties, including superagonism and biased agonism.

In addition, the interactions between µOR and δOR at the single cell level are examined. Insights from these studies will provide information that can be used to (i) advance the design and development of compounds with equivalent and high binding affinity for both the µOR and the

δOR, (ii) identify lead compounds for in vivo analysis, and (iii) further understand the mechanism of action of bifunctional opioid peptidomimetics at the opioid receptors and how this might explain the beneficial pharmacology generated by such compounds.

20 Chapter II

Characterization of Bifunctional Opioid Peptidomimetics

Introduction

Traditional opioid drugs, such as morphine and fentanyl, bind to and activate the μ Opioid

Receptor (μOR) with limited pharmacology mediated via the three other types of opioid receptor

(δOR, κOR, and NOP). Activation of the μOR is responsible for analgesia and the unwanted side

effects of clinically used opioids. Opioids that act as bifunctional compounds, by binding to both

μOR and δOR simultaneously and activating the μOR (agonism) while inhibiting the δOR

(antagonism), may offer clear advantages in terms of their side-effect profile (Ananthan, 2008;

Schiller, 2010). Several studies have produced convincing data that led to the hypothesis that inhibition of the δOR could mitigate the development of tolerance to the analgesics effects of opioids, the development of physical dependence, and withdrawal effects associated with chronic

opioid treatment (Abdelhamid et al., 1991; Hepburn et al., 1997). These studies provided the

rationale for the design, synthesis, and characterization of peptides, and then peptidomimetics, that

are bifunctional μOR agonist / δOR antagonist opioid compounds.

The design of the bifunctional opioid peptidomimetics was based on prior analysis of the similarities and differences between the active and inactive receptor states of μOR, δOR, and κOR

(Mosberg and Fowler, 2002; Pogozheva, Przydzial and Mosberg, 2008). Using refined ligand-

receptor homology models, as well as computational modeling, it was hypothesized that the ligand

binding pocket in the active δOR conformational state is smaller than the ligand binding pocket in

21 the inactive conformational state (Purington et al., 2011). Peptidomimetic analogues with bulky

substituents could thus be accommodated into the larger pocket of the inactive, but not active, δOR

conformation. Therefore, analogues with larger substituents were predicted to be δOR antagonists.

It was also predicted from homology models that the active μOR binding pocket is larger than the active δOR binding pocket. Thus, peptidomimetic analogues with bulky substituents may produce

μOR agonism in addition to δOR antagonism. Also taken into consideration were insights from the publications on the solved crystal structures of both of the inactive receptor states of the μOR,

δOR, and κOR (Granier et al., 2012; Manglik et al., 2012; Wu et al., 2012; Fenalti et al., 2014,

2015) and the active receptor states of the μOR and κOR (Huang et al., 2015; Che et al., 2018).

Moreover, the strategy in the development of bifunctional peptidomimetic analogues was to (i) use receptor homology models and (ii) integrate structure-activity relationship (SAR) data from the bifunctional peptides to design novel peptidomimetics, (iii) synthesize analogues with precise pharmacological properties, and (iv) pharmacologically characterize them to elucidate the structural features that govern bifunctional peptidomimetic-receptor binding and activity. The characterization of bifunctional peptidomimetic analogues involved the evaluation of their receptor binding affinity (Ki) and relative efficacy (i.e. potency (EC50) and percent stimulation of

GTPγ35S binding compared to a standard agonist) at the μOR and the δOR.

The lead bifunctional peptidomimetic for this endeavor was compound 1 (4R) (Mosberg et al., 2013), also shown in Figure 4. The tetrahydroquinoline (THQ) core of compound 1 served as a suitable scaffold because it allowed for an extensive variety of structural modifications. Various chemical substitutions were made at the R1 and R2 positions of compound 1 to further elucidate

the chemical properties that drive ligand binding and G protein activation at each of the opioid

receptors. Two key features of the bifunctional peptides were preserved in the design of the

22 peptidomimetics, the 2’,6’-Dimethyl-L-tyrosine (DMT) moiety and the benzyl side chain, which

are highlighted in blue in Figure 2F. As was discussed previously, our goal is to develop potent,

μOR agonist / δOR antagonist peptidomimetic analogues with subnanomolar binding affinities for

both the μOR and the δOR. To achieve this goal, a variety of chemical substitutions were made at

R1 as well as at R2 and R3 of the THQ core of compound 1 (Figure 4). The data and results

presented in this chapter were generated and analyzed in collaboration with Dr. Henry Mosberg

and the medicinal chemists in his lab (please see footnote1), who were responsible for the design

and synthesis of the numerous peptidomimetic analogues, some of which are discussed in this

chapter and dissertation. Extensive SAR data have been published in the following scientific

articles on which I am a co-author; (Bender, Griggs, Anand, et al., 2015; Harland et al., 2015,

2016; Anand et al., 2018; Nastase et al., 2018, 2019; Henry et al., 2019).

Figure 4. Positions for Substitution of the THQ core of compound 1. 1 through 8 is the numbering scheme for the tetrahydroquinoline (THQ) core of the molecule. Substitutions to the THQ core were made at the C6 position (R1), the N1 position (R2), and the C8 position (R3). This lead peptidomimetic has been represented in the literature as 7a (Wang et al., 1998) and as compound 1 (Mosberg et al., 2013).

1 Aaron M. Bender (AMB), Aubrie A. Harland (AAH), Anthony F. Nastase (AFN), Deanna J. Montgomery (DJM), and Sean P. Henry (SPH) are the graduate student medicinal chemists from the lab of Dr. Henry Mosberg who are responsible for the synthesis of the peptidomimetic analogues presented in this dissertation. 23 Results

μOR SAR: C6 position Substitutions (R1)

Bifunctional opioid peptidomimetic analogues of compound 1 were evaluated first for their

binding affinity and relative efficacy at the μOR. Analogues with a benzyl substituent in the C6

position (R1) and carbonyl groups at the N1 position (R2) demonstrate high binding affinity for the

μOR, , in the subnanomolar to low single-digit range (Table 1). These analogues of compound 1

are potent agonists at the μOR, with an EC50 in the low single-digit range, however, there is some

variability in their relative efficacy. AMB42 (53 ± 3.2 %) is a partial agonist and AMB48 (96 ±

5.1 %) is a full agonist at the μOR, whereas compound 2 (76 ± 4.1 %) is similar to the unsubstituted

amine and lead peptidomimetic, compound 1 (81 ± 2.0 %).

The addition of the larger 2-naphthyl substituent, in place of a benzyl substituent, at R1 did

not significantly change μOR binding or efficacy characteristics (Table 2), albeit there was a small increase in the potency of the peptidomimetic analogues. Results from replacement of the benzyl with 2-methyl-tetrahydroisoquinoline (2-methyl-THIQ) at R1 significantly changed μOR efficacy

characteristics and are shown in Table 3. Assessed by maximal stimulation of GTPγ35S binding,

three of these peptidomimetic analogues demonstrate high μOR potency (subnanomolar EC50) and

two of them demonstrate an increase in relative efficacy compared to DAMGO (i.e. GTPγ35S

binding of >100 % of DAMGO). For example, the stimulation of GTPγ35S binding by AMB39,

with the unsubstituted amine and 2-methyl-THIQ group at R1 is 110 ± 5.5 % compared to

DAMGO. Whereas the stimulation of GTPγ35S binding by compound 3, with the unsubstituted amine and 2-naphthyl group at R1 (Table 2) is less at 96 ± 3.0 % compared to DAMGO, the

stimulation of GTPγ35S binding for compound 1, with the unsubstituted amine and benzyl group

35 at R1 (Table 1) is even less at 81 ± 2.0 %. In addition, the stimulation of GTPγ S binding by

24 AFN42 (110 ± 4.7 %), which has a 2-methyl-THIQ at R1, shows a similar increase in efficacy

compared to its corresponding (R2) analogue AFN40 (102 ± 4.0 %), which has a 2-naphthyl group

at R1 (Table 4). Thus, the 2-methyl-THIQ substitution at R1 demonstrates a significant effect on

μOR agonism, whereby analogues with the 2-methyl-THIQ substitution at R1 are more potent and

efficacious than analogues with 2-naphthyl and benzyl group substitutions.

35 It is interesting that the potency of AFN42 (EC50 = 0.08 ± 0.03 nM) to stimulate GTPγ S

binding is not significantly different from the affinity of AFN42 (0.11 ± 0.02 nM) at the μOR.

From the numerous bifunctional opioid peptidomimetics that have been examined, this feature has

rarely been observed. In general, μOR binding affinity (Ki) is at least 10-fold higher (lower nM)

than μOR potency (EC50). For example, AMB39 has a μOR Ki of 0.03 ± 0.01 nM and a μOR EC50

of 0.35 ± 0.11 nM. This difference, whereby AMB39 has a 10-fold higher binding affinity relative

to its potency is a trend that has been observed for nearly all of the peptidomimetic analogues. The

+ discrepancy between Ki and EC50 is due to the presence of Na ions and guanine nucleotide in the

buffer that is used for the GTPγ35S assays; Na+ ions and guanine nucleotide (i.e. GTPγS) decrease

3 the binding affinity (Ki) of opioid agonists (Lee et al., 1999). [ H]-DPN competition binding assays

have traditionally been performed in 50 mM Tris buffer, without Na+ ions and guanine nucleotide,

and thus the binding affinity for agonists is usually higher than the potency of the agonists for

35 stimulating GTPγ S binding. AFN42 is unique for its equivalent Ki and EC50, which implies that

it demonstrates an insensitivity to sodium and or guanine nucleotide, and this feature is further investigated in Chapter III.

25

Table 1. Effects of Benzyl R1 Substitutions and Carbonyl Groups at R2

3 Table 1. Effects of Benzyl R1 Substitutions and Carbonyl Groups at R2. Affinity (Ki) values were obtained from [ H]-diprenorphine competition binding assays, 35 and Potency (EC50) and Efficacy (% Max) values were obtained from stimulation of GTPγ S binding experiments, both using C6 cell membrane preparations expressing either μOR or δOR. All values are in nanomolar concentration (nM) and are represented as the mean ± standard error of the mean (SEM) of three separate assays performed in duplicate. * Peptidomimetic analogue 1 is in the literature as 1 (4R) from (Mosberg et al., 2013). Peptidomimetic analogue 2 is in the literature as 4b from (Harland et al., 2015). AMB48, AMB42, and AAH40 are in the literature as 4c, 4d, and 4h, respectively, from (Harland et al., 2016).

26 δOR SAR: C6 position Substitutions (R1)

Bifunctional opioid peptidomimetic analogues of compound 1 were then evaluated for their

binding affinity and relative efficacy at the δOR. The principal finding is that additional

35 hydrophobic bulk at R1 completely eliminated the analogues ability to stimulate GTPγ S binding,

35 (Table 2). Whereas analogues with the bulky 2-naphthyl at R1 stimulated GTPγ S binding less

than 10% compared to the standard δOR agonist DPDPE, the corresponding analogues with the

smaller benzyl group at R1 (Table 1) displayed a range of δOR partial agonism, producing

anywhere from 16–65% stimulation of GTPγ35S binding compared to DPDPE. The smallest

peptidomimetic, compound 1, weakly stimulates GTPγ35S binding (16% of DPDPE), whereas the

R2 substituted analogues produced a greater level of G protein activation at δOR; AMB48 (65 %),

AMB42 (47 %), AAH40 (42 %). These data suggest that bulky modifications to R1 in the form of

2-naphthyl group, either clash with residues of the active state of the δOR or interact favorably

with residues of the inactive state of the δOR. Peptidomimetics with less bulky modifications to

R1 interact favorably with residues of the active state of the δOR, and moreover R2 substitutions

contribute to an increase in δOR efficacy for these less bulky analogues.

The effect of C6 substitutions on δOR affinity is variable. The δOR Ki of compound 2,

which has a benzyl group at R1 and an acetyl group at R2 is 1.8 ± 0.1 nM, whereas the δOR Ki of

the corresponding analogue AAH8, which has a 2-naphthyl group at R1 and also an acetyl group

at R2, is 10-fold higher with a δOR affinity of 0.2 ± 0.02 nM. Similar results were seen in the comparison of other analogues in Table 1 to their corresponding analogues in Table 2. Consider

AMB48 (with a C6 benzyl substituent) compared to the bulkier AAH36 with a 2-naphthyl group.

Here, the δOR Ki of AMB48 was 1.1 ± 0.3 nM, whereas the δOR Ki of AAH36 was 0.2 ± 0.1 nM, again suggesting that the larger 2-naphthyl group at R1 may improve δOR affinity, at least for

27 certain analogues. For compound 1 and the bulkier compound 3, neither of which had an R2

substitution, this trend did not hold true, as both peptidomimetics had a similar δOR affinity of

approximately 10 nM. This trend also did not hold true for the pair AMB42 (1.0 ± 0.4 nM) and the

bulkier AAH37 (0.8 ± 0.1 nM), or the pair AAH40 (0.5 ± 0.2 nM) and the bulkier AAH33 (1.0 ±

0.6 nM). Thus, it is inadequate to conclude that C6 substitutions and additional bulk at R1 alone

demonstrates a significant impact on δOR affinity.

δOR SAR: N1 Substitutions (R2)

The effect of N1 substitutions on δOR affinity, however, it consistent in that they increase δOR affinity. The δOR binding affinities of the unsubstituted amine analogues, such as compound 1 and compound 3, are significantly lower compared to their corresponding R2 substituted analogues. Shown in Table 1, the N-propionyl (AMB48) and N-butyryl (AMB42) substituted

analogues have δOR binding affinities of 1.1 ± 0.3 nM and 1.0 ± 0.4 nM, respectively, compared

to the unsubstituted amine analogue, compound 1, which has a δOR binding affinity of 9.4 ± 0.3

nM. Similar results were observed for the analogues in Table 2, in which the N-propionyl (AAH36)

and N-butyryl (AAH37) substituted analogues have δOR binding affinities of 0.2 ± 0.1 nM and

0.8 ± 0.1 nM, respectively, compared to the unsubstituted amine analogue, compound 3, which has a δOR binding affinity of 10 ± 2.0 nM. Moreover, as demonstrated by the peptidomimetic analogues in Table 4, regardless of the R1 substituent, analogues with R2 substitutions (i.e. mesyl

group in Table 4) consistently show a higher and even subnanomolar δOR affinity compared to

the R2 unsubstituted amine analogues.

A significant increase in δOR Ki value for R2 substituted analogues may contribute to an

increase in δOR efficacy (Table 1). Stimulation of GTPγ35S binding at δOR increases for

28 compound 2 (26 ± 3.0 %), AMB48 (65 ± 1.6 %), AMB42 (47 ± 3.3 %) and AAH40 (42 ± 6.1 %) compared to compound 1 (16 ± 2.0 %). This finding that an increase in δOR affinity translated to

an increase in δOR efficacy was not apparent in the peptidomimetic analogues (Tables 2-4), due

to bulky R1 substituents producing δOR antagonism.

μOR SAR: N1 Substitutions (R2)

All of the opioid peptidomimetic analogues in Table 1, Table 2 and Table 3 demonstrate similar

high binding affinity at the μOR. These data may be suggestive of a decrease in μOR Ki as the size of the R2 substituent increases. Shown in Table 3, the μOR Ki of the benzamide substituted

analogue, DJM45 (0.37 ± 0.11 nM), decreased compared to the unsubstituted amine analogue,

AMB39 (0.03 ± 0.01 nM). Similar results are observed in Table 2; the μOR Ki of AAH36 (0.28 ±

0.05 nM) and AAH33 (0.32 ± 0.08 nM) decreased compared to the μOR Ki of the unsubstituted

amine, compound 3 (0.08 ± 0.01 nM). However, the presence or size of the R2 substituent did not

change the μOR Ki for any analogue with the smaller benzyl group at the R1 position (Table 1).

The effect of substitutions at R2 on μOR agonism was minimal, with the majority of

analogues maintaining a similar level of agonism as the lead peptidomimetic, compound 1. The

compounds in Table 3 were observed to be full μOR agonists, despite a decrease in μOR affinity.

Peptidomimetic analogues with the mesyl group substitution at R2 were extremely potent (EC50 =

0.08 – 0.26 nM) and efficacious (96 – 110% stimulation of GTPγ35S binding compared to

DAMGO) at μOR (Table 4). Overall, substitutions at R2 did not significantly decrease μOR

potency or efficacy, and effects on binding affinity were minimal as the analogues all displayed

subnanomolar affinities that were not so different from the lead peptidomimetic.

29

Table 2. Effects of the 2-naphthyl R1 Substitution and Carbonyl Groups at R2

3 Table 2. Effects of the 2-naphthyl R1 Substitution and Carbonyl Groups at R2. Affinity (Ki) values were obtained from [ H]-diprenorphine competition binding 35 assays, and Potency (EC50) and Efficacy (% Max) values were obtained from stimulation of GTPγ S binding experiments, both using C6 cell membrane preparations expressing either μOR or δOR. All values are in nanomolar concentration (nM) and are represented as the mean ± standard error of the mean (SEM) of three separate assays performed in duplicate.* Peptidomimetic analogue 3 is in the literature as 15c from (Mosberg et al., 2013). AAH8 is in the literature as 14a in (Harland et al., 2015) and as 15b from (Harland et al., 2016). AAH36, AAH37, and AAH33 are in the literature as 15c, 15d, and 15e from (Harland et al., 2016).

30

Table 3. R1 2-methyl-THIQ Substitutions Increase μOR Agonism and R2 Substitutions Improve μOR / δOR Affinity Balance

Table 3. R1 2-methyl-THIQ Substitutions Increase μOR Agonism and R2 Substitutions Improve μOR / δOR Affinity Balance. Affinity (Ki) values were 3 35 obtained from [ H]-diprenorphine competition binding assays, and Potency (EC50) and Efficacy (% Max) values were obtained from stimulation of GTPγ S binding experiments, both using C6 cell membrane preparations expressing either μOR or δOR. All values are in nanomolar concentration (nM) and are represented as the mean ± standard error of the mean (SEM) of three separate assays performed in duplicate.* AMB39, AMB47, AFN37, DJM45, and AFN42 are in (Nastase et al., 2019) as 1D, 2D, 3D, 5D and 4D, respectively.

31

Table 4. R2 Mesyl Substitutions Increase μOR Agonism and Improve μOR / δOR Affinity Balance

3 Table 4. R2 Mesyl Substitutions Increase μOR Agonism and Improve μOR / δOR Affinity Balance. Affinity (Ki) values were obtained from [ H]-diprenorphine 35 competition binding assays, and Potency (EC50) and Efficacy (% Max) values were obtained from stimulation of GTPγ S binding experiments, both using C6 cell membrane preparations expressing either μOR or δOR. All values are in nanomolar concentration (nM) and are represented as the mean ± standard error of the mean (SEM) of three separate assays performed in duplicate. * Peptidomimetic analogue 1 is in the literature as 1 (4R) from (Mosberg et al., 2013). AFN38, AFN40, AFN42, AFN43 are in (Nastase et al., 2019) as 4A, 4B, 4D, and 4C, respectively.

32 C8 position Substitutions (R3)

To further explore the hypothesis that for analogues with a benzyl side chain, an increase in δOR affinity leads to an increase in δOR agonism, modifications were made at position C8 of the tetrahydroquinoline (THQ) core of the lead peptidomimetic analogue, (R3 substitutions; see Table

5) (Nastase et al., 2018). This series of analogues, all with a benzyl group at R1, allowed us to

examine how substituents at position C8 modulate δOR binding and in particular δOR agonism.

Analogues with alkyl substitutions at R3 produce variable levels of δOR agonism, as determined

by stimulation of GTPγ35S binding compared to DPDPE; (AFN18 = 71 ± 3.1 %), > (AFN35 = 48

± 1.6 %), > (AFN7 = 36 ± 2.6 %), > (AFN8 = 29 ± 0.7 %). These results show that an increase in the acyl chain length correlates with a decrease in the level of δOR agonism (Table 5).

In contrast to the alkyl substituted analogues, analogues with carbonyl substitutions produce δOR antagonism, again implying the larger the substituted analogue, the better chance at producing δOR antagonism. AFN20, AFN22, and AFN44 do not activate the δOR as determined by stimulation of GTPγ35S binding compared to DPDPE. Surprisingly, these three analogues

showed slightly decreased μOR agonism; (AFN22 = 73 ± 4.5 %) > (AFN20 = 71 ± 2.6 %) >

(AFN44 = 56 ± 4.9 %). This decrease in μOR agonism is independent from changes in μOR

affinity. In summary, by analyzing the SAR data from both the N1 substituted and C8 substituted

series of analogues, analogues with carbonyl moieties at R2 and R3 show high affinity δOR antagonism. The results from this study on the effects of C8 substitution lend additional support to the idea that substitutions at this region of the molecule modulate δOR binding and δOR efficacy characteristics.

33

Table 5. Bulky C8 Carbonyl Substitutions Maintain δOR Antagonist Profile

Table 5. Bulky C8 Carbonyl Substitutions Maintain δOR Antagonist Profile. Affinity (Ki) values were obtained 3 from [ H]-diprenorphine competition binding assays, and Potency (EC50) and Efficacy (Max %) values were obtained from GTPγ35S binding experiments, both using C6 cell membrane preparations expressing either μOR or δOR. All values are in nanomolar concentration (nM) and are represented as the mean ± standard error of the mean (SEM) of three separate assays performed in duplicate. * Peptidomimetic analogue 1, AFN18, AFN35, AFN7, AFN8, AFN44, AFN20, and AFN22 are from (Nastase et al., 2018) as 1, 7b, 7c, 7d, 7f, 7m, 7n, 7p, respectively.

34

Figure 5. In Vitro SAR Summary. (A) Structure of the lead peptidomimetic compound 1, the R1 substitutions at position C6 that produce μOR agonism and δOR antagonism, and the R2 substitutions at position N1 that increase δOR affinity. Structure of the bifunctional μOR agonist / δOR antagonist peptidomimetic, AAH8, which is described in the literature as 14a in (Harland et al., 2015), 15b from (Harland et al., 2016), and 2B from (Nastase et al., 2019).

In Vitro δOR Antagonism

GTPγ35S binding assays were performed using the δOR agonist DPDPE and a saturating

concentration of peptidomimetic analogue to determine if the peptidomimetics were functional

antagonists. Antagonist affinity constants (Ke) were determined for AAH8 (1.8 ± 0.1 nM), AMB46

(95 ± 17 nM), and AMB47 (4.4 ± 0.4 nM) (Table 6). In addition, All peptidomimetics produced a

right-shift in the concentration-response curve of DPDPE and were reported as functional δOR

antagonists in vitro (Bender, Griggs, Anand, et al., 2015; Harland et al., 2015). It should be noted that these Ke values are lower affinity than the Ki values from radioligand competition binding

assays, which is due to the presence of sodium (Na+) ions [100 mM] and guanine nucleotide GDP

[30 μM] in the buffer. To support the validity of the Ke data, radioligand competition binding assays were performed using the Na+ containing buffer with guanine nucleotide that is traditionally

35 used for stimulation of GTPγ S binding assays. The δOR Ki for AAH8 was 1.1 ± 0.4 nM, for

35 AMB46 it was 83 ± 9.7 nM, and for AMB47 it was 3.5 ± 0.4 nM. Therefore, whereas the δOR Ki

+ for AMB46 is 83 ± 9.7 nM in Na containing buffer (and reflects the Ke of 95 ± 17 nM), it was 15

± 5.3 nM in the standard 50 mM Tris buffer (Table 6). These data show that AAH8 and AMB47 are high affinity δOR antagonists, whereas AMB46 is a δOR antagonist with a lower affinity compared to the other two peptidomimetics, and importantly, their binding affinities in the Na+

containing buffer are similar to their Ke antagonist affinity constants. Notably, AMB46 is very similar to AMB47 at the μOR; AMB46 has a Ki of 0.12 ± 0.05 nM, an EC50 of 2.08 ± 0.91 nM,

and it is a full agonist (99 % ± 4 %).

Table 6. In Vitro δOR Antagonism of Lead Peptidomimetics

Table 6. In Vitro δOR Antagonism of Lead Peptidomimetics. Radioligand competition binding assays were performed in 50 mM Tris buffer or in buffer that included 50 mM Tris, 100 mM NaCl, 5 mM MgCl2, 1 mM EDTA, and 10 μM cold GTPγS (free nucleotide), and stimulation of GTPγ35S binding (% stimulation) experiments were performed in buffer that included 50 mM Tris, 100 mM NaCl, 5 mM MgCl2, 1 mM EDTA. All experiments were performed using C6 δOR membranes (20-40) μg membrane protein. DPDPE was the δOR agonist used in the assays intended to determine antagonist affinity constants (Ke values) for the three peptidomimetics shown above (AAH8 on the left, AMB46 in the center, and AMB47 on the right). These data are published in (Anand et al., 2018)

36 Discussion

In Vitro SAR Exploration

As was discussed earlier, the goal for elucidating the structure-activity relationship (SAR) data is to develop potent, μOR agonist / δOR antagonist analogues with subnanomolar binding affinities for both μOR and δOR. To achieve this goal a variety of chemical substitutions were made at R1

as well as at R2 and R3 of the THQ core of compound 1. The SAR data that are presented in the

five tables in the results section are simplified and summarized in Figure 5, which has been adapted

from a figure published in Nastase et al., 2019. Highlighted first is that a reduction in δOR efficacy,

or δOR antagonism, is a result of the bulky substituents at R1, such as the 2-naphthyl or 2-methyl-

THIQ. Highlighted second is that an increase in δOR affinity is a result of carbonyl substitutions at R2, such as an acetyl group. These types of substitutions at R2 were critical in producing

analogues with a subnanomolar δOR binding affinity.

In addition to increasing δOR affinity, substitutions at R2 also controlled the degree of δOR agonism, which is shown most clearly in the series of analogues with a benzyl group at R1 (Table

1 and Table 5). The SAR data from the N1 substitutions show that even a slight increase in the

size of the carbonyl group can significantly alter δOR agonism. Shown in Table 5, an increase in

the acyl chain length at position C8 correlates with a decrease in the level of δOR agonism, and

furthermore, bulky carbonyl groups produce δOR antagonism. Together these results suggest that

substitutions to the THQ core at R2 and R3 not only control δOR affinity but also δOR agonism

and that this region of the molecule is key in determining the degree of δOR binding and activity

for smaller analogues that lack a bulky substituent at R1.

R2 substituted analogues with the bulky 2-naphthyl group at R1 (Table 2) and R2 substituted

analogues with the bulky 2-methyl-THIQ at R1 (Table 3) exhibited a high δOR affinity but no δOR

37 agonism, and this was in stark contrast to R2 substituted analogues with the less bulky benzyl group at R1 (Table 1) and analogues with smaller alkyl substitutions at R3, both of which produced δOR

agonism. This agrees with the hypothesis, predicted from homology models and conformational

docking, that analogues with bulky substituents can be accommodated into the larger pocket of the inactive δOR conformation but not into the corresponding smaller pocket in the active δOR

conformation. Thus, analogues with the larger substituents, 2-naphthyl and 2-methyl-THIQ, at R1

stabilize the inactive δOR conformation to produce δOR antagonism, whereas the smaller analogues with the benzyl at R1 are able to fit into and thus stabilize the active δOR conformation

and produce δOR agonism.

These data are interesting because it appears that a combination of a bulky substituent at

R1 and any type of substituent at R2 produced high affinity δOR antagonism. It is also interesting

to note that bulky carbonyl R3 substitutions (at position C8 of the THQ) were able to produce δOR

antagonism even without a bulky substituent at R1. This suggests that modifications to the

secondary amine (N1) of the THQ may play a primary role in docking the ligand into the binding

pocket, and modifications to position 6 play a secondary role and dictate whether the analogue

produces agonism or antagonism based on the size of the substituent at R1. The combined effect

of both a bulky R1 substitution and a carbonyl R2 substitution was an improvement in δOR affinity, leading to a more balanced affinity profile. A balanced affinity profile is thought to be key because it eliminates selectivity for the μOR that is inherent in many traditional opioid analgesics. Prior dogma dictates that high affinity and highly selective μOR agonists are the suitable analgesics, and this is partially correct, as μOR agonism produces robust analgesia. However, selective μOR agonists produce side effects, and it may be that non-selective opioids are more desirable.

Moreover, a bifunctional peptidomimetic with a balanced affinity profile is able to function both

38 as a μOR agonist and as a δOR antagonist, equivalently, and this is hypothesized to be the integral

pharmacodynamic profile that equips the bifunctional peptidomimetic analogues with an improved

side-effect profile.

Overall, the results from these in vitro studies provide valuable information in regard to the optimal design of the μOR agonist / δOR antagonist pharmacological profile. 2-methyl-THIQ

substitutions at R1 (Table 3) that produce full μOR efficacy, compounded with mesyl group

substitutions at R2 (Table 4) that produce subnanomolar potency for G protein activation, led to

the analogue that incorporates both of these substituents. AFN42 produces μOR agonism and δOR

antagonism with a very high affinity for both receptors, and it exhibits a unique property that

prompted further examination into its “superagonism”, which is a term for classifying agonists that

are significantly more potent and efficacious than the standard agonist, and in this case DAMGO.

Relationship Between In Vitro SAR and In Vivo Behavioral Effects

In addition to using in vitro assays to investigate the molecular pharmacology of newly synthesized

bifunctional opioid peptidomimetics, the antinociceptive activities of the peptidomimetics were

examined in vivo. Antinociception was evaluated using the mouse warm water tail withdrawal

(WWTW) assay to investigate acute and chronic antinociceptive activity and antinociceptive tolerance of a small set of lead bifunctional opioid peptidomimetics. The behavioral evaluation of

AAH8, AMB46, and AMB47 was performed in comparison to morphine; the chemical structures and physiochemical properties of each are shown in Figure 9 and Figure 9. The in vivo experiments were performed by Dr. Jessica Anand working with Dr. Jutkiewicz and Dr. Mosberg, and the complete results are presented in Anand et al., 2018 and additionally summarized in a comprehensive review on bifunctional opioid ligands (Anand and Montgomery, 2018).

39 To measure the development of tolerance to the antinociceptive effects of the opioid

compounds, a repeated administration paradigm was used. Mice were administered either

morphine, AAH8, AMB46, and AMB47, or saline daily in a dose escalation procedure. Chronic

treatment with either AAH8 or AMB47 produced no significant shift in the dose-effect curves

(Figure 6A and Figure 6B). Thus, mice failed to develop antinociceptive tolerance to AAH8 and

AMB47 in this paradigm. Chronic treatment with both AMB46 and morphine produced a 3-fold

and significant rightward shift in the dose-effect curves (Figure 6C and Figure 6D). Thus, mice

developed tolerance to the antinociceptive effects of AMB46 and morphine. From the evaluation

of the peptidomimetics in vivo, it is likely that the δOR still plays a role in the mitigation of opioid

tolerance, as it appears δOR inhibition decreases the severity of the development of antinociceptive

tolerance. Other physiochemical factors could play a role in the variability that was observed

between the peptidomimetics; however, the cLogP values of AMB46 and AMB47 are very similar,

suggesting at least this pharmacokinetic parameter does not contribute.

From the in vitro data shown in Table 6, δOR antagonism could explain the lack of

development of tolerance that was seen with AAH8 and AMB47. The δOR affinities (Ki) and

antagonist affinity constants (Ke) of AAH8 and AMB47 are much lower than that of AMB46.

Therefore, AMB46 is a much weaker δOR antagonist compared to the other two analogues.

Although AMB46 did not stimulate GTPγ35S binding compared to DPDPE in C6 cell membranes,

+ there were discrepancies between the Ki in 50 mM Tris buffer and the Ki in 100 mM Na buffer.

Na+ ions and guanine nucleotide cause a shift in affinity, as is commonly seen with agonists

(Livingston and Traynor, 2014) and this indicates the peptidomimetics may have a degree of δOR agonism. In regard to AMB46, either its low δOR affinity or weak δOR agonism are credible factors that may have contributed to its antinociceptive tolerance and explain the in vivo results

40 (Figure 6C). On the contrary, the high δOR affinity of AAH8 and AMB47, even in the presence

of Na+ ions and guanine nucleotide, may explain their lack of antinociceptive tolerance following their chronic administration in vivo (Figure 6A and Figure 6B). Based on this comparison, it seems

logical that a greater level of δOR inhibition was responsible for the decreases in the severity of

the development of antinociceptive tolerance.

Figure 6. In Vivo Behavioral Effects of Bifunctional Opioid Peptidomimetics. Five days of chronic escalating treatment with (A) AAH8 or (B) AMB47 (10–50 mg/kg i.p., twice daily) treatment i.p. produces no shift in the dose– effect curve in wild-type BL6 mice. Five days of chronic escalating (C) AMB46 or (D) Morphine (10–50 mg/kg i.p., twice daily) treatment i.p.), but not saline produces a significant threefold rightward shift in the dose–effect curve in wild-type black 6 (BL6) mice. * P<0.05, significantly different from data from day 1. Data shown are mean ± SEM for all groups (n = 6 for each group) (Anand et al., 2018).

41 In summary, the results presented in this chapter provide continual support that bifunctional

compounds can be designed with differential binding and efficacy profiles at the μOR and δOR.

Several moieties were found to be essential in tailoring these binding and efficacy profiles to achieve our preferred pharmacodynamic profile. Based on their pharmacological profile, several compounds were identified as leads and further investigated in vivo for their ability to produce antinociception and antinociceptive tolerance in mouse models of acute pain. Such in vivo behavioral analysis led to the identification of several promising compounds with a robust and prolonged in vivo antinociceptive activity profile comparable to morphine, and notably, such compounds also demonstrated an improved side effect profile compared to morphine. Overall, the production of a large dataset documenting the affinity, potency, and efficacy parameters of these novel bifunctional opioid peptidomimetics, and the insights gleaned from analyzing the structure- activity relationships, provide valuable information for future drug discovery and development.

First, to advance the design and development of such compounds with equivalent high affinity for both the μOR and the δOR, and second, to advance the development of compounds for subsequent in vivo behavioral evaluation.

42 Methods

Materials and Reagents

The bifunctional opioid peptidomimetic analogues were synthesized using methodology described

in (Bender, Griggs, Anand, et al., 2015; Harland et al., 2015; Nastase et al., 2019). All chemicals,

unless otherwise specified, were purchased from Sigma Aldrich (St. Louis, MO). Radiolabeled

[3H]-diprenorphine (DPN) and radiolabeled guanosine‐5′‐O‐(3‐[35S]thio)-triphosphate (GTPγ35S)

were purchased from PerkinElmer Life Sciences (Waltham, MA). All tissue culture medium,

penicillin-streptomycin, geneticin (G148), trypsin, and fetal bovine serum were purchased from

Invitrogen (Carlsbad, CA).

Cell Lines and Membrane Preparation

The tissue culture and maintenance of C6 rat glioma cells stably transfected with rat μ Opioid

Receptor (μOR) or rat δ Opioid Receptor (δOR) was performed as previously described (Clark et

al., 2003). Cells were grown to confluence at 37 °C in 5% CO2 in Dulbecco’s modified Eagle

medium (DMEM) containing 10% fetal bovine serum (FBS) and 5% penicillin/streptomycin and

400 ug/mL G418. C6 membranes were prepared by washing confluent cells three times with phosphate buffered saline (pH 7.4), which were then detached using harvesting buffer (20 mM

HEPES, 150 mM NaCl, 0.68 mM EDTA, pH7.4) and pelleted by centrifugation at 200 x g for 3 min at room temperature. The pellet was resuspended in ice-cold 50mM Tris (pH 7.4) and homogenized using a Tissue Tearor (Biospec Products, Inc., Bartlesville, OK). The homogenate was centrifuged at 20000 x g at 4 C for 20 min, after which the pellet was then resuspended, homogenized, and centrifuged once more. The final pellet was resuspended in 50 mM Tris (pH

7.4) using a glass dounce homogenizer and frozen in aliquots at 80 °C. Protein concentration was

43 determined by performing a BCA protein assay (Thermo Scientific Pierce, Waltham, MA) using

bovine serum albumin as the standard.

Radioligand Binding Assays

[3H]-diprenorphine (DPN) competition binding assays were performed in 500μL reactions in 96- well plates using the cell membrane homogenates described above. Briefly, cell membranes [20-

40 μg] were incubated in 50 mM Tris buffer containing 0.2 nM [3H]-DPN and various concentrations of peptidomimetic analogue for 1 hour in a shaking water bath at room temperature.

Nonspecific binding was determined using 10 µM naloxone. The reaction was terminated by vacuum filtration onto glass microfiber GF/C filters using a tissue harvester (Brandel,

Gaithersburg, MD) and washed with 50 mM Tris buffer. Filters were dried and, following the addition of EcoLume scintillation cocktail, bound radioactivity was quantified in a Wallac 1450

MicroBeta Liquid Scintillation and Luminescence Counter (PerkinElmer, Waltham, MA).

Radioligand binding assays in Na+ containing buffer were performed in 200 μL reactions in 96-

well plates using the cell membrane homogenates described above Briefly, cell membranes [20-

40 μg] were incubated in Na+ containing assay buffer (50 mM Tris, 100 mM NaCl, 5 mM MgCl2,

1 mM EDTA, pH 7.4) containing 10 μM cold GTPγS, and various concentrations of

peptidomimetic analogue for 90 minutes in a shaking water bath at room temperature. Nonspecific

binding was determined using 10 µM naloxone. The reaction was terminated by vacuum filtration

onto glass microfiber GF/C filters using a tissue harvester and washed with Na+ containing buffer,

and the filters were processed as described above.

44 Stimulation of GTPγ35S Binding Assays

Stimulation of GTPγ35S binding experiments were performed in 200μL reactions in 96-well plates

using cell membrane homogenates prepared as described above. Briefly, cell membranes [10 μg]

were incubated in Na+ containing buffer [50 mM Tris, 100 mM NaCl, 5 mM MgCl2, 1 mM EDTA,

pH 7.4] containing 0.1 nM guanosine‐5′‐O‐(3‐[35S]thio)triphosphate (GTPγ35S) and 30 μM

guanosine 5’-diphosphate (GDP) and various concentrations of peptidomimetic analogue for 1

hour in a shaking water bath at 25 °C. Basal stimulation of GTPγ35S binding was determined by

incubation in the absence of any ligand and maximal stimulation of was defined by using 10 μM

of the standard opioid agonist compounds, DAMGO ([D-Ala2,N-MePhe4,Gly-ol]-enkephalin) for

μOR, DPDPE (D-Pen2,5-enkephalin) for δOR, or SNC80 [(+)-4-[(αR)-α-((2S,5R)-4-Allyl-2,5-

dimethyl-1-piperazinyl)-3-methoxybenzyl]-N,N-diethylbenzamide] for δOR. The reaction was

terminated by vacuum filtration onto glass microfiber GF/C filters using a tissue harvester and

washed with Na+ containing buffer, and filters were processed as described above.

35 For antagonist affinity (Ke) determination, stimulation of GTPγ S binding by SNC80 and by

SNC80 in the presence of a single concentration of an δOR antagonist was measured in C6 δOR

membranes. The difference between the EC50 of SNC80 alone and the EC50 of SNC80 in the presence of the δOR antagonist represents the shift in concentration response. The reaction was terminated by vacuum filtration onto glass microfiber GF/C filters using a tissue harvester and washed with Na+ containing buffer, and filters were processed as described above. The antagonist affinity constant was calculated as Ke = (concentration of δOR antagonist) / (concentration- response shift − 1).

45 In Vivo Antinociception Assays

Antinociceptive effects were evaluated in the mouse warm water tail withdrawal (WWTW) assay.

Withdrawal latencies were determined by briefly placing a mouse into a cylindrical plastic restrainer and immersing 2−3 cm of the tail tip into a water bath at 50 °C. The latency to tail withdrawal was recorded with a maximum cutoff time of 20 s to prevent tissue damage.

Antinociceptive effects were determined using a cumulative dosing procedure. Each mouse received an injection of saline i.p., and then 30 min later baseline withdrawal latencies were recorded. Following baseline determinations, cumulative doses of each test compound (1, 3.2, and

10 mg/kg) were administered i.p. at 30 min intervals. After 30 min of each injection, tail withdrawal latency was measured, as described above. To determine the duration of

antinociception, baseline latencies were determined after which animals were administered a 10

mg/kg bolus injection of test compound i.p. Latency to tail withdrawal was then determined at 5,

15, and 30 min after injections and every 30 min thereafter until latencies returned to baseline

values. For chronic escalating treatment studies, mice were treated for five days with opioid drug

(10-50 mg/kg, twice daily) and on day six cumulative doses of each test compound (1, 3.2, and 10

mg/kg) were administered i.p. at 30 min intervals, and after 30 min of each injection, tail

withdrawal latencies were measured, as described above.

Data Analysis

Graphpad Prism software (San Diego, CA) was used for all statistical analysis, and data were

graphed as individual experiments for analyses unless otherwise stated. Binding affinity (Ki)

values were calculated using the Cheng-Prusoff equation via nonlinear regression analysis from at

least three separate binding assays performed in duplicate. Potency (EC50) and relative efficacy (%

46 stimulation compared to a standard opioid agonist) was determined via nonlinear regression analysis from at least three separate binding assays performed in duplicate. For in vivo antinociception assays, ED50 values were also calculated using nonlinear regression analysis.

47 Appendix to Chapter II

Interaction between μOR and δOR at the single cell level

Introduction

Various preclinical studies have shown that the administration of a δOR antagonist in

combination with morphine reduces the degree of morphine tolerance and dependence

(Abdelhamid et al., 1991; Hepburn et al., 1997; Martin et al., 2000; Chefer and Shippenberg,

2009). In addition to these reports that assessed the pharmacological effects of the administration

of δOR-selective antagonists on the development of morphine tolerance and dependence in vivo, there have been several studies that examined receptor expression in vivo and the data show that

µOR and δOR populations are localized to the same neurons (Wang et al., 2010; Erbs et al., 2015).

A recent study determined that both μORs and δORs are expressed in neurons of the lateral parabrachial nucleus, amygdala, anterior cingulate cortex, pontine nucleus, and lateral reticular nucleus (D Wang et al., 2018). There have been numerous reports on the hetero-dimerization of

μOR and δOR; some reports have suggested that the μOR-δOR heterodimer exhibits distinct signaling properties compared to the monomeric opioid receptors (Gomes et al., 2000, 2011;

George et al., 2002) and that activation of δOR in μOR-δOR complexes on the surface of dorsal root ganglion neurons leads to a internalization and codegradation of both receptors (He et al.,

2011). Collectively, these findings propose a role for δOR in the regulation of μOR effects and provide evidence for a direct interaction between the two receptor types in certain cell types,

further supporting preliminary evidence for molecular ‘cross-talk’ between the µOR and the δOR.

48 One prevailing hypothesis is that δOR serves as a negative regulator of μOR signaling, and that δOR inhibition would afford an increase in ligand potency and efficacy at the μOR. There are some published in vivo data that support this hypothesis, in which the administration of the δOR antagonist naltriben (NTB) significantly enhanced the potency of the μOR agonist [Dmt1]-

DALDA in opioid tolerant mice (Zhao, 2003). Additional studies have reported that δOR antagonists increase the potency and efficacy of μOR agonists in cells co-expressing both receptors and that there is positive cooperativity between the two receptor types (Gomes et al., 2000; Devi et al., 2004). Our hypothesis is that δOR functions as a regulator of μOR signaling at the level of the single cell, and in addition, that the inhibition of δOR could reduce the development of μOR tolerance. SH-SY5Y human neuroblastoma cells were used, and this is an ideal model system because they have native expression levels of µORs and δORs and also a physiological ratio of receptor to G protein and other intracellular proteins involved in signal transduction processes

(Elliott, Guo and Traynor, 1997).

Methods

To selectively target both μOR and δOR to isolate the specific contribution from each receptor toward the total signaling output, the standard agonist DAMGO was used for activating

μOR at a concentration of 100 nM (Traynor and Nahorski, 1995), and to selectively target δOR, the δOR-selective inverse agonist ICI 174,864 was chosen because of its very high selectivity for

δOR over μOR and used for inhibiting δOR at a concentration of 1000 nM (Szekeres and Traynor,

1997). This allowed for the study of the µOR-agonist tolerance effect using DAMGO and an investigation into the role of the δOR in regulating μOR tolerance at the single cell level. SH-

SY5Y cells were grown to confluence and harvested into membrane preparations, as previously

49 described (Alt et al., 2001). Stimulation of GTPγ35S binding by DAMGO in SH-SY5Y cell membranes was first measured in the absence or presence of the δOR-selective antagonist ICI

174,864 [1000 nM] to determine the potency and maximal stimulation of DAMGO.

Results

The EC50 from the concentration response curve for DAMGO alone (117 ± 6 nM) was not

different from the EC50 from the curve for DAMGO in the presence of 1000 nM ICI 174,864 (139

± 21 nM) (Figure A1A; dashed line). These data demonstrate that there is no direct interference to

the ability of DAMGO to activate the μOR in the presence of ICI 174,864. SH-SY5Y cells were

then treated with ICI 174,864 for 24 hours prior to the harvesting of the cell membranes, after

which stimulation of GTPγ35S binding was performed using DAMGO (Figure A1B). Again, there

was no significant alteration in the concentration-response curve of DAMGO (EC50 = 132 ± 21

nM). Thus, cells treated with ICI 714,864 for 24 hours show the same agonist response to DAMGO

in the ability to stimulate GTPγ35S binding as vehicle-treated cells. Next, SH-SY5Y cells were treated with DAMGO [100 nM] for 24 hours, to induce μOR tolerance (Figure A1C). As expected, there was a 50% reduction in the response to DAMGO; stimulation of GTPγ35S was 51 ± 3.8 %

compared to vehicle treated cells, as well as a rightward shift in the concentration response (EC50

= 289 ± 57 nM). Finally, SH-SY5Y cells were co-treated with DAMGO and ICI 174,864 for 24

hours prior to the harvesting of cell membranes. There was a similar reduction in the response by

DAMGO (53 ± 7.2 % compared to vehicle treated cells) and an EC50 of 302 ± 85 nM (Figure

A1D). Thus, ICI 174,864 failed to rescue DAMGO-mediated μOR tolerance. In summary, these

results suggest that chronic administration of ICI 174,864 and inhibition of the δOR in SH-SY5Y

cells did not modify the functional ability of μOR to bind agonist and activate G protein signaling.

50 Discussion

The data from these experiments indicate that δOR inhibition does not play a role in μOR tolerance at the level of the single cell. There are some caveats to this study and the results should be interpreted accordingly. First, G protein activation was the exclusive measure of receptor activation, and further downstream signaling could have been evaluated to determine if chronic inhibition of δOR prevented μOR-mediated tolerance. Preliminary experiments using CHO-Flag-

μOR-Myc-δOR cell membranes produced data similar to those in Figure A1.

One recent study found that few spinal nociceptive neurons co-express μOR and δOR, and the two receptors internalize and function independently when they are co-expressed in certain neurons (Dong Wang et al., 2018). This report supports the results of my experiments. A reasonable conclusion is that δOR antagonism and its beneficial effect on μOR (i.e. reduction in the development of opioid tolerance) may be at the systems level and not at the level of the single cell. Indeed, earlier studies suggested an indirect interaction between these two receptors, either at the intracellular level or systems level (Traynor and Elliott, 1993).

Thus, questions still remain in regard to how the beneficial effects of δOR inhibition on

μOR signaling result, whether they are from indirect or direct interactions between μOR and δOR located on separate neurons or from intracellular processes at the single cell level. Could a δOR antagonist such as ICI 174,864 modulate intracellular processes such as μOR trafficking or the recruitment of arrestin proteins? Certainly, there are many intracellular processes that work in tandem to first produce and then counteract opioid tolerance and other opioid-mediated effects. In summary, the molecular mechanisms that drive opioid tolerance are not clearly defined, and it is still unknown exactly how the development of opioid tolerance is seemingly attenuated via δOR antagonism in rodent models.

51

Figure A1. Inhibition of δOR Does Not Prevent DAMGO-mediated µOR Tolerance in SH-SY5Y cells. Stimulation of GTPγ35S binding was performed using SH-SY5Y cell membranes and various concentrations of DAMGO. Cell membrane homogenates were prepared from cells that were (A) untreated, (B) treated with ICI 174,864 [1000 nM] for 24 hours, (C) treated with DAMGO [100 nM] for 24 hours, or (D) cells that were treated with DAMGO [100 nM] in combination with ICI 174,864 [1000 nM] for 24 hours. Experiments were performed in quadruplicate (n ≥ 4) and data points are represented as mean ± SEM.

52 Chapter III

Agonism of Bifunctional Opioid Peptidomimetics

Introduction

Efficacy was first defined to discriminate partial agonists from full agonists, antagonists,

and inverse agonists (Stephenson, 1956), where different drugs exhibit varying capacities to

produce a response. The efficacy of an agonist acting at a GPCR describes its ability to shift an

unoccupied inactive receptor state (R) or occupied inactive receptor state (DR) to any number of

active receptor states (DR*) (Ehlert, 2018). Every molecule that binds to a receptor has an ability

to stabilize a unique conformation along a spectrum of receptor states. Indeed, the idea that one agonist may stabilize one active state receptor conformation (DR*), while another agonist may stabilize an entirely different active state receptor conformation (DR**), is known in the literature as biased agonism (Rankovic, Brust and Bohn, 2016). Proteins are not static, so the inactive receptor state (R) spontaneously transitions to an active state (R*) on occasion; this is referred to as constitutive activity. Efficacy not only represents the ratio of the amount of active occupied receptor states (DR*) to the total number of occupied receptor states, but it also takes into consideration the time that the active state of a receptor is occupied by an agonist. Thus, efficacy encompasses the pharmacodynamics of a ligand and its biological activity at the single-receptor level and also its physiological and behavioral response over time, and thus is a rather challenging parameter to assess. Efficacy measurements from G protein activation assays (i.e. stimulation of

GTPγ35S binding) are dependent on and can change in accordance with the different cell systems

53 and experimental conditions. For example, modifications to the protocol in which the buffer

components, incubation time, cell lines, and temperature can vary, and all influence signaling

output and artificially alter the apparent efficacy. To circumvent some of these limitations, a

standard agonist is incorporated into each stimulation of GTPγ35S binding assay and thus a

determination of ‘relative efficacy’ can be achieved.

The relative efficacy of a compound can be estimated by comparing affinity from

radioligand competition binding assays determined under different experimental conditions (Pert

and Snyder, 1976) It is known from early studies that Na+ ions and guanine nucleotide decreases

the binding affinity (Ki) of opioid agonists and that there is a strong correlation between the

+ efficacy of an opioid agonist and the shift in Ki of the agonist by Na ions and guanine nucleotide

(Lee et al., 1999). In particular, higher efficacy ligands show greater shifts in affinity upon the addition of Na+ ions and guanine nucleotide, and conversely, antagonist compounds show no

affinity shift (Livingston and Traynor, 2014). Radioligand binding assays performed in the absence and presence of Na+ ions and guanine nucleotide afford the determination of an affinity shift, which represents an appropriate measure of efficacy despite the absence of any G protein signaling output

(Livingston et al., 2018). Measuring affinity shift is an acceptable approach to quantify efficacy

while also addressing the problem of signaling bias since it does not depend on a GPCR signaling

35 output. From the potency (EC50) and relative efficacy (maximal stimulation of GTPγ S binding

compared to a standard agonist) derived from concentration-response curves, a ( ) can be 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 derived (Kenakin, 2017). This logarithmic measurement is valuable because it𝐿𝐿𝐿𝐿𝐿𝐿 reduces relative

efficacy down to a single number. Assuming certain qualifications are met (e.g. Hill coefficient of

one for a given concentration-response curve), the ( ) of a standard compound (in our case 𝑀𝑀𝑀𝑀𝑀𝑀 𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50 ( ) ( DAMGO) is subtracted from the of a test compound. This produces the 50) 𝑀𝑀𝑀𝑀𝑥𝑥 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 ∆ 𝐸𝐸𝐸𝐸 𝐿𝐿𝐿𝐿𝐿𝐿 54 𝐿𝐿𝐿𝐿𝐿𝐿 that accurately reflects the efficacy and affinity of a test compound in a system independent

( manner. System effects are effectively cancelled, thus 50) reflects an ‘index of agonism’ 𝑀𝑀𝑀𝑀𝑀𝑀 ∆ 𝐸𝐸𝐸𝐸 as discussed by Kenakin (Kenakin, 2017). 𝐿𝐿𝐿𝐿𝐿𝐿

In this chapter, several methods were used to evaluate agonism in a set of opioid

compounds. The extraordinarily high potency of AFN42 at the μOR, which was found to be 1000

times more potent than DAMGO at stimulating GTPγ35S binding, prompted a closer look into the

pharmacodynamic properties of several of the peptidomimetics; these are AAH8, AMB46,

AMB47, AMB67, and AFN42 and they are shown in Figure 8Figure 8. These peptidomimetics

were also chosen due to differences in their physiochemical properties, including molecular

weight, topological Polar Surface Area (tPSA), and partition coefficient between n-octanol and water (cLogP). DAMGO, fentanyl, morphine, and two high efficacy compounds, ( and

BU72), were selected for comparison with the peptidomimetics (Figure 9Figure 9). Radioligand

competition binding assays were performed in the absence or presence of Na+ ions and guanine

nucleotide to measure agonist affinity (Ki) shifts, and experiments were also performed using three different μOR expressing cell lines (C6, CHO, and SH-SY5Y) to further define the pharmacology of bifunctional opioid peptidomimetics. Stimulation of GTPγ35S binding assays were performed

to measure the relative efficacy of the peptidomimetics compared to DAMGO, after which

( 50) values were calculated to determine the degree of agonism for the opioid compounds. 𝑀𝑀𝑀𝑀𝑀𝑀 ∆ 𝐸𝐸𝐸𝐸 Finally,𝐿𝐿𝐿𝐿𝐿𝐿 since efficacy can vary with downstream signaling measures, we compared the ability of the peptidomimetics to activate G protein versus β-arrestin pathways. This also allowed determination of a bias factor (i.e. preference for G protein signaling over β-arrestin signaling, or vice versa.

55 For many years, our understanding of opioid pharmacology was that the effects of opioid-

mediated μOR activation were mediated by the activity of the heterotrimeric G protein subunit and

its target proteins, such as adenylyl cyclase and calcium and potassium ion channels (Al-Hasani

and Bruchas, 2011). Following discoveries by Dr. Laura Bohn and colleagues, consensus on the

extent of μOR signaling has changed (Bohn et al., 1999; Raehal, Walker and Bohn, 2005). In

experiments using β-arrestin 2 knockout mice, there was a significant increase in the potency of

the antinociceptive effect of morphine, a reduction in tolerance development, and a significant

decrease in side effects associated with morphine administration, including constipation and respiratory depression. This led to a proposed model that certain opioid effects, including constipation and respiratory depression, are a result of signaling via a β-arrestin pathway, as opposed to a G protein pathway. Therefore, the selective activation of the G protein component of opioid receptor signaling may be preferable to the activation of the β-arrestin signaling component, in regard to the clinical efficacy and safety of opioids.

Since these discoveries, opioid ligands have been designed to selectively activate the G protein signaling component of the μOR. These ligands are termed “G protein-biased agonists” and have been in both preclinical and clinical development. TRV130, also known as

(Figure 7A), was the first example of a G protein-biased opioid agonist. In animal models, TRV130 produced antinociception with significantly less respiratory depression and constipation compared to morphine (DeWire et al., 2013). Although these preclinical data were considered compelling, other data showed that the drug produced abuse-related reinforcing effects following its prolonged administration in various animal models (Altarifi et al., 2017; Austin Zamarripa et al., 2018).

Results were reported in phase I trials were considered appropriate (Soergel et al., 2014), however, the phase 3 trial results were unsatisfactory (S., K. and F., 2018). The FDA declined to approve

56 Oliceridine for the treatment of moderate to severe pain due to inadequate safety data supporting

its proposed dosing and concerns that higher doses could lead to QT prolongation in patients.

Figure 7. Chemical Structures of G protein-biased Agonists. (A) TRV130 (DeWire et al., 2013) (B) PZM21 (Manglik et al., 2016) (C) SR17018 (Schmid et al., 2017).

Despite the initial failure of TRV130 and current debate about the promise of G protein-

biased agonists, they have been favorably reviewed, (Siuda et al., 2017) and numerous other efforts are ongoing. For example, PZM21 (Figure 7B) demonstrated G protein bias over β-arrestin recruitment and also showed promising in vivo results (Manglik et al., 2016). However, PZM21 demonstrated robust respiratory depression effects in a dose-dependent manner in additional experiments (Disney et al., 2018). Furthermore, tolerance developed to the antinociceptive effect of PZM21 over the course of three days but not to the respiratory depression effects, raising questions about its potential clinical safety (Disney et al., 2018). The negative data thus far for

TRV130 and PZM21 could be due to unknown factors that are separate from their G protein-bias properties, or perhaps the compounds did not exhibit enough of a degree of separation between G

57 protein- and β-arrestin signaling pathways to achieve a clinically significant outcome. Other G

protein-biased agonists are currently being evaluated in preclinical studies (Váradi et al., 2016;

Bohn and Aubé, 2017; Schmid et al., 2017; Kennedy et al., 2018). In one study SR17018 (Figure

7C) showed a much greater degree of G protein- and β-arrestin separation (Schmid et al., 2017).

This study was important for its thorough evaluation of SAR data to produce bias factors for a

number of novel analogues of SR17018, which will serve as valuable pharmacological tools to

understand the complex nature of biased agonism at the opioid receptors.

Due to AFN42 demonstrating an exceptional potency and superagonism for μOR-mediated activation of G proteins, we wondered if it would be biased. Also due to several of the bifunctional opioid peptidomimetics producing mixed results in vivo, we wondered if biased agonism could explain the discrepancies between AAH8 and AMB46 in regard to the development of antinociceptive tolerance. Therefore, we selected a few peptidomimetics to investigate their relative potential for β-arrestin recruitment as compared to G protein activation. β-arrestin recruitment assays were performed using CHO cells that expressed either μOR or δOR (see

Methods). The endogenous peptide Met-enkephalin was chosen as the reference ligand to which data were normalized, and BU72 and morphine were also examined for comparison to the peptidomimetics. The efficacy (maximal recruitment of β-arrestin compared to the standard agonist, Met-enkephalin), potency (EC50), and subsequently ( ) were calculated from the 𝑀𝑀𝑀𝑀𝑀𝑀 𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50 concentration-response curves of each ligand at μOR or δOR. ∆∆ ( ) and the antilog ‘bias 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 factors’ were calculated and plotted for each ligand at μOR as well as𝐿𝐿𝐿𝐿𝐿𝐿 δOR. Additionally, β-arrestin

recruitment assays were performed in the presence of pertussis toxin (PTX) to determine the

requirement of G proteins in recruitment of β-arrestin proteins to the opioid receptor and also to

probe the degree of G protein-independent signaling at μOR following its activation by the

58 bifunctional opioid peptidomimetics. The objective of this study is to determine if there is a relationship between the pharmacodynamic properties of the peptidomimetics and the in vivo findings.

Figure 8.Chemical Structures and Physiochemical Properties of a Small Series of Bifunctional Opioid Peptidomimetics. The chemical structure, molecular weight, topological Polar Surface Area (tPSA), and partition coefficient (cLogP) are provided for each of the following bifunctional opioid peptidomimetics; (A) AAH8 (B) AMB46 (C) AMB47 (D) AMB67 and (E) AFN42.

59

Figure 9. Chemical Structures and Physiochemical Properties of Several Classical and Highly Efficacious Opioid Agonists. The chemical structure, molecular weight, topological Polar Surface Area (tPSA), and partition coefficient (cLogP) are provided for the following opioid agonists; (A) DAMGO, (B) Fentanyl, (C) Morphine, (D) BU72 and (E) Etorphine.

60 Results

Stimulation of GTPγ35S Binding in C6 μOR Cell Membranes

GTPγ35S binding assays were performed to measure the relative efficacy of the bifunctional opioid

peptidomimetics compared to DAMGO at μOR. Stimulation of GTPγ35S binding in C6 μOR

membranes by the five bifunctional peptidomimetics was highly potent (EC50 range from 0.08 nM for AFN42 to 2.08 nM for AMB46). These values are at least 100 times more potent compared to the stimulation of GTPγ35S binding by fentanyl, morphine, and the standard agonist, DAMGO

(EC50’s = 132 – 290 nM). From the concentration-response curves and data shown in Figure 10A,

the peptidomimetics demonstrated potencies that were similar to those seen with the highly potent

opioid ligands BU72 (0.34 ± 0.1 nM) and etorphine (2.61 ± 0.08 nM).

In addition to their high potencies, the peptidomimetics also demonstrated high relative

efficacies, and fractional maximal responses2 (Max) as compared to DAMGO are described as follows. AMB46 (Max = 0.99 ± 0.04 nM) and AMB47 (Max = 1.00 ± 0.03 nM), as compared to

BU72 (Max = 1.03 ± 0.02 nM) and etorphine (Max = 0.99 ± 0.03 nM), all produced the maximal response as shown with DAMGO (Max = 1.02 ± 0.01 nM) in the stimulation of GTPγ35S binding

assay. However, AAH8 (Max = 0.89 ± 0.02 nM), > AMB67 (Max = 0.80 ± 0.07 nM), > fentanyl

(Max = 0.77 ± 0.05 nM), > morphine (Max = 0.60 ± 0.02 nM) produced sub-maximal response compared to DAMGO. AFN42 exhibited the greatest maximal response in the stimulation of

35 GTPγ S binding assay (Max = 1.10 ± 0.05 nM) and had a very high potency (EC50 = 0.08 ± 0.0

nM) in comparison to the other opioid agonists and bifunctional peptidomimetics at μOR. Relative

agonism parameters were determined for each opioid agonist (Figure 10B). The relative agonism for AFN42 (10.23 ± 0.24) was greater than that of DAMGO (6.89 ± 0.03). Since these parameters

2 Stimulation of GTPγ35S binding data (% of DAMGO) were converted to a ‘fractional maximal response’ whereby the reference agonist DAMGO = 1.0 61 are on a logarithmic scale, the difference in relative agonism between AFN42 and DAMGO (10.23

– 6.89 = 3.33) represents an over 2,000-fold difference (2,137-fold). The data for each opioid in the series were normalized to DAMGO, whereby the relative agonism of DAMGO was subtracted from that of each opioid agonist to calculate a ( ) (Table 7). With the ( ) of 𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 𝐸𝐸𝐸𝐸50 fentanyl at –0.43 and morphine at –0.40, the agonism∆𝐿𝐿𝐿𝐿𝐿𝐿 of these two compounds is observably∆𝐿𝐿𝐿𝐿𝐿𝐿 less

than that of DAMGO. Conversely, the relative agonism of AFN42 (3.33), > BU72 (2.62), > AMB47

(2.38), > AAH8 (2.29), > AMB46 (2.07), > etorphine (1.82), > AMB67 (1.76). These compounds

show at least a 100-fold greater degree of relative agonism compared to DAMGO (Figure 10B)

Figure 10. Relative Efficacy of Opioid Agonists and Peptidomimetics at μOR. (A) Stimulation of GTPγ35S binding at the μOR was performed using membranes from C6 cells with various concentrations of opioid agonists. Maximal stimulation of GTPγ35S binding was produced by 10 μM of the reference agonist, DAMGO. (B) Max (calculated as a fractional response compared to 10 μM DAMGO, from concentration-response curves) and EC50 values were determined from concentration-response curves for the opioid agonists DAMGO, fentanyl, morphine, etorphine, and BU72, and for the bifunctional opioid peptidomimetics AFN42, AMB47, AMB46, AAH8, and AMB67. The relative efficacy of each opioid agonist is represented in a bar graph and reduced to a single number via the calculation of Log (Max/EC50) from the data from (A).

62 Table 7. Relative Efficacy, Potency, and an Index of Agonism of Opioids at μOR

Table 7. Relative Efficacy, Potency, and an Index of Agonism of Opioids at μOR. Stimulation of GTPγ35S binding at the μOR was performed using membranes from C6 cells with various concentrations of opioid agonists. Maximal stimulation of GTPγ35S binding was determined by 10 μM of the reference agonist, DAMGO. 35 The fractional maximal response of agonist-mediated stimulation of GTPγ S binding (Max) and effective concentrations (EC50) were determined by nonlinear regression analysis of the concentration-response curves shown in Figure 10. The Max and EC50 values were used to calculate an index of agonism [ ( )] 𝑀𝑀𝑀𝑀𝑀𝑀 that was normalized to DAMGO (i.e. ∆Log (Max/EC50)) for each opioid compound. 𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50

63 Affinity Shifts at the μOR in C6 μOR Cell Membranes

Radioligand competition binding assays were performed in the absence or presence of Na+ ions

and guanine nucleotide to measure agonist affinity (Ki) shifts. DAMGO (Ki = 5.63 ± 1.33 nM), fentanyl (5.54 ± 1.70 nM), and morphine (3.25 ± 1.01 nM) showed an approximately 10-fold lower

μOR affinity compared to etorphine (0.48 ± 0.11 nM) and BU72 (0.21 ± 0.04 nM) when [3H]-

diprenorphine binding was performed using C6 μOR membranes in 50 mM Tris buffer (Table 8).

The peptidomimetics all exhibited μOR binding affinities comparable to etorphine and BU72, with

AAH8 (0.04 ± 0.01 nM) demonstrating the highest μOR affinity in the set of opioid agonists.

Subsequently, [3H]-diprenorphine binding was performed in Na+ containing buffers for the set of

opioid agonists to obtain affinity shift data. represents the shift in affinity of a ligand 100 mM Na+ 50 mM Tris + � � by GTPγS [10 μM] and Na ions [100 mM], whereas represents the shift in affinity 300 mM Na+ 50 mM Tris by GTPγS [10 μM] and Na+ ions [300 mM]. � �

As shown in Table 8, the binding affinities of DAMGO (Ki = 754 ± 130 nM), fentanyl (726

± 27.5 nM), and morphine (438 ± 98.9 nM) for μOR all decrease in the presence of nucleotide and

[100 mM] Na+ ions, producing shifts in affinity of at least 100-fold. There is also a decrease in

μOR affinity and at least a 10-fold shift in affinity for etorphine (8.21 ± 2.77 nM; 17-fold shift),

AMB67 (3.42 ± 1.21 nM; 23-fold shift), AAH8 (0.66 ± 0.21 nM; 15-fold shift), and AMB46 (1.74

± 0.47 nM; 14.6). Moreover, a smaller shift in affinity was observed for AMB47 (0.51 ± 0.15 nM;

4-fold shift) and BU72 (0.42 ± 0.09 nM; 2-fold shift), however, there was no shift in affinity in

Na+ [100 mM] and GTPγS [10 μM] containing buffer for AFN42 (0.12 ± 0.02 nM; 1.2-fold shift),

emphasizing that AFN42 is insensitive to Na+ ions and guanine nucleotide.

The opioid antagonist diprenorphine (DPN) demonstrated no shift in affinity under these

different buffer conditions (Table 8). The shifts in affinity in the presence of [300 mM] Na+ and

64 [10 μM] GTPγS followed a similar trend, whereby the fold shift for DAMGO (472-fold) >

morphine (255-fold) > fentanyl (154-fold) > etorphine (50-fold) = AAH8 (50-fold) > AMB46 (24-

fold) = AMB47 (20-fold) > BU72 (5.9-fold) = AFN42 (5.3-fold) (Table 8). BU72 and etorphine, which were determined to be relatively insensitive to Na+ ions, show a reduced response to positive

allosteric modulators that function by allosteric displacement of Na+ ions (Livingston and Traynor,

2014). It was predicted therefore that AFN42 would also be insensitive to the positive allosteric

modulator, BMS-986122. Thus, [3H]-DPN competition binding was performed in Na+ [100 mM]

buffer in the presence of 10 μM BMS-986122 (Figure 11C). The Ki of AFN42 did not change in

the presence of BMS-986122, supporting this prediction and also adding further support that

AFN42 is relatively insensitive to the effect of sodium.

The Ki data shown in Table 8 were determined from the concentration-effect curves,

provided in Figure 12 and Figure 12. To determine accurate Ki values and by extension produce

accurate affinity shift data, a precise Kd of diprenorphine was obtained from saturation binding

experiments in the various buffers. The Kd of diprenorphine at μOR was determined to be 0.15 nM

in buffer that included 50 mM Tris [data not shown], 0.30 nM in buffer that included 50 mM Tris,

100 mM NaCl, 5 mM MgCl2, 1 mM EDTA, 10 µM GTPγS (a representative experiment is shown

in Figure 12M), and 0.50 nM in buffer that included 50 mM Tris, 300 mM NaCl, 5 mM MgCl2, 1

mM EDTA, and 10 µM GTPγS [data not shown]. To rule out any potential effect of Mg2+ ions on affinity shifts and to confirm that the effect was due to Na+ ions and GTPγS, [3H]-diprenorphine

binding was performed with the agonist DAMGO in buffer that included Mg+ ions but excluded

+ + Na ions and GTPγS. In buffer that only included Mg ions there was no change in the Ki of

DAMGO compared to the affinity of DAMGO in 50 mM Tris buffer (Figure 12L).

65 + Table 8. Binding Affinity (Ki) of Opioid Ligands and the Affinity Shift by Na ions and Guanine Nucleotide at μOR

+ Table 8. Binding Affinity (Ki) of Opioid Ligands and the Affinity Shift by Na ions and Guanine Nucleotide at μOR. To measure the shifts in receptor affinity, [3H]-diprenorphine competition binding assays were performed using 0.2 – 0.5 nM radioligand and the following buffers: 50 mM Tris buffer (column 1), 50 mM Tris, 100 mM NaCl, 5 mM MgCl2 1 mM EDTA, and 10 μM GTPγS (column 2) or 50 mM Tris, 300 mM NaCl, 5 mM MgCl2 1 mM EDTA, and 10 μM GTPγS (column 3). Column 4 represents the affinity shift by 100 mM Na+ buffer and column 5 represents the affinity shift by 300 mM Na+ buffer. Data are represented as mean ± SEM and all experiments were performed in duplicate or quadruplicate (n ≥ 3).

66

Figure 11. Inhibition of [3H]-diprenorphine Binding by Opioid Ligands at μOR. Radioligand competition binding assays were performed using 0.2 – 0.5 nM [3H]-diprenorphine, 20 μg membrane protein, and various concentrations opioid ligands; (A) DAMGO (B) BU72 (C) AFN42 (± 10 μM BMS-986122) and (D) AAH8, in the following buffers; 50 mM Tris (circles), 50 mM Tris, 100 mM NaCl, 5 mM MgCl2 1 mM EDTA, 10 μM GTPγS (squares), or 50 mM Tris, 300 mM NaCl, 5 mM MgCl2 1 mM EDTA, 10 μM GTPγS (triangles). Data are represented as mean ± SEM and all experiments were performed in duplicate or quadruplicate (n ≥ 3).

67

Figure 12. Inhibition of [3H]-diprenorphine Binding by Opioid Ligands at μOR (cont’d). Radioligand competition binding assays were performed using 0.2 – 0.5 nM [3H]-diprenorphine, 20 μg membrane protein, and various concentrations opioid ligands; (E) etorphine (F) fentanyl or (G) morphine (H) AMB46 (I) AMB47 (J) AMB67 and (K) diprenorphine in the following buffers; 50 mM Tris (circles), 50 mM Tris, 100 mM NaCl, 5 mM MgCl2 1 mM EDTA, 10 μM GTPγS (squares), and 50 mM Tris, 300 mM NaCl, 5 mM MgCl2 1 mM EDTA, 10 μM GTPγS (triangles). (L) Radioligand competition binding assays were performed using DAMGO in 50 mM Tris buffer that excluded Na+ ions and guanine nucleotide but included Mg2+ ions [5 mM]. (M) Representative figure of a saturation binding experiment, performed using C6 μOR membranes in buffer that included 50 mM Tris, 100 mM NaCl, 5 mM MgCl2, 1 mM EDTA, and 10 µM GTPγS. Saturation binding was also performed in C6 δOR membranes using the 100 mM Na+ buffer, in C6 μOR and C6 δOR membranes using 50 mM Tris buffer, and in C6 μOR and C6 δOR membranes using the 300 mM Na+ buffer [data not shown]. Data are represented as mean ± SEM and all experiments were performed in duplicate or quadruplicate (n ≥ 3).

68 Agonism Correlation Plots

This series of opioid peptidomimetics demonstrated a remarkable degree of μOR agonism in

stimulation of GTPγ35S binding and several of the compounds demonstrated a significantly decreased sensitivity to the presence of Na+ ions and nucleotide (i.e. GTPγS) in the buffer. As

shown in Figure 13, these two indirect measures of agonism were found to be significantly

correlated for both the [100 mM] Na+ buffer conditions (r2 = 0.89, p-value < 0.0001) and the [300 mM] Na+ buffer conditions (r2 = 0.77, p-value = 0.0008). This significant positive correlation confirms that these two experimental methods are both predictive measures of agonism, which is supportive of previously published data from our lab (Livingston and Traynor, 2014).

Ehlert’s efficacy formula, [ = 0.5 (Max)(1 + )], is another common quantitative 𝐾𝐾i 𝐸𝐸𝐸𝐸50 approach used to measure the efficacy𝐸𝐸 of opioid drugs (Ehlert, 1985; Griffin et al., 2007;

3 Livingston et al., 2018). This parameter incorporates the Ki calculated from the [ H]-diprenorphine

binding assay with the EC50 and relative efficacy (i.e. Max) determined from stimulation of

GTPγ35S binding. As shown in Figure 14, there was no significant correlation (r2 = 0.15, p-value

= 0.27) between the affinity shift due to Na+ ions and GTPγS and this particular measure of efficacy

for the set of opioid compounds. Using Ehlert’s efficacy formula, there was much less difference

between the compounds; DAMGO (3.27) was estimated to be the most efficacious compound in

the set of opioid compounds, followed by etorphine (2.05). The eight other compounds had

efficacy values between 0.86 (AAH8) and 1.40 (AMB67). As for this measure of efficacy, all of

the peptidomimetics demonstrated a lower relative efficacy compared to DAMGO, which is in

stark contrast to the data that was analyzed using ∆Log (Max/EC50), whereby all of the

peptidomimetics demonstrated a higher relative efficacy compared to DAMGO.

69

Figure 13. Correlation of Agonism (∆Log (Max/EC50)) with Affinity Shift by Na+ and Guanine Nucleotide at 35 μOR. ∆Log (Max/EC50) values were determined from stimulation of GTPγ S binding data with DAMGO used as the + 3 reference agonist (Figure 10), Log (Ki, 100 mM or 300 mM Na / Ki, 50 mM Tris) values were determined from [ H]- diprenorphine binding data (Figure 11and Figure 12), and data were plotted for each of the following ligands; DAMGO, AFN42, BU72, AMB47, AMB46, AAH8, Etorphine, AMB67, Fentanyl, and Morphine. (A) represents the [100 mM] Na+ buffer; r2 = 0.8854, p**** = < 0.0001 and (B) represents the [300 mM] Na+ buffer; r2 = 0.7728, p*** = 0.0008.

Figure 14. Correlation of Ehlert’s Efficacy with Affinity Shift by Na+ ions and Guanine Nucleotide at μOR. DAMGO (E = 3.27) > etorphine (E = 2.05) > AMB67 (E = 1.40) = fentanyl (E = 1.35) = AFN42 (E = 1.32) > BU72 (E = 1.05) = AMB47 (E = 0.98) > AAH8 (E = 0.86) = AMB46 (E = 0.91) = morphine (E = 0.93). r2 = 0.15, p = 0.27.

70 Kinetics of Binding at μOR in C6 μOR Cell Membranes

Next, we sought to determine if the extraordinary potency and efficacy of AFN42 was reflected in

its pharmacokinetic properties at the level of the μOR. The indirect “Mahan Motulsky method”

(see Materials and Methods) was performed to determine the kinetics of AFN42, AAH8, DAMGO

and BU72, the latter of which was previously shown to be a slowly dissociating ligand from μOR

(Livingston et al., 2018). In these experiments, the half-life of dissociation of BU72 was

determined to be 48 min in the presence of [3H]-diprenorphine (Figure 15A), whereas the half-life

of dissociation of DAMGO (Figure 15B) was estimated to be 1 min and is in agreement with previous data from our lab (Livingston et al., 2018). The half-life of dissociation of AFN42 (Figure

15C) was determined to be 10 min. However, the half-life of dissociation of AAH8 was not able to be accurately determined with the concentrations [0.1 nM] and [0.3 nM] of AAH8 that were used in the experiments (Figure 15D). As shown by the graph, AAH8 [0.1 nM] did not displace the binding of [3H]-diprenorphine at the μOR, and as such the half-life of dissociation was not able to be determined using the kinetics of competitive binding regression model. It is possible that the very high affinity and unique structure of AAH8, with the 2-naphthyl group, precluded [3H]-

diprenorphine from disassociating from the receptor, thus causing experimental problems.

71

Figure 15. Kinetics of Competitive Binding at the μOR. Radioligand competition binding was performed in 50 mM Tris buffer using 0.2 nM [3H]-diprenorphine and 20 μg C6 μOR membranes in the absence of agonist (circles) or in the presence of 0.1 nM (squares) or 0.3 nM (triangles) BU72 (A), AFN42 (C), or AAH8 (D), and 3 nM (red squares) or 10 nM (red triangles) DAMGO (B) for various time points over 8 hours. Data were analyzed using the kinetics of competitive binding nonlinear regression in GraphPad Prism and represent the mean ± SEM (n = 3).

72 Stimulation of GTPγ35S Binding and Affinity Shifts in C6 δOR Cell Membranes

As reported in previous studies (Bender, Griggs, Anand, et al., 2015; Harland et al., 2015, 2016;

Nastase et al., 2018), the majority of bifunctional opioid peptidomimetics were found to be antagonists at the δOR as determined by the lack of stimulation of GTPγ35S binding in C6 δOR

membranes compared to the standard agonist, DPDPE. To corroborate the findings that the peptidomimetics were δOR antagonists, antagonist affinity constants (Ke) were determined for

AMB46 (Ke = 78 nM), AMB47 (Ke = 4.6 nM), and AAH8 (Ke = 2.0 nM) (Table 6). From the discrepancies in the affinity values between experiments performed in 50 mM Tris buffer and the

+ Ke experiments performed in buffer containing Na ions and nucleotide, we sought to determine if

the peptidomimetics demonstrated an affinity shift at the δOR. As was previously discussed,

affinity shift due to Na+ ions and nucleotide is a predictive measure of agonism, and therefore,

radioligand competition binding experiments were performed in buffer containing Na+ ions, either at [100 mM] or [300 mM] and also with GTPγS [10 μM].

Saturation binding experiments were first performed as control experiments to calculate the appropriate Kd of diprenorphine for each buffer condition. The Kd of diprenorphine was

determined to be 0.45 nM at δOR in buffer that included 50 mM Tris, 0.45 nM at δOR in buffer that included 50 mM Tris, 100 mM NaCl, 5 mM MgCl2, 1 mM EDTA, 10 µM GTPγS, and 0.55 nM at δOR in buffer that included 50 mM Tris, 300 mM NaCl, 5 mM MgCl2, 1 mM EDTA, and

10 µM GTPγS. The minor differences in the Kd values between the conditions ruled out potential

for error in the calculation of Ki values, and thereby affinity shift data, for the set of opioid compounds.

From radioligand competition binding experiments in 50 mM Tris buffer using δOR

membranes, AAH8 (Ki = 0.23 ± 0.02 nM ) had the lowest Ki value, whereas AMB46 (15 ± 5.25

73 + nM) had the highest Ki value (Table 9). The shifts in affinity due to Na [100 mM] and GTPγS [10

μM] were as follows: DPDPE (20–fold) > AMB67 (6.3–fold) = BU72 (5.9–fold) > AAH8 (4.5–

fold) = AMB46 (5.5–fold) > AMB47 (3.2–fold) > AFN42 (0.8–fold). The shifts in affinity due to

Na+ [300 mM] and GTPγS [10 μM] followed a slightly different trend and were as follows; DPDPE

(30–fold) > AAH8 (13–fold) > BU72 (11–fold) > AMB67 (7.8–fold) > AMB47 (3.6–fold) >

AMB46 (2.7–fold) > AFN42 (1.1–fold). Notably, in an experiment in which SNC80 was used as

the standard agonist, AFN42 produced a rightward-shift in the concentration-response curve

Figure 16A and produced an antagonist affinity constant (Ke) of 0.85 nM, providing evidence that

AFN42 is a high affinity δOR antagonist. Moreover, AFN42 was the only peptidomimetic in the

series that failed to show an affinity shift, even in [300 mM] Na+ conditions (Figure 16B).

Excluding AFN42, the peptidomimetics appeared to exhibit partial δOR agonism, in regard to their

extent of affinity shift relative to DPDPE.

Figure 16. AFN42 is a δOR Antagonist and is Insensitive to the Na+ Effect on Binding Affinity. (A) Stimulation of GTPγ35S binding in the absence (open squares) or presence of AFN42 [1 nM] (closed squares) with various concentrations of SNC80, or GTPγ35S binding with various concentrations of AFN42 alone (open circles). (B) [3H]- diprenorphine binding was performed using 20-40 μg C6 δOR membrane protein, 0.2 nM [3H]-diprenorphine, and various concentrations of AFN42 in the buffers previously described. All data are represented as mean ± SEM and all experiments were performed in duplicate or quadruplicate (n ≥ 3).

74 Table 9. Binding Affinity of Opioid Ligands and the Affinity Shift by Na+ ions and Guanine Nucleotide in C6 δOR membranes

+ Table 9. Binding Affinity of Opioid Ligands and the Affinity Shift by Na ions and Guanine Nucleotide at δOR. Affinity shifts represent the difference in Ki values determined from competition binding assays that were performed using the following buffer conditions; 50 mM Tris buffer (column 1), 50 mM Tris, 100 mM NaCl, 5 mM MgCl2 1 mM EDTA, and 10 μM GTPγS (column 2) or 50 mM Tris, 300 mM NaCl, 5 mM MgCl2 1 mM EDTA, and 10 μM GTPγS (column 3). Column 4 represents the affinity shift by 100 mM Na+ buffer and column 5 represents the affinity shift by 300 mM Na+ buffer. Data are represented as mean ± SEM and all experiments were performed in duplicate or quadruplicate (n ≥ 3).

75 Stimulation of GTPγ35S Binding in CHO δOR Cell Membranes

To examine for potential δOR agonism of these compounds, higher δOR expressing cells were used. CHO cells that express the human opioid receptor demonstrate a high sensitivity to agonism because they have been transfected to stably express higher levels of the δOR. This is useful for the investigation of very low-efficacy compounds, such as buprenorphine. For buprenorphine,

35 stimulation of GTPγ S binding was 24% of the full agonist, DAMGO with an EC50 of 0.24 nM in

C6 μOR membranes, whereas the stimulation of GTPγ35S binding was 59% of DAMGO with an

EC50 = 0.17 nM in CHO μOR membranes [data not shown]. These data support that the efficacy

of buprenorphine depends on the cell line and permit the use of CHO δOR cells to evaluate the

bifunctional opioid peptidomimetics.

In CHO δOR cell membranes, the full δOR agonists Met-enkephalin and BU72, and the partial agonist, morphine, were compared using the GTPγ35S binding assay (Figure 17A). The

relative efficacy (maximal stimulation compared to Met-enkephalin) of BU72, morphine, and the

peptidomimetics in rank order were as follows; BU72 (96 ± 2.0 %) > morphine (60 ± 12 %) AAH8

(46 ± 4.2 %) = AMB67 (45 ± 2.6 %) = AMB47 (36 ± 2.8 %) > AFN42 (35 ± 1.0 %) > AMB46

(23 ± 1.0 %) (Figure 17B). Although the EC50 for each of the compounds varied, with potencies

for the peptidomimetics in the single digit nanomolar range, the peptidomimetics were all found

to be partial agonists. In addition, relative agonism measurements were calculated from the

concentration-response curves and given in (Figure 17B).

76

Figure 17. Bifunctional Peptidomimetics are Partial Agonists in CHO δOR Membranes. (A)The bifunctional peptidomimetics AFN42, AMB47, AMB46, AAH8, and AMB67 are partial agonists compared to the full agonists, Met-enkephalin (black) and BU72 (blue), and demonstrate a potent EC50 for the stimulation 35 of GTPγ S binding in the range of 1-10 nM, compared to the weak potency of morphine (red), the EC50 of which was over 1000 nM in CHO δOR membranes. (B) Efficacy (Max), potency (EC50), Log (Max/EC50) and ∆Log (Max/EC50) data were derived from the concentration-response curves from (A) for all of the opioid compounds.

77 Stimulation of GTPγ35S Binding in SH-SY5Y Cell Membranes

To further define the pharmacology of the bifunctional opioid peptidomimetics, the relative

efficacies of the compounds were measured in the SH-SY5Y human neuroblastoma cells. Whereas

CHO cells were highly sensitive to weak-efficacy agonists, the SH-SY5Y human neuroblastoma

cell line has a low endogenous level of opioid receptor expression, and so would be expected to

provide a larger separation between high and low efficacy compounds. SH-SY5Y cells express both μOR and δOR, with about twice as many μORs compared to δORs (Alt et al., 2001). The

35 maximal stimulation of GTPγ S binding of the δOR full agonist SNC80 was 34 % (with an EC50

of 30 nM) compared to the full μOR agonist DAMGO (Figure 18A; Table 10), supporting evidence that SH-SY5Y cells exhibit lower expression levels of δORs compared to μORs (Alt et al., 2001).

The potencies of DAMGO, fentanyl, and morphine (EC50’s between 115 nM – 189 nM)

and BU72, AFN42, AMB47, AMB46, and AAH8 (EC50’s between 0.11 nM – 0.37 nM) for

stimulation of GTPγ35S binding in SH-SY5Y membranes (Figure 18A; Table 10) were very similar

to those seen in C6 μOR membranes. Maximal stimulation of GTPγ35S binding compared to

DAMGO by fentanyl (76 ± 3 %), morphine (79 ± 2 %), and AAH8 (82 ± 5 %) were sub-maximal.

However, in comparison to C6 μOR membranes, maximal stimulation of GTPγ35S binding of

morphine increased by 20 % which is possibly due to its binding and activity at the δORs in the

SH-SY5Y cells (Alt et al., 2001). In SH-SY5Y membranes, similar increases in percent maximal

stimulation compared to DAMGO were seen with the δOR agonist BU72 (121 ± 6 %), AMB47

(111 ± 4 %), and AFN42 (109 ± 2 %). Calculated from the concentration-response curves and

shown in Table 10, AFN42 demonstrated the greatest index of agonism compared to the other

compounds in SH-SY5Y cell membranes, with a ( ) of 3.19. Moreover, this degree of 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 relative agonism is comparable to that observed in ∆C6𝐿𝐿𝐿𝐿𝐿𝐿 μOR membranes (Table 7).

78 The relative agonism values from data in Figure 18A were AFN42 (3.19) > AMB47

(2.98) > BU72 (2.87) > AMB46 (2.74) > AAH8 (2.56) > morphine (– 0.13) > fentanyl (–0.27).

Shorter incubation times provide a greater separation of response between agonists, thus, AFN42,

BU72, and DAMGO were also evaluated in experiments in which bound GTPγ35S was quantified

after 15 minutes (Figure 18B). GTPγ35S binding for AFN42 was 118 % compared to DAMGO

(100 %), whereas BU72 was 136 % compared to DAMGO (Table 10). The EC50 of AFN42 was

the same at 15 minutes as it was at 1 hour, which suggests that AFN42 demonstrates a rapid

kinetics of opioid receptor binding and activation of G protein signaling.

Figure 18. Stimulation of GTPγ35S Binding by Opioid Agonists in SH-SY5Y Membranes. Stimulation of GTPγ35S binding assays used 10 μM DAMGO as the standard agonist, 20-30 μg SH-SY5Y membrane protein, 30 μM GDP, 35 35 and 0.1 nM GTPγ S in the following buffer; 50 mM Tris, 100 mM NaCl, 5 mM MgCl2, and 1 mM EDTA). GTPγ S was included in the assay for a total of 1 hour (A) or 15 min (B) before the reaction was terminated and the samples were counted. Experiments were performed in quadruplicate and data are represented as mean ± SEM (n = 5).

79 Table 10. Stimulation of GTPγ35S Binding by Opioid Agonists in SH-SY5Y Membranes

Table 10. Stimulation of GTPγ35S Binding by Opioid Agonists in SH-SY5Y Membranes. Stimulation of GTPγ35S binding assays used 10 μM DAMGO as the 35 standard agonist, 20-30 μg SH-SY5Y membrane protein, 30 μM GDP, and 0.1 nM GTPγ S in the following buffer; 50 mM Tris, 100 mM NaCl, 5 mM MgCl2, and 1 mM EDTA). GTPγ35S was included in the assay for a total of 1 hour or 15 minutes before the reaction was terminated and samples were counted. Experiments were performed in quadruplicate and data are represented as mean ± SEM (n = 5).

80 Biased Agonism of Bifunctional Opioid Peptidomimetics at the μOR

Several of the bifunctional opioid peptidomimetics were seen to be extremely potent at stimulating

β-arrestin recruitment at the μOR (Figure 20A). AFN42 (EC50 = 0.0005 ± 0.22 nM; relative

agonism = 12.36) was most potent, followed by AMB47 (EC50 = 0.001 ± 0.08 nM; relative

agonism = 11.93), > AMB46 (EC50 = 0.006 ± 0.25 nM; relative agonism = 11.24), > AAH8 (EC50

= 0.05 ± 0.08 nM; relative agonism = 10.21) (Table 11). The peptidomimetic AMB67 (EC50 = 21

± 0.06 nM; relative agonism = 7.65), which is structurally distinct from the other four peptidomimetics (Figure 8), was significantly less potent and similar to the potency of the reference ligand, Met-enkephalin (EC50 = 25 ± 0.11 nM) with a relative agonism of 7.38. As for

the two control ligands, morphine was a partial agonist for β-arrestin recruitment and demonstrated a weaker potency (EC50 = 100 ± 0.06 nM; relative agonism = 6.79) compared to Met-enkephalin,

whereas BU72 was a full agonist with a 10-fold greater potency for β-arrestin recruitment (EC50 =

2.8 ± 0.02 nM; relative agonism = 8.60).

( 35 To determine bias factors, 50) data from stimulation of GTPγ S binding 𝑀𝑀𝑀𝑀𝑀𝑀 ∆ 𝐸𝐸𝐸𝐸 experiments, performed using C6 μOR𝐿𝐿𝐿𝐿𝐿𝐿 cell membranes with DAMGO as the internal reference

( ligand (Figure 10), and 50) data from concentration-response curves (Figure 20A) were 𝑀𝑀𝑀𝑀𝑀𝑀 ∆𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸 ( used to determine 50) values. Bias factors were calculated using the formula 𝑀𝑀𝑀𝑀𝑀𝑀 ∆∆𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸 [10 𝑀𝑀𝑀𝑀𝑀𝑀 ] and presented in Table 11. AMB67 (bias = 10) was the most G protein-biased ∆∆𝐿𝐿𝐿𝐿𝐿𝐿�𝐸𝐸𝐸𝐸50� ligand that was evaluated, followed by BU72 (bias = 5). Morphine was found not to be biased toward one signaling pathway over another (bias = –3) and so should be considered neutral and

comparable to Met-enkephalin. However, three of the peptidomimetics that were evaluated were

significantly biased toward β-arrestin; AMB47 (bias = –724), AMB46 (bias = –302), and AFN42

(bias = –213). Although structurally very similar, AAH8 (bias = –17) is much less biased

81 compared to AMB47. Nevertheless, AAH8 is significantly biased toward β-arrestin recruitment over G protein activation at the μOR.

Biased Agonism of Bifunctional Opioid Peptidomimetics at the δOR

The peptidomimetics were partial agonists as measured by the recruitment of β-arrestin at the δOR, however, they were much less potent than at the μOR (Figure 20B). The potency for stimulation

of β-arrestin recruitment for AAH8 (EC50 = 0.09 ± 0.08 nM) and AFN42 (EC50 = 0.09 ± 0.07 nM)

were the highest and roughly equivalent to the potency of BU72 (EC50 = 0.17 ± 0.04 nM). The potency of Met-enkephalin (EC50 = 4.7 ± 0.03 nM) and AMB47 (EC50 = 7.7 ± 0.05 nM) were

similar and greater than that of AMB67 (EC50 = 167 ± 0.04 nM), AMB46 (EC50 = 196 ± 0.04 nM)

and morphine (EC50 = 2230 ± 0.08 nM), which barely stimulates β-arrestin recruitment to the δOR

(Table 12). To determine bias factors, ( ) data from stimulation of GTPγ35S binding 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 experiments performed using CHO δOR∆ 𝐿𝐿𝐿𝐿𝐿𝐿cell membranes with Met-enkephalin as the internal reference ligand (Figure 17), and ( ) data from concentration-response curves (Figure 𝑀𝑀𝑀𝑀𝑀𝑀 ∆𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50 20B) were used to determine these ( ) values and the δOR bias factors. The 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 peptidomimetics AMB46 (bias = 49) and∆ AMB67∆𝐿𝐿𝐿𝐿𝐿𝐿 (bias = 37) were the most G protein-bias ligands

at δOR, and interestingly, AAH8 (bias = –17) was the only ligand that was biased toward β-arrestin

recruitment. Moreover, the potency of AAH8 (EC50 = 0.09 ± 0.08 nM) was comparable to that of

the full δOR agonist, BU72 (EC50 = 0.17 ± 0.04 nM).

Role of G proteins in β-arrestin Recruitment

β-arrestin recruitment assays were further performed using CHO μOR cells that were treated with

100 ng/mL pertussis toxin (PTX) for 6 hours, and the results were plotted in Figure 19 and

82 compared to control conditions for each opioid compound (Figure 20A). The maximal response

for all opioid compounds decreased by roughly 40% following PTX treatment. The potency for

stimulation of β-arrestin recruitment following PTX treatment for AFN42 (EC50 = 0.0013 ± 0.28

nM) was unchanged compared to control conditions (EC50 = 0.0005 ± 0.07 nM; Table 12). Similar

results were observed for AAH8 (EC50 = 0.07 ± 0.05 nM) and AMB67 (EC50 = 18.2 ± 0.11 nM).

The results for Met-enkephalin (EC50 = 848 ± 0.03 nM) were similar to morphine (EC50 = 1385 ±

0.16 nM), AMB46 (EC50 = 0.17 ± 0.12 nM), and AMB47 (EC50 = 0.07 ± 0.23 nM) in that their

potency for β-arrestin recruitment decreased compared to control conditions. The outlier was

BU72 (EC50 = 0.15 nM), which showed an increase in potency for β-arrestin recruitment following

PTX treatment (from 2.8 nM).

Figure 19. β-Arrestin Recruitment of Opioid Ligands Following Pertussis Toxin Treatment at the μOR. Concentration-response curves showing β-arrestin recruitment at the μOR in CHO PathHunter cells that were treated with 100 ng/mL pertussis toxin (PTX) for 6 hours. The DiscoveRx β-galactosidase enzyme fragment complementation assay was performed for the following opioid compounds; Met-enkephalin, BU72, morphine, and the bifunctional peptidomimetics; AFN42, AMB47, AMB46, AAH8, and AMB67. Data were normalized to a maximal concentration of Met-enkephalin in cells that were not treated with PTX.

83

Figure 20. β-Arrestin Recruitment of Opioid Ligands at the μOR and δOR and Biased Agonism Plots. β-Arrestin recruitment at the μOR in CHO PathHunter cells (A) or at the δOR in CHO PathHunter cells (B) using the DiscoveRx β-galactosidase enzyme fragment complementation assay, for the following opioid compounds; Met-enkephalin, BU72, morphine, and the bifunctional peptidomimetics; AFN42, AMB47, AMB46, AAH8, and AMB67. Met-enkephalin produces a maximal response for β-Arrestin recruitment and was used as the standard agonist reference ligand, thus data were normalized to a maximal concentration of 35 Met-enkephalin. (C) ∆∆Log (Max/EC50) values were plotted and calculated from ∆Log (Max/EC50) data from stimulation of GTPγ S binding experiments performed using C6 μOR cell membranes with DAMGO as the internal reference ligand, and from ∆Log (Max/EC50) data calculated from concentration-response 35 curves from panel A. (D) ∆∆Log (Max/EC50) values were plotted and calculated from ∆Log (Max/EC50) data from stimulation of GTPγ S binding experiments performed using CHO δOR cell membranes using Met-enkephalin as the internal reference ligand, and from ∆Log (Max/EC50) data calculated from concentration- response curves from panel B.

84 Table 11. Bias Factors for Opioid Compounds at the μOR

35 Table 11. Bias Factors for Opioid Compounds at the μOR. EC50 and Log (Max/EC50) data were produced from stimulation of GTPγ S binding experiments performed using C6 μOR cell membranes. 10 μM DAMGO was used as the standard agonist to produce the maximal response (Max = 1.0). EC50 and Log (Max/EC50) data were produced from β-arrestin recruitment assays, performed using CHO μOR cells. Here, 10 μM Met-enkephalin was used as the standard agonist to produce the maximal response (Max = 1). The ∆Log (Max/EC50) data [not shown] was normalized to the standard agonists for each assay; ∆∆Log (Max/EC50) values represent the difference between the ∆Log (Max/EC50) data from each assay, which were used to determine the bias factor for each opioid. Thus, bias factors 35 were calculated from the Log (Max/EC50) data from both stimulation of GTPγ S binding and β-arrestin recruitment assays.

85 Table 12. Bias Factors for Opioid Compounds at the δOR

35 Table 12. Bias Factors for Opioid Compounds at the δOR. EC50 and Log (Max/EC50) data were produced from stimulation of GTPγ S binding experiments performed using C6 δOR cell membranes. 10 μM Met-enkephalin was used as the standard agonist to produce the maximal response (Max = 1.0). EC50 and Log (Max/EC50) data were produced from β-arrestin recruitment assays performed using CHO δOR cells. Again, 10 μM Met-enkephalin was used as the standard agonist to produce the maximal response (Max = 1). The ∆Log (Max/EC50) data [not shown] was normalized to Met-enkephalin for each assay; ∆∆Log (Max/EC50) values represent the difference between the ∆Log (Max/EC50) data from each assay, which were used to determine the bias factor for each opioid. Thus, bias factors 35 were calculated from the Log (Max/EC50) data from both stimulation of GTPγ S binding and β-arrestin recruitment assays.

86

Discussion

In this chapter, various experiments were implemented for a set of bifunctional opioid

peptidomimetics and classical opioid agonists to evaluate the agonism of the peptidomimetics: (i)

[3H]-DPN competition binding assays were performed in C6 μOR and C6 δOR membranes in the

absence or presence of Na+ ions [100 mM] or [300 mM] and GTPγS [10 μM] to measure affinity

35 (Ki) shifts and estimate agonist efficacy, (ii) stimulation of GTPγ S binding assays were

performed in four different cell lines (C6 μOR, C6 δOR, CHO δOR, and SH-SY5Y) to assess

relative efficacy and potency after which ( ) values were calculated to establish an index 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 of agonism, and (iii) the “Mahan Motulsky∆𝐿𝐿𝐿𝐿𝐿𝐿 method” was performed to determine the kinetics of binding at the μOR. Collectively, the data in this chapter represent a comprehensive evaluation of several bifunctional peptidomimetics for their agonism at both the μOR and the δOR.

Stimulation of GTPγ35S binding in C6 μOR and CHO δOR membranes revealed that all

five peptidomimetics were characterized as potent, full μOR agonists and potent, partial δOR

agonists, respectively. As these data were gathered in C6 and CHO cells that overexpress the rat

and human opioid receptors, respectively, the peptidomimetics were then evaluated in membranes

from SH-SY5Y cells that express endogenous levels of human μOR and δOR. In the stimulation

of GTPγ35S binding experiments using SH-SY5Y cell membranes, the peptidomimetics displayed

equivalent or greater agonism compared to BU72 (Table 10) according to ( ) values. 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 However, the maximal stimulation of BU72 was highest at 121 % compared to∆ 𝐿𝐿𝐿𝐿𝐿𝐿DAMGO and this

is likely due to BU72 acting as a full δOR agonist (Neilan et al., 2004). The peptidomimetics

demonstrate potent (but weak efficacy) δOR partial agonism compared to BU72, which is apparent via the concentration-response curves in CHO cells overexpressing δOR (Figure 17). Thus, it is

87 possible they did not appreciably activate G proteins downstream of the δOR in the SH-SY5Y cell membranes. In GTPγ35S binding experiments that were performed in SH-SY5Y cell membranes for 15 minutes, as opposed to one hour, there was an increase in the degree of agonism of AFN42 relative to BU72 Table 10. It is already known from the literature that BU72 has slow kinetics

(Neilan et al., 2004). These data substantiate AFN42 having faster kinetics than BU72, which was previously determined using the Mahan Motulsky method (Figure 15C).

AFN42 was the most potent compound examined with a 10-fold higher potency than

BU72, etorphine, and the other peptidomimetics, as well as a 1000-fold higher potency than

DAMGO, fentanyl, and morphine (Figure 10 and Table 7). AFN42 is the “least sensitive” opioid compound to Na+ ions and guanine nucleotide in competition binding experiments. It is generally accepted that the binding of agonists is most sensitive to Na+ ions and guanine nucleotide and that antagonists are insensitive. This is because Na+ ions stabilize a low affinity inactive state of the receptor (Livingston et al., 2018). In spite of this, the potent and efficacious agonist AFN42 appears insensitive to Na+ ions. AFN42 has the same affinity for the inactive state and active state receptor. One explanation is that AFN42 has such a high affinity for the receptor that Na+ ions are unable to bind to the receptor to function as negative allosteric modulators, returning the receptor

35 to its inactive state. Furthermore, its Ki is the same as its EC50 from GTPγ S binding experiments, suggesting that it binds to inactive and active state receptors with the same high affinity and immediately “locks” the receptor in an active state and initiates rapid and robust G protein signaling. Therefore, an affinity shift due to the presence of Na+ ions and guanine nucleotide is not reliable predictor of efficacy when dealing with very potent compounds (Livingston et al., 2018), in contrast to previously accepted ideas (Pert and Snyder, 1976). This feature likely explains the

88 lack of a significant positive correlation between Ehlert’s efficacy and sensitivity to Na+ ions for

the peptidomimetics (Figure 14).

The data from the β-arrestin recruitment experiments confirm that morphine lacks G

protein-arrestin bias, as previously reported (Schmid et al., 2017). Its potency for stimulating β-

arrestin recruitment at μOR is 100-fold greater than that at δOR, which is in line with its greater

potency and efficacy to stimulate G protein signaling at μOR compared to δOR. Data from our

GTPγ35S binding experiments report that morphine is a partial agonist with a high affinity for μOR

but a weak agonist with low affinity for δOR (Anand et al., 2018). An unsuspecting observation

from these studies is that functional G proteins are not absolutely necessary for β-arrestin

recruitment at the μOR, as shown in Figure 19. Pertussis toxin (PTX) modifies opioid receptor signaling via its ADP-ribosylation of Gαi/o proteins and so inhibits their interaction with cell

membrane receptors. PTX prevents heterotrimeric G protein coupling to the receptors and thus

allows the measurement of G protein-independent receptor signaling. Following PTX

administration, there was a sub-maximal response and in fact the potency of BU72 for stimulating

β-arrestin recruitment increased. The potency for AMB67, AAH8, AFN42 and morphine did not

significantly change, whereas the potency for AMB47 decreased the most, followed by AMB46.

Although G proteins are not absolutely necessary for β-arrestin recruitment at the μOR, a few β-

arrestin-biased peptidomimetics (i.e. AMB46 and AMB47) seemed to depend on G proteins for

their high potency for β-arrestin recruitment because PTX treatment significantly attenuated their

potencies. However, for the peptidomimetics that were G protein-biased, the potency for β-arrestin

recruitment either did not change or it improved as was seen with BU72. Perhaps β-arrestin

recruitment is inhibited by the presence of G proteins, and once that physical interaction is eliminated, β-arrestin proteins are able to bind with higher potency. It is also reasonable to

89 speculate that extremely potent opioid agonists, such as AFN42 and BU72, stabilize a distinct

receptor state that engages β-arrestin proteins and circumvents canonical receptor signal

transduction (i.e. G protein activation leading to kinase-mediated phosphorylation of the receptor).

The lead peptidomimetic AAH8, has a balanced binding affinity for both μOR and δOR

35 and an equivalent potency for the stimulation of GTPγ S binding at μOR and δOR (μOR EC50 =

0.71 ± 0.2 nM / δOR EC50 = 3.36 ± 1.1 nM). The results from the β-arrestin recruitment

experiments (without PTX treatment) also show that it is an agonist at both receptors (μOR EC50

= 0.05 ± 0.08 nM / δOR EC50 = 0.09 ± 0.08 nM). It is β-arrestin biased by 17-fold at both μOR

and at δOR as determined by its calculated negative ( ) values. In stark contrast, 𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸50 AMB46, AMB47, and AFN42, as measured by β-arrestin∆∆𝐿𝐿𝐿𝐿𝐿𝐿 recruitment, are all significantly β- arrestin-biased with very high potencies. These three peptidomimetics are also very selective for

μOR over δOR. For example, the potency of AMB46 is 0.006 ± 0.07 nM at μOR and 196 ± 0.04

nM at δOR, which represents a 10,000-fold selectivity for the μOR over the δOR. AMB67 has a

modest 10-fold selectivity for μOR over δOR and is G protein biased at both receptors.

Structurally, AMB67 lacks the THQ core that is characteristic of the other peptidomimetics; the

nitrogen atom is replaced with a sulfur atom. Accordingly, the THQ core may be a molecular

feature of the peptidomimetics that is responsible for their significant bias for β-arrestin

recruitment, and further comparison of the two structures could be fruitful in exploring biased

agonism.

In summary of the peptidomimetics, AMB67 demonstrates significant G protein bias,

AAH8 has a lack of μOR / δOR selectivity and a high potency for receptor activation, and AMB46

and AMB47 exhibit high μOR selectivity and β-arrestin bias; together these findings may elucidate

some of the differences that were seen in vivo. Antinociceptive tolerance did not develop following

90 chronic administration of AAH8, AMB47 and AMB67, however, antinociceptive tolerance did

develop following chronic administration of AMB46 in mice. To explain the in vivo findings with

the recent in vitro results, it might be most beneficial to compare AMB46 with AAH8. In β-arrestin

recruitment assays using CHO δOR cells, AMB46 has a weak potency (EC50 = 196 nM) compared

to AAH8 (EC50 = 0.09 nM), which is nearly a full agonist for β-arrestin recruitment at the δOR.

Furthermore, AAH8 is a potent partial agonist in CHO cells and a high affinity, functional

antagonist in C6 cells for G protein activation. In support of the promising in vivo data for AAH8 and negative in vivo data for AMB46, the ideal pharmacological profile could be a potent agonist at μOR, a high affinity antagonist at δOR, and perhaps an agonist for β-arrestin recruitment at

δOR. Indeed, AMB46 is a very selective μOR full agonist with a weak δOR affinity, and it also exhibits preference for G protein signaling over β-arrestin signaling at the δOR. Thus, it is conceivable that the sought-after compound would be biased for G protein signaling at μOR and

also biased for β-arrestin recruitment and signaling at δOR, although of course, additional studies

will be necessary to verify this pharmacological model.

The results and discussions within this dissertation have hopefully provided some insight

into the biology of the mu and delta opioid receptors. I have demonstrated that several bifunctional

opioid peptidomimetics are significantly biased for β-arrestin recruitment over G protein signaling

and could be useful as pharmacological tools to explore the nature of biased agonism. G protein-

biased μOR agonists are poised to be safer analgesics with greater clinical efficacy compared to

traditional opioid analgesics, however, this research in still in progress. Our thorough in vitro and

in vivo characterization of our bifunctional opioid peptidomimetics should provide valuable

information for other researchers with the common goal of developing improved opioid analgesics.

91 Methods

Materials and Reagents

AAH8 was synthesized by Aubrie A. Harland, AMB46, AMB47, and AMB67 were synthesized

by Aaron M. Bender, and AFN42 was synthesized by Anthony F. Nastase (Bender, Griggs, Anand,

et al., 2015; Harland et al., 2015; Nastase et al., 2019). BMS 986122 was synthesized as previously described (Burford et al., 2015). All chemicals, unless otherwise specified, were purchased from

Sigma Aldrich (St. Louis, MO). Radiolabeled [3H]-diprenorphine (DPN) and radiolabeled

guanosine‐5′‐O‐(3‐[35S]thio)-triphosphate (GTPγ35S) were purchased from PerkinElmer Life

Sciences (Waltham, MA). All tissue culture medium, penicillin-streptomycin, geneticin (G148), trypsin, and fetal bovine serum were purchased from Invitrogen (Carlsbad, CA). PathHunter detection reagents were acquired from DiscoveRx (Freemont, CA).

Cell Lines and Membrane Preparation

The tissue culture and maintenance of C6 rat glioma cells stably transfected with rat μ Opioid

Receptor (μOR) or rat δ Opioid Receptor (δOR) was performed as previously described (Clark et al., 2003). C6 cells were grown to confluence at 37 °C in 5 % CO2 in Dulbecco’s modified Eagle

media (DMEM) containing 10 % fetal bovine serum (FBS), 5 % penicillin/streptomycin (P/S), and

400 ug/mL G418.

CHO cells stably expressing wild‐type human μOR (CHO μOR) or stably expressing wild‐type human δOR (CHO δOR) were grown to confluence at 37 °C in 5 % CO2 in Dulbecco’s modified

Eagle media (DMEM) containing 10 % fetal bovine serum (FBS), 1 % penicillin/streptomycin

(P/S), and 400 ug/mL G418.

92 CHO PathHunter cells expressing enzyme acceptor (EA)-tagged β-arrestin 2 and ProLink (PK)- tagged μOR (CHO-OPRM1) were purchased from DiscoveRx (Freemont, CA). Cells were grown in F-12 media (Invitrogen 11765), containing 10 % FBS, Hygromycin 300 μg/mL, Geneticin

(G418) 800 μg/mL and maintained at 37 °C in a humidified incubator containing 5 % CO2.

SH-SY5Y (ATCC Catalog # CRL-2226) cells were grown in Dulbecco’s modified Eagle medium

(DMEM) containing 10% fetal bovine serum (FBS), 1% penicillin/streptomycin (P/S) maintained at 37 °C in a humidified incubator containing 5 % CO2.

Cell membrane homogenates (C6, CHO, and SH-SY5Y) were prepared by washing confluent cells three times with phosphate buffered saline (pH 7.4), after which they were then detached using harvesting buffer (20 mM HEPES, 150 mM NaCl, 0.68 mM EDTA, pH7.4) and pelleted by centrifugation at 200 x g for 3 min at room temperature. The pellet was resuspended in ice-cold 50 mM Tris buffer (pH 7.4) and homogenized using a Tissue Tearor (Biospec Products, Inc.,

Bartlesville, OK). The cell lysate was centrifuged at 20000 x g at 4 °C for 20 minutes, after which the pellet was then resuspended, homogenized, and centrifuged once more at 20000 x g at 4 C for

20 minutes. The final pellet was resuspended in 50 mM Tris (pH 7.4) using a glass dounce homogenizer and then dispensed into aliquots and stored at -80 °C. The final protein concentration was determined by performing a BCA protein assay (Thermo Scientific Pierce, Waltham, MA) using bovine serum albumin as the standard.

93 Radioligand Binding Assays

[3H]-diprenorphine (DPN) competition binding assays were performed in 500μL reactions in 96-

well plates using the cell membrane homogenates described above. Briefly, cell membranes [20-

40 μg] were incubated in 50 mM Tris buffer containing 0.2 nM [3H]-DPN, and various concentrations of peptidomimetic analogue for 1 hour in a shaking water bath at room temperature.

Nonspecific binding was determined using 10 µM naloxone. The reaction was terminated by vacuum filtration onto glass microfiber GF/C filters using a tissue harvester (Brandel,

Gaithersburg, MD) and washed with 50 mM Tris buffer. Filters were dried and, following the addition of EcoLume scintillation cocktail, bound radioactivity was quantified in a Wallac 1450

MicroBeta Liquid Scintillation and Luminescence Counter (PerkinElmer, Waltham, MA).

Radioligand binding assays in Na+ containing buffer were performed in 200 μL reactions in 96-

well plates using the cell membrane homogenates described above Briefly, cell membranes [20-

40 μg] were incubated in Na+ containing buffers (50 mM Tris, 100 mM NaCl, 5 mM MgCl2, 1

mM EDTA, pH 7.4) or (50 mM Tris, 300 mM NaCl, 5 mM MgCl2, 1 mM EDTA, pH 7.4)

containing 10 μM cold GTPγS, and various concentrations of peptidomimetic analogue for 90

minutes in a shaking water bath at room temperature. Nonspecific binding was determined using

10 µM naloxone. The reaction was terminated by vacuum filtration onto glass microfiber GF/C

filters using a tissue harvester and washed with Na+ containing buffer, and the filters were

processed as described above.

Radioligand binding using the Mahan Motulsky method was performed using three different

concentrations of cold ligand that were added to a 96-well plate containing 0.2 nM 3H-

94 diprenorphine in 500 μL assay buffer (50 mM Tris, pH 7.4). 30 μg membrane protein were added

to the plate stepwise, and samples were incubated for various time-points (5, 15, 20, 30, 60, 90,

120, 240 and 480 minutes) before terminated and counted as described above. Nonspecific binding

was determined using 10 µM naloxone. Data were analyzed using the kinetics of competitive

binding nonlinear regression in GraphPad Prism.

Stimulation of GTPγ35S Binding Assays

Stimulation of GTPγ35S binding experiments were performed in 200μL reactions in 96-well plates

using cell membrane homogenates prepared as described above. Briefly, cell membranes [10 μg]

were incubated in Na+ containing buffer [50 mM Tris, 100 mM NaCl, 5 mM MgCl2, 1 mM EDTA,

pH 7.4] containing 0.1 nM guanosine‐5′‐O‐(3‐[35S]thio)triphosphate (GTPγ35S) and 30 μM

guanosine 5’-diphosphate (GDP) and various concentrations of peptidomimetic analogue for 1

hour in a shaking water bath at 25 °C. Basal stimulation of GTPγ35S binding was determined by

incubation in the absence of any ligand and maximal stimulation of was defined by using 10 μM

of the standard opioid agonist compounds, DAMGO ([D-Ala2,N-MePhe4,Gly-ol]-enkephalin) for

μOR, DPDPE (D-Pen2,5-enkephalin) for δOR, or SNC80 [(+)-4-[(αR)-α-((2S,5R)-4-Allyl-2,5-

dimethyl-1-piperazinyl)-3-methoxybenzyl]-N,N-diethylbenzamide] for δOR. The reaction was

terminated by vacuum filtration onto glass microfiber GF/C filters using a tissue harvester and

washed with Na+ containing buffer, and filters were processed as described above.

PathHunter β-Arrestin Assays

Confluent flasks of CHO-OPRM1 cells were harvested with TrypLE Express and resuspended in

F-12 media supplemented with 10 % FBS and 25 mM HEPES, at a density of 6.67 x 105 cells /mL

95 and plated (3 μL / well) into white solid TC-treated 1536-well plates (Corning, NY). In parallel,

100 ng/mL pertussis toxin (PTX) was administered to CHO-OPRM1 cells for 6 hours. Plates were then incubated overnight at 37 C in a 5% CO2 humidified incubator. The next day, increasing concentrations of test peptidomimetics (40 nL of 100 x final concentration in 100% DMSO) were added to separate rows of the assay plates by acoustic dispense using an Echo-550 (Labcyte,

Sunnyvale, CA) from Echo-qualified 1536-well source plates (Labcyte). Next, 1 μL of increasing concentrations of DAMGO (4 x final concentration in assay buffer) were added to separate columns of the assay plates containing cells. Plates were covered with a lid and incubated at room temperature for 90 min. Incubations were terminated by the addition of 2 μL PathHunter Reagent

(DiscoveRx). One hour later luminescence was detected using a Viewlux imaging plate reader

(PerkinElmer).

Data Analysis

Graphpad Prism software (San Diego, CA) was used for all statistical analysis. Binding affinity

(Ki) values were calculated using the Cheng-Prusoff equation, potency (EC50), relative efficacy (%

stimulation compared to a standard opioid agonist). β-arrestin recruitment analysis was performed

using a nonlinear regression and bias calculations were performed as described by (Kenakin,

2017). Briefly, individual curves were used to calculate the ( ). The difference in 𝑀𝑀𝑀𝑀𝑀𝑀 𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50 ( ) values between β-arrestin recruitment and stimulation of GTPγ35S binding afforded a 𝑀𝑀𝑀𝑀𝑀𝑀 𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50 ( ) for each compound. Differences between the ( ) values for the reference 𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀 ∆𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50 ∆𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50 ligand and each opioid compound were calculated to give a ( ), the antilog of which 𝑀𝑀𝑀𝑀𝑀𝑀 ∆∆𝐿𝐿𝐿𝐿𝐿𝐿 𝐸𝐸𝐸𝐸50

[10 𝑀𝑀𝑀𝑀𝑀𝑀 ] is the bias factor. ∆∆𝐿𝐿𝐿𝐿𝐿𝐿�𝐸𝐸𝐸𝐸50�

96 Chapter IV

Conclusions and Future Directions

Opioid addiction is a severe burden in the United States. Every day it is estimated that more than 130 people die from an opioid overdose (Jalal et al., 2018). The solutions to prevent fatal overdoses, such as Narcan®, and to treat opioid-use disorders, such as Suboxone®, are finally reaching the people in need. However, these two drugs are short-term pharmacological solutions to the various problems that are associated with opioids. Two of these problems are explicit: (i)

opioid-use disorder and (ii) chronic pain, and the link between the two is also well known. The

number of people with chronic pain is staggering; estimates from various surveys range anywhere

from 50 million to 120 million people (Nahin, 2015; Dahlhamer et al., 2018). Unfortunately, nearly

every opioid analgesic comes with side effects that include abuse liability. Thus, the development

of a non-addictive opioid analgesic would be a giant leap forward in pain management.

Historically, drug discovery and development of opioid analgesics has largely favored

highly selective ligands for the mu opioid receptor (μOR). To be effective, opioid analgesics must

activate this receptor, which is responsible for analgesia. However, this receptor is also responsible

for side effects like tolerance, physical dependence, and respiratory depression. Recent strategies

in opioid drug development have focused on non-selective opioid compounds (i.e. bifunctional opioid compounds that not only target μOR but also other opioid receptors) and G protein-biased

compounds that target the μOR, due to a growing body of preclinical evidence that suggests that

97 these compounds may provide significant therapeutic benefit compared to traditional opioid analgesics.

This dissertation reports the pharmacological evaluation of several novel opioid compounds and identifies a few lead compounds that exhibit an improved side-effect profile.

These novel opioid compounds, termed ‘bifunctional opioid peptidomimetics’, take key elements

of opioid peptides that are vital for activity and incorporate small molecule-like features to provide

for bioavailability, blood brain barrier permeability and increased duration of action. Lead

compounds produced acute antinociceptive activity comparable to that of morphine, following

peripheral administration in mice. Notably, compared to mice that were treated chronically with

morphine, mice that were treated chronically with the lead compounds showed a significant

reduction in the development of antinociceptive tolerance. Based on a number of studies in the

literature, the emergence of numerous bifunctional opioid compounds in preclinical development,

and results from our laboratories, bifunctional opioid compounds may show promise as analgesics.

The results presented in chapter II highlight the main structure-activity relationship (SAR) insights that served as a guide for how the peptidomimetic analogues were optimized and selected for in vivo analysis. From in vitro evaluation we identified several lead peptidomimetic analogues, such as AAH8, that has a high affinity for both the μOR and the δOR. Specifically, we identified novel chemical group substituents that were responsible for the peptidomimetics having a subnanomolar affinity for both the μOR and δOR (i.e. the N-acetyl group in AAH8). It is worth re-emphasizing that a balanced affinity profile, in which the compounds demonstrate no binding selectivity for μOR over δOR, seems to be key for the reduced side effect profile that was observed in vivo (i.e. AMB46 versus AMB47). In addition, the 2-naphthyl (AAH8) and 2-methyl-THIQ

98 (AFN42) were bulky substituents that were important for the analogues to produce δOR

antagonism.

The in-depth characterization of the small set of peptidomimetics in chapter III is

significant for a few reasons. The various methods and cell lines that were used to precisely

quantify the agonism of the peptidomimetics at both the μOR and the δOR allowed for a deeper

understanding of the mechanism of action of bifunctional compounds at the opioid receptors and

how this might explain the beneficial pharmacology generated by such compounds. The analogue

AFN42, which is 1000-fold more potent than DAMGO and is insensitive to Na+ ions and guanine

nucleotide, is considered a ‘superagonist’ at the μOR. Its superagonism at μOR is comparable to

that of BU72, which was used by crystallographers to solve the active state crystal structure of the

μOR. AFN42, demonstrates potential as a very high affinity ligand that could be used to solve the

active state conformation of the μOR bound to a “peptidomimetic”. If this were to be achieved, it could provide additional insight into how the μOR interacts with peptide-like compounds, such as

DAMGO, and this would be key in understanding the nature of endogenous μOR signaling.

The μOR agonist / δOR antagonist pharmacological profile may be promising for the development of bifunctional opioid compounds as non-addictive analgesics. With that being said, several of the bifunctional μOR agonist / δOR antagonist peptidomimetics are significantly biased

99 for β-arrestin recruitment over G protein signaling, including a lead peptidomimetic (AMB47) that did not produce the development of antinociceptive tolerance or physical dependence in our in vivo behavioral evaluation paradigm. This may raise some criticism to the hypotheses’ that β- arrestin signaling is bad and G protein-biased compounds are good. Moreover, our β-arrestin biased and G protein-biased bifunctional opioid peptidomimetics are innovative pharmacological tools that can be utilized to investigate the relatively new concept of biased agonism that is still actively being discussed and defined. Although only a small series of peptidomimetics were examined thus far for their ability to recruit β-arrestin, we could identify and assess other compounds of interest and in the process create a library of compounds with bias factors that would be available for comparative analysis. Since numerous peptidomimetics have already been characterized in vitro for their G protein activation and in vivo for their behavioral evaluation, we could carry out a β-arrestin recruitment study and then compare the bias factors to the in vivo behavioral effects. The results of this study may lend support to, or dispute, the emerging hypothesis that G protein-biased compounds demonstrate greater therapeutic potential and a reduced side effect profile. In conclusion, these μOR agonist / δOR antagonist bifunctional opioid peptidomimetics are not only beneficial pharmacological tools that should be used to explore the promise of biased agonism, but they are worthy contenders in the next generation of opioid analgesics.

Another future direction is to follow up on the studies that sought to examine the role of the δOR in the development of μOR tolerance using SH-SY5Y cells. Additional experiments could be performed to examine if δOR expression levels change following chronic ICI 174,864 treatment and sustained δOR antagonism. Saturation binding experiments could be performed using [3H]-

DPN and the μOR antagonist CTAP to precisely measure δOR expression. Also, stimulation of

100 GTPγ35S binding by the δOR-selective compound SNC80 can be performed to assess changes in

δOR expression and function. Additionally, confocal microscopy using a different cell line (i.e.

CHO-Flag-μOR-Myc-δOR cells) may be a valuable technique to quantify the ratio of μOR and

δOR on the plasma membrane and measure the extent of μOR and δOR internalization following

ICI 174,864 treatment, DAMGO treatment, or combination treatment. Collectively, these

experiments should be adequate to address limitations of the experiments presented in Figure A1.

Initial attempts using CRISPR-Cas9 were unsuccessful in knocking out the δOR in SH-

SY5Y cells. This alternative molecular biology approach would complement the pharmacological approach of using the selective δOR inverse agonist ICI 174,864 to antagonize the receptor, and also demonstrate how the absence of δOR signaling influences μOR signaling and μOR-agonist tolerance. I anticipate the results of the CRISPR-Cas9 δOR knockout experiments to mimic that which was observed in in Figure A1. If the opposite were to be observed, whereby stimulation of

GTPγ35S binding by DAMGO in tolerant SH-SY5Y cell membranes were to increase in δOR

knockout conditions, this would (i) support the current literature that suggests δOR directly

interacts with the μOR (George et al., 2002) and (ii) support the proposed mechanism of action by

which δOR antagonism reduces the development of μOR tolerance at the level of the single cell

as opposed to at the systems level. In parallel to using CRISPR-Cas9 to knockout δOR, Furchgott

experiments and analysis is a simple method to decrease the number of functional μORs in the SH-

SY5Y cells and provide a more equivalent ratio of μORs and δORs. Following treatment with beta-Funaltrexamine (B-FNA) in SH-SY5Y cells, the aforementioned experiments could be performed to explore the molecular interactions between μOR and δOR at a more stoichiometric level and further assess how δOR inhibition may regulate μOR-mediated G protein signaling.

101 Moving in a drug development direction, pursuing the pharmacodynamic profile of AFN42

may lead to the generation of opioid analgesics with an improvement in clinical efficacy. AFN42,

with such a high degree of μOR agonism is likely to produce analgesia, however, it may produce

respiratory depression. Indeed, AFN42 is a δOR antagonist and previous studies have shown that

δOR antagonism lessens respiratory depressive effects (Freye, Latasch and Portoghese, 1992; Su,

McNutt and Chang, 1998). Experiments assessing the respiratory depressive effects of the

bifunctional peptidomimetics have not been performed, but they are the logical next steps forward.

Several analogues have been synthesized and characterized, at least in vitro, that are partial agonists at μOR and could be safer (Ridzwan et al., 2012). For example, the properties of buprenorphine that afford its safety profile, namely its μOR partial agonism, lend credibility to the potential safety of bifunctional μOR partial agonist / δOR antagonist peptidomimetics. A valuable set of experiments would be to (i) compare very potent, balanced affinity bifunctional peptidomimetics with varying degrees of μOR agonism in vitro and in vivo and (ii) determine their potential as safe opioid analgesics by testing their ability to suppress respiratory function in vivo.

Additionally, the lack of significant δOR efficacy in the peptidomimetics could be an important safety feature as well, as some δOR full agonists such as SNC80 produce convulsions in animal models (Jutkiewicz, 2005; Chu Sin Chung et al., 2015). Therefore, a potent μOR partial agonist /

δOR antagonist profile or a potent μOR partial agonist / δOR partial agonist profile both appear to be promising avenues to pursue with regard to safety and clinical utility. In conclusion, the overarching goal of our studies is to optimize the properties of our bifunctional peptidomimetic compounds and ultimately develop safe and non-addictive opioid analgesics that may mitigate the risks of opioid misuse, addiction, and overdose.

102 Bibliography

Abdelhamid, E. E. et al. (1991) ‘Selective blockage of delta opioid receptors prevents the development of morphine tolerance and dependence in mice’, J Pharmacol Exp Ther, 258(1), pp. 299–303. Available at: https://www.ncbi.nlm.nih.gov/pubmed/1649297.

Ahrnsbrak, R. et al. (2017) ‘Key Substance Use and Mental Health Indicators in the United States: Results from the 2016 National Survey on Drug Use and Health’, Substance Abuse and Mental Health Services Administration. doi: 10.1016/j.drugalcdep.2016.10.042.

Al-Hasani, R. and Bruchas, M. R. (2011) ‘Molecular mechanisms of opioid receptor-dependent signaling and behavior’, Anesthesiology. doi: 10.1097/ALN.0b013e318238bba6.

Alt, A et al. (2001) ‘Stimulation of guanosine-5’-o-(3-[35S]thio)triphosphate binding in digitonin- permeabilized C6 rat glioma cells: evidence for an organized association of mu-opioid receptors and G protein.’, The Journal of pharmacology and experimental therapeutics.

Altarifi, A. A. et al. (2017) ‘Effects of acute and repeated treatment with the biased mu opioid receptor agonist TRV130 (oliceridine) on measures of antinociception, gastrointestinal function, and abuse liability in rodents’, Journal of Psychopharmacology. doi: 10.1177/0269881116689257.

Anand, J. P. et al. (2016) ‘The behavioral effects of a mixed efficacy antinociceptive peptide, VRP26, following chronic administration in mice’, Psychopharmacology. doi: 10.1007/s00213- 016-4296-8.

Anand, J. P. et al. (2018) ‘In vivo effects of μ-opioid receptor agonist/δ-opioid receptor antagonist peptidomimetics following acute and repeated administration’, British Journal of Pharmacology. doi: 10.1111/bph.14148.

Anand, J. P. and Montgomery, D. (2018) ‘Multifunctional opioid ligands’, in Handbook of Experimental Pharmacology. doi: 10.1007/164_2018_104.

Ananthan, S. (2008) ‘Opioid ligands with mixed μ/δ opioid receptor interactions: An emerging approach to novel analgesics’, in Drug Addiction: From Basic Research to Therapy. doi: 10.1007/978-0-387-76678-2_23.

Angst, M. S. and Clark, J. D. (2006) ‘Opioid-induced hyperalgesia: a qualitative systematic review.’, Anesthesiology.

103 Austin Zamarripa, C. et al. (2018) ‘The G-protein biased mu-opioid agonist, TRV130, produces reinforcing and antinociceptive effects that are comparable to oxycodone in rats’, Drug and Dependence. doi: 10.1016/j.drugalcdep.2018.08.002.

Balboni, G. et al. (2010) ‘Evolution of the Bifunctional Lead mu Agonist / delta Antagonist Containing the Dmt-Tic Opioid Pharmacophore’, ACS Chem Neurosci, 1(2), pp. 155–164. doi: 10.1021/cn900025j.

Barshop, K. and Staller, K. (2017) ‘Eluxadoline in irritable bowel syndrome with diarrhea: rationale, evidence and place in therapy’, Therapeutic Advances in Chronic Disease. doi: 10.1177/2040622317714389.

Beckett, A. H. and Casy, A. F. (1954) ‘SYNTHETIC ANALGESICS: STEREOCHEMICAL CONSIDERATIONS’, Journal of Pharmacy and Pharmacology. doi: 10.1111/j.2042- 7158.1954.tb11033.x.

Bender, A. M., Griggs, N. W., Anand, J. P., et al. (2015) ‘Asymmetric Synthesis and in Vitro and in Vivo Activity of Tetrahydroquinolines Featuring a Diverse Set of Polar Substitutions at the 6 Position as Mixed-Efficacy μ Opioid Receptor/δ Opioid Receptor Ligands’, ACS Chemical Neuroscience. doi: 10.1021/acschemneuro.5b00100.

Bender, A. M., Griggs, N. W., Gao, C., et al. (2015) ‘Rapid Synthesis of Boc-2′,6′-dimethyl- l - tyrosine and Derivatives and Incorporation into Opioid Peptidomimetics’, ACS Medicinal Chemistry Letters, 6(12), pp. 1199–1203. doi: 10.1021/acsmedchemlett.5b00344.

Besse, D. et al. (1990) ‘Pre- and postsynaptic distribution of μ, δ and κ opioid receptors in the superficial layers of the cervical dorsal horn of the rat spinal cord’, Brain Research. doi: 10.1016/0006-8993(90)91519-M.

Blackburn, T. P. et al. (1988) ‘Autoradiographic localization of delta opiate receptors in rat and human brain’, Neuroscience. doi: 10.1016/0306-4522(88)90283-7.

Bohn, L. M. et al. (1999) ‘Enhanced morphine analgesia in mice lacking β-arrestin 2’, Science. doi: 10.1126/science.286.5449.2495.

Bohn, L. M. and Aubé, J. (2017) ‘Seeking (and Finding) Biased Ligands of the Kappa Opioid Receptor’, ACS Medicinal Chemistry Letters. doi: 10.1021/acsmedchemlett.7b00224.

Breslin, H. J. et al. (2012) ‘Identification of a dual δ or antagonist/μ or agonist as a potential therapeutic for diarrhea-predominant Irritable Bowel Syndrome (IBS-d)’, Bioorganic and Medicinal Chemistry Letters. doi: 10.1016/j.bmcl.2012.05.042.

Brook, K., Bennett, J. and Desai, S. P. (2017) ‘The Chemical History of Morphine: An 8000-year Journey, from Resin to de-novo Synthesis’, Journal of Anesthesia History. doi: 10.1016/j.janh.2017.02.001.

104 Burford, N. T. et al. (2015) ‘Discovery, synthesis, and molecular pharmacology of selective positive allosteric modulators of the delta-opioid receptor’, J Med Chem, 58(10), pp. 4220–4229. doi: 10.1021/acs.jmedchem.5b00007.

Cahill, C. M. et al. (2016) ‘Allostatic Mechanisms of Opioid Tolerance Beyond Desensitization and Downregulation’, Trends in Pharmacological Sciences. doi: 10.1016/j.tips.2016.08.002. Chavkin, C. and Martinez, D. (2015) ‘Kappa Antagonist JDTic in Phase 1 Clinical Trial’, Neuropsychopharmacology. doi: 10.1038/npp.2015.74.

Che, T. et al. (2018) ‘Structure of the Nanobody-Stabilized Active State of the Kappa Opioid Receptor’, Cell. doi: 10.1016/j.cell.2017.12.011.

Chefer, V. I. and Shippenberg, T. S. (2009) ‘Augmentation of morphine-induced sensitization but reduction in morphine tolerance and reward in delta-opioid receptor knockout mice’, Neuropsychopharmacology. doi: 10.1038/npp.2008.128.

Chen, K. Y., Chen, L. and Mao, J. (2014) ‘Buprenorphine-naloxone therapy in pain management’, Anesthesiology. doi: 10.1097/ALN.0000000000000170.

Chen, Y. et al. (1994) ‘Molecular cloning and functional expression of a mu opioid receptor from rat brain’, Regulatory Peptides. doi: 10.1016/0167-0115(94)90214-3.

Chu Sin Chung, P. et al. (2015) ‘Delta opioid receptors expressed in forebrain GABAergic neurons are responsible for SNC80-induced seizures’, Behavioural Brain Research. doi: 10.1016/j.bbr.2014.10.029.

Clark, J. A. et al. (1986) ‘[D-Pen2,D-Pen5]enkephalin (DPDPE): a δ-selective enkephalin with low affinity for μ1opiate binding sites’, European Journal of Pharmacology. doi: 10.1016/0014- 2999(86)90784-3.

Clark, M. J. et al. (2003) ‘Endogenous RGS protein action modulates μ-opioid signaling through Gα o : Effects on adenylyl cyclase, extracellular signal-regulated kinases, and intracellular calcium pathways’, Journal of Biological Chemistry. doi: 10.1074/jbc.M208885200.

Dahlhamer, J. et al. (2018) ‘Prevalence of Chronic Pain and High-Impact Chronic Pain Among Adults - United States, 2016.’, MMWR. Morbidity and mortality weekly report. doi: 10.15585/mmwr.mm6736a2.

Devi, L. A. et al. (2004) ‘A role for heterodimerization of μ and δ opiate receptors in enhancing morphine analgesia’, Proceedings of the National Academy of Sciences. doi: 10.1073/pnas.0307601101.

DeWire, S. M. et al. (2013) ‘A G protein-biased ligand at the mu-opioid receptor is potently analgesic with reduced gastrointestinal and respiratory dysfunction compared with morphine’, J Pharmacol Exp Ther, 344(3), pp. 708–717. doi: 10.1124/jpet.112.201616.

105 Disney, A. et al. (2018) ‘The novel μ-opioid receptor agonist PZM21 depresses respiration and induces tolerance to antinociception’, British Journal of Pharmacology. doi: 10.1111/bph.14224.

Ehlert, F. J. (1985) ‘The relationship between muscarinic receptor occupancy and adenylate cyclase inhibition in the rabbit myocardium.’, Molecular Pharmacology.

Ehlert, F. J. (2018) ‘Analysis of Biased Agonism’, in Progress in Molecular Biology and Translational Science. doi: 10.1016/bs.pmbts.2018.08.001.

Elliott, J., Guo, L. and Traynor, J. R. (1997) ‘Tolerance to μ-opioid agonists in human neuroblastoma SH-SY5Y cells as determined by changes in guanosine-5’-O-(3-[35S]- thio)triphosphate binding’, British Journal of Pharmacology. doi: 10.1038/sj.bjp.0701253.

Elzey, M. J., Barden, S. M. and Edwards, E. S. (2016) ‘Patient Characteristics and Outcomes in Unintentional, Non-fatal Prescription Opioid Overdoses: A Systematic Review’, Pain Physician.

Erbs, E. et al. (2015) ‘A mu-delta opioid receptor brain atlas reveals neuronal co-occurrence in subcortical networks’, Brain Struct Funct, 220(2), pp. 677–702. doi: 10.1007/s00429-014-0717- 9.

Evans, C. J. et al. (1992) ‘Cloning of a delta opioid receptor by functional expression’, Science. doi: 10.1126/science.1335167.

Fenalti, G. et al. (2014) ‘Molecular control of δ-opioid receptor signalling’, Nature. doi: 10.1038/nature12944.

Fenalti, G. et al. (2015) ‘Structural basis for bifunctional peptide recognition at human δ-opioid receptor’, Nature Structural and Molecular Biology. doi: 10.1038/nsmb.2965.

Freye, E., Latasch, L. and Portoghese, P. S. (1992a) ‘The delta receptor is involved in sufentanil- induced respiratory depression--opioid subreceptors mediate different effects’, Eur J Anaesthesiol.

Freye, E., Latasch, L. and Portoghese, P. S. (1992b) ‘The delta receptor is involved in sufentanil- induced respiratory depression--opioid subreceptors mediate different effects’, European journal of anaesthesiology.

Fujita, W. et al. (2014) ‘Molecular characterization of eluxadoline as a potential ligand targeting mu-delta opioid receptor heteromers’, Biochemical Pharmacology. doi: 10.1016/j.bcp.2014.09.015.

Fundytus, M. E. et al. (1995) ‘Attenuation of morphine tolerance and dependence with the highly selective δ-opioid receptor antagonist TIPP[ψ]’, European Journal of Pharmacology. doi: 10.1016/0014-2999(95)00554-X.

Gaskin, D. J. and Richard, P. (2012) ‘The economic costs of pain in the United States’, Journal of Pain. doi: 10.1016/j.jpain.2012.03.009.

106

George, S. R. et al. (2002) ‘Oligomerization of μ- and δ-Opioid Receptors’, Journal of Biological Chemistry. doi: 10.1074/jbc.m000345200.

Gilbert, P. E. and Martin, W. R. (1976) ‘The effects of morphine and -like drugs in the nondependent, morphine-dependent and -dependent chronic spinal dog.’, The Journal of pharmacology and experimental therapeutics.

Gomes, I. et al. (2000) ‘Heterodimerization of mu and delta opioid receptors: A role in opiate synergy.’, The Journal of neuroscience : the official journal of the Society for Neuroscience.

Gomes, I. et al. (2011) ‘G Protein-Coupled Receptor Heteromerization: A Role in Allosteric Modulation of Ligand Binding’, Molecular Pharmacology. doi: 10.1124/mol.110.070847.

Granier, S. et al. (2012) ‘Structure of the δ-opioid receptor bound to naltrindole’, Nature. doi: 10.1038/nature11111.

Griffin, M. T. et al. (2007) ‘Estimation of Agonist Activity at G Protein-Coupled Receptors: Analysis of M2 Muscarinic Receptor Signaling through Gi/o,Gs, and G15’, Journal of Pharmacology and Experimental Therapeutics. doi: 10.1124/jpet.107.120857.

Günther, T. et al. (2018) ‘Targeting multiple opioid receptors – improved analgesics with reduced side effects?’, British Journal of Pharmacology. doi: 10.1111/bph.13809.

Hansen, D. W. et al. (1992) ‘Systemic Analgesic Activity and δ-Opioid Selectivity in [2,6- Dimethyl-Tyr1, D-Pen2, D-Pen5]enkephalin’, Journal of Medicinal Chemistry. doi: 10.1021/jm00082a008.

Harland, A. A. et al. (2015) ‘Further Optimization and Evaluation of Bioavailable, Mixed-Efficacy μ-Opioid Receptor (MOR) Agonists/δ-Opioid Receptor (DOR) Antagonists: Balancing MOR and DOR Affinities’, Journal of Medicinal Chemistry. doi: 10.1021/acs.jmedchem.5b01270.

Harland, A. A. et al. (2016) ‘Effects of N-Substitutions on the Tetrahydroquinoline (THQ) Core of Mixed-Efficacy μ-Opioid Receptor (MOR)/δ-Opioid Receptor (DOR) Ligands’, Journal of Medicinal Chemistry. doi: 10.1021/acs.jmedchem.6b00308.

Hayhurst, C. J. and Durieux, M. E. (2016) ‘Differential Opioid Tolerance and Opioid-induced Hyperalgesia: A Clinical Reality’, Anesthesiology. doi: 10.1097/ALN.0000000000000963.

He, S. Q. et al. (2011) ‘Facilitation of μ-Opioid Receptor Activity by Preventing δ-Opioid Receptor-Mediated Codegradation’, Neuron. doi: 10.1016/j.neuron.2010.12.001.

Healy, J. R. et al. (2013) ‘Synthesis, modeling, and pharmacological evaluation of UMB 425, a mixed mu agonist/delta antagonist opioid analgesic with reduced tolerance liabilities’, ACS Chem Neurosci, 4(9), pp. 1256–1266. doi: 10.1021/cn4000428.

107 Healy, J. R. et al. (2017) ‘Benzylideneoxymorphone: A new lead for development of bifunctional mu/delta opioid receptor ligands’, Bioorganic and Medicinal Chemistry Letters, 27(3), pp. 666– 669. doi: 10.1016/j.bmcl.2016.11.057.

Hepburn, M. J. et al. (1997) ‘Differential effects of naltrindole on morphine-induced tolerance and physical dependence in rats’, J Pharmacol Exp Ther, 281(3), pp. 1350–1356. Available at: https://www.ncbi.nlm.nih.gov/pubmed/9190871.

Holzer, P. (2008) ‘New approaches to the treatment of opioid-induced constipation’, European Review for Medical and Pharmacological Sciences.

Huang, W. et al. (2015) ‘Structural insights into μ-opioid receptor activation’, Nature. doi: 10.1038/nature14886.

Jalal, H. et al. (2018) ‘Changing dynamics of the drug overdose epidemic in the United States from 1979 through 2016’, Science. doi: 10.1126/science.aau1184.

Jutkiewicz, E. M. (2005) ‘Differential Behavioral Tolerance to the δ-Opioid Agonist SNC80 ([(+)- 4-[( R)- -[(2S,5R)-2,5-Dimethyl-4-(2-propenyl)-1-piperazinyl]-(3-methoxyphenyl)methyl]-N,N- diethylbenzamide) in Sprague-Dawley Rats’, Journal of Pharmacology and Experimental Therapeutics. doi: 10.1124/jpet.105.088831.

Kenakin, T. (2017) ‘A Scale of Agonism and Allosteric Modulation for Assessment of Selectivity, Bias, and Receptor Mutation’, Molecular Pharmacology. doi: 10.1124/mol.117.108787.

Kennedy, N. M. et al. (2018) ‘Optimization of a Series of Mu Opioid Receptor (MOR) Agonists with High G Protein Signaling Bias’, J Med Chem. doi: 10.1021/acs.jmedchem.8b01136.

Kest, B. et al. (1996) ‘An antisense oligodeoxynucleotide to the delta opioid receptor (DOR-1) inhibits morphine tolerance and acute dependence in mice’, Brain Research Bulletin. doi: 10.1016/0361-9230(95)02092-6.

Kieffer, B. L. and Gavériaux-Ruff, C. (2002) ‘Exploring the opioid system by gene knockout’, Progress in Neurobiology. doi: 10.1016/S0301-0082(02)00008-4.

Kolodny, A. et al. (2015) The Prescription Opioid and Heroin Crisis: A Public Health Approach to an Epidemic of Addiction, SSRN. doi: 10.1146/annurev-publhealth-031914-122957.

Lee, K. O. et al. (1999) ‘Differential binding properties of oripavines at cloned μ- and δ-opioid receptors’, European Journal of Pharmacology. doi: 10.1016/S0014-2999(99)00460-4.

Lembo, A. J. et al. (2016) ‘Eluxadoline for Irritable Bowel Syndrome with Diarrhea.’, The New England journal of medicine. doi: 10.1056/NEJMoa1505180.

Levio, S. and Cash, B. D. (2017) ‘The place of eluxadoline in the management of irritable bowel syndrome with diarrhea’, Therapeutic Advances in Gastroenterology. doi:

108 10.1177/1756283X17721152.

Lewis, J. and Husbands, S. (2005) ‘The Orvinols and Related Opioids - High Affinity Ligands with Diverse Efficacy Profiles’, Current Pharmaceutical Design. doi: 10.2174/1381612043453027.

Livingston, K. E. et al. (2018) ‘Measuring ligand efficacy at the mu-opioid receptor using a conformational biosensor’, eLife. doi: 10.7554/eLife.32499.

Livingston, K. E. and Traynor, J. R. (2014) ‘ Disruption of the Na + ion binding site as a mechanism for positive allosteric modulation of the mu-opioid receptor ’, Proceedings of the National Academy of Sciences. doi: 10.1073/pnas.1415013111.

Lord, J. A. H. et al. (1977) ‘Endogenous opioid peptides: Multiple agonists and receptors’, Nature. doi: 10.1038/267495a0.

Manglik, A. et al. (2012) ‘Crystal structure of the μ-opioid receptor bound to a morphinan antagonist’, Nature. doi: 10.1038/nature10954.

Manglik, A. et al. (2016) ‘Structure-based discovery of opioid analgesics with reduced side effects’, Nature, 537(7619), pp. 185–190. doi: 10.1038/nature19112.

Mansour, A. et al. (1994) ‘Mu, delta, and kappa opioid receptor mRNA expression in the rat CNS: An in situ hybridization study’, Journal of Comparative Neurology. doi: 10.1002/cne.903500307.

Mao, J. (2002) ‘Opioid-induced abnormal pain sensitivity: implications in clinical opioid therapy’, Pain. doi: 10.1016/s0304-3959(02)00422-0.

Martin, T. J. et al. (2000) ‘Antagonism of delta(2)-opioid receptors by naltrindole-5’- isothiocyanate attenuates heroin self-administration but not antinociception in rats.’, The Journal of pharmacology and experimental therapeutics.

McFadyen, I. J. et al. (2000) ‘Tetrapeptide derivatives of [D-Pen(2),D-Pen(5)]-enkephalin (DPDPE) lacking an N-terminal tyrosine residue are agonists at the mu-opioid receptor’, The Journal of Pharmacology and Experimental Therapeutics. doi: 10.1016/j.neubiorev.2017.07.003.

Mollereau, C. et al. (1994) ‘ORL1, a novel member of the opioid receptor family. Cloning, functional expression and localization’, FEBS Letters. doi: 10.1016/0014-5793(94)80235-1.

Morphy, R. and Rankovic, Z. (2006) ‘The physicochemical challenges of designing multiple ligands’, Journal of Medicinal Chemistry. doi: 10.1021/jm0603015.

Mosberg, H. I. et al. (1983) ‘Cyclic penicillamine containing enkephalin analogs display profound delta receptor selectivities’, Life Sciences. doi: 10.1016/0024-3205(83)90538-6.

Mosberg, H. I. et al. (2013) ‘Opioid peptidomimetics: Leads for the design of bioavailable mixed

109 efficacy μ opioid receptor (MOR) agonist/δ opioid receptor (DOR) antagonist ligands’, Journal of Medicinal Chemistry, 56(5), pp. 2139–2149. doi: 10.1021/jm400050y.

Mosberg, H. I. et al. (2014) ‘Development of a bioavailable μ opioid receptor (MOPr) agonist, δ opioid receptor (DOPr) antagonist peptide that evokes antinociception without development of acute tolerance’, Journal of Medicinal Chemistry. doi: 10.1021/jm5002088.

Mosberg, H. I. and Fowler, C. B. (2002) ‘Development and validation of opioid ligand-receptor interaction models: The structural basis of mu vs. delta selectivity’, Journal of Peptide Research. doi: 10.1034/j.1399-3011.2002.21061.x.

Nahin, R. L. (2015) ‘Estimates of Pain Prevalence and Severity in Adults: United States, 2012’, Journal of Pain. doi: 10.1016/j.jpain.2015.05.002.

Nastase, A. F. et al. (2018) ‘Synthesis and Pharmacological Evaluation of Novel C-8 Substituted Tetrahydroquinolines as Balanced-Affinity Mu/Delta Opioid Ligands for the Treatment of Pain’, ACS Chemical Neuroscience. doi: 10.1021/acschemneuro.8b00139.

Neilan, C. L. et al. (2004) ‘Characterization of the complex morphinan derivative BU72 as a high efficacy, long-lasting mu-opioid receptor agonist’, European Journal of Pharmacology. doi: 10.1016/j.ejphar.2004.07.097.

Pert, C. B. and Snyder, S. H. (1976) ‘Opiate receptor binding-enhancement by opiate administration in vivo’, Biochemical Pharmacology. doi: 10.1016/0006-2952(76)90157-X.

Pert, C. and Snyder, S. (1973) ‘Opiate Receptor: Demonstration in Nervous Tissue 1973_Pert.pdf’, Science.

Pogozheva, I. D., Przydzial, M. J. and Mosberg, H. I. (2008) ‘Homology modeling of opioid receptor-ligand complexes using experimental constraints’, in Drug Addiction: From Basic Research to Therapy. doi: 10.1007/978-0-387-76678-2_33.

Portoghese, P. S. (1965) ‘A New Concept on the Mode of Interaction of Analgesics with Receptors’, Journal of Medicinal Chemistry. doi: 10.1021/jm00329a013.

Purington, L. C. et al. (2011) ‘Development and in vitro characterization of a novel bifunctional μ-Agonist/δ-Antagonist opioid tetrapeptide’, ACS Chemical Biology. doi: 10.1021/cb200263q.

Raehal, K. M., Walker, J. K. and Bohn, L. M. (2005) ‘Morphine side effects in beta-arrestin 2 knockout mice’, J. Pharmacol. Exper. Ther.

Rankovic, Z., Brust, T. F. and Bohn, L. M. (2016) ‘Biased agonism: An emerging paradigm in GPCR drug discovery’, Bioorganic and Medicinal Chemistry Letters. doi: 10.1016/j.bmcl.2015.12.024.

Ridzwan, I. E. et al. (2012) ‘A single compound alternative to A Buprenorphine/Naltrexone

110 combination to prevent to drug addiction’, British Journal of Clinical Pharmacology.

Ross, E. M. and Wilkie, T. M. (2002) ‘GTPase-Activating Proteins for Heterotrimeric G Proteins: Regulators of G Protein Signaling (RGS) and RGS-Like Proteins’, Annual Review of Biochemistry. doi: 10.1146/annurev.biochem.69.1.795.

S., B., K., C. and F., S. (2018) ‘Safety of oliceridine, a G protein-biased ligand at the μ-opioid receptor, in patients with moderate-to-severe acute pain after colorectal surgery: Results from a Phase-3, open-label study’, Diseases of the Colon and Rectum. doi: 10.1097/DCR.0000000000001104 LK

Schiller, P. W., Fundytus, M. E., et al. (1999) ‘The opioid mu agonist/delta antagonist DIPP- NH(2)[Psi] produces a potent analgesic effect, no physical dependence, and less tolerance than morphine in rats’, J Med Chem, 42(18), pp. 3520–3526. doi: 10.1021/jm980724+.

Schiller, P. W., Weltrowska, G., et al. (1999) ‘The TIPP opioid peptide family: development of delta antagonists, delta agonists, and mixed mu agonist/delta antagonists’, Biopolymers, 51(6), pp. 411–425. doi: 10.1002/(SICI)1097-0282(1999)51:6<411::AID-BIP4>3.0.CO;2-Z.

Schiller, P. W. (2010) ‘Bi- or multifunctional opioid peptide drugs’, Life Sciences. doi: 10.1016/j.lfs.2009.02.025.

Schmid, C. L. et al. (2017) ‘Bias Factor and Therapeutic Window Correlate to Predict Safer Opioid Analgesics’, Cell. doi: 10.1016/j.cell.2017.10.035.

Shiotani, K. et al. (2007) ‘Design and synthesis of opioidmimetics containing 2′,6′-dimethyl-l- tyrosine and a pyrazinone-ring platform’, Bioorganic and Medicinal Chemistry Letters. doi: 10.1016/j.bmcl.2007.08.058.

Shippenberg, T. S., Chefer, V. I. and Thompson, A. C. (2009) ‘Delta-Opioid Receptor Antagonists Prevent Sensitization to the Conditioned Rewarding Effects of Morphine’, Biological Psychiatry. doi: 10.1016/j.biopsych.2008.09.009.

Siuda, E. R. et al. (2017) ‘Biased mu-opioid receptor ligands: a promising new generation of pain therapeutics’, Current Opinion in Pharmacology. doi: 10.1016/j.coph.2016.11.007.

Soergel, D. G. et al. (2014) ‘First clinical experience with TRV130: Pharmacokinetics and pharmacodynamics in healthy volunteers’, Journal of Clinical Pharmacology. doi: 10.1002/jcph.207.

Spahn, V. and Stein, C. (2017) ‘Targeting delta opioid receptors for pain treatment: drugs in phase I and II clinical development’, Expert Opinion on Investigational Drugs. doi: 10.1080/13543784.2017.1275562.

STEPHENSON, R. P. (1956) ‘A modification of receptor theory.’, British journal of pharmacology and chemotherapy. doi: 10.1111/j.1476-5381.1956.tb00006.x.

111

Su, Y. F., McNutt, R. W. and Chang, K. J. (1998) ‘Delta-opioid ligands reverse alfentanil-induced respiratory depression but not antinociception’, J Pharmacol Exp Ther.

Szekeres, P. G. and Traynor, J. R. (1997) ‘Delta opioid modulation of the binding of guanosine- 5’-O-(3-[35S]thio)triphosphate to NG108-15 cell membranes: characterization of agonist and inverse agonist effects.’, The Journal of pharmacology and experimental therapeutics.

The Lancet (2017) ‘The opioid crisis in the USA: a public health emergency’, The Lancet. doi: 10.1016/s0140-6736(17)32808-8.

Traynor, J. R. and Elliott, J. (1993) ‘δ-Opioid receptor subtypes and cross-talk with μ-receptors’, Trends in Pharmacological Sciences. doi: 10.1016/0165-6147(93)90068-U.

Traynor, J. R. and Nahorski, S. R. (1995) ‘Modulation by mu-opioid agonists of guanosine-5’-O- (3-[35S]thio)triphosphate binding to membranes from human neuroblastoma SH-SY5Y cells.’, Molecular pharmacology.

Trescot, A. M. et al. (2008) ‘Effectiveness of opioids in the treatment of chronic non-cancer pain.’, Pain physician.

Váradi, A. et al. (2016) ‘Mitragynine/Corynantheidine Pseudoindoxyls As Opioid Analgesics with Mu Agonism and Delta Antagonism, Which Do Not Recruit β-Arrestin-2’, Journal of Medicinal Chemistry. doi: 10.1021/acs.jmedchem.6b00748.

Volkow, N. D. and Collins, F. S. (2017) ‘The Role of Science in Addressing the Opioid Crisis’, N Engl J Med, 377(4), pp. 391–394. doi: 10.1056/NEJMsr1706626.

Wang, C. et al. (1998) ‘Design of a high affinity peptidomimetic opioid agonist from peptide pharmacophore models’, Bioorganic and Medicinal Chemistry Letters. doi: 10.1016/S0960- 894X(98)00472-7.

Wang, D et al. (2018) ‘Functional Divergence of Delta and Mu Opioid Receptor Organization in CNS Pain Circuits’, Neuron, 98(1), pp. 90-108 e5. doi: 10.1016/j.neuron.2018.03.002.

Wang, Dong et al. (2018) ‘Functional Divergence of Delta and Mu Opioid Receptor Organization in CNS Pain Circuits’, Neuron. doi: 10.1016/j.neuron.2018.03.002.

Wang, H.-B. et al. (2010) ‘Coexpression of δ- and μ-opioid receptors in nociceptive sensory neurons’, Proceedings of the National Academy of Sciences. doi: 10.1073/pnas.1008382107.

Weber, L., Yeomans, D. C. and Tzabazis, A. (2017) ‘Opioid-induced hyperalgesia in clinical anesthesia practice: what has remained from theoretical concepts and experimental studies?’, Current Opinion in Anaesthesiology. doi: 10.1097/ACO.0000000000000485.

Williams, J. T. et al. (2013) ‘Regulation of MOR-Opioid Receptors: Desensitization,

112 Phosphorylation, Internalization, and Tolerance’, Pharmacological Reviews. doi: 10.1124/pr.112.005942.

Wu, H. et al. (2012) ‘Structure of the human κ-opioid receptor in complex with JDTic’, Nature. doi: 10.1038/nature10939.

Yasuda, K. et al. (1993) ‘Cloning and functional comparison of kappa and delta opioid receptors from mouse brain.’, Proceedings of the National Academy of Sciences. doi: 10.1073/pnas.90.14.6736.

Zadina, J. E. et al. (1997) ‘A potent and selective endogenous agonist for the μ-opiate receptor’, Nature. doi: 10.1038/386499a0.

Zhao, G.-M. (2003) ‘Profound Spinal Tolerance after Repeated Exposure to a Highly Selective micro-Opioid Peptide Agonist: Role of delta -Opioid Receptors’, Journal of Pharmacology and Experimental Therapeutics. doi: 10.1124/jpet.302.1.188.

Zhu, Y. et al. (1999) ‘Retention of supraspinal delta-like analgesia and loss of morphine tolerance in delta opioid receptor knockout mice’, Neuron, 24(1), pp. 243–252. Available at: https://www.ncbi.nlm.nih.gov/pubmed/10677041.

113