The Un-Design and Design of Insulin: Structural Evolution

Home , 1TYM, Hyperproinsulinemia, Inhalable insulin, Methionine

THE UN-DESIGN AND DESIGN OF INSULIN: STRUCTURAL EVOLUTION

WITH APPLICATION TO THERAPEUTIC DESIGN

NISCHAY K. REGE

Submitted in partial fulfillment of the requirements for the degree of Doctor of

Philosophy

Department of Biochemistry

Dissertation Advisor: Dr. Michael A. Weiss

CASE WESTERN RESERVE UNIVERSITY

August, 2018

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the thesis/dissertation of

Nischay K. Rege

candidate for the degree of Doctor of Philosophy*.

Committee Chair

Paul Carey

Committee Member

Michael Weiss

Committee Member

Faramarz Ismail-Beigi

Committee Member

George Dubyak

Date of Defense:

June 26th, 2018

*We also certify that written approval has been obtained for any proprietary

material contained therein.

Dedication

This thesis is dedicated to my mother, Dipti, whose constant love and faith have never

failed, to my father, Kiran, who taught me of the virtue of curiosity, and to my wife,

Shipra, whose kindness and companionship have given me enough strength for eight

lifetimes.

Table of Contents

Dedication ...... i Table of Contents ...... ii List of Tables ...... v List of Figures ...... vii Acknowledgements ...... x List of Abbreviations ...... xiv Abstract ...... xvi Chapter 1: Introduction ...... 1 1.1 History of insulin research ...... 1 1.2 Purpose of Dissertation ...... 3 1.3 Insulin Structure ...... 6 1.4 Insulin Oligomerization and Storage ...... 8 1.5 Insulin Biosynthesis and Secretion: ...... 15 1.6 Proinsulin Folding:...... 20 1.7 Insulin Fibrillation ...... 23 1.8 Insulin Physiology ...... 32 1.9 Insulin Receptor Structure ...... 35 1.10 Insulin-IR Complexation ...... 37 1.11 Diabetes Mellitus ...... 43 1.12 Development of Clinical Insulin Analogs: ...... 46 1.13 Conclusion ...... 54 Chapter 2: Toxic Protein Misfolding Constrains the Evolution of Insulin Sequences ...... 58 2.1 Chapter Summary ...... 58 2.2 Introduction ...... 59 2.3 Results: ...... 67 2.4 Discussion ...... 93 2.5 Methods and materials ...... 105 Chapter 3: Identification of a Minimal Prodomain in Insulin: A Weak Hydrogen Bond is Critical for Proinsulin Folding, but Dispensible in the Mature Hormone ...... 111 3.1 Chapter Summary ...... 111 3.2 Introduction ...... 112 Results ...... 117 ii

3.4 Discussion...... 135 3.5 Methods...... 142 Chapter 4: Protein Evolution at the Edge of Non-Foldability: Impaired Biosynthesis of an Insulin Variant with Native Structure and Function ...... 148 4.1 Chapter Summary ...... 148 4.2 Introduction ...... 149 4.3 Results ...... 160 4.4 Discussion ...... 194 4.5 Methods ...... 206 Chapter 5: Structure-Based Stabilization of Insulin as a Therapeutic Protein Assembly via Enhanced Aromatic-Aromatic Interactions ...... 211 5.1 Chapter Summary ...... 211 5.2 Introduction ...... 212 5.3 Results ...... 217 5.4 Discussion ...... 262 5.5 Methods ...... 292 5.6 Endnotes ...... 297 Chapter 6: 4SS-Insulin ...... 299 6.1 Chapter Summary ...... 299 6.2 Introduction ...... 300 6.3 Results ...... 305 6.4 Discussion ...... 327 6.5 Methods ...... 333 6.6 Endnotes ...... 336 Chapter 7: Towards a Glucose-Responsive Insulin Analog ...... 337 7.1 Chapter Summary ...... 337 7.2 Summary of previous GRI Research ...... 337 7.2 Design of Glucose-Responsive Insulin Analogs Containing Internal Glucose-Sensing Motifs ...... 365 7-3 Results ...... 367 7-4 Discussion ...... 394 7-5 Methods...... 396 Chapter 8: Continuing Work, Future Directions, and Conclusions ...... 399 8.1 Summary of Dissertation ...... 399 8.2 Continuing Work and Future Directions ...... 401

iii

8.2 Concluding Remarks ...... 439 Bibliography ...... 441

List of Tables

Table 2-1. Biological Function of Insulin Analogs ...... 73 Table 2-2. Stabilities of Insulin Analogs ...... 80

Table 3-1. Stability of A21-Variant Insulin Analogs ...... 127 2+ Table 3-2. Half-lives of Co -Substituted R6 Hexamers of Insulin Analogs ...... 130 Table 3-3. In Vitro Affinities of A21-Variants for IR ...... 134

Table 4-1. Thermodynamic stabilities of insulin analogs...... 165 Table 4-2a. Thermodynamic stabilities of insulin analogs at 37°C ...... 168 Table 4-2b. Thermodynamic stabilities of insulin analogs at 4°C ...... 168 Table 4-3. Comparison of observed NOEs involving PheB24 of WT (DKP-insulin) and the analog TyrB24...... 178 Table 4-4. Self-association properties of insulin analogs ...... 181 Table 4-5 Fibrillation lag times of insulin analogs...... 187

Table 5-1a. Interaction energy between the B26 residue of an energy minimized model of TrpB26 insulin relative to native TyrB26 in the context of a TRf dimer...... 221 Table 5-1b. Local energy-minimized model of TrpB26 insulin relative to native TyrB26 where polar atoms of Tyrosine and Tryptophan rings are replaced with nonpolar synthetic residues ...... 222 Table 5-1c. Interaction energy between the B26 residue of a local energy-minimized B26 B26 model of Trp insulin relative to native Tyr in the context of an R2 dimer...... 223 Table 5-1d. Interaction energy between the B26 residue of a local energy-minimized model of B26 B26 Trp insulin relative to native Tyr in the context of an T2 dimer...... 224 Table 5-2. Self-association properties of insulin analogs...... 229 Table 5-3. Data collection and refinement statistics pertaining to the crystal structure of TrpB26, OrnB29-insulin...... 245 Table 5-4. RMSD values of main-chain and side-chains of the T-state protomer of the B26 B29 crystal structures of Trp , Orn -insulin and selected WT T6 and T3R3 hexamers ...... 250 Table 5-5. Comparison of χ1 and χ2 dihedral-angle values of the T protomer of the B26 B29 Trp , Orn -insulin crystal structure to the B26 side chains of WT T6 or T3R3 insulin crystal structures ...... 251 Table 5-6. Comparison of χ1 and χ2 dihedral angle values of the R protomomer of B26 B29 f Trp , Orn -insulin crystal structure to the B26 side chains of WT R6 or R insulin crystal structures ...... 252 Table 5-7. Thermodynamic stabilities of insulin analogs...... 254 Table 5-8. Comparison of predicted proton-proton distances involving TrpB26 from TrpB26, OrnB29-insulin crystal structure and TrpB26-associated NOEs in TrpB26 lispro ...... 260 Table 5-9. Comparison of predicted proton-proton distances involving TyrB26 from WT crystal structure and TyrB26-associated NOEs from lispro NMR spectrum...... 261 Table 5-10a. Energy of non-bonded interaction of TyrB26 with local aromatic residues ...... 271 Table 5-10b. Energy of non-bonded interaction of TrpB26 with neighboring aromatic residues ...... 272 Table 5-11. Summary of previous designs of basal insulin analogs ...... 273 Table 5-12a. Intercentroid distances and angles between TyrB26 and local aromatic residues ...... 281 Table 5-12b. Intercentroid distances and angles between TyrB26 and local aromatic residues ...... 281

Table 5-13a. Thermodynamic stabilization associated with the B26 side chain during folding of insulin analogs ...... 282 Table 5-13b. Thermodynamic stabilization associated with the B26 side chain during dimerization of insulin analogs ...... 282 Table 5-14. Native-like crystal structures of R6 hexamers formed by mutant insulin...... 284

Table 6-1 In Vitro Affinities of 4SS Analogs for IR...... 310 Table 6-2 Summary of Modifications in 4SS Analogs...... 314

Table 7-1 Summary of GRI Technologies ...... 348 Table 7-2. Glucose-Dependent Hexamer Dissociation Rate of GRI Candidates...... 381

Table 8-1. In vitro IR-Affinities of Destabilized Insulin Analogs ...... 410 Table 8-2. Half-Lives of R6 Insulin Hexamers of Gly-Inserted Insulin Analogs ...... 415 Table 8-4. Hexamer Dissociation Half-Lives of Non-Standard B26 Variants ...... 420 Table 8-5. Thermodynamic Stabilities of Non-Standard B26 Variants ...... 421 Table 8-6 In vitro IR-Affinity of B24-Methylated Analogs ...... 436

List of Figures

Figure 1-1. Structure of the insulin monomer ...... 9 Figure 1-2. Structural representation of the transition between the “closed” and “open” conformations of insulin...... 12 Figure 1-3. Structure of insulin oligomers...... 16 Figure 1-4. Effects of association-state on the pharmacokinetics of insuin...... 18 Figure 1-5. Schematic of the biosynthesis, folding, and secretion of proinsulin...... 24 Figure 1-6. Schematic of insulin fibrillation in relation to folding and oligomeriztion...... 28 Figure 1-7 Complexation of Insulin with site 1 of IR...... 39 Figure 1-8. Structural summary of rapid-acting insulin analogs...... 50 Figure 1-9. Historical strategies used to create basal insulin analogs ...... 55

Figure 2-1 Sequence and binding mode of insulin...... 65 Figure 2-2 Cartoon representation of the experimental design utilized to test the register shift paradigm...... 68 Figure 2-3 In vitro competition IR binding assays...... 71 Figure 2-4 In vivo assessment of biological activity in streptozotocin-induced diabetic rats...... 75 Figure 2-5 Assessment of secondary structural stability of analogs by circular dichroism spectroscopy...... 78 Figure 2-6. Assessment of fibrillation lag times of insulin analogs...... 82 Figure 2-7 Visible absorption spectra of cobalt-stabilized hexamers and kinetics of metal ion release...... 85 Figure 2-8 HPLC size-exclusion chromatography and hormone self-assembly...... 88 Figure 2-9. In vivo assessment of proinsulin folding, misfolding, and secretion...... 91

Figure 3-1. AsnA21-GlyB23 H-Bond within the Structure of Insulin ...... 115 Figure 3-2. Impact of foldability of A21-B23 hydrogen bond on in vitro proinsulin folding. .... 119 Figure 3-3. Proinsulin folding experiments conducted in cells...... 122 Figure 3-4: Circular dichroism spectroscopy of amidated, monomeric insulin analogs...... 125 Figure 3-5. Hexamer Dissociation Assays of des-A21 and Control Analogs ...... 128 Figure 3-6. Biological Activity of des-A21-KP-amide...... 132 Figure 3-7 NMR Spectroscopy of a Des-A21 Insulin Analog in Relation to that of its Native Counterpart...... 136

Figure 4-1. Schematic of the role of PheB24 in proinsulin structure...... 150 Figure 4-2 Phylogenetic Tree of Vertebrate Evolution ...... 152 Figure 4-3. Alignment of Vertebrate Insulin and IGF Sequences ...... 156 Figure 4-5. Biological activity of TyrB24, OrnB29-insulin in relation to control analogs...... 161 Figure 4-6. Circular dichroism studies of TyrB24 insulin analogs in relation to controls...... 163 Figure 4-7. Circular Dichroism of Control Insulin Analogs ...... 166 Figure 4-8. CD Spectrum of Diluted TyrB24, OrnB25 Insulin...... 169 Figure 4-9 NMR studies of B24Tyr-DKP insulin analog...... 174 Figure 4-10 Solution Structure of a Monomeric TyrB24 Insulin Analog ...... 176

vii

Figure 4-11. Size-exclusion chromatography of WT-insulin and analogs...... 179 Figure 4-12. Hexamer dissociation of insulin analogs ...... 183 Figure 4-13. Dot plot of lag time to fibril formation of insulin analogs...... 185 Figure 4-14. Folding and secretion properties of proinsulin variants in a human cell line...... 189 Figure 4-15. Assessment of trans-dominance of proinsulin variants...... 192 Figure 4-16. Intracellular distribution of human proinsulin wild type (WT), or mutations of human proinsulin PheB24 with Cys, Ser, or Tyr substitutions...... 195

Figure 5-1 Structural overview of insulin...... 215 Figure 5-2. Comparison of Ab initio electrostatics and CHARMM parameters of Tyr and Trp side chains...... 219 Figure 5-3. Molecular simulations of aromatic interactions in the insulin dimer...... 225 Figure 5-4. Hexamer dissociation of TrpB26 analog...... 227 Figure 5-5 Size-exclusion chromatography of TrpB26 hexamer...... 230 Figure 5-6. Size exclusion chromatography (SEC)...... 232 Figure 5-7 Pharmacology of TrpB26 analog...... 236 Figure 5-8. Effects of Zn2+ on PD profile of WT insulin...... 238 Figure 5-9. In vivo potency of pI-shifted insulin analogs...... 241 Figure 5-10 Effects of Zn2+ on the PK/PD Profile of pI-shifted insulin analogs...... 242 Figure 5-11. Comparison of the structure of TrpB26, OrnB29-insulin to a collection of WT insulin structures...... 246 Figure 5-12 Crystal structure of TrpB26, OrnB29-insulin...... 248 Figure 5-13 Thermodynamic Stability of TrpB26, OrnB29-insulin...... 255 Figure 5-14. Homonuclear 2D NMR of a TrpB26 analog...... 258 Figure 5-15. Ab initio calculations of energy of interaction between pairs of isolated aromatic molecules...... 263 Figure 5-16 Binding surface of TyrB26 on an IR fragment...... 266 Figure 5-17. Comparison of B26 crevice within TRf dimers of TrpB26, OrnB29-insulin to WT (1TRZ)...... 277

Figure 6-1 Structures of Wild-Type Insulin and Stabilized Analogs...... 303 Figure 6-2 Thermodynamic Stability of 4SS Analogs ...... 307 Figure 6-3 Activity of 4SS-GluB29 Analogs ...... 312 Figure 6-4 Circular Dichroism 4SS-SCIs...... 316 Figure 6-5. NMR-Monitored Amide-Proton Exchange of 4SS-SCI-2 ...... 318 Figure 6-6. In Vivo Potency of 4SS-SCIs...... 320 Figure 6-7. Activity of SCI Analogs After Incubation at High Temperatures...... 323 Figure 6-8. Activity of Insulin Analogs after Extraction from PLGA Matrices...... 325 Figure 6-9. In Vitro and In Vivo Elution of 4SS-SCI-3 from PLGA Matrix ...... 328

Figure 7-1 Ribbon representation of structure of ultra-stable insulin analogs...... 345 Figure 7-2 Schematic of phenylboronic acid-monosaccharide complexation...... 350 Figure 7-3: Cartoon representation of polymer-based GRIs...... 353

viii

Figure 7-4 Cartoon representation of derivatization of the insulin molecule in GRI technology...... 359 Figure 7-5 Structural representation of insulin IR binding...... 363 Figure 7-6. Schematic of GRI Design ...... 369 Figure 7-7. Scheme for PBA-F Coupling to the N-terminus of the Insulin A chain ...... 371 Figure 7-8. Activity of Prototype GRI Candidate Analogs in Diabetic Rats ...... 375 Figure 7-9 Glucose-Dependent Hexamer Dissociation Rates of Prototype GRIs ...... 378 Figure 7-10. Activity of Candidate GRI in Non-Diabetic Rats ...... 382 Figure 7-11 Activity of Positional DOPA Variants in Diabetic and Non-Diabetic Rats...... 385 Figure 7-12 Activity of GRI Candidates with Aromatic Diols At LysB28 in Diabetic and Non-Diabetic Rats ...... 387 Figure 7-13. Comparison of Activity of GRIs in Diabetic and Non-Diabetic Rats ...... 389 Figure 7-14 Dose-Dependent Biological Activity of B28 Gallic Acid-Insulin ...... 392

Figure 8-1. Redox Re-Folding of des-A21 SCI-1 ...... 403 Figure 8-2 Biological Activity of Destabilized Insulin Analogs ...... 408 Figure 8-3. Hexamer Dissociation Rates of Gly-Inserted Analogs ...... 413 Figure 8-5 Self-Association Properties of B26-Variant Insulin Analogs ...... 418 Figure 8-6. PD Profiles of High-Concentration (u100) TrpB26 Analogs in Diabetic Rats ...... 425 Figure 8-7. Binding of SCIs and WT Insulin in the Insulin-μIR Co-Crystal Structure ...... 431 Figure 8-8 Biological Activity of B24-Methylated Analogs ...... 437

Acknowledgements

I would like to acknowledge a number of people without whom my training as a scientist would not have been possible.

I thank my advisor, Dr. Michael Weiss, for is dedication to scientific inquiry and for instilling in me a deep appreciation for all aspects of science ranging from classical physics to clinical medicine and for always keeping me excited about my work.

I thank Dr. Nelson Phillips who provided me with critical mentorship throughout my

PhD. Dr. Phillips has taught me many skills including scientific writing, poster making, oral presentation, experimental design, and how to fix a number of the older machines in the lab. Dr. Phillips is a mentor worthy of emulation for future academics.

I thank Dr. Faramarz Ismail-Beigi who, in spite of a hectic schedule as a clinician and as a scientist, always made the time to sit down with me and discuss some of our most complex projects. Conversations with Dr. Beigi always exposed me to new perspectives on sciene and helpe me tie even our most fundamental lines of inquiry to clinical medicine. Dr. Beigi and his team, which includes Kelley Carr, Alisar Tustan, Mamuni

Swain, Paul Macklis, and Rachel Grabowski, were my most critical collaborators; they spend a tremendous amount of time conducting the many animal experiments described in this dissertation.

I would like to acknowledge the contributions of my Thesis Committee, which comprises

Dr. Weiss and Dr. Ismail-Beigi along with Dr. Paul Carey and Dr. George Dubyak. I thank them for their guidance through the PhD process, continuous advocacy on my behalf, and dedication to ensuring that I received rigorous scientific training.

I would like to specifically acknowledge two collaborators within our laboratory group for their work on topics relating to this dissertation. I thank Dr. Nalinda Wickramasinghe, who conducted the NMR experiments and MM calculations decribed in this thesis; he often worked into the early hours of the morning completing the many studies that were being conducted in parallel. I would also like to thank Dr. Balamurugan Dhayalan, who collaborated closely with me on preliminary experiments described in Chapter 7. I learned a tremendous amount about peptide chemistry from him.

The projects described in this dissertation represent collaborative efforts. These studies would not have been possible with critical input and experimentation by collaborators. I thank them for their contributions as follows.

Chapter 2: Dr. Ming Liu and Dr. Peter Arvan conducted cell-based assays assessing the foldability, secretion, and UPR activation associated with B24 proinsulin variants in cells.

Dr. Ismail-Beigi and his team carried out experiments in rats.

Chapter 3: Dr. Leena Hataaja, Dr. Ming, Liu, and Dr. Peter Arvan conducted live-cell imaging assays on cells expressing fluorescently labeled proinsulin. Dr. Michal Avital-

Shmilovici created A21 proinsulin variants by native chemical ligation and performed in vitro folding experiments. Dr. Nalinda Wickramasinghe collected NMR spectra and provided guidance on data-processing and ongoing structure-determination methods. Dr.

Ismail-Beigi and his team carried out experiments in rats. The staff of Thermalin

Diabetes®, LLC provided isotopically labeled semisynthetic precursors of insulin.

Chapter 4: Dr. Ming Liu and Dr. Peter Arvan conducted cell-based assays characterizing proinsulin folding and secretion and whole-cell imaging. Dr. Nalinda Wickramasinghe

xi conducted NMR and structure-determination experiments. Dr. Ismail-Beigi and his team carried out experiments in rats. The staff of Thermalin Diabetes®, LLC provided isotopically labeled semisynthetic precursors of insulin.

Chapter 5: Dr. Nalinda Wickramasinghe conducted NMR experiments and MM calculations. Dr. Vivien Yee for collecting the x-ray crystallography data and teaching me how to solve crystal structures. Dr. Ismail-Beigi and his team carried out experiments in rats.

Chapter 6: The staff of Thermalin Diabetes®, LLC provided the 4SS-DOI precursor and

4SS-SCI proteins. Parker Lee and Dr. Jon Pokorski conducted polymer fabrication experiments. Dr. Ismail-Beigi and his team carried out experiments in rats and measured elution of insulin from polymer fabrications in vitro.

Chapter 7: Dr. Ismail-Beigi and his team carried out experiments in rats. The staff of

Thermalin Diabetes®, LLC provided the “TK-insulin” precursor and a subset of the insulin analogs tested.

I would also like to thank our collaborators Dr. John Menting, Dr. Michael Lawrence, and Dr. Brian Smith for stimulating conversations about structural basis of insulin-IR binding and related molecular dynamics experiments. Dr. Lawrence’s group also provided me with a number of figures used in this dissertation. I thank Dr. Krystel El

Hage and Dr. Markus Meuwly for teaching me about halogen-based interactions within proteins and for collaborating with our group on future computational simulations of our novel insulin analogs.

xii

I would like to acknowledge my colleagues in the Weiss group for their continued support and chameraderie: Dr. Yanwu Yang, Dr. Yen-Shan Chen, Dr. Michael Glidden,

Dr. Sanghamitra Bhattacharya, Homa Phillips, and Matthew Jirka. I extend a special

“thank you” to my undergraduate student, Brian O’Rourke, for his dedication to research.

I would like to extend my thanks to the staff of Thermalin Diabetes®, LLC, who provided us with a number of reagents and provided cordial advice and guidance on our ongoing studies. Finally, I would like to acknowledge former lab members. I thank Dr. Jonathan

Whittaker and Linda Whittaker for teaching me how to do in vitro IR-bining assays. I thank Dr. Vijay Pandyarajan, who mentored me through my rotation and my early days as a graduate student and continues to advise me on my progress through the convoluted pathway of physician-scientist training. I thank Dr. Joseph Racca, who provided me with invaluable guidance even though we were not direct scientific collaborators. Both Dr.

Pandyarajan and Dr. Racca remain my close friends today.

xiii

List of Abbreviations αCT α-helix within the C-terminal region of the insulin receptor α subunit θ ellipticity measured by circular dichrosim µIR domain-minimized insulin receptor comprising a fragment of L1 and CR domains and soluble αCT peptide 4SS insulin analog containing an engineered disulfide linkage between reisdues A10 and B4 Akt protein kinase B BGL blood-glucose level CD circular dichroism CGM continuous glucose monitoring DesDi LysB28 insulin analog lacking residues B29-B30, single-chain forms contain a peptide linkage between B28 and A1 DKP insulin analog with AspB10, LysB28, ProB29 modifications DM diabetes mellitus EEC entropy-enthalpy compensation ER endoplasmic reticulum ETF edge-to-face (interactions) GIR glucose-infusion rate GRI glucose-responsive insulin HSQC heteronuclear single quantum coherence (NMR spectroscopy) IGF-1/2 insulin-like growth factors 1 and 2 IGF-1/2R insulin-like growth factor 1/2 receptors IRT insulin-replacement therapy IV intravenous KP insulin lispro (LysB28, ProB29) L1 leucine-rich domain of IR MALDI-TOF matrix-assisted laser desorption ionization-time of flight (mass spectrometry) MD molecular dynamics MIDY mutant Ins-gene diabetes of the young MM molecular mechanics MODY maturity-onset diabetes of the young MPI mini-proinsulin (single chain insulin with LysB29-GlyA1 peptide bond) NOE nuclear overhouser enhancement NOESY nuclear overhouser effect spectroscopy NPH/NPL neutral protamine Hagedorn/neutral protamine lispro PBA(-F) Phenylboronic acid, (4-carboxy-3-fluoro-phenylboronic acid) PD pharmacodynamics PDB protein data bank PDI protein disulfide isomerase PEG poly-ethylene glycol PIP porcine insulin precursor (single chain insulin with “AK” 2-AA linker) PK pharmacokinetics PLGA poly-lactic-co-glycolic acid PNDM permanent neonatal onset diabetes mellitus

xiv

RMSD root-mean-square deviation RP-HPLC reverse-phase high performance liquid chromatography SQ subcutaneous SCI single chain insulin STZ streptozotocin (also spelled streptozoicin) T1DM type 1 diabetes mellitus T2DM type 2 diabetes mellitus TOCSY total correlation spectroscopy (NMR) TK tyrosine kinase WT wild-type

The Un-Design and Design of Insulin: Structural Evolution with Application to

Therapeutic Design

Abstract

NISCHAY K. REGE

Insulin is a peptide hormone that is the primary regulator of glucose homeostasis in vertebrates. Insulin is secreted by the endocrine pancreas in response to increased interstitial glucose levels; insulin initiates the uptake of glucose by peripheral tissues.

Since its first use in 1921, insulin has been the primary treatment for the metabolic condition known as Type I Diabetes Mellitus (T1DM; caused by the absolute lack of insulin), and a component in the treatment of Type II Diabetes Mellitus (T2DM; caused by the relative lack of insulin in relation to peripheral insulin resistance). The ubiquitous clinical use of insulin has led to investigation of its structure-function relationships. Such studies have uncovered the rich evolutionary history of insulin and its usefulness as a model molecule. Indeed, the study of insulin has revealed a number of concepts of protein structure/function relationships and the in vitro and biosynthesis of proteins.

This dissertation is a continuation of the nearly century-old field of insulin biochemistry. The first section of the thesis examines the evolutionary constraints responsible for the conservation of several structural features within the insulin B chain as they relate to the biosynthesis, stability, and biological activity of the hormone.

Concepts expounded in these studies may be generalized to the evolution and folding

xvi process of globular proteins as a class. Furthermore, such studies may be used to inform the design of therapeutic insulin analogs as exemplified in the second part of the dissertation. This section demonstrates how conserved structural features of insulin may be exploited and modified to produce favorable therapeutic effects even if such modifications would be unfavorable in the context of vertebrate physiology. This approach underscores the importance and usefulness of a multidisciplinary approach to the study of insulin both as a model molecule and as a therapeutic agent.

xvii

Chapter 1: Introduction

1.1 History of insulin research

Although insulin was not discovered until 1921, its role in the metabolic conditions known as (Type I Diabetes Mellitus) and Type II Diabetes Mellitus (T1DM and T2DM, respectively) has been studied since antiquity. The earliest description of a hallmark symptom of diabetes, polyuria, was found in an Egyptian scroll, known as the Ebers

Papyrus, dating back to c. 2000 B.C.E (Karamanou, Protogerou, Tsoucalas, Androutsos,

& Poulakou-Rebelakou, 2016). Later sources, found in India, China, and the Middle East over a period ranging from c 500 B.C.E. to c 250 C.E. described the primary symptoms of diabetes, including polyuria, polydipsia, and glucosuria and noted that the condition manifested in the wealthier social strata of these ancient societies (Das & Shah, 2011).

The Greek physician Areteus, who was able to accurately diagnose the signs and symptoms associated with T2DM, who first named the condition “diabetes.” This name was a reference to the production of large volumes of urine by diabetics—the root word of diabetes, diabo, means “to run through” in classical Greek. The term “mellitus” was added in the 17th century as a reference to the “honeying” of patients’ urine (Karamanou, et al., 2016).

The 19th century brought considerable advances to the understanding of DM that led to the discovery of insulin. French physician Claude Bernard discovered the glycogenic capability of the liver and indicated that DM was a result of dysregulated glucose homeostasis (Ahmed, 2002). In 1889, the role of the pancreas in diabetes was discovered by German physician Oskar Minkowski, who observed that excision of the pancreas caused diabetes in dogs. Minkowski proposed that it was internal secretion of a hormone

1 by the pancreas that was responsible for maintaining glucose homeostasis. However, he was unable to successfully prove the existence of such secretion. Nevertheless, the theory gained popularity and the yet undiscovered hormone was named “insulin” by physicians in the early 20th century (Bliss, 1982).

Insulin was first used clinically and isolated in large quantities in the 1920s by the team of Banting, Best, Collip, and McLeod. After experimenting with a variety of islet extracts from the ligated canine pancreas, Banting and colleagues demonstrated the ability of such an extract from the beef pancreas to successfully lower blood glucose levels and prevent ketoacidosis in a 14 year-old T1DM patient named Leonard

Thompson. Within the next decade, the Lilly corporation began to produce insulin in large quantities, using a newly developed method of isoelectric precipitation to increase yields from pancreatic extracts (Bliss, 1982).

Advances in biochemistry and molecular biology led to significant advances in the development of insulin as a therapeutic agent. The primary sequence of the mature insulin hormone was determined by Frederick Sanger and colleagues in 1955; insulin, the first protein to be sequenced, was revealed to contain two polypeptide chains (Sanger,

1959; Sanger & Thompson, 1952). This landmark discovery, along with the determination of the crystal structure of the zinc-coordinated insulin hexamer in 1969 by

Dorothy Hodgkin and colleagues paved the way for the structure-based design of insulin analogs (Harding, Hodgkin, Kennedy, O'Conor, & Weitzmann, 1966). The development of recombinant DNA technology also allowed human insulin to be synthesized in the laboratory setting from E. coli and yeast and circumvented the need to digest animal pancreases to produce insulin. The production of insulin from recombinant DNA was

2 first reported at the City of Hope Hospital in Duarte, California in 1978 (Goeddel et al.,

1979).

Since the 1980’s, insulin research has focused on improving the pharmacological properties of insulin. Structure-based design of insulin analogs has led to the development of rapid- and long-acting insulins to improve glycemic control in patients with T1DM and a subset of those with T2DM. The first such analog released commercially was the rapid-acting insulin lispro (Humalog®), which was released by Eli

Lilly in 1997 (Brems et al., 1992; R. K. Campbell, Campbell, & White, 1996). Whereas a number of insulin analogs have entered the market since the release of Humalog, present- day investigators continue to seek ways to improve the efficacy and pharmacology of such analogs. Modern understanding of the allostery of receptor-ligand interaction, as described by Leftkowitz and colleagues for the G protein-coupled receptor (GPCR), has motivated the investigation of biased- or tissue-selective insulin analogs (P. Kolb et al.,

2009; Shojaee-Moradie et al., 2000; S. G. Vienberg et al., 2011). Furthermore, advanced knowledge of the pathophysiology of DM has led many groups to investigate fusion peptides that are chimeric constructs of insulin and other peptide hormones, such as GLP-

1 agonists (Michael Rosario DeFelippis, DiMarchi, & Ng, 2007), for improved therapy for DM.

1.2 Purpose of Dissertation

The canonical view of protein evolution suggests that the physiological function of a protein is the primary determinant of its structure. However, such a view fails to account for the implicit constraints of the ability of a protein to adopt and maintain its native conformation (C. M. 14685248 Dobson, 2003). The evolution of insulin, a long-studied

3 model of protein structure, exemplifies this concept: the structural evolution of insulin exhibits interplay between its physiological function, engagement and activation of the insulin receptor (IR), its efficient biosynthesis, and its structural stability in relation to proteotoxic aggregation (J. M. Conlon, 2001; Stefani & Dobson, 2003). Indeed, a number of mutations that affect the foldability of proinsulin, the prohormone precursor to insulin, through the native pairing of the three canonical disulfide linkages of the prohormone have been shown to cause forms diabetes termed mutant INS-gene diabetes of the youth (MIDY) or permanent neonatal diabetes mellitus (PNDM) (Edghill et al.,

2008).

The first section of this dissertation examines the role of conserved structural features within the C-terminal region of the insulin B chain—a region that plays a critical role in the biological activity, stability, and foldability of insulin (Ming Liu et al., 2015; Vijay

Pandyarajan et al., 2014b). This study examines four structural elements within this region: a conserved β-turn spanning residues B20-B23, the “lynchpin” residue PheB24, and a conserved—although weak—hydrogen bond between the amide proton of AsnA21 and GlyB23 (Chu, Wang, Burke, Chanley, & Katsoyannis, 1987a). Findings suggest that a number of these features have evolved to optimize the foldability of proinsulin and maintain the stability of the mature insulin molecule during storage in the secretory granules of pancreatic β-cells. Indeed, results suggest that the allostery of insulin-IR complexation may accommodate a number of small perturbations to the active conformation of insulin, and for this reason, the evolutionary constraints of receptor binding on insulin structure are relaxed compared to those of foldability and stability.

Such findings may also shed light into the evolutionary history of insulin as a disordered

4 peptide. Furthermore, the results discussed in this thesis provide further insight into the pre-oxidative folding pathway of insulin suggesting conformations that may be required for proper disulfide pairing in the nascent fold of proinsulin.

The ubiquitous use of insulin in the treatment of T1DM and in select populations of

T2DM (Salsali & Nathan, 2006) patients has motivated research into the development of therapeutic insulin analogs (J. P. Mayer, Zhang, & DiMarchi, 2007). Indeed, the development of the rapid-acting insulin analog insulin lispro was among the earliest therapeutic applications of structure-based protein engineering (Brems, et al., 1992).

Continuing research has focused on developing insulin analogs with improved pharmacokinetic/pharmacodynamic (PK/PD) profiles. That is, investigators seek to create rapid-acting analogs with faster onset of activity with a rapid return to baseline (fast on, fast off), basal insulin analogs with a longer duration of activity with reduced peaks, and analogs that are more stable. Among the highest goals of the insulin research community is the development of glucose-responsive insulin analogs (GRIs), which are insulin delivery systems that provide blood-glucose lowering activity that is proportionate to the glycemic state of the patient (Zaykov, Mayer, & DiMarchi, 2016).

The second section of this thesis describes the development of next-generation insulin analogs through the dissection of the structure of insulin from a fundamental perspective.

Chapter 5 describes the development of a long-acting insulin analog through the stabilization of the insulin hexamer. This was achieved by improving aromatic-aromatic interactions at the insulin dimer interface through the substitution of residue TyrB26 by

Trp. The TrpB26 substitution may be used to further prolong the activity of basal insulin analogs currently in clinical use. Chapter 6 describes the development of an ultra-stable

5 insulin analog with novel biological properties through the use of two independent stabilizing mechanisms. These analogs termed “4SS-SCIs,” contain an engineered fourth disulfide linkage between residues A10 and B4 (T. N. Vinther et al., 2013) and a foreshortened, 6-8 residue C-domain spanning the C-terminal B chain of insulin and the

N-terminal A chain (Q. X. Hua et al., 2008). The increased stability of such analogs allows them to be fabricated into slow-release PLGA-based polymer fabrications, a novel method for insulin delivery. Finally, chapter 7 describes the early stages of the development of a GRI through the exploitation of a conformational switch in insulin that allows it to bind to IR.

1.3 Insulin Structure

Insulin is a peptide hormone that comprises two polypeptide chains, a 21 residue A chain and a 30 residue B chain. The two chains are connected via two interchain disulfide linkages between residues A7 and B7 and A20 and B19. A third, intra-chain disulfide bridge is found between residues A6 and A11(E. N. Baker et al., 1988) (Figure 1-1A,B).

The secondary structure of insulin is primarily α helical. The free insulin monomer contains three helices spanning residues A1-A8, A12-A19, and B8-B20. The hormone also contains two type I β-turns spanning residues A5-A8 and residues B20-B23 (Figure

1-1B). A β strand that forms anti-parallel hydrogen bonds in the insulin dimer is found between residues B24 and B28 (E. N. Baker, et al., 1988).

The integrity of the hydrophobic core of insulin is a key determinant of the stability of the globular structure of the hormone. This core comprises the side chains of residues

ValB12, LeuB15, LeuB11, and TyrB16 of the B chain and residues IleA2, ValA3, and TyrA19 of the A chain (Fig. 1-1C). Mutation of the residues that comprise the core of insulin reduce

6 the thermodynamic stability of the insulin monomer to varying degrees (M. Liu et al.,

2009; Weiss, Nakagawa, et al., 2002). Such mutations have also been shown to impact the ability of proinsulin, the prohormone precursor of insulin, to fold into its native state

(Kristensen et al., 1997).

The C-terminal B chain of insulin is responsible for stabilizing the core of the hormone. The side chains of residues PheB24 and TyrB26 serve as aromatic “anchors” that pack against the hydrophobic side chains within the insulin core providing structural stability by filling cavities that would result in packing defects within the region and protecting it from exposure to solvent (Fig. 1-1C) (Pandyarajan et al., 2016; Vijay

Pandyarajan, et al., 2014b). These residues are critical to the thermodynamic and physical stability of insulin, the foldability of its precursor, and its ability to engage IR (Ming Liu, et al., 2015).

The partial unfolding of the B20-B23 β-turn is responsible for a conformational

“switch” that allows insulin to engage IR (J. G. Menting et al., 2013; Y. Yang, Petkova, et al., 2010). In the classical insulin structure, determined by x-ray crystallography, the

B20-B23 β-turn retains

its canonical hairpin conformation: glycine residues B20 and B23 display ϕ angles of

55.9° and 84.1°, respectively (Nakagawa et al., 2006). To engage the insulin receptor, this turn must partially unfold: a 10° rotation about GlyB20 and a 50° rotation about PheB24 occurs so that the C-terminal B chain of insulin may detach from the core of the protein.

This exposes residues B24 and B26, which occupy specific binding sites on IR (Figure 1-

2) (J.G. Menting et al., 2014).

Although it is necessary for insulin-IR complexation, the “open” conformation of insulin is highly unstable in relation to the “closed” storage conformation. Indeed, mutations of anchor residues of the C-terminal B chain (PheB24 and TyrB26 (Pandyarajan, et al., 2016; Vijay Pandyarajan, et al., 2014b)) or β-turn glycine residues B20 and B23

(Nakagawa, et al., 2006) that interfere with the closed conformation of insulin have been shown to decrease the thermodynamic stability of the hormone and to increase its propensity for proteotoxic misfolding (Brange, Andersen, Laursen, Meyn, &

Rasmussen, 1997). For this reason, the conformational switch in the B20-B23 region of insulin is critical for maintaining a biologically active yet stable hormone (Brange, et al.,

1997; J.G. Menting, et al., 2014).

1.4 Insulin Oligomerization and Storage

Due to its central role in the glucose homeostasis, vertebrate insulin must be translated and stored in large quantities in pancreatic β cells; for this reason, insulin has evolved the ability to form oligomers (dimers and zinc-coordinated hexamers) to stabilize its storage conformation (G. Dodson & Steiner, 1998). The C-terminal B chain of insulin comprises a majority of the dimer interface (E. N. Baker, et al., 1988; Blundell, Dodson, Hodgkin,

& Mercola, 1972). Anti-parallel hydrogen bonds form between the backbone amide and carbonyl groups of PheB24 and TyrB26 of each constituent monomer. Furthermore, residue

ProB28 forms hydrophobic interactions across the dimer interface with GlyB23 of the opposite monomer (Ciszak et al., 1995). The interface is further stabilized by weakly polar interactions between aromatic

residues TyrB16, PheB24, TyrB26, and their symmetry-related mates. These dimer-related interactions dampen the conformational fluctuations of the C-terminal B chain and

Figure 1-1. Structure of the insulin monomer (A) The amino acid sequence of human insulin comprises a 21-reside A chain and a 30-residue B chain. The three canonical disulfide linkages are indicated by black lines and secondary structural features are labeled. (B) The 3D structure of insulin (PDB ID: 4INS (E. N. Baker, et al., 1988)).

Secondary structural features are color-coded as in (A). (C) The hydrophobic core of insulin comprises residues IleA2, ValA3, TyrA19, ValB12, LeuB15, LeuB11, and TyrB16, shown as sticks. Residues PheB24 (teal sticks) and TyrB26 anchor the C-terminal strand of the insulin B chain within the hydrophobic core of insulin.

10 prevent insulin from adopting its open conformation (Figure 1-3A). This prevents insulin from forming proteotoxic aggregates during storage, when insulin concentrations are high

(Kd of dimerization is 5 μM (Attri, Fernandez, & Minton, 2010a)), but allows it to freely adopt its open conformation after secretion (dilution).

Insulin is further stabilized during storage through the formation of zinc-coordinated hexamers. A hexamer comprises three dimeric subunits that coordinate with two divalent zinc ion via HisB10 (Carpenter & Wilcox, 2014). When stored in β cells, the hexamer has also been shown to bind Ca2+ ions via residue GluB13 (Howell, Tyhurst, Duvefelt,

Andersson, & Hellerstrom, 1978; Steiner, Kemmler, Clark, Oyer, & Rubenstein, 1972;

Storm & Dunn, 1985). The classical crystallographic structure of the insulin hexamer

(solved by Hodkin, et al) (Harding, et al., 1966) is referred to as the T6 hexamer. The defining characteristics of this hexameric state are the octahedral coordination of Zn2+ by three HisB10 residues and three water molecules and that residues B1-B8 adopt a “random coil” secondary structure (Figure 1-3B) (Ramesh & Bradbury, 1986) (corresponding closely to the solution structure of the insulin monomer as determined by NMR spectroscopy (Q. X. Hua & Weiss, 1990; Olsen, Ludvigsen, & Kaarsholm, 1996)).

An alternate form of the insulin hexamer, designated R6, is formed in the presence of

2+ phenolic ligands. The R6 hexamer is defined by the tetrahedral coordination of Zn by three HisB10 residues and a chloride ion and by residues B1-B8, which adopt an α-helical conformation (Figure 1-3C) (U. Derewenda et al., 1989; Roy et al., 1989) . The transition between the T- and R- states is mediated by GlyB8, which adopts a right-handed (positive)

ϕ angle in the T6 hexamer. Phenolic ligands bind preferentially to the R6 hexamer in which GlyB8 adopts a left-handed (negative) ϕ angle, thus allowing residues B1-B8 to

Figure 1-2. Structural representation of the transition between the “closed” and

“open” conformations of insulin.(A) The insulin B chain (central α-helix shown in black) undergoes a conformational change via changes in chirality of residues GlyB20 and

GlyB23 that allows insulin to engage IR. The conformation of the C-terminal B chain of insulin in its “closed” state is shown in brown. The conformational transition to the open state involves a 10° rotation about residue B23 (conformation shown in blue) and a 50° rotation about residue B24 (conformation shown in green). An inferior view is shown below. (B) The “closed to open” conformational switch allows insulin to bind to IR. In its closed state, the insulin B chain (brown) encounters a steric clash with the αCT domain of IR (purple). In its open conformation, insulin may properly intercalate with IR, exposing the residues of the C-terminal B chain (green) and N-terminal A chain so that they may interact with their respective binding surfaces on IR (three residues of the αCT domain that interact with insulin are shown as purple sticks). This figure has been adapted from (J.G. Menting, et al., 2014) with the permission of the authors.

adopt a helical conformation (Michael A Weiss, 2009) Figure 1-3D,E). The presence of phenolic ligands shifts the allosteric equilibrium of the insulin hexamer toward the R state (Duane T Birnbaum, Dodd, Saxberg, Varshavsky, & Beals, 1996). An intermediate

f conformation of the insulin hexamer, designated T3R 3, has been identified in which one monomer of each of the three constituent dimers adopts the T state whereas the other monomer adopts an Rf (R-frayed) state in which residues B4-B8 are in an α-helical conformation and residues B1-B4 are in a random coil conformation. This hexamer may also be visualized as a trimer coordinating one Zn2+ ion adopting the T-state, and the trimer coordinating the other Zn2+ ion adopting the Rf-state (Ciszak, et al., 1995). Insulin hexamers associate into microcrystal arrays in secretory granules, further protecting the hormone from degradation through aggregation or proteolysis. For this reason, the conformational state of the insulin hexamer in vivo remains unknown (Steiner, Hallund,

Rubenstein, Cho, & Bayliss, 1968; Steiner, et al., 1972).

The ability of insulin to hexamerize was utilized to stabilize the hormone in pharmaceutical formulations. Insulin is capable of forming amyloid fibrils in vials, greatly reducing the shelf-life of insulin products. For this reason, insulin formulations contain zinc and phenol, thus favoring the formation of R6 hexamers. Indeed, oligomerization has been shown to greatly dampen the conformational fluctuations of the insulin hormone, which contribute to the formation of pre-fibrillar aggregation states (J.

Dong, Wan, Popov, Carey, & Weiss, 2003; Jacoby, Hua, Stern, Frank, & Weiss, 1996;

Wollmer et al., 1987). The R6 hexamer has been shown to be especially stable (R6 >

f T3R 3 >T6) (Figure 1-4A).

The association state of insulin analogs can affect their PK/PD profiles: the absorption of insulin from its subcutaneous injection site into systemic circulation is inversely proportional to the size of its association state. That is, insulin monomers are rapidly absorbed from the bloodstream whereas the absorption of dimers from the bloodstream is

100 to 1000 fold slower. Absorption of hexamers is negligible in relation to that of dimers and monomers (M. Berger, Cuppers, Hegner, Jorgens, & Berchtold, 1982) (Figure

1-4B). For this reason, insulin analogs that form stable hexamers tend to have a protracted

PK/PD profile (Brange & Vølund, 1999). Strategies for destabilizing insulin hexamers have been utilized to create rapid acting, postprandial insulin analogs, which will be discussed in a later section (Brange et al., 1988).

1.5 Insulin Biosynthesis and Secretion:

Insulin is translated as a single chain preprohormone, preproinsulin. Proinsulin is co- translationally translocated into the rough endoplasmic reticulum (rER). Preproinsulin is processed into proinsulin after the cleavage of the N-terminal 23-amino acid signal peptide by a signal peptidase (Steiner, 1990). Proinsulin folds into its native conformation in the ER lumen following a number of critical steps that are discussed below. The prohormone is then transported to the Golgi apparatus where prohormone convertases

PC1, PC3, and PC2 cleave the 55 amino acid C domain via the di-basic sites at the C- terminus of the B domain and the N-terminus of the A domain (Steiner, 1998; Tanaka et al., 1996). The basic residues are then removed by prohormone convertase E (Haataja et al., 2013). The processed insulin molecule and C-peptide are then packaged into glucose- regulated secretory vesicles. An influx of Zn2+ ions into the vesicles (mediated by the

Figure 1-3. Structure of insulin oligomers. (A) Ribbon diagram of an insulin dimer extracted from the crystal structure of a T6 hexamer (PDB ID: 4INS). The dimer interface is composed of the C-terminal B chain of insulin. An expanded view of the interface is shown in the inset; residues that comprise the interface are shown as sticks, hydrogen bonds formed across the interface are shown as yellow dashed lines. Individual residues contributing to the dimer interface are color-coded as follows, TyrB16, blue; PheB24, red;

PheB25, purple; TyrB26, cyan; residues B28-B30, orange. (B) Ribbon representation of an insulin hexamer in the T6 state (PDB ID: 4INS). The A chain is colored black and the B chain is colored light orange. The six HisB10 residues that coordinate a Zn2+ ion (pink sphere) are shown as sticks. The eight N-terminal residues of the insulin B chain, which adopt a coil conformation in the T-state are highlighted in blue. (C) Ribbon representation of an insulin hexamer in the R6 state (PDB ID: 1ZNJ). The A chain is colored black and the B chain is colored light orange. The six HisB10 residues that coordinate a Zn2+ ion (pink sphere) are shown as sticks. The eight N-terminal residues of the insulin B chain, which adopt a coil conformation in the T-state are highlighted in blue. (D) Ribbon representation of an insulin hexamer in the R6 state. The A chain is colored black and the B chain is colored light blue. Residues B1-B8, which adopt an α- helical conformation in the R-state are colored blue. Phenol molecules are depicted as red sticks. (E) Comparison of the orientation of the torsional conformation of residue GlyB8 in the T-state (orange sticks) and the R-state (blue sticks). (F) Expanded view of residue

GlyB8, color code as in (D): the differences in ϕ angle are indicated by the black arrow.

Figure 1-4. Effects of association-state on the pharmacokinetics of insuin. (A)

Comparison of the three states of the insulin hexamer (PDB IDs are as follows, T6:4INS,

f T3R : 1TRZ, R6: 1ZNJ). Protomers in the R-state are shown as blue ribbon and those in the T-state are shown as green ribbon. Phenol molecule are shown in ball representation

(dark gray). (B) Schematic of the dissociation of R6 hexamers after subcutaneous

f injection. Minimal amounts of R6, T3R , and T6 hexamers are absorbed from the subcutaneous depot into capillaries, whereas the rate of absorption of insulin monomers exceeds that of dimers 100-1000 fold.

ZnT8 transporter) initiates the hexamerization of insulin and subsequent microcrystallization (Lemaire et al., 2009).

1.6 Proinsulin Folding:

The folding process of proinsulin is illustrative of the energy landscape model of protein folding (Dill & Chan, 1997). This model represents the process of protein folding as both a thermodynamic and kinetic process. The “folding landscape” of a protein is visualized as a funnel in which the stem of the funnel represents the conformation of the protein corresponding to the thermodynamic minimum of the system: this is typically the native folded state. The slope of the funnel corresponds to the kinetics of the folding process: the steeper the slope, the more rapid the rate of folding of the protein. The nascent polypeptide itself is visualized as a marble rolling towards the stem of the funnel from its lip. The polypeptide may take a number of pathways to achieve its native conformation, but its progression to the folded state is guided kinetically by the formation of folding nuclei through local hydrogen bonding and secondary structure formation.

Local minima and maxima, visualized as “peaks” and “valleys” in the funnel, represent metastable folding intermediates. Such intermediates may be on pathway—meaning that they are necessary checkpoints for the peptide to achieve its native fold— or off pathway, meaning that the peptide may be trapped in a conformation that is not conducive to native folding (Zhang & Goldenberg, 1993). Such kinetic traps can limit the folding efficiency of proteins; these kinetic barriers must be overcome by the action of catalysts, such as molecular or protein chaperones (M. Mayer, Kies, Kammermeier, & Buchner, 2000).

The folding pathway of proinsulin is readily visualized in the context of a folding landscape. The native pairing of the three disulfide linkages of proinsulin represents the

20 major kinetic barrier to the folding of the globular prohormone: proinsulin and mature insulin will readily re-fold into their native conformation after chemical denaturation as long as the disulfide linkages remain intact. The formation of each of the disulfide linkages is a stepwise process. Pairwise mutational analysis of the disulfide linkages revealed that the formation of the A20-B19 disulfide bridge is essential for the native pairing of the remaining cysteine residues (Q. X. Hua, Narhi, et al., 1996). The A7-B7 disulfide linkage is also necessary for proinsulin to achieve its native conformation and for the proper processing and secretion of the prohormone (Q. X. Hua et al., 2001).

However, the order of formation of the A7-B7 and A6-A11 cystines is not important for proper folding (Qiao, Guo, & Feng, 2001). The A6-A11 linkage is dispensable for the synthesis of a functional insulin molecule, but makes contributions to the stability of the secondary and tertiary structure of the hormone (M. Liu, Li, Cavener, & Arvan, 2005).

The efficiency with which proinsulin is able to adopt its native conformation increases with the formation of each disulfide linkage. This process is referred to as “landscape maturation,” which can be visualized as the landscape funnel of insulin becoming steeper after each cystine is formed. The formation of off-pathway disulfide isomers represents a major kinetic trap in the folding landscape of proinsulin (Q. X. Hua et al., 1995).

Efficiency of disulfide formation is maintained, in part, by ER chaperones such as protein disulfide isomerase (Rajpal, Schuiki, Liu, Volchuk, & Arvan, 2012) and ER- oxidoreductases (Wright et al., 2013).

Clinical mutations that disrupt the proinsulin folding pathway are responsible for mutant proinsulin syndrome, which manifests itself clinically as permanent neonatal onset diabetes (PNDM) and mutant INS diabetes of the young (MIDY). PNDM is

21 diagnosed within the first six months of life, and is associated with severe proinsulin folding defects (Stoy et al., 2007), whereas MIDY typically manifests in the third decade of life and is associated with a milder folding defect (M. Liu, Haataja, et al., 2010). In both cases, endoreticular stress in pancreatic β cells caused by the accumulation of unfolded proinsulin leads to the activation of the unfolded protein response (UPR).

Prolonged activation of the UPR, in turn, initiates apoptosis of β cells, which causes DM.

For this reason, both conditions exhibit autosomal dominant inheritance (Weiss, 2013b).

The study of INS-gene mutations causing MIDY and PNDM has provided insight into the nebulous folding pathway of insulin. PNDM mutations fall into two major classes.

The first class comprises mutations that introduce ectopic cysteine residues to proinsulin or eliminate its native cysteines; the introduction of an unpaired cysteine residue to proinsulin disrupts the native pairing of the disulfide bridges, resulting a severe folding defect that causes PNDM. The second class of mutations involves non-cysteine-based substitutions of

conserved residues (Stoy et al., 2010). Such mutations are thought to affect the preoxidative folding of proinsulin; that is, they prevent the formation of structural motifs that direct the native formation of disulfide linkages in the partially folded protein. It is this class of mutation, along with mutations associated with MIDY, that serve as

“experiments of nature,” indicating structural features that are critical to the foldability of proinsulin (Q. X. Hua, Mayer, Jia, Zhang, & Weiss, 2006).

Biological and biophysical characterization of non-cysteine proinsulin mutations suggested mechanisms by which such mutations impaired proinsulin folding. Proline- or glycine-associated mutations of the residues neighboring native cysteines, such

LeuB6→Pro, GlyB8 →Ser, ValB18 →Gly, prevent native disulfide formation by alternating the torsional conformation of the proinsulin backbone; such mutations result in PNDM

(Edghill, et al., 2008; Stoy, et al., 2010). The integrity of the globular core of proinsulin identified as a critical step in preoxidative folding. A clinical mutation in which LeuB15, a central residue in the core of mature proinsulin, is substituted by His results in PNDM

(Kristensen, et al., 1997; Stoy, et al., 2007). Insulin analogs containing mutations TyrA19 and LeuA16, which are also core residues, also displayed reduced folding-efficiency in vitro, further underscoring the hypothesis (Du & Tang, 1998; Kristensen, et al., 1997;

Weiss, Nakagawa, et al., 2002). Moreover, mutations that affect the disposition of core residues also result in PNDM: one such mutation, GlyB23→Val, disrupts the Type I β-turn spanning residues B20-B23 and prevents

PheB24 from occupying its native position within the core of insulin (Edghill, et al., 2008).

1.7 Insulin Fibrillation

The canonical representation protein of folding depicts the native state of a protein as the thermodynamic minimum of a folding landscape, yet this representation fails to take into account thermodynamic benefits associated with aggregation (C. M. 14685248

Dobson, 2003). Indeed, both non-specific and specific aggregates (such as oligomers or crystals) may represent thermodynamic states with lower free energies than that of the native state (C. M. Dobson, 2001). Among such aggregates are amyloids, which are polymeric, cross β-sheet assemblies that have been identified in organismal biology, typically as pathologic features. Amyloids form from partially unfolded proteins that form non-specific aggregates known as prefibrillar species or amyloid nuclei. Large, ordered amyloid polymers form spontaneously from such nuclei as a consequence of their

Figure 1-5. Schematic of the biosynthesis, folding, and secretion of proinsulin. (A) schematic of proinsulin trafficking within a cell. (B) An expanded schematic of proinsulin biosynthesis and transport within organelles. Steps are labeled as follows (a) translation in the rough endoplasmic reticulum (ER), (b) translocation into the ER with cleavage of a signal sequence (M. Liu, Hodish, Rhodes, & Arvan, 2007), (c) oxidative folding in the ER with the assistance of oxidoreductases to achieve native disulfide pairing (M. Liu, et al., 2007), (d) quality control (associated with chaperone binding or

ER-associated degradation) and (e) export to the Golgi apparatus, (f) processing of the dibasic sites at the BC and CA junctions to yield the mature hormone and C peptide; (g) trafficking to glucose-regulated secretory vesicles; and (h) zinc-dependent assembly and

(i) microcrystallization (G. Dodson & Steiner, 1998). (C) Folding landscape of proinsulin. The thermodynamic folding landscape of proinsulin undergoes the process of landscape maturation with the successive formation of each disulfide linkage. The formation of the A19-B20 disulfide linkage is the critical step necessary for the oxidative folding of the rest of the hormone. Formation of off-pathway disulfide isomers are visualized as kinetic traps in the second of the three landscapes depicted.

25 thermodynamic stability (C. M. Dobson, 1999). For this reason, amyloids are not readily degraded. Individual amyloid polymers further associate with one another in an antiparallel fashion to form long, insoluble fibrils and plaques (Jimenez, Tennent, Pepys,

& Saibil, 2001).

Insulin is a model for studying the constraints of amyloid formation on protein structure. Although insulin forms amyloid fibrils in vials, insulin-based amyloids are not observed in human pathophysiology. Structural analysis of insulin has led to the identification of a number of adaptations that prevent formation of pre-fibrillar species.

One such adaptation is the B20-B23 β-turn of insulin, which allows residues PheB24 and

TyrB24 to pack against the hydrophobic core of insulin. This not only stabilizes the globular structure of the hormone but also protects the hydrophobic core from solvent exposure, which may lead to non-specific aggregation of multiple insulin molecules

(Brange, et al., 1997; Vijay Pandyarajan, et al., 2014b). Another adaptation is the ability of insulin to oligomerize; the conformational fluctuations of the insulin molecule are successively dampened in dimeric and hexameric association states (Chiti & Dobson,

2006). This reduces the potential for local unfolding of insulin particularly at the C- terminal B chain, a critical step in the nucleation of fibrils. This motif comprises a bulk of the insulin dimer interface (J. Dong, et al., 2003).

Biochemical studies have led to a deeper understanding of the factors contributing to insulin fibrillation. Low pH and temperatures exceeding 25°C, were found to favor fibril formation; this is largely the result of decreased dimer and hexamer formation as well as the partial unfolding of the insulin monomer with increased temperature (Weiss, 2012).

Whereas typical proteins tend to form aggregates more rapidly at increased

26 concentrations, the fibrillation lag time of insulin increases at high concentrations at which dimer formation is favored. Similarly, an increase in ionic strength favors formation of oligomers and can prolong the fibrillation of lag time of insulin, but may increase fibrillation propensity under conditions that disfavor oligomerizaiton (such as low pH). Agitation of insulin solutions, storage in vials composed of polar materials

(such as glass), and presence of air-liquid interfaces have all been shown to cause rapid fibrillation (Brange, et al., 1997).

The process of insulin fibril formation is pertinent to its use as a therapeutic protein.

Fibrillation of insulin analogs limits their shelf-life and necessitates their refrigeration

(Weiss, 2013a). Furthermore, the subcutaneous injection of insulin fibrils has been associated with inflammation (Swift, Hawkins, Richards, & Gregory, 2002) at the injection site and is hypothesized to contribute to development of immune response against insulin (Uchio, Baudyš, Liu, Song, & Kim, 1999). This has led to the investigation of techniques to prevent the fibrillation of insulin in storage conditions.

Storage of insulin buffered solutions (pH 7.4) in the presence of zinc and phenol, which promote formation of R6 hexamers, prolongs the shelf-life of insulin. Furthermore, the presence of glycerol or surfactants in formulations mitigates the pro-amyloid effects of agitation and air-liquid interfaces (Brange & Langkjaer, 1997; B. S. Chang, Kendrick, &

Carpenter, 1996). Experimental fibrillation inhibitors, peptides or small molecules that prevent the formation of prefibrillar aggregates, have also been developed. Such inhibitors have included antibodies (Alam, Siddiqi, Chturvedi, & Khan, 2017), small molecules (Mukherjee, Jana, & Chatterjee, 2018), or insulin-mimetic peptides (Gibson &

Murphy, 2006) that were hypothesized to bind to regions of insulin (typically the central

Figure 1-6. Schematic of insulin fibrillation in relation to folding and oligomeriztion.The insulin monomer is depicted as a black triangle with the B20-B23 β- turn represented as a gray circle, the aromatic triplet of residues (B24-B26) is represented as a red rectangle, and residues B27-B30 are shown as a gray rectangle. The process of fibril formation begins with the partial unfolding of the insulin monomer. The native state is protected by classical self-assembly (highlighted in green). Disassembly leads to an equilibrium between native and partially folded monomers (parallelogram, with red-and- gray line). The putative partial fold may either adopt its IR-binding conformation (black

¾ circle), unfold completely as an off-pathway event (as labeled), or aggregate to form a nucleus en route to a protofilament (labeled “toxic misfolding”).

29 helix of the B chain or the C-terminal B chain) and prevent the partial unfolding required for fibrillation. However, these strategies were unsuccessful.

The clinical interest in engineering safeguards against amyloid fibril formation in pharmaceutical formulations as well as in vivo, in diseases such as Parkinson’s disease or systemic amyloidosis, has motivated characterization of the structure of amyloid fibrils and their precursors (C. M. 12563307 Dobson, 2003). Nevertheless, the structure of mature insulin fibrils remains nebulous and that of prefibrillar nuclei remains unknown.

Whereas naturally ocurring amyloid plaques are refractory to crystallization, the crystal structure of short fibrils formed from synthetic peptides revealed a pattern of stacked parallel β-sheets . The side-chains of the amino acid residues constituting the fibril were oriented perpendicular to the long axis of the fiber and packed against the side-chains of a neighboring fibril in an anti-parallel fashion forming a multi-fiber structure (Fitzpatrick et al., 2013). Study of insulin amyloids by cryo-EM tomography confirmed that mature insulin fibrils contain multiple fibers that pack against one another to form a larger strand that may be visualized by transmission electron microscopy (Jimenez et al., 2002).

Models of the molecular orientation of insulin within a mature fiber have been proposed based on these findings. Each model envisions a linear fiber formed by the insulin B chain with the residues constituting the central helix in native insulin constituting a bulk of the body of the fiber; the insulin A chain forms a fiber that is parallel to that formed by the B chain. However, the proposed orientation of the A chain fibril varies from model to model. Whereas some models describe a curved fiber in which the A6-A11 disulfide linkage causes a bend in the β-sheet formed from the A chain (forming from a protofilament in which both A- and B chains adopt a stacked, triangular conformation),

30 others describe a linear fiber in which the segment between CysA6 and CysA11 forms a kink that does not participate in the hydrogen bonding of the sheet (Jimenez, et al., 2002).

The limited structural knowledge of the insulin fibril has nevertheless been utilized in two strategies to create fibrillation-resistant insulin analogs. The first such strategy was used to develop a class of single chain insulin analogs (SCIs), that contained foreshortened C domains (analogous to the C domain of proinsulin) (Q. X. Hua, et al.,

2008). SCIs used as precursors in the synthesis of two-chain insulin that contained very short C domains (2 amino acids, in the case of porcine insulin precursor) or that were missing C domains entirely (miniproinsulin, MPI, which contained a peptide linkage between LysB29 and GlyA1) exhibited increased thermodynamic stability and were resistant to fibrillation (Neumoin, Mares, Lerch-Bader, Bader, & Zerbe, 2007). However, these insulin analogs were biologically inert; they were incapable of adopting the active conformation of insulin in which the C-terminal B chain detaches from the core of the molecule to engage IR. This contrasted with proinsulin (containing a 35 amino acid C domain), which was biologically active, but was capable of forming fibrils (Huang,

Dong, Phillips, Carey, & Weiss, 2005). Further investigation of C-domain length identified a “sweet spot” length of 4-8 amino acids that prevented the insulin polypeptide form adopting the elongated conformation required for fibril formation, yet allowed enough flexibility for insulin to bind to IR with native affinity (Q. X. Hua, et al., 2008).

The second strategy for developing a fibrillation-resistant insulin analog was the introduction of an ectopic cysteine residues at A10 (native Ile) and B4 (native Gln) that form a fourth disulfide linkage that recapitulates the stabilizing action of the A11-B4 hydrogen bond in native insulin (T. N. Vinther, et al., 2013). The fourth disulfide linkage

31 restricts the conformation of the core of the insulin hormone: the N-terminal B chain, which must be extended in order to form β sheets in fibrils, is anchored to the central loop of the A chain. Because the ectopic cysteines do not affect residues on the receptor binding surface of insulin, “4SS” insulin analogs are highly biologically active (Tine N

Vinther, Kjeldsen, Jensen, & Hubálek, 2015).

1.8 Insulin Physiology

Insulin is a peptide hormone that serves as a primary regulator of glucose homeostasis in vertebrates. Insulin is secreted by pancreatic β cells, which are located in the islets of

Langerhans in the endocrine pancreas, in response to increased interstitial glucose levels

(Gerich, Charles, & Grodsky, 1976). The most widely known action of insulin is its initiation of the uptake of glucose by peripheral tissues including the liver, muscle, and adipose tissue. Insulin also suppresses gluconeogenesis in the liver and initiates glycogen synthesis in the liver and skeletal muscle. Insulin also initiates the synthesis of fatty acids and inhibits their oxidation. In addition to its metabolic action, insulin is also associated with mitogencity. Insulin initiates transcription of genes associated with cell growth and division (P. Kurtzhals et al., 2000). Secondary targets of insulin include the brain, spleen, and endothelium, which respond to mitogenic insulin signaling (Glendorf, Sorensen,

Nishimura, Pettersson, & Kjeldsen, 2008). Insulin is counter-regulated by glucagon, a peptide hormone secreted by pancreatic α-cells that increases blood-glucose concentrations. Glucagon plays a critical role in preventing hypoglycemia in non-diabetic individuals. The combined action of insulin and glucagon allows the endocrine pancreas to maintain fasting blood glucose levels between 80-100 mg/dl (4.4-5.6 mM) (J. J. Holst et al., 2017).

Increased interstitial glucose levels initiate the secretion of insulin by pancreatic β cells. Insulin is secreted through the pancreatic duct into the portal vein, from where it travels to the liver. The liver is the most insulin-sensitive tissue: the liver consumes 70% of all insulin secreted by the endocrine pancreas (M. D. Michael et al., 2000). Insulin reaches its remaining target tissues after passing through the liver, which clears 35-50% of insulin on first pass (Chap et al., 1987)\, into the hepatic vein and entering systemic circulation. Insulin is primarily removed from the bloodstream through receptor mediated clearance in which insulin is and its molecular target, the insulin receptor (IR), are internalized as a complex and degraded within the cytoplasm of target cells (M. D.

Michael, et al., 2000). Alternatively, insulin may be cleared renally. Insulin is filtered at the glomerulus where it may be excreted (<1% of insulin) or re-absorbed and degraded by proximal tubular cells. This process is distinct from IR-mediated clearance of insulin by tubular cells in the distal nephron, which are a hormonal target (Rabkin, Ryan, &

Duckworth, 1984).

Insulin secretion is well-regulated at the cellular level. Increased interstitial glucose- levels are detected as a function of glucose import by pancreatic β cells through the insulin-independent glucose transporter, Glut2. The influx of glucose leads to an increase in intracellular ATP synthesis, which in turn inhibits the export of potassium ions via

K/ATP channels. The resulting depolarization of the cellular membrane triggers the opening of voltage-gated Ca2+ channels. The influx of calcium into the β-cell initiates transport of insulin-containing secretory vesicles to the cell membrane and their subsequent fusion and secretion of insulin (Rorsman & Renström, 2003). In addition to glucose insulin secretion is stimulated by mannose, fructose, fatty acids, and certain

33 amino acids, such as leucine. However, it may be noted that unlike other nutrients, fatty acids and fructose do not stimulate the biosynthesis of proinsulin and have been linked to depletion of insulin stores within β-cells (Uchizono, Alarcon, Wicksteed, Marsh, &

Rhodes, 2007). The pancreatic hormone glucagon and the intestinal hormones gastric inhibitory peptide (GIP) and glucagon-like peptide 1 (GLP-1) also stimulate insulin secretion and biosynthesis. GIP and GLP-1 are responsible for the incretin effect, which refers to the second “long” phase of insulin secretion that is triggered by absorption of nutrients by the small intestine (Nauck, Stöckmann, Ebert, & Creutzfeldt, 1986). Insulin is stored within secretory granules for approximately 5 days. Insulin that is not secreted is degraded through autophagy and replaced with newly synthesized insulin. In this way, a store of functional insulin is maintained within pancreatic β cells (Boland, Rhodes, &

Grimsby, 2017).

Insulin initiates the uptake of glucose by its primary targets by binding to its molecular target, the IR. Upon complexation, IR, a receptor tyrosine kinase, undergoes a conformational change that initiates its autophosphorylation. Insulin activates two major signaling pathways. The primary pathway, which is associated with metabolic signaling, is transduced via insulin receptor substrate (IRS) adapter proteins through the PI-3/AKT pathway (Saltiel & Pessin, 2002). This pathway initiates the translocation of the sugar transporter Glut4 to the surface of target cells, thus facilitating the influx of glucose from the blood-stream into the cell. Additionally, this pathway initiates the synthesis of fatty acids and glycogen and inhibits fatty acid oxidation and glycogenolysis; gluconeogenesis is also inhibited in the liver. The second major signaling pathway associated with IR is the MEK/MAP/ERK pathway, which is associated with cell growth and mitogenicity

(Saltiel & Pessin, 2002). This pathway is associated with related hormones insulin-like growth factor I and insulin-like growth factor II (IGF-I and IGF-II, respectively) and is thought to originate from their shared evolutionary origins as hormones regulating both metabolism and growth (J. M. Conlon, 2001).

1.9 Insulin Receptor Structure

IR is the molecular target of insulin. It comprises four subunits, two extracellular α subunits and two transmembrane β subunits that are held together by a series of disulfide linkages: the α subunits contain the elements of the receptor that interact with insulin, whereas the β subunits contain the tyrosine kinase (TK) domain. The IR transcript is translated as a single protein that is proteolytically cleaved into separate α and β subunits that re-associate and dimerize with another pair of α and β subunits to form the mature receptor (Lawrence, McKern, & Ward, 2007).

The structure of the IR ectodomain has been determined at low resolution revealing a number of features pertinent to its biological activity (Croll et al., 2016). The ectodomain of IR comprises six subdomains. From N-terminus to C-terminus, these domains are as follows: a leucine-rich domain (L1), a cysteine-rich domain (CR), a second leucine-rich domain (L2), and three fibronectin-III-like domains, FNIII-1, FNIII-2, and FNIII-3. The cleavage site between the α and β subunits is found within the FNIII-2 domain; the C- terminal half of the FNIII-2 domain and the entire FNIII-3 domain are therefore contained within the β subunit. The C-terminal 120 residues of the α subunit are referred to as the insert domain (ID), which extends outward from FNIII-2 and terminates in a C- terminal α-helix (αCT). This region, which is largely disordered, contains the alternative splicing site that differentiates the A- and B isoforms of the IR. A 12 amino acid loop that

35 is present in the insert domain of IR-B; this segment, referred to as exon 11, is excised from the transcript of IR-A (Croll, et al., 2016).

Each monomer comprising the IR holoreceptor contains two distinct bindings sites (P.

De Meyts, 2015). Site 1, often termed the “classical site,” comprises the main β-helix of the L1domain of one α subunit and the αCT of the opposite α subunit (αCT’) (Huang et al., 2004; J.G. Menting, et al., 2014). This site binds insulin with nanomolar affinity. Site

2, termed the “low affinity site” comprises the C-terminal L2 and FNIII-1 domains and binds insulin with high-nanomolar to micormolar affinity. This site remains poorly characterized. High-affinity insulin-IR complexation requires binding to both sites. EM and fluorescence-based studies revealed a conformational change in IR upon insulin complexation (P. De Meyts, 2015). MD simulations further suggested that this transition involves the movement of αCT across L1 as insulin engages site 1. This fluctuation leads to a movement of the L1 domain towards L2 and FNIII-1 and directs the bound insulin molecule into site 2’ (site 2 of the IR monomer opposite to the L1 domain where the insulin initially bound). The conformational change associated with site 2 binding is propagated to the β-subunits of IR via Cys547 and activates the TK domain. Mutational analysis has revealed that the conformational change associated with high-affinity insulin-IR binding requires the transmembrane anchors of the β-subunits (Whittaker,

Garcia, Yu, & Mynarcik, 1994; Ye et al., 2017). IR also exhibits negative cooperativity between site 1 (and associated site 2’) and site 1’ (and associated site 2’). Insulin engaging site 1’ decreases the affinity of site 1 for insulin and destabilizes the insulin- bound conformation of IR (P. De Meyts, Roth, Neville, Gavin, & Lesniak, 1973).

1.10 Insulin-IR Complexation

The residues of insulin that interact with IR have been identified through mutational analysis and photocrosslinking studies. The residues that interact with the classical binding site of IR are predominantly found near the N-terminus of the insulin A chain

(residues A1-A3) or the C-terminus of the insulin B-chain (residues B24-B26). Several of these residues are buried within the core of insulin in its closed, storage form. A recent co-crystal structure of insulin bound to a fragment of IR containing L1, CR, and partial

L2 domains along with a solubilized αCT peptide (μIR), revealed that insulin in its receptor-bound form adopts an “open” conformation in which the C-terminal B chain detaches from the hydrophobic core of insulin via the partial unfolding of the B20-B23 β- turn. This exposes residues A1, A3, B24, and B26 and allows insulin to bind IR (J.G.

Menting, et al., 2014).

The three invariant N-terminal residues of insulin, GlyA1, IleA2, and ValA3, have been shown to interact with αCT (Xu, Hu, Chu, Wang, et al., 2004). Each of these residues has a strict steric or stereochemical requirement for high-affinity IR binding. The A1 position accommodates only amino acids capable of adopting right-handed conformations (D- amino acids or the achiral glycine) (Cosmatos, Okada, & Katsoyannis, 1976), whereas positions A2 and A3 have strict steric requirements. Substitution of IleA2 and ValA3 with naturally-occurring or non-standard aliphatic residues differing by a singly methylene group greatly reduces the ability of the variant analog to bind IR (J. P. Mayer, et al.,

2007); this is underscored by the clinical mutation insulin Wakayama, which contains a

ValA3→Leu mutation that results in an inactive insulin (Nanjo et al., 1986). The insulin-

μIR co-crystal structure revealed the intercalation of a histidine side chain of αCT with

37 the cavity created by the side chains of insulin residues A1-A3 (J.G. Menting, et al.,

2014).

The role of the aromatic triplet (comprising residues PheB24, PheB25, and TyrB26) in insulin-IR complexation has been determined to great detail through mutational analysis and x-ray crystallography. PheB24 binds within a hydrophobic pocket formed by the central β-sheet of the L1β-helix that was visualized in the insulin-μIR crystal structure

(J.G. Menting, et al., 2014). Through mutational analysis, this pocket was shown to accommodate phenylalanine or large aliphatic residues (such as Leu, Ile, Met, or cyclehexylalanine) (Vijay Pandyarajan, et al., 2014b). Although the B25 binding surface was poorly defined in the insulin-μIR structure (J.G. Menting, et al., 2014), the residue was shown to interact with αCT, a finding confirmed by photocrosslinking (Xu, Hu, Chu,

Huang, et al., 2004). The B25 binding site is thought to accommodate aromatic amino acids of varying size: substitution of PheB25 by Tyr, Trp, or the non-standard 2- naphtylalanine produces analogs with native-like IR affinity (Nakagawa & Tager, 1986).

The binding surface of the B26 residue, which was determined to lie within L1 by photocrosslinking (Xu, Hu, Chu, Huang, et al., 2004), was also poorly defined in the insulin-μIR crystal structure (J.G. Menting, et al., 2014). The constraints of insulin-IR complexation on the character of the amino acid residue at the B26 position are less rigid than those on the B24 or B25 positions. Substitution of TyrB26 by small, polar, charged, or aromatic residues produces insulin analogs with native-like or ehnhanced affinity

(Pandyarajan, et al., 2016). Although the B26 residue has been shown to be dispensable in the context of a truncated, amidated insulin analog (des-[B26-B30]-pentapeptide insulin-amide, DPA) (Mirmira, Nakagawa, & Tager, 1991), substitutions of large

Figure 1-7 Complexation of Insulin with site 1 of IR. (A) Crystal structure of the IR ectodomain (determined to 3.4 Å).The anterior monomer is depicted as a cyan tube. The individual subdomains of the α-subunit are labeled in cyan. The α-subunit of the posterior monomer is depicted as a white surface with the αCT highlighted in purple. The L1 domain of the anterior monomer and the αCT domain of the posterior monomer comprise the classical insulin binding site on IR (site 1, indicated by the black arrow). A second such site is present in the IR ectodomain; it constitutes the L1 domain of the posterior monomer and the αCT of the anterior monomer (labeled site 1’). (B) Crystal structure of insulin bound to a fragment of IR (μIR) highlighting the interactions between the aromatic triplet of residues (green sticks) and the L1domain (blue surface) and αCT

(purple surface) of IR at site 1. The rest of the insulin B chain is shown as black ribbon and the A chain is shown as yellow ribbon. (C) An alternate view of the insulin-μIR complex. Figure adapted from (J.G. Menting, et al., 2014) with the permission of the authors.

40 aliphatic residues at the B26 position attenuate IR affinity in the context of full-length insulin (Pandyarajan, et al., 2016).

Several residues within the core of the insulin hormone have also been shown to be conserved for their role in binding to IR. Residues ValB12 and TyrB16, which are part of the central helix of the insulin B chain, form critical contacts with the L1 domain (Huang, et al., 2004). Like that of ValA3, the binding site of ValB12 accommodates specific branched-side chains. Substitution of ValB12 by similar residues such as Thr or Ile results in a 20-fold decrease in receptor affinity, whereas the non-standard alloisoleucine at B12 retains ~80% of biological activity (Nakagawa, Tager, & Steiner, 2000). The contact between TyrB16 and L1 is less critical for insulin-IR complexation with AlaA16-insulin retaining 33% IR affinity. Photocrosslinking studies have suggested a transient contact between ThrA8, the C-cap residue of the A1-A8 helix, and αCT. Mutational analysis has revealed that although adequate, substitution ThrA8 by His or Gln increases the affinity of insulin for IR (Weiss et al., 2001). However, the details of the A8-IR contact are not well resolved, and cannot rationalize findings from mutational analyses (Xu et al., 2009).

The topography of site 2 of IR is not well characterized, thus structure/function relationships at the interface of insulin and this binding site are poorly understood.

Residues thought to interact with site 2 include SerA12, LeuA13, GluA17, HisB10, GluB13, and

LeuB17 (P. De Meyts, 2004). Whereas those residues that interact with the classical binding site of IR exhibit significant overlap with those that are responsible for insulin dimerization, the residues that interact with site 2 display significant overlap with the hexamer-forming surface of insulin (P. De Meyts, 2015). Structural models of insulin-IR complexation have suggested that site 2 binding is the key event in the activation of the

TK domain and initiation of the insulin-associated signal-transduction pathway (P. De

Meyts, 2000). Indeed, mutations at site 2-interacting residues such as HisB10→Asp, have been shown to affect not only the potency of insulin, but also to bias its signaling: AspB10 insulin has been shown to activate the mitogenic pathway associated with IR to a greater extent than WT insulin (Hansen, Kurtzhals, Jensen, Dejgaard, & Russell-Jones, 2011).

Studies of this mutation have also suggested that site 2 interaction may govern the kinetics of receptor binding and rate of internalization of IR as well (Ribel, Hougaard,

Drejer, & Sorensen, 1990).

The details of the receptor-bound conformation of insulin remain unknown. Whereas chiral substitution of residues GlyB20 and GlyB23 by L- and D-alanine suggested that the orientation and integrity of the B20-B23 β turn were critical for directing residues B24-

B26 onto their specific binding surfaces; this hypothesis was later confirmed by crystallographic structures. Chiral substitution of the invariant GlyB8 by D-serine attenuated the biological activity of insulin, whereas analogs with L-serine at the B8 position retained substantial activity. Such a finding suggested that the orientation of the

N-terminal insulin B chain is of importance to the achieving the receptor-bound conformation of insulin in spite of the dispensable nature of the individual residues (Q. X.

Hua et al., 2006). Whereas it was hypothesized that the R-state (helical) configuration of the N-terminal B chain may be a requirement for insulin-IR complexation, analogs that cannot achieve the R-state, such as 4SS-insulin retain high biological activity suggesting that the R-state per se does not reflect the IR-bound conformation of insulin (T. N.

Vinther, et al., 2013).

1.11 Diabetes Mellitus

T1DM is distinguished by the absolute lack of insulin. The root cause of the condition is the autoimmune destruction of pancreatic β-cells. “Insulitis” or infiltration of the central islet by immune cells. Islet destruction may be mediated by CD4+ T-lymphocytes, autoantibody-producing B-lymphocytes, or activation of innate immune response. T1DM typically manifests in the second decade of life after loss of about 50% of β cells in the endocrine pancreas (Bluestone, Herold, & Eisenbarth, 2010). Although a number of genes causing T1DM have been identified, primarily in the HLA region, the most prevalent is known as IDDM1, which is found in the MHC-II-coding region of chromosome 6. T1DM typically presents in the first or second decade of life (Noble &

Erlich, 2012). Whereas a majority of cases of T1DM manifest in the second decade of life, individuals with latent autoimmune diabetes of adults (LADA) develop T1DM during adulthood (Naik, Brooks-Worrell, & Palmer, 2009).

The pathogenesis of T2DM is more complex than that of T1DM. There are two major components to T2DM: the loss of sensitivity of peripheral tissues to insulin signaling, termed insulin resistance, and the inability of pancreatic β cells to synthesize and secrete the quantity of insulin required to maintain adequate blood-glucose levels (S. E. Kahn,

Cooper, & Del Prato, 2014). The incidence of T2DM is associated with obesity, dyslipidemia, and cardiovascular disease. The early stages of T2DM are defined by hyperinsulinemia, increased levels of C-peptide in the blood, and increased islet mass resulting from the upregulation of insulin biosynthesis and secretion. In later stages of the disease, β-cells develop secretory defects exhibiting decreased glucose-stimulated insulin secretion and resistance to incretins (Nauck, et al., 1986); eventually this results in

43 insulin-dependent diabetes as β-cell dysfunction progresses to its end stage (S. E. Kahn, et al., 2014).

Whereas insulin resistance (as assessed using the statistic HOMA-IR) may vary significantly even between non-diabetic individuals, only two mutations to the IR gene have been identified that contribute to T2DM (Longo et al., 2002; Semple, Savage,

Cochran, Gorden, & O'rahilly, 2011). Instead, a majority of the genetic causes of T2DM have been shown to affect the ability of pancreatic β-cells to produce insulin (S. Kahn,

2003). Genetic factors that contribute to the risk of developing T2DM include mutations to potassium channels (affecting the glucose-mediated insulin secretion), incretin genes

(affecting the stimulation of insulin secretion by the gut), and genes affecting the proliferation and growth of pancreatic β-cells (Frayling, 2007).

A subset of etiologies of T2DM are referred to as maturity onset diabetes of the young

(MODY). These condtions typically manifest in the third decade of life and caused by monogenic mutations. The genes associated with the most common forms of MODY include those encoding transcription factors expressed in liver cells (HNF4α, HNF1α, and

HNF1β) and GCK, which encodes glucokinase (Stanescu, Hughes, Kaplan, Stanley, &

De León, 2012). Milder forms of MIDY also manifest clinically as MODY (M. Liu,

Haataja, et al., 2010). In all, MODY contributes to approximately 5% of T2DM cases

(Fajans & Bell, 2011).

Independent of etiology, T1 and T2DM cause hyperglycemia, which is the primary cause of morbidity in diabetic individuals. Acutely, hyperglycemia can cause nausea, vomiting, and confusion. The most severe acute consequence of hyperglycemia, which occurs in patients with T1DM (Kerl, 2001) and a subset of those with T2DM (Westphal,

1996), is diabetic ketoacidosis (DKA). DKA results from the accumulation of unmetabolized ketone bodies produced through fatty acid oxidation; this condition can lead to coma and death (Kerl, 2001). Chronically elevated blood-glucose levels have been associated with microvascular disease, which is responsible for renal failure, neuropathy in distal appendages, and retinopathy. “Macrovascular” complications, such as circulation defects and increased cardiovascular mortality, have also been associated with hyperglycemia in patients with T1 and T2DM (Kilpatrick, Rigby, & Atkin, 2009).

Elevated blood-glucose levels have also been associated with the formation of cataracts and increased susceptibility to infection (Association, 2016).

Glycemic control is the central focus of therapy for DM. This is accomplished in

T1DM using insulin-replacement therapy (IRT) in which exogenous insulin is administered to patients to attempt to mimic the insulin secretion patterns of the endocrine pancreas. Intensive control of blood-glucose levels through IRT was shown to decrease the risk of microvascular disease in prospective clinical trials (DCCT) (DCCT,

1993a) and to have long-term benefits to cardiovascular morbidity and mortality (Lachin,

Orchard, Nathan, & Group, 2014). Glycemic control in T2DM is typically maintained using oral agents that increase insulin sensitivity, such as metformin or PPARγ agonists, or through the administration of insulin secretagogues such as sulfonylureas or GLP-1 agonists. Intensive insulin-based therapy of T2DM has been shown to be effective in a subset of patients (UKPDS, 1998), but may still increase the overall risk of cardiovascular morbidity (Margolis et al., 2014). Recent evidence has also suggested that the administration of basal doses of insulin early in the course of T2DM may have

45 protective effects on β-cell function improving recovery from T2DM; this occurs through a phenomenon termed “β-cell rest (Brown & Rother, 2008).”

The use of insulin to achieve glycemic control poses the risk of hypoglycemia, which itself may have clinical consequences as severe as those of hyperglycemia. Acutely, hypoglycemia may result in altered mental status, and in severe instances (BGL <60 mg/dl) seizures. loss of consciousness, and death. Long-term recurrence of even mild to moderate hypoglycemic episodes has been associated with early onset dementia (Philip E

Cryer, 1997). For this reason, approach to developing insulin-based regimens for glycemic control must be designed to reduce the number of hyper- hypoglycemic excursions (P. E. Cryer, 2002). Indeed, long-term evaluation of hypoglycemic episodes in the intensive arm of DCCT revealed that the long- and short-term risks associated with hypoglycemia may outweigh the benefits of intensive control (Gubitosi-Klug et al.,

2017). For this reason, therapeutic targets for glycemic control in patients with both

T1DM and T2DM have been relaxed to BGL of 140 mg/dL (corresponding to a glycated hemoglobin percentage of 7.5%) from the original 100 mg/dL (6.0%) (Karges et al.,

2014).

1.12 Development of Clinical Insulin Analogs:

The development of insulin analogs for clinical use has been motivated by the requirement for strict glycemic control in patients with DM. In T1DM such control is achieved by IRT, in which exogenous insulin is administered to act in place of insulin secreted by the endocrine pancreas, which secretes basal amounts of insulin during fasting periods with short bursts of secretion during feeding periods (stimulated by both incretins and increased BGL). This pattern is simulated by the administration of long-

46 acting insulin analogs that provide a basal level of insulin activity over a 12-48-hour period, and rapid acting insulin analogs that are administered during mealtimes (S. Garg,

Ampudia-Blasco, & Pfohl, 2010).

Rapid-acting insulin analogs were developed by engineering structural modifications to the hormone that de-stabilized its oligomers. As previously discussed, the primary factor limiting onset of the biological activity of pharmacologically administered insulin is the absorption of the hormone from its subcutaneous injection site; the rate of absorption of insulin is inversely proportional to its association state—the insulin monomer is absorbed more rapidly than dimers, hexamers, or intermediate species. A number of engineering strategies introduced mutations at the dimer interface of insulin

(predominantly its C-terminal B chain) that were expected to disrupt its native packing

(Brange & Vølund, 1999).

Insulin lispro (trade name Humalog®) was the first commercially available rapid- acting insulin analog. This analog contains an interposition of residues ProB28 and

LysB29, a modification that was based on the sequence of the monomeric IGF-I hormone

(Brems, et al., 1992). Crystallographic studies revealed that such an interposition eliminated a stabilizing interaction in the insulin dimer between ProB28 of one monomeric constituent and GlyB23 of the other (Ciszak, et al., 1995). This resulted in a 300-fold decrease in association constant of the insulin dimer (Attri, Fernandez, & Minton, 2010b;

Carpenter & Wilcox, 2014) and a 2-3 fold increase in hexamer dissociation rate in relation to WT insulin (D. T. Birnbaum, Kilcomons, DeFelippis, & Beals, 1997).

An alternative strategy utilized the introduction of repulsive electrostatic interactions at the dimer interface of insulin to decrease the stability of insulin oligomers. This

47 strategy led to the development of a series of analogs containing acidic amino-acid substitutions at positions B16, B25, B26, B27, B28, and B29, each of which were expected to repel residue GluB22 across the dimer interface and thereby de-stabilize the assembly. Analogs with substitutions of TyrB16 and PheB25 by acidic amino acids exhibited reduced biological activity, whereas substitution of TyrB26 by glutamic acid reduced the stability of the variant analog (Brange, et al., 1988; J. Markussen et al.,

2003). However, studies of analogs with acidic substitutions at the B28 and B29 positions led to the development of two commercially available rapid-acting insulin analogs.

Insulin aspart (trade name Novolog®) contains a substitution ProB28 by Asp, which, in addition to introducing repulsive electrostatics at the dimer interface, eliminates the

ProB28-GlyB23’ interaction. Insulin glulisine (trade name Apidra®) contains a substitution of LysB29 by Glu and an additional substitution of AsnB3 by Lys. Although this analog self-associates more readily than both insulins lispro and aspart, its structural stability permits its formulation in the absence of zinc (S. K. Garg, Ellis, & Ulrich, 2005). The analog is administered in a monomer-dimer equilibrium and is rapidly absorbed from the subcutaneous depot (Helms & Kelley, 2009).

The chemical and physical stability of insulin is a limiting factor in the development of rapid-acting insulin analogs. Although a number of insulin analogs with more rapid hexamer and dimer dissociation rates than the three clinical agents have been identified, their susceptibility to aggregation, precipitation, and chemical degradation has prevented their development into clinical analogs (Brange, et al., 1988; Mirmira, et al., 1991).

Indeed, even commercially available rapid-acting insulin analogs are especially susceptible to fibrillation and chemical degradation as a consequence of their diminished

48 ability to form protective oligomers even in the presence of zinc and phenol (Pandyarajan

& Weiss, 2012).

The development of long-acting insulin analogs through the stabilization of the insulin hexamer proved more challenging than its converse strategy. Chiral stabilization of the

B8 R6 insulin hexamer by the substitution of Gly (the “pivot point” defining the T- and R- states of the insulin hexamer) by Ser resulted in decreased foldability and destabilization of the tertiary structure of the variant analog (Q. X. Hua, Nakagawa, et al., 2006).

Attempts to stabilize the insulin hexamer through hydrophobic interactions also proved unsuccessful. Introduction of hydrophobic residues at the insulin dimer interface resulted in the aggregation and precipitation of the variant analogs, which limited their utility in clinical formulations (J. M. Markussen, Hougaard, Ribel, Sørensen, & Sørensen, 1987).

An insulin analog containing a substitution of GluB13 by Gln, which eliminated a repulsive electrostatic interaction in the insulin hexamer, was shown to form stable, long lived hexamers that were capable of associating even in the absence of zinc. However, the Glu→Gln substitution reduced the biological activity of the analog and thus limited its viability as a therapeutic agent (G. A. Bentley et al., 1992; J. Markussen et al., 1988).

Clinical long-acting insulin analogs were developed using strategies independent of the insulin hexamer. Early methods to protract the biological activity of insulin involved the formation of crystalline precipitates of insulin with positively charged proteins.

Insulin lente and ultralente were developed through the co-crystallization of porcine insulin with protamine that were slow to dissociate after subcutaneous injection

(Hagedorn, Jensen, Krarup, & Wodstrup, 1936; Wagner, Diez, Schulze-Briese, &

Schluckebier, 2009). The duration of action of the two analogs was 24h and 36h,

Figure 1-8. Structural summary of rapid-acting insulin analogs (A) Representation of modifications to the sequence of insulin to create rapid-acting insulin analogs. Insulin lispro (developed by analogy to insulin-like growth factor-I [IGF-I]), highlighted the

ProB28/GlyD23 dimer contact (highlighted in sky blue in panels B and C) (D. T. Birnbaum, et al., 1997). Other designs introduced polar or charged substitutions at the dimer interface. Among these were substitution of ThrB27, TyrB26, or TyrB16 by Glu (Brange &

Vølund, 1999; G. G. Dodson, Dodson, Turkenburg, & Bing, 1993) shown as black circles in A). Clinical analogs insulin glulisine (Apidra®) and insulin aspart (Novolog®) contain acidic residues at positions B29 (green) and B28 (yellow), respectively. These residues were hypothesized to repel residue GluD21, impairing formation of dimers and higher- order oligomers (Becker, Frick, Burger, Potgieter, & Scholtz, 2005; Tim Heise et al.,

2017). (B) The native residues of the positions listed above are highlighted in a 3D structure of an insulin dimer (PDB entry 5INS with color code as above). (C) Expanded view of panel B with residue positions as labeled.

51 respectively. Both of these analogs, which differ mainly in the size and structure of their crystalline precipitates have been clinically replaced by intermediate insulins neutral protamine hagedorn (NPH) and neutral protamine lispro (NPL), which have a maximum duration of activity of 12h and were shown to be less immunogenic than their predecessors (Owens, 2011). The utilization of slow-dissociating precipitates in the subcutaneous depot to protract insulin activity was further refined in the development of insulin glargine (trade name Lantus®). Insulin glargine contains an “Arg-Arg” extension at the C-terminus of its B chain (ArgB31, ArgB32); the introduction of these basic residues shifts the isoelectric point (pI) of the analog to near physiologic pH (7.4). This causes insulin glargine, which is formulated in solution at acidic pH, to form a slow-dissociating precipitate after injection into the subcutaneous space. Insulin glargine provides consistent insulin activity that lasts from 18-24h (Gillies, Figgitt, & Lamb, 2000).

The acylation of insulin has also been used to create basal insulin analogs. Insulin detemir (trade name Levemir®) was the first such insulin analog available commercially.

Insulin detemir comprises des-B30 insulin that is acylated at the ε-amino group of LysB29 with myristic acid. This modification effects the pharmacokinetic profile of insulin through two mechanisms. First, the alkyl chain of the myristic acid moiety forms inter- hexameric interactions after injection into the subcutaneous depot, retarding the absorption of insulin into systemic circulation (J Markussen et al., 1996). Second, the insulin forms a “circulating depot” by binding to albumin via the myristic acid moiety, preventing the hormone from absorption into peripheral tissues where it may engage IR

(Havelund et al., 2004). The duration of activity of insulin detemir is 9-14h (Owens,

2011). The use of acylation to protract the biological activity of insulin was further

52 refined in the development of insulin degludec (trade name Tresiba®). Insulin dgludec, like detemir, is a des-B30 insulin analog with an acyl modification at the ε-amino group of LysB29. However, rather than a myristyl moiety, insulin degludec is acylated by hexadodecanedioic acid via a γ-L-glutamyl spacer (Gough, Harris, Woo, & Davies,

2013). This modification allows insulin degludec to bind with greater affinity to albumin, creating a more stable circulating depot. Moreover, insulin degludec forms a stable subcutaneous depot through the formation of novel multihexameric assemblies within the subcutaneous space. Such assemblies form as a result of interaction between the distal carboxyl group of the hexadodecanedioic acid moiety and the zinc ion coordinated by a neighboring insulin hexamer (Steensgaard et al., 2013). The duration of activity of insulin degludec has been reported to last up to 42h as a result of these two sequestration mechanisms (T. Heise, Nosek, Bottcher, Hastrup, & Haahr, 2012).

Although insulin analogs currently in clinical use provide adequate glycemic control, insulin analogs with improved and more predictable PK/PD profiles are expected to improve clinical outcomes and avoid adverse effects of insulin-based therapy. The development of basal insulin analogs with further protracted activity are expected to improve adherence to insulin regimens by reducing the number of injections required.

Furthermore, elimination of “peaks” in the PD profile of basal analogs is expected to reduce hypoglycemic episodes, particularly overnight. Ongoing research has focused on the development of polymer-encapsulated insulins that would provide slow release of the hormone over extended periods of time. The development of rapid acting insulin analogs with faster onset and reduced “tailing,” is expected to reduce hyperglycemia at mealtimes and reduce hypoglycemic episodes during the “tail” phase of insulin action (Zaykov, et

53 al., 2016). Such analogs would be particularly useful in the context of continuous glucose monitoring (CGM) insulin pumps (Lv et al., 2015). Observation of delayed absorption of even monomeric insulin analogs in patients, particularly those with increased BMI has motivated development of insulin analogs with intrinsically improved absorption profiles (T Heise et al., 2015). Insulin Fi-Asp, a formulation of insulin aspart containing histamine, has exhibited improved absorption charcateristics in relation to typical formulations of Novolog®, this has motivated studies of insulin formulations that may further improve PK profiles of the hormone (Tim Heise, Hövelmann, et al., 2017).

1.13 Conclusion

The ubiquitous use of insulin in the treatment of T1DM has motivated research of the structure, function, and evolution of the hormone. Such research uncovered insulin’s rich evolutionary history and complex structure-function relationships making it an ideal model molecule for studying fundamental biophysical properties of proteins. Study of insulin from a translational perspective led to a number of breakthroughs in basic protein biochemistry and structural biology and vice versa. This dissertation highlights the contributions of insulin research to fundamental protein biochemistry and evolutionary biology and the ability of such research to contribute to the development of therapeutic analogs.

This chapter has highlighted the evolutionary constraints on the insulin hormone as they relate to its metabolic function, physical and chemical stability, and its ability to be synthesized in an efficient manner. The following chapters expand on this concept through studies interrogating structural features that are conserved primarily due to their role in the foldability and stability of the insulin protein. Such studies provide key

Figure 1-9. Historical strategies used to create basal insulin analogs A. Sites in insulin modified in long-acting insulins. Insulin glargine (Lantus®; purple) represents the most successful product (Gillies, et al., 2000) with pI shifted towards neutrality due to a basic extension of the B chain (J. M. Markussen, et al., 1987). A second class of basal insulins are derivatized with long acyl chains at LysB29 (I. Jonassen et al., 2012; Peter

Kurtzhals et al., 1995). These include insulin detemir (Levemir®; green circles) and degludec (Tresiba®; magenta circles) (I. B. Jonassen et al., 2010),(J. P. Mayer, et al.,

2007). These mechanisms are orthogonal to efforts to enhance the intrinsic stability of the insulin hexamer itself. Amino acid residues forming the dimer and hexamer interfaces of insulin in orange and light blue, respectively. HisB10, which coordinates with divalent zinc ions in hexamers, is highlighted in pink. B. Dimer- and hexamer interfaces in the zinc insulin hexamer (T6; color-coded as above). Analogs, designated “hydrophobic insulins,” were engineered to contain hydrophobic substitutions at dimer or hexamer interfaces in an effort to stabilize the hexamer. Another class of analogs introduced hydrophobic substitutions to the surface of insulin hexamers to create non-specific multihexameric precipitates (J. M. Markussen, et al., 1987). C. Electrostatic stabilization of the insulin hexamer: view of the negatively charged ring formed by the six GluB13 side chains

(highlighted green). An unobstructed view is shown in panel D. The most successful effort to stabilize the insulin hexamer involved substitution of GluB13 by Gln, which eliminated repulsion within the Glu residues. However, the GlnB13 analog had reduced biological activity (G. A. Bentley, et al., 1992).

56 information about the folding pathway of the insulin hormone that may be generalized to the broad class of globular proteins. This chapter has also highlighted the need for clinical analogs with improved PK/PD profiles and signaling capabilities; development of current clinical insulin analogs has revealed that engineered modifications that may be unfavorable to insulin in the context of vertebrate physiology may be clinically useful.

The second part of this dissertation highlights how such features (identified through the study of insulin as a model protein) may be exploited in the development of therapeutic analogs including basal insulin analogs and glucose responsive insulins (GRIs). Future investigation of the novel findings described in this dissertation may also uncover novel concepts regarding insulin-IR binding and associated signaling pathways.

Chapter 2: Toxic Protein Misfolding Constrains the Evolution of Insulin Sequences

2.1 Chapter Summary

The sequence of a globular protein encodes not only functional structures (collectively designated the native state), but also their foldability: i.e., a conformational search once efficient and robust to misfolding. Insights into molecular determinants of foldability have been provided by mutations associated with diseases of toxic misfolding. Of particular interest are residues that are conserved yet dispensable once the native state has been reached. Here, we have exploited the mutant proinsulin syndrome (a major cause of permanent neonatal-onset diabetes mellitus; DM) to highlight toxic misfolding as an implicit evolutionary constraint. Our studies focused on an invariant aromatic motif in the

B chain (PheB24-PheB25-TyrB26) with complementary functions in native self-assembly and receptor binding. A novel class of mutations provided evidence that insulin is able to bind to the receptor in two different modes, distinguished by a register-shift within this motif. Such “register-shift” mutations were found to be active in receptor binding but exquisitely susceptible to fibrillation in vitro and associated with impaired foldability in mammalian cell culture. Expression of the corresponding variant proinsulin led to induction of endoreticular (ER) stress, a general feature of the mutant proinsulin syndrome. The variant insulin (with sequence GlyB24-PheB25-TyrB26) was nonetheless equipotent with wild-type insulin in treating DM in a rat model. Our results suggest that the register-shifted mode of receptor binding has been disallowed by the implicit threat of toxic misfolding. This hidden constraint, not apparent in crystal structures of insulin or its model receptor complexes, limits sequence variation among vertebrate insulins.

2.2 Introduction

The structure and function of a globular protein are determined by its sequence. Yet the informational content of protein sequences must also contain determinants of folding efficiency, providing implicit safeguards against toxic misfolding (C. M. 14685248

Dobson, 2003). Such safeguards may be inapparent once the native state has been reached (Simkovsky & King, 2006; Thomas, Becka, Sargent, Yu, & King, 1990). In this study we demonstrate that insulin can in principle exhibit two modes of receptor binding, but only one is associated with protection from toxic misfolding. An invariant aromatic residue functions explicitly at the receptor interface and implicitly to ensure foldability.

Mutation of this residue in proinsulin enhances the risk of β-cell dysfunction and diabetes mellitus (DM) due to endoreticular (ER) stress (M. Liu, Hodish, et al., 2010).

The physical properties of extant biological proteins are unrepresentative of polypeptides as a general class of heteropolymers (Pedersen & Otzen, 2008). Whereas the thermodynamic ground state of a polypeptide may be an amyloid (C. M. Dobson, 1999), evolutionary winnowing of biological sequences has, in almost all cases, established large kinetic barriers protecting the nascent chain from this and other forms of aggregation-coupled misfolding (S. Betts & J. King, 1999; Westermark et al., 2005).

This principle is honored in the breach by human proteotoxic diseases: rare monogenic amyloidogenic syndromes (Selkoe, 2003; Stefani & Dobson, 2003) (such as the mutant lysozyme syndrome (Settembre et al., 2007)) and common neurodegenerative diseases of the elderly (Murphy, 2002). Studies of these exceptional cases (involving fewer than 20 of the >104 members of the human proteome (Monsellier & Chiti, 2007)) have stimulated

59 interest in what sequence- or structural features safeguard almost all encoded proteins from toxic misfolding (Fink, 2005).

Our study was motivated by the hypothesis that residues may function as critical safeguards but be dispensable in the context of the native protein (Simkovsky & King,

2006; Thomas, et al., 1990). Two classes of safeguards have been identified (P. S. Kim &

Baldwin, 1982). The first, known as prodomains, function within the primary translation product but are removed in the biosynthesis of the mature protein (Subbian, Williamson,

& Shinde, 2015). An example is provided by the connecting (C) domain of proinsulin, which markedly favors native disulfide pairing (Steiner & Oyer, 1967) relative to classical chain combination of isolated A- and B chains (Katsoyannis et al., 1967; C. C.

Wang & Tsou, 1991). Such internal catalysis enables native folding of proinsulin despite the intrinsic amyloidogenic properties of A- and B-chain segments (Eisenberg &

McLachlan, 1986). The second class of safeguards contains residues within the mature protein required for its foldability but otherwise dispensable for function once the native state has been reached (Thomas, et al., 1990). Overlooked in functional screens of protein analogs, such residues may nonetheless be conserved due to their cryptic roles in erecting kinetic barriers to toxic aggregation (Devlin et al., 2006; Fink, 2005).

Insulin provides a general model for studies of protein evolution (Brange, 2000; Y.

Yang, Petkova, et al., 2010). Biological selection is strict due to its essential role in metabolic homeostasis (J. M. Conlon, 2001). Indeed, large quantities of the hormone must be expressed and stored in pancreatic β-cells (the concentration of insulin in secretory granules can be as high as 12 mM (Weiss, 2013b)). The foldability of proinsulin is nonetheless only precariously maintained: the majority of human cell lines

60 cannot efficiently fold proinsulin (Hartley et al., 2010; M. Liu, Hodish, et al., 2010), highlighting the specialized milieu of the β-cell. Even in β-cells over-expression of proinsulin as a physiologic response to peripheral insulin resistance (e.g., in the context of obesity (Greenfield & Campbell, 2004)) can induce endoplasmic-reticular (ER) stress

(J. Sun et al., 2015), ultimately leading to β-cell dysfunction and death (S. Kahn, 2003).

This process appears central to the natural history of Type 2 DM (T2DM) (S. E. Kahn, et al., 2014). The current pandemic of obesity and T2DM reflects recent societal changes beyond the scope of hominid evolution (Zimmet, Alberti, & Shaw, 2001).

Given that even wild-type (WT) proinsulin lies near the border of ER stress (Weiss,

2013b), it is not surprising that mutations in the insulin gene (INS) might impair foldability in the absence of insulin resistance (Raile et al., 2011; Stoy, et al., 2007).

Indeed, such mutations define a monogenic cause of DM, collectively designated the mutant proinsulin syndrome (also known as mutant INS diabetes of the young [MIDY]

(Vinik & Bell, 1988)). The prototype of such mutations, CysA7Tyr, was first identified as an autosomal diabetes locus in the Akita mouse (Izumi et al., 2003) and then observed in human neonatal-onset DM (N. Herbach et al., 2007). Diverse clinical mutations remove or introduce a cysteine, leading in either case to an odd number of thiol groups

(Edghill, et al., 2008). The variant proinsulin must therefore contain an unpaired cysteine, in principle mediating aberrant disulfide interchange and formation of inter-molecular disulfide bridges (M. Liu, et al., 2005). Such a variant proinsulin interferes with the folding, trafficking and secretion of the WT protein (M. Liu, Haataja, et al., 2010), leading in the first year of life to unremitting ER stress, β-cell dysfunction, apoptosis and permanent DM (M. Liu, et al., 2007).

Of particular interest are DM-associated mutations in proinsulin not involving cysteine.

Whereas patients with Cys-related mutations invariably present with neonatal DM, non-

Cys-related mutations are associated with ages of onset ranging from neonatal to the third decade of life; the latter pedigrees typically exhibit incomplete penetrance (Ming Liu, et al., 2015). These mutations presumably perturb, to varying extents, key structural contacts stabilizing on-pathway folding intermediates (Weiss, 2013b). An example of a severe mutation is GlyB8Ser, in which chiral perturbation of the B8 ϕ ?dihedral angle markedly impairs the efficiency of disulfide pairing (Q. X. Hua, Nakagawa, et al., 2006;

Nakagawa, Zhao, Hua, & Weiss, 1998; M. A. Weiss, 2009b). The present study was broadly stimulated by a mild mutation, PheB24Ser (S. E. Shoelson, Polonsky, Zeidler,

Rubenstein, & Tager, 1984b). This variant was originally described as an insulinopathy

(designated insulin Los Angeles (S. Shoelson, Fickova, et al., 1983)) due to its isolation in the proband’s serum—implying at least partial preservation of prohormone processing, trafficking and secretion (S. E. Shoelson, et al., 1984b). SerB8-insulin and SerB24-insulin each exhibit decreased but non-negligible receptor-binding affinity (reduced by ca. two- and tenfold, respectively) (Q. X. Hua, Nakagawa, et al., 2006; Nakagawa, et al., 1998; S.

Shoelson, Haneda, et al., 1983). Such decrements are unlikely to explain the phenotypes of heterozygous patients. Instead, the relative severities of their associated clinical phenotypes correlate with extent of ER stress induced on expression of the respective variant proinsulin in β-cell lines (M. Liu, Haataja, et al., 2010).

PheB24, invariant among vertebrate insulins and insulin-like growth factors (IGFs), anchors the B20-B23 β-turn in the free hormone (Vijay Pandyarajan, et al., 2014b), and directly contacts the IR (J.G. Menting, et al., 2014). In the classical structure of insulin

62 this turn enables the C-terminal B-chain β-strand (residues B24-B28) to pack against the conserved side chains of IleA2, LeuB11, and LeuB15. These contacts seal the hydrophobic core of the α-helical domain. On insulin self-assembly the β-strand forms a dimer-related antiparallel β-sheet (E. N. Baker, et al., 1988; Vijay Pandyarajan, et al., 2014b). On binding to the primary hormone-binding elements of the IR, a conformational change ensues: rotation of the B20-B23 β-turn and adjoining PheB24-PheB24 element leads to detachment of the C-terminal B-chain β-strand from the core. The β-strand itself packs in a groove between IR elements L1 and αCT: the side chain of PheB24 packs within a nonpolar pocket whereas the side chains of PheB25 and TyrB26 occupy more peripheral sites (J.G. Menting, et al., 2014). Although the B24-B25-B26 triple of aromatic amino acids is a hallmark of the vertebrate insulin family (J. M. Conlon, 2000; J. M. Conlon,

2001; Mirmira, et al., 1991), this motif is not present in invertebrate homologs (Chan &

Steiner, 2000; Grönke, Clarke, Broughton, Andrews, & Partridge, 2010).

Mutational analysis of the B24-B26 aromatic triplet uncovered distinct and specific side-chain determinants of IR binding. The B24-binding pocket (defined by residues in

L1 and the central B-chain α-helix) optimally accepts Phe but can also accommodate cyclohexanylalanine or branched aliphatic side-chain; Tyr, His and Trp are disfavored

(Mirmira & Tager, 1989; Vijay Pandyarajan et al., 2014a). The B25-binding cleft in the

αCT domain requires aromatic side chains without significant distinction among Phe, Tyr or Trp (Mirmira & Tager, 1991; Nakagawa & Tager, 1986). Despite the conservation of

TyrB26 (Phe among IGFs) (Pandyarajan, et al., 2016), the solvent-exposed B26-binding surface accommodates diverse charged, polar or aromatic side chains (Mirmira, et al.,

1991; Pandyarajan, et al., 2016). The striking preservation of the B24-B26 aromatic

63 triplet among vertebrate insulins and IGFs—conserved for more than 500 million years

(Chan & Steiner, 2000)—is believed to represent intersecting evolutionary constraints imposed by function, foldability, assembly and stability.

The present study has addressed a long-standing anomaly: the native activity of

GlyB24-insulin (Mirmira, et al., 1991; Mirmira & Tager, 1989). A putative framework was provided by the register-shift model (Figure 2-1B): an alternative receptor-binding mode of insulin in which residue B24 reorients as part of a non-canonical five-residue chain reversal (B20-B24), in turn enabling residues B25-B27 to engage the respective B24-,

B25-, and B26-binding pockets of the IR (Vijay Pandyarajan, et al., 2014b; Žáková et al.,

2013). This paradigm posits that the introduction of GlyB24 (or D-amino-acid substitutions at B24 (Kobayashi et al., 1982; Mirmira & Tager, 1989)) destabilizes the classical four-residue β-turn (B20-B23). The proposed register shift would (i) preserve insertion of a Phe within the B24-binding pocket (PheB25), (ii) deploy a well-tolerated Tyr within the B25-binding pocket (TyrB26), and (iii) exploit the promiscuity of the solvated

B26-binding surface to accommodate a small polar side chain (ThrB27) (Figure 2-1C).

Our results provide mutational evidence in support of the register-shift model and suggest that this alternative mode of binding has been excluded in the evolution of the vertebrate insulin family by the susceptibility of such active yet unstable insulin analogs to toxic protein misfolding.

As suggested in previous studies, GlyB24 is expected to have defects in foldability and thermodynamic stability. The absence of a stabilizing side-chain at the B24 position, along with the torsional flexibility of the ectopic glycine residue are expected to increase the propensity for unfolding of the B20-B23 β turn

Figure 2-1 Sequence and binding mode of insulin. (A) The mature insulin hormone comprises two polypeptide chains, designated “A” and “B” that are connected by two disulfide bridges spanning residues A7-B7 and residues A20-B19; a third intra-chain disulfide bridge spans residues A6-A11. The “aromatic triplet” of residues comprises residues PheB24, PheB25 (red), and TyrB26 (blue). Insulin analogs in which PheB24 is substituted for Gly have been shown to have native-like biological activity and affinity for IR. Insulin analogs created using semisynthetic methods are created in the context of

OrnB29 insulin, in which LysB29 is substituted for a non-standard amino acid, ornithine.

This substitution eliminates the tryptic site between residues B29 and B30. (B) Cartoon representation of insulin-IR binding. The C-terminal B-chain of insulin detaches from the hydrophobic core of the molecule and intercalates between the αCT and L1 domains of

IR with the aromatic triplet of residues occupying specific binding surfaces on αCT and

L1, the binding surfaces of B24, B25, and B26 are indicated by orange, purple, and blue- gray half circles, respectively. The proposed mechanism of GlyB24 insulin binding to IR suggests a “register shift” in which the GlyB24 substitution directs PheB25 into the B24 binding pocket TyrB26 onto the B25 binding surface with ThrB27 occupying the B26 binding surface as indicated. (C) View of B24-B26 binding surfaces in insulin-IR co- crystal structure. A co-crystal structure of insulin bound to a fragment of IR that comprised the L1 and CR domains and a soluble αCT peptide (side and top view, left and right, respectively). The side chain of PheB24 was shown to occupy a well-defined pocket in L1 whereas PheB25 occupied a cleft in αCT and TyrB26 occupied a solvent-exposed binding surface on L1.

(Q. X. Hua, Shoelson, & Weiss, 1992; Olsen, et al., 1996). This change in conformational dynamics is expected to decrease fibrillation lag time of GlyB24 insulin and reduce its folding efficiency in vivo.

2.3 Results:

A series of mutant analogs were designed to validate the “register shift” model of

GlyB24 insulin-IR binding. These analogs exploited the requirements of the B24 and B25 binding surfaces of IR for specific side chains. Whereas the B24 binding pocket has strict size and geometric requirements for side chain occupancy, analogs with β-branched aliphatic or small aromatic residues (Leu, Cha, or Phe) at the B24 position are expected to retain native-like affinity for IR. Aromatic side chains (Phe, Tyr, and Trp) are required for engagement of the B25 binding surface. If a register shift were to take place in the context of a GlyB24 mutation, analogs with GlyB24, Leu/ChaB25 substitutions would be expected to retain high biological activity and receptor affinity relative to B25 mutants containing the native PheB24. Conversely, a GlyB24 analog with a TyrB25 substitution would be expected to have attenuated biological activity and receptor affinity as the larger Tyrosine side chain is directed towards the conformationally rigid, compact B24 binding pocket rather than the B25 surface, where the mutation would retain native-like activity (Figure 2-2).

To this end, a series of analogs containing GlyB24 substitutions along with ChaB25,

LeuB25, or TyrB25 were created semisynthetically (Inouye et al., 1979) along with control analogs with single substitutions of Cha, Leu, or Tyr at either the B24 or B25 position.

To eliminate the tryptic site between LysB29 and ThrB30, a substitution of LysB29 for the

Figure 2-2 Cartoon representation of the experimental design utilized to test the register shift paradigm. The aromatic triplet of residues in the C-terminal B-chain of insulin is critical for insulin-IR complexation. Residues PheB24, PheB25, and TyrB26 must occupy specialized binding surfaces (denoted by orange, purple, and blue-gray semicircles) on the αCT and L1 domains for proper IR engagement (shown in the inset).

In the “register shift” model, the ectopic glycine residue at the B24 position would allow

PheB25 to occupy the B24 binding pocket, whereas TyrB26 occupies the B25 binding surface and ThrB27 occupies the B26 binding surface (shown below the inset as indicated by the black arrow). This model may be validated using a series of position-specific mutations at the B24 and B25 positions. Whereas branched-chain aliphatic amino acids, such as leucine or cyclohexyl alanine (Cha), are functional at the B24 position, they are non-functional at the B25 position. For this reason, as indicated in “Model 1”, the register shift model would predict that an analog with GlyB24 and Cha or LeuB25 substitutions would be biologically active due to direction of the Cha or Leu side chain (denoted by the yellow circle labeled “B”) into the B24 binding pocket (orange semicircle) rather than onto the B25 binding surface (purple semicircle). Conversely, tyrosine is functional at the

B25 position but not the B24 position (as shown in the top two rows of “Model 2”). An analog with a GlyB24 and a TyrB25 mutation is predicted to be biologically inactive due to direction of the Tyr side chain towards the B24 binding pocket where it is unable to properly intercalate (bottom row of “Model 2”).

non-standard amino acid, ornithine (Orn), was introduced to all analogs, allowing

analogs to be amenable to trypsin-catalyzed semisynthesis.

In Vitro Receptor Binding

In vitro results provided evidence supporting the register shift model of IR binding

(Figure 2-3; Table 2-1); these results are summarized in Table 1. GlyB24 OrnB29 insulin

showed native-like affinity for IR compared to OrnB29 insulin and wild-type human

insulin. The introduction of the GlyB24 mutation to Cha or LeuB25 analogs was able to

rescue the IR affinity in relation their PheB24 counterparts, suggesting that the

introduction of the glycine residue may displace the B25 side chain from the B25 to the

B24 binding surface on IR. The IR affinity of GlyB24 TyrB25 was attenuated in relation to

PheB24TyrB25. This result suggests that the GlyB24 mutation is unlikely to cause global

changes in the insulin-IR binding complex increasing the affinity of the analog- a

hypothesis that has been proposed in previous studies (J.G. Menting, et al., 2014;

Mirmira, et al., 1991).

Assessment of Biological Activity

To assess the physiological relevance of the in vitro receptor-binding data, the biological activity of the mutant insulin analogs was tested by intravenous injection into rats rendered diabetic by streptozocin (Figure 2-4; Table 2-1). Such studies are further motivated by previous evidence indicating discrepancies between in vitro receptor-binding and the biological profiles of insulin analogs (Ribel, et al., 1990), which may be affected

by such factors as clearance of the analog, ability of the analog to initiate signaling pathways, and differences between the conformation of the analog in the bloodstream and those in buffered solutions.

Figure 2-3 In vitro competition IR binding assays. Affinities of B24 and B25 insulin analogs along with controls were determined in vitro, raw data are shown as colored points with smooth lines showing fit dissociation curves, KD values were calculated from fit curves, they are summarized in Table 1. A. Dissociation curves and raw data of wild- type HI (black squares), OrnB29 insulin (blue squares), and GlyB24 OrnB29 insulin (red squares); B. ChaB24OrnB29 yellow squares, ChaB25OrnB29 (orange squares) GlyB24 ChaB25

OrnB29 (brown squares); C. LeuB24OrnB29 (lavender squares), LeuB25OrnB29 (cyan squares)

GlyB24 LeuB25 OrnB29 (crimson squares), D. TyrB24OrnB29(green squares), TyrB25OrnB29

(blue squares) GlyB24 TyrB25 OrnB29 (purple squares).

Table 2-1. Biological Function of Insulin Analogs

Analog % AUC KD (N) (nm) LeuB24-OrnB29 α 61 ± 13 (9) 0.13 ± 0.03

LeuB25-OrnB29 * 78 ± 11 (9) 4.9 ± 2.5

GlyB24-LeuB25-OrnB29 α 62 ± 13 (9) 0.48 ± 0.10

ChaB24-OrnB29 α 53 ± 14 (12) 0.07 ± 0.05

ChaB25-OrnB29 * 80 ± 10 (10) 11.00 ± 2.4

GlyB24-ChaB25-OrnB29 α 63 ± 12 (15) 0.12 ± 0.02

TyrB24-OrnB29 α 68 ± 8 (10) 1.73 ± 0.57

TyrB25-OrnB29 * 45 ± 10 (10) 0.03 ± 0.005

GlyB24-TyrB25-OrnB29 α 72 ± 12 (10) 0.95 ± 0.19

GlyB24-OrnB29 61 ± 12 (10) 0.06 ± 0.02

OrnB29 61 ± 12 (35) 0.07 ± 0.04

HI 52 ± 10 (22) 0.08 ± 0.03 *Displayed statistically significant difference from B24 analog as determined by two- sided students’ T-test with a significance level of 0.05 α Displayed statistically significant difference from B25 analog as determined by two- sided students’ T-test with a significance level of 0.05

Rodent assays showed results consistent with those obtained in vitro. GlyB24 OrnB29 insulin displayed biological activity equivalent to that of OrnB29 insulin, whereas wild-type human insulin displayed slightly increased in vivo potency. The introduction of the GlyB24 to the minimally active Cha and LeuB25 insulin analogs rescued the biological activity of the analogs. The double mutant analogs displayed biological activities similar to those of

Cha or LeuB24 insulins. Conversely, addition of GlyB24 to TyrB25 insulin, an analog that displayed increased potency, attenuated the biological activity of the analog. The glucose- clearance profile of GlyB24 TyrB25 insulin was similar to that of TyrB24 insulin, which also showed diminished biological activity. Biological potencies of analogs, quantified by

AUC, are summarized in Table 1.

Assessment of Thremodynamic Stability

Circular dichroism (CD) spectroscopy-monitored guanidine titrations were used to interrogate the impact of the GlyB24 analog on the global secondary structure of the insulin hormone. It was hypothesized that the introduction of a glycine in the place of PheB24 would cause a discernable perturbation in the secondary structure of the globular insulin hormone, causing the mutation to be disfavored evolutionarily. Additionally, such perturbation would limit the application of GlyB24 insulin as a therapeutic analog.

To this end, the thermodynamic stability of functional GlyB24 analogs in relation to their parent analogs was assessed. Native OrnB29 insulin displayed the greatest stability of

B24 surveyed analogs (ΔGu of 3.6 kcal/mol). The substitution of Phe for Cha or Leu led to a

B24 B24 significant loss of stability: Cha and Leu analogs displayed ΔGu values of 2.5 kcal/mol and ΔGu of 2.2 kcal/mol, respectively. Such results are in accordance with

B24 B29 B24 previously reported data. Gly Orn insulin displayed a ΔGu similar to that of its Leu

Figure 2-4 In vivo assessment of biological activity in streptozotocin-induced diabetic rats. Left hand panels display averaged curves of glucose-lowering as a fraction of initial blood-glucose level; error bars represent SEM. Right hand panels display box plots of fractional area under the curve (AUC) of glucose-lowering curves. This statistic was calculated over 240 minutes of time with maintenance of initial blood-glucose level corresponding to 1.0. The lower and upper edges of each box represent the first and third quartiles, respectively, whereas the whiskers represent median 5th and 95th percentiles.

Median values are represented as gray lines within each box and mean values are represented as white lines. Individual data points are represented as gray diamonds.

Curves correspond to analogs as follows: A,B. Diluent vehicle (gray), human insulin

(black), OrnB29 (blue), GlyB24OrnB29 (red); C,D. ChaB24OrnB29, ChaB25OrnB29, GlyB24

ChaB25OrnB29; E,F. LeuB24OrnB29, LeuB25OrnB29, GlyB24 LeuB25OrnB29;. G,H.

TyrB24OrnB29, TyrB25OrnB29, GlyB24 TyrB25OrnB29. AOC values of potency are listed in

Table 1.

76 counterpart (2.3 kcal/mol). Substitutions of PheB24 for non-aromatic residues has been reported to impact the global thermodynamic stability of the insulin molecule as a result of the loss of aromatic-aromatic interactions with TyrB26 and through suboptimal packing of smaller side-chains against the hydrophobic core of the protein (Ratnaparkhi &

B24 Varadarajan, 2000). The similar ΔGu values of Leu and Gly insulin suggests the primary defect in GlyB24 stability is the result of the loss of the PheB24 side chain.

ChaB25 and LeuB25 mutations at B25 displayed decreased thermodynamic stability compared to the wild type; both analogs had greater thermodynamic stability than their

B25 B25 B24 counterparts: Cha displayed a ΔGu of 2.7 kcal/mol and Leu displayed a ΔGu of

2.9 kcal/mol. Unlike PheB24, PheB25 does not interact with the core of the insulin protein. It is exposed to solvent in the context of monomeric insulin. The decrease in thermodynamic stability between Cha and LeuB25 analogs is likely the result of the increased hydrophobic character of the solvent-exposed B25 side chains (Jørgensen, Kristensen, Led, &

Balschmidt, 1992).

The introduction of GlyB24 in the context of Cha and LeuB25 leads to a dramatic loss of thermodynamic stability, GlyB24 LeuB25 displayed a thermodynamic stability of 1.7 kcal/mol whereas GlyB24 ChaB25 displayed a thermodynamic stability of 2.1 kcal/mol.

These data are consistent with the properties of GlyB24 and Cha and LeuB25 analogs. The marked reduction in the stability of GlyB24 LeuB25 may suggest a synergistic effect, rather than an additive one, of the GlyB24 and LeuB25 mutations (Figure 2-5, Table 2-2).

Amyloid Fibrillation

The ability of functional GlyB24 analogs to form fibrils compared to that of their B24 and B25 single mutant counterparts was tested. Each of the three functional GlyB24 analogs

Figure 2-5 Assessment of secondary structural stability of analogs by circular dichroism spectroscopy. A-C. Circular dichroism spectra of insulin analogs. Circular dichroism spectroscopy was used to assess the secondary structural stability of insulin analogs, far UV wavelength scans were used to assess global secondary structure of functional GlyB24 insulins and their parent analogs. D-F. Unfolding of analogs as a result of increasing concentration of guanidine HCl denaturant was monitored using CD signal at 222 nm. Resulting curves were fit to two-state unfolding models to determine free energy of unfolding, as summarized in Table 2. Wavelength scans (left) are shown in the left-hand panels and guanidine titration curves are shown on the right. Spectra correspond to analogs as follows: A,D. HI, black; OrnB29, blue; GlyB24 OrnB29. B,E. ChaB24 OrnB29, yellow; ChaB25 OrnB29, orange; GlyB24 ChaB25 OrnB29, brown. C,F. LeuB24 OrnB29, lavender; LeuB25 OrnB29, dark cyan; GlyB24 LeuB25 OrnB29, green.

Table 2-2. Stabilities of Insulin Analogs

a b Analog ΔGu Cmid m fibrillation lag (kcal mol-1) (M) (kcal mol-1 M-1) timec (days (N)) LeuB24-OrnB29 2.2 ± 0.1 4.4 ± 0.09 0.5 ± 0.01 4.7 ± 0.5 (3) LeuB25-OrnB29 2.9 ± 0.1 4.8 ± 0.16 0.6 ± 0.02 14.0 ± 1.7 (3) GlyB24-LeuB25-OrnB29 1.7 ± 0.1 4.3 ± 0.32 0.4 ± 0.03 3.0 ± 0.1 d (3) ChaB24-OrnB29 2.5 ± 0.1 5.0 ± 0.30 0.5 ± 0.03 7.0 ± 1.3 (3) ChaB25-OrnB29 2.7 ± 0.1 4.5 ± 0.08 0.6 ± 0.01 5.0 ± 0.1 d (3) GlyB24-ChaB25-OrnB29 2.1 ± 0.1 4.2 ± 0.25 0.5 ± 0.03 4.3 ± 0.6 (3) GlyB24-OrnB29 2.3 ± 0.1 4.6 ± 0.28 0.5 ± 0.03 3.0 ± 0.1 d (3) OrnB29 3.6 ± 0.2 4.9 ± 0.21 0.7 ± 0.03 7.8 ± 2.2 d (6) HI 3.4 ± 0.1 5.0 ± 0.20 0.7 ± 0.03 3.4 ± 1.7 (5) aThermodynamic parameters were inferred from CD-detected guanidine denaturation data by application of a two-state model. bThe m-value (slope Δ(G)/Δ(M)) correlates with extent of hydrophobic surfaces exposed on denaturation. cFibrillation lag times pertain to zinc-free wild-type insulin and analogs (in a monomer-dimer equilibrium) ; each protein was made 60 µM in phosphate- buffered saline (pH 7.4). A twofold increase over baseline in ThT fluorescence provided a criterion for onset of fibrillation. dAll individual samples in this set exhibited the same lag time of 2 days. As the method employed in this study could not distinguish fibril lag times with a resolution given in hours, some variance was added to the data by adding ± 0.1 to individual data points in order to obtain standard deviations.

(GlyB24 OrnB29, GlyB24 ChaB25 OrnB29, and GlyB24 LeuB25 OrnB29) showed foreshortened fibrillation lag times compared to native PheB24 or Cha or LeuB24 analogs. This effect is particularly pronounced between the GlyB24 OrnB29 and the native OrnB29 analog, which displayed fibrillation lag times of 3 days and 9 days, respectively. Substitution of PheB25 for Leu delayed the fibrillation lag time of OrnB29 insulin whereas substitution of the residue with Cha decreased lag time.

The results suggest that the impact of the GlyB24 mutation on fibrillation lag time is more dramatic in relation to LeuB24 or ChaB24 substitutions, indicating that GlyB24 renders insulin particularly susceptible to fibrillation in relation to other B24 substitutions. This may occur as a result of the lack of side chain of glycine, preventing the B24 residue to properly anchor the C-terminal B-chain to the hydrophobic core of insulin, or due to the increased torsional flexibility of glycine, which may disrupt the B20-B23 β-turn (Figure 2-

6; Table 2-2).

Assessment of R6 Hexamer Stability by EDTA-Cobalt Extraction Assays

The role of the C-terminal B-chain of insulin in the monomer-monomer interface of the insulin dimer motivated the study of the impact of GlyB24 on insulin oligomers. EDTA-

Cobalt extraction assays were performed to determine the stability of metal-ion-

B24 coordinated, phenol-stabilized (R6) insulin hexamers. Gly insulin was found to have a

B29 markedly reduced hexamer lifetime (t1/2 = 72 s) compared to that of the parent Orn (492 s), wild-type HI (460 s), and the rapidly-dissociating insulin analog lispro (KP) (278 s).

Furthermore, the magnitude of the 574 nm absorbance band that is associated with tetrahedrallycoordinated Co2+ ions was attenuated in the GlyB24 sample compared to HI,

OrnB29, and KP samples. These data suggest that GlyB24 insulin is impaired in its ability to

Figure 2-6. Assessment of fibrillation lag times of insulin analogs. A dot plot of lag time (days) to fibril formation of functional GlyB24 insulin analogs and their parent analogs. Onset of fibrillation was defined by a 2-fold enhancement of thioflavin T fluorescence. Mean fibrillation lag times are shown as black horizontal lines. Fibrillation lag time values are summarized in table 2.

83 form R6 hexamers, as evidenced by the attenuation of the 574nm absorbance band, and those hexamers that do form are short-lived. The destabilization of the C-terminal B-chain, which comprises a majority of the monomer-monomer interface in the insulin dimer and hexamer, is the likely cause of the weak GlyB24 hexamerization potential of GlyB24 insulin

(Figure 2-7).

Assessment of Self-association by Size-Exclusion Chromatography

The propensity of GlyB24 insulin and control analogs to self-associate in the absence of zinc or phenol was determined using size-exclusion chromatography (SEC). Insulin analogs were injected onto an Zenix 150-SEC column as described (ref). Predicted molecular weights are indicated in Figure 2-8B. GlyB24 displayed a largely monomeric association state compared to the parent OrnB29 analog and wild-type HI with the predicted molecular mass of GlyB24. The elution profile of KP represented an association state that was more monomeric than that of GlyB24 insulin.

To assess association state of analogs in the presence of zinc and phenolic ligands, insulin analogs were formulated in buffer containing zinc and phenol and injected onto the

SEC column with the mobile phase supplemented with 0.3 mM ZnCl2 and 50 mM cyclohexanol. Under these conditions, HI displayed the highest-order association state of

49000 Da. The elution profile of OrnB29 contained two distinct peaks corresponding to two distinct populations with molecular weights of 46800 Da and 31700 Da. KP eluted primarily in a single peak corresponding to a species with a molecular weight 32300 Da with a smaller peak corresponding to an association state similar to that of HI (47900 Da).

GlyB24 insulin showed a markedly decreased potential for self-association eluting as a

Figure 2-7 Visible absorption spectra of cobalt-stabilized hexamers and kinetics of metal ion release. A, Co2+ d-d bands of human insulin (black), OrnB29 insulin (blue), insulin Lispro, or KP (green), and GlyB24Orn29 insulin (red) near 574 nm provide a

B24 29 signature of the R6 (or Rf ) hexameric state. The amplitude of Gly Orn was attenuated in relation to human insulin, OrnB29 insulin, and KP. Au, absorption units. B, sequestration of divalent cobalt ions from insulin analogs by EDTA; human insulin

(black), OrnB29 insulin (blue), insulin Lispro, or KP (green), and GlyB24Orn29 insulin

(red). Half-life values of analogs are summarized in the upper right corner.

86 small peak corresponding to a molecular weight of 30900 Da with a larger, diffuse peak corresponding to a monomer-like association state (3310 Da) (Figure 2-8).

Although the molecular weights of oligomeric insulin species did not correspond absolutely to masses of insulin hexamers, dimers, or monomers, their relative elution times provided evidence of differences in their association states. Whereas HI displayed a modest increase in self-association and stability of higher-order oligomeric species

(presumable hexameric in nature) compared OrnB29 insulin; insulin lispro (KP) had a significantly smaller association state, corresponding to tetramers or other intermediate species associated with the dissociation of hexamers. GlyB24 insulin has a markedly reduced association state compared to the rapidly dissociating KP with a majority of the analog eluting as a monomer with a minor peak on the chromatogram corresponding to an intermediate associated with hexamer dissociation.

Assessment of Folding and Secretion in Cells

The consequences of the biophysical and biochemical properties of GlyB24 insulin on the synthesis and secretion of the hormone were assessed in vivo. Constructs expressing wild-type proinsulin or proinsulin variants were transfected into HEK 293T cells. Proteins were radioactively labeled with S35-labeled cysteine in a 30-minute pulse-chase experiment. The foldability of proinsulin variants compared to wild-type proinsulin was assessed through native, non-reducing SDS-PAGE analysis of cell lysates. Whereas wild- type proinsulin was determined to fold most efficiently, GlyB24 insulin displayed a statistically significant decrease in folding efficiency. Compared to the well-studied clinical mutant proinsulin, SerB24 Proinsulin (also known as insulin Los Angeles), GlyB24 proinsulin had a modest decrease in folding efficiency that was not found to be statistically

Figure 2-8 HPLC size-exclusion chromatography and hormone self-assembly. A, elution profiles of the various analogs in the absence of zinc or phenolic ligand are shown on the left: human insulin (black), OrnB29 insulin (blue), insulin Lispro, or KP (green), and GlyB24Orn29 insulin (red). Molecular weight of peaks was determined using a linearized plot of log molecular weight versus Ve/Vo of different molecular mass standard proteins. Ve is the elution volume of each protein, and V0 is the column’s void volume. Calibration: thyroglobulin (669 kDa, Vo), bovine serum albumin (67 kDa), ovalbumin (45 kDa), ribonuclease A (13.7 kDa), cytochrome C (12.3 kDa), and a synthetic peptide (1.2 kDa). The line represents a linear fit, that was used to determine the molar mass of insulin elution peaks as summarized in the table. B, HPLC SEC in the presence of zinc and cyclohexanol. Proteins were fractionated as in A with inclusion of

50 mM cyclohexanol and 0.3 mM ZnCl2 using a superdex column; the column was recalibrated in this buffer with the following standards: thyroglobulin (669 kDa, V0) bovine serum albumin (67 kDa) ovalbumin (45 kDa), carbonic anhydrase (29 kDa), ribonuclease A (13.7 kDa), cytochrome C (12.3 kDa), and synthetic peptides (3.6 and 1.2 kDa).

89 significant. To assess the ability of variant proinsulin analogs to be packaged into secretory granules and exocytosed by cells, the amounts of radiolabeled proinsulin in cell lysates were compared to those in the cell media. GlyB24 insulin displayed a 25% secretion efficiency compared to wild-type insulin. This value was reduced compared to SerB24 insulin, which displayed a ~40% secretion efficiency in relation to wild-type proinsulin.

The potential for misfolding for each proinsulin variant was determined through transfection of proinsulin-expressing constructs in HEK 293T Cells containing a BiP- luciferase fusion protein. BiP is a chaperone protein associated with the degradation pathway of mis-folded proteins, and its expression was used to determine the level of misfolding of proinsulin. Whereas expression GlyB24 proinsulin initiated greater expression of BiP-luciferase compared to wild-type insulin, the degree of BiP expression was similar to that of SerB24 proinsulin (Figure 2-9).

The results of in vivo insulin folding and secretion suggest a folding and processing defect associated with GlyB24 insulin. However, folding defect is not so severe as to completely prevent the mutation from being expressed and secreted in pancreatic β-cells.

This is evidenced through the modest decrease in foldability and secretability of GlyB24 insulin compared to that of SerB24 proinsulin, a clinically-occurring mutation that causes hyperinsulinism (in which increased concentrations of SerB24 insulin are detected in the bloodstream) that may progress to diabetes. Furthermore, GlyB24 proinsulin displayed a misfolding potential virtually identical to that of SerB24, suggesting that a GlyB24 mutation occurring clinically was unlikely to cause neonatal diabetes or be an embryonic lethal phenotype.

Figure 2-9. In vivo assessment of proinsulin folding, misfolding, and secretion. The ability of mutant proinsulins to fold in cell lines and their efficiency of secretion were tested in HEK293T cells. A. comparison of natively folded proinsulin of mutant insulin analogs compared to wild-type proinsulin. Efficiency of proinsulin folding was assessed using non-reducing PAGE. Efficiency of folding was determined by comparing the intensity of bands corresponding to natively folded proinsulin. B. Secretability of proinsulin mutants was determined using a pulse-chase experiment with S35 radiotracer.

Comparison of insulin signal in lysed cells and in cell media was used to determine the efficiency of secretion. Data were normalized against the efficiency of wild type proinsulin. C. Insulin mis-folding was determined using the ER-stress related protein, which was placed in a luciferase-based reporter construct. The increase in translation of the BiP-Luciferase reporter over wild-type proinsulin was used to assess the degree of misfolding of mutant proinsulins.

2.4 Discussion

Insulin Structure:

Insulin is a primarily α-helical, globular protein that contains two polypeptide chains designated A and B. In its precursor form, proinsulin, the C-terminus of the B chain is connected to the N-terminus of the A chain by 30-35 residue “C domain.” (E. N. Baker, et al., 1988). Proinsulin undergoes oxidative folding in the ER of pancreatic β-cells

(Winter, Lilie, & Rudolph, 2002) in which three disulfide bridges between residues A7-

B7, A19-B20, and A6-A11 are formed (M. Liu, et al., 2005). The formation of these disulfide bridges serves as the critical kinetic barrier between folded and disordered conformations of proinsulin (M. Liu, Haataja, et al., 2010; M. Liu, et al., 2005).

Proinsulin is converted into mature insulin through the cleavage of the C domain by prohormone convertases (Steiner, Bell, & Rubenstein, 2006).

Although insulin readily refolds after denaturation, it remains susceptible to the formation of ordered aggregates via misfolding when stored at high concentrations. This has been observed in pharmacological formulations. For this reason, Insulin adopts a specialized storage conformation in secretory granules that allows the hormone to self- associate into dimers and zinc-coordinated hexamers (Attri, et al., 2010a, 2010b).

Hexamers then crystallize with calcium to form microcrystals within the secretory granules (G. Dodson & Steiner, 1998; Michael, Carroll, Swift, & Steiner, 1987). These storage conditions dampen the conformational structure of insulin that may nucleate cross-β polymer assembly (J. Dong, et al., 2003; Huang, Maiti, Phillips, Carey, & Weiss,

2006). An insulin ortholog that has lost self-association capability, that of octadon degus,

has been shown to form amyloid plaques in pancreatic β-cells that lead to the

development of DM (Hellman et al., 1990; Nishi & Steiner, 1990).

The broad conservation of the primary sequence of insulin represents 500 million years of selection (Steiner, Chan, Welsh, & Kwok, 1985). Vertebrate insulins have evolved within a system of signaling molecules that includes the insulin-like growth factor (IGF) and relaxin families (P. De Meyts, 2002; Schwabe, Gowan, & Reinig, 1982); the divergence of structures and molecular targets of each of the three major classes of hormone is thought to represent the increasing complexity of metabolic and growth pathways through the evolutionary process (Leibush et al., 1998).

Insulin is the primary hormone controlling carbohydrate metabolism. For this reason, large quantities of insulin must be readily available to meet the varying demand for glucose uptake (Hedeskov, 1980). The structure of modern vertebrate insulin reflects a variety of unique requirements within the insulin family: i) the hormone must bind to the insulin receptor with high affinity and specificity; ii) the insulin protein must fold efficiently to limit stress to the endoplasmic reticulum; iii) the molecule must be resistant to local unfolding or misfolding; iv) the molecule must be amenable to storage at high concentrations allowing a large reserve of active hormone to be sequestered in secretory granules within β-cells (J. M. Conlon, 2001).

Phylogenetic analysis of the insulin family of hormones has led to the hypothesis that insulin evolved from a relatively disordered peptide. A small number of conformations of such a peptide were likely capable of binding with high affinity to its molecular target. As signaling pathways associated with metabolim diverged from those associated with growth, co-evolution between the hormone and its receptor led to a further limitation in the

94 number of active conformations to those that were able to provide specific as well as high- affinity binding. This led to the development of distinct receptor-ligand systems of insulin and IGF. Although the divergence of signaling systems a winnowing of conformational flexibility of insulin, a non-unique number of conformations are permissible by the allostery of IR. This is evidenced, in part, by the high degree of cross-reactivity between insulin and IGF systems. For this reason, receptor affinity and specificity does not represent the primary driving force behind the divergence of insulin structure from that of less-ordered molecules in the insulin family.

Naturally occurring insulins reveal impact of foldability on insulin structure.

Clinical mutations that have been identified in the insulin gene underscore the evolutionary constraints of folding on insulin structure. A subset of these mutations involves residues at the hormone-receptor interface: PheB24Ser (insulin Los Angeles) (S.

E. Shoelson, Polonsky, Zeidler, Rubenstein, & Tager, 1984a) and PheB25Leu (insulin

Chicago) are clinical mutations of the aromatic triplet, whereas ValA3Leu (insulin

Wakayama) involves the IR-binding region of the N-terminal A-chain of insulin (Nanjo, et al., 1986). This class of mutations leads to familial hyperinsulinism that progresses to the development of DM during adulthood as a result of ER stress associated with producing increased amounts of insulin with deficient folding efficiency. The clinical syndrome associated with such mutations is known as mutant INS diabetes of the young (MIDY).

Another subset of mutations impacts proinsulin foldability more severely. This subset, which includes mutations such as LeuB11Pro and GlyB23 Val, results in neonatal diabetes as a result of dysfunction in the endoplasmic reticula of pancreatic β cells

(Edghill, et al., 2008).

Insulin orthologs of the hystricomorpha suborder of rodents demonstrate the physiological consequences of the proteotoxicty of mature insulin (Horuk et al., 1979).

Hystricomorphs have acquired a variety of mutations to the INS gene that produce monomeric, although foldable, insulin molecules. Many hystricomorphs, such as guinea pigs, have compensated for the inability of their insulin analogs to oligomerize by storing smaller quantities of insulin. However, others, such as the South American mouse octadon degus, has sown the propensity to develop adult-onset diabetes as the result of the formation of amyloid aggregates forming in its pancreatic β-cells (Edwards, 2009;

Hellman, et al., 1990; Jekl, Hauptman, & Knotek, 2011).

Identification of mechanism of GlyB24 insulin activity and validation of microreceptor co-crystal structure.

Given the strict functional requirement of small aromatic or branched-chain amino

acid residues at the B24 position, it would be expected that that insulin analogs with

glycine substitutions at the B24 position would be biologically inactive with low affinity

for IR. However, studies have reported native biological activity and IR affinity of GlyB24

insulin analogs (J.G. Menting, et al., 2014; Mirmira, et al., 1991; V. Pandyarajan et al.,

2014). Although such reports have been corroborated by a variety of studies, the current

structural models of insulin-IR binding have been unable to account for the activity of

glycine at the B24 position.

The recent co-crystal structure of insulin bound to µIR, although resolved to only 3.5 Å, provides a number of key insights into insulin-IR interaction. First, it reveals a conformational switch that is required for insulin-IR interaction: the C-terminal B-chain separates from the hydrophobic core of insulin via the unfolding of a β-turn formed

96 between residues B20 and B23 exposing amino acid residues that interact with IR in the N- terminal A-chain and in the C-terminal B-chain. Second, the detached C-terminal B-chain was revealed to intercalate between αCT and L1: residues B24, B25, and B26 intercalate with specific binding surfaces. Finally, although the IR binding surfaces of PheB25 and

TyrB26 were poorly visualized, a well-defined, rigid hydrophobic pocket in L1 was visualized as the binding site of the B24 side-chain (J.G. Menting, et al., 2014).

This model was validated by mutational screening analysis of the B24 and B26 positions. Such screening indicated that the B24 binding pocket in L1 accommodates branched-chain aliphatic amino acid or methionine substitutions at position B24 in addition to the native phenylalanine (Vijay Pandyarajan, et al., 2014a). The B26 binding surface on L1 can accommodate a variety of small, polar, or charged residues at the B26 position (Pandyarajan, et al., 2016). The B26 is dispensable for receptor-binding in the context of truncated analogs [des-B26-B30, B25-amide]-insulin (Mirmira, et al., 1991). A variety of mutational analyses of the B25 position have strongly suggested a requirement for an aromatic residue at B25 for insulin-IR binding (Mirmira & Tager, 1991; Nakagawa

& Tager, 1986).

The native-like biological activity and affinity for IR of GlyB24 was reported in several previous studies; these findings appeared to contradict the established role of the B24 side chain in insulin-IR binding (Mirmira, et al., 1991; Vijay Pandyarajan, et al., 2014a). The

“register shift” hypothesis, in which PheB25 occupies the B24 binding pocket in the contest of GlyB24 insulin with TyrB26 and ThrB27 occupying the binding surfaces of B25 and B26, respectively, is a model that predicts native-like biological activity while maintaining the occupancy requirement for the binding surfaces of residues B24-B26.

The register shift model was validated by observing the biological activity and IR affinity of binding surface-specific mutations at the B25 position in the context of GlyB24 insulin analogs. The ability of the GlyB24 substitution to restore the biological activity of non-functional B25 mutations that are active at the B24 position, such as Leu or Cha, while attenuating the activity of functional B25 substitutions that are not functional at the

B24 position, such as Tyr, provides concrete evidence of the register-shift phenomenon taking place in the context of receptor-binding.

The results also address an alternative hypothesis describing the native-like activity of

GlyB24 insulin analogs. This hypothesis suggests that the introduction of a PheB24  Gly substitution results in an increase in conformational flexibility and entropy of the C- terminal B-chain allowing the insulin molecule to adopt a more thermodynamically favorable conformation in the insulin-IR complex causing the bound state of insulin to be favored over its free state. The attenuation of the biological activity and IR affinity of

TyrB25 insulin by the introduction PheB24  Gly substitution suggests that any increase in entropy or conformational flexibility of the C-terminal- B-chain of insulin is not sufficient to compensate for the loss of critical B24 side chain or the presence of an unfavorable side-chain at the B25 position; that is, GlyB24 insulin analogs are only functional if the B25 position contains an amino acid that may engage the B24 binding pocket of IR.

Whereas the register shift model of insulin-IR binding was previously described in the context of D-amino acids at the B24 position (Žáková, et al., 2013), demonstrating a shift of residue B26 onto the B25 binding surface, this study reveals that the achiral glycine also initiates a register shift in the context of insulin-IR binding. For this reason, the occurrence of the register shift in the context of glycine allows the phenomenon to be assessed in the

98 context of insulin evolution. The biophysical and biological consequences of GlyB24 may be used to further the previously described hypothesis that insulin structure represents evolutionary pressure optimizing insulin-IR engagement and the structural stability of insulin.

GlyB24 Impacts the stability of the native fold of insulin.

The C-terminal region of the B-chain plays a critical role in the maintenance of the native structure of insulin, prevention of amyloid fibril formation, and oligomerization.

The sidechain of PheB24 packs against the hydrophobic core of the insulin that is formed by residues including LeuB11, ValB12, and LeuB15, stabilizing the “closed” storage form of the insulin molecule. PheB24 further stabilizes the closed conformation through an aromatic interaction with TyrB26. The packed C-terminal B-chain comprises a bulk of the monomer- monomer interface of the insulin dimer. Anchor residues PheB24 and TyrB26 hydrogen bonds with the TyrB26 and PheB24 of the opposite monomer, respectively, forming an antiparallel β-strand.

GlyB24 insulin was expected to have decreased thermodynamic and physical stability as a consequence of two properties of the ectopic glycine residue. First, the lack of an aromatic side chain at the B24 position reduces the thermodynamic favorability the closed conformation through the loss of hydrophobic interactions between residue B24 and the hydrophobic residues of the central B-chain helix and through the loss of the potential aromatic interaction with residue B26 (S. Burley & Petsko, 1986a; S. K. Burley & Petsko,

1985; El Hage et al., 2016; V. Pandyarajan, et al., 2014). Second, the introduction of a conformationally flexible glycine residue at B24 perturbs the B20-B23 β-turn, which is well-ordered in the closed monomeric and oligomeric states of the protein. The native

99 structure incurs an entropic penalty as a result of such a perturbation as a result of increased exposure of the hydrophobic core of the protein (Brange, et al., 1997; Xu et al.,

2002).

Biophysical analysis of the GlyB24 analogs in relation to their PheB24, LeuB24, or ChaB24 counterparts was able to confirm the former hypothesis. GlyB24 insulin displayed a reduced thermodynamic stability compared to wild type, OrnB29, and ChaB24 insulin analogs.

However, the thermodynamic stability of the analog was similar to that of LeuB24 insulin, suggesting that the torsional flexibility of GlyB24 does not significantly contribute to the destabilization the global secondary structure of GlyB24 insulin. GlyB24 analogs displayed markedly reduced fibrillation lag times compared to their Phe, Cha, or Leu counterparts, suggesting the introduction of glycine at the B24 position may increase the propensity of analogs to locally unfold, aggregate, and misfold beyond that of other B24-substituted analogs. This finding may indicate that although the native or “ground” structure of GlyB24 insulin may not be significantly perturbed compared to other B24-substituted analogs, the ectopic glycine may cause an increase in conformational excursions of the C-terminal B- chain that may nucleate fibril formation.

In addition to its ability to destabilize the secondary structure of insulin, GlyB24 was found to destabilize insulin oligomers. EDTA-Cobalt experiments indicated a severe

B24 B29 reduction in half-life of phenol stabilized, zinc-coordinated (R6) hexamers of Gly Orn insulin compared to that of native OrnB29 insulin, wild-type insulin (HI), and insulin lispro

(KP). Furthermore, GlyB24 insulin was found to have diminished capacity to hexamerize and dimerize by size-exclusion chromatography compared to control analogs.

100

Because the C-terminal B-chain of insulin comprises a majority of the monomer- monomer interface of the insulin dimer, these results strengthen the hypothesis that the C- terminal B-chain separates more frequently from the core of the insulin molecule more frequently in the context of GlyB24 analogs than in that of native PheB24 analogs. Whereas previous studies have noted the destabilization of insulin hexamers as a result of substitutions at the B24 position (Vijay Pandyarajan, et al., 2014a), the dramatic decrease in hexamer stability may also be attributed to the increased dynamicity of the C-terminal

B-chain that results from the chiral flexibility of the ectopic glycine residue. As noted by

Ludvigsen, et al, the change in positioning of the C-terminal B-chain of GlyB24 insulin leads to the inability to form hydrogen bonds between B24 and B26 on the monomer- monomer interface.

GlyB24 decreases efficiency of insulin folding and secretion.

In vivo experiments suggest impaired oxidative folding of GlyB24 insulin analog that, in turn, leads to a decrease in secretion of proteolytically processed insulin. Such findings corroborate previous studies that suggest that the integrity of the C-terminal B-chain is critical for efficient oxidative folding of insulin (Steiner, Park, Stay, Philipson, & Bell,

2009). Although secretion of GlyB24 insulin is significantly decreased compared to wild- type insulin, the efficiency of secretion is comparable to the clinical mutant analog SerB24 insulin (Insulin Los Angeles), which has been detected at high levels in the blood-stream of patients bearing this mutation. This suggests that the inability of GlyB24 insulin to form stable dimers or zinc-coordinated hexamers does not entirely preclude the molecule from being exocytosed through secretory granules. The results of pulse-chase experiments suggest that self-association is not necessary for the proper packaging of insulin into

101 secretory granules, a property previously observed in the setting of the clinical mutant

Insulin Providence (HisB10  Asp) (Gruppuso et al., 1984).

Physiological and evolutionary consequences of GlyB24.

The structural and biological properties of GlyB24 insulin underscore the evolutionary optimization of the insulin molecule. Although GlyB24 insulin is fully potent and displays native affinity for IR, the detabilization of the tertiary structure of the molecule and its reduced tendency to oligomerize prevent the mutation from becoming an observable phenotype in nature. As previously noted, a variety of proteins have developed adaptations that prevent the formation of poly-β assemblies that lead to proteotoxic amyloid formation.

Insulin has been shown to be stored at concentrations ranging from 70-120 mM in secretory granules within pancreatic β-cells (Rorsman & Renström, 2003); for this reason, even slight perturbations in structure that may contribute to the formation of non-specific aggregates are strongly disfavored (Bucciantini et al., 2002). Hexamer formation, which further stabilizes the native conformation of the insulin molecule, also contributes to the ability to store insulin at millimolar concentrations.

Mutations affecting the B20-B23 β-turn of insulin have been shown to affect the ability of the insulin precursor, proinsulin, to fold into its mature structure. A clinical mutation at the B23 position, Gly  Val, causes neonatal diabetes by initiating unfolded protein responses in the endoplasmic reticula of pancreatic β cells; the prolonged activation of this pathway leads to cell death (Scheuner & Kaufman, 2008). The decreased secondary structural stability of GlyB24 suggested an overall decrease in folding efficiency of the mutant.

102

GlyB24 proinsulin showed comparable levels of misfolding to SerB24 proinsulin as determined by BiP expression in transiently transfected cell lines. This result suggests that although GlyB24 may lead to impaired proinsulin folding, the defect is more likely to manifest itself phenotypically as maturity onset diabetes (similarly to SerB24) rather than neonatal diabetes (as is the case for ValB23). The impaired foldability of GlyB24 proinsulin is likely among the reasons the variant has been disfavored evolutionarily.

The in vivo assessment of GlyB24 proinsulin foldability further validates the structural characterization of the analog. The comparable levels of misfolding of GlyB24 and SerB24 proinsulin suggest a native-like structural “ground state” of both mutants. A complete destabilization or elimination of the C-terminal B-chain and the B20-B23 β-turn, such as that expected in ValB23 proinsulin, would be expected to lead to severe protein misfolding with significantly greater BiP expression than that activated by either GlyB24 or SerB24 analogs.

Implications on the active conformation of insulin.

The structure of the GlyB24 analog reveals that the B20-B23 β-turn is not the exclusive secondary structural motif that is able to facilitate insulin-IR binding. Previous studies have indicated that replacing GlyB20 with an alanine, altering the conformation of the turn, leads to increased biological activity accompanied by decreased thermodynamic stability.

Such substitutions are observed in nature among the monomeric insulins expressed by the mammalian suborder hystricomorpha. Observations regarding GlyB24 suggest that the torsional flexibility of glycine residues, which can attain positive φ angles, is required to properly direct residues B24 and B25 onto their respective binding surfaces on IR. The

103 introduction of a glycine at B24 allows residues B25-B27 to occupy the B24-B26 binding surfaces by attaining a similar right-handed conformation.

The perturbed structure and decreased stability of hystricomorph insulins combined with our findings regarding GlyB24 insulin analogs indicates that the B20-B23 β-turn represents an optimum structural motif for maintaining biological activity through the presence of a glycine residue at B23, while maintaining the stability of the native insulin structure by protecting the hydrophobic core of the protein and providing an interface for oligomerization. The β-turn represents a structural motif that is necessary for aromatic residues B24 and B26 occupying their native positions in the closed conformation of insulin. The high physiological demand for insulin among larger mammals has led to the strict conservation of this turn, which is a key structural feature in allowing insulin to be stored at millimolar concentrations in secretory granules.

Summary of Findings.

Our studies of a glycine substitution in insulin at position B24 revealed an alternative binding mode of the analog in which residues B25-B27 occupy the binding surfaces on IR that are occupied by residues B24-B27 in the context of native insulin-IR binding. This

“register shift” phenomenon confers native-like biological activity and receptor affinity to

GlyB24 insulin analogs. GlyB24 analogs exhibit perturbed structure in which the conserved

B20-B23 β-turn is disordered and the C-terminal B-chain has increased dynamaticity. This perturbed structure causes an overall decrease in thermodynamic stability of the analog, reduced oligomerization capability, and increased tendency to aggregate into amyloid species.

104

Our findings imply that the B20-B23 β-turn is sufficient, but not necessary for insulin-

IR binding, but represents the optimum secondary structure for the stabilization of the native insulin structure. The induced-fit model for insulin-IR engagement requires an amino acid residue that can acquire a positive φ-angle just upstream of the aromatic triplet of residues for successful insulin-IR engagement indicating some flexibility in possible active conformations of insulin at the C-terminal region of its B-chain.

2.5 Methods and materials

Preparation of Insulin Analogs

Analogs were prepared by trypsin-catalyzed semi-synthesis using an insulin fragment,

des-octapeptide (B23-B30)-insulin, and modified octapeptides as described (Inouye,

Watanabe, Tochino, Kobayashi, & Shigeta, 1981). The fragment was made by tryptic

cleavage of human insulin and purified by reverse phase HPLC. Modified octapeptides

were prepared by solid phase synthetic methods (Barany & Merrifield, 1980). The

resulting 51-residue insulin analogs were purified by preparative reverse phase C4 HPLC

(Higgins Analytical Inc., Proto 300 C4 ,10 uM, 250 x 20 mm), and their purity was

assessed by analytical reverse phase C4 HPLC (Higgins Analytical Inc., Proto 300 C4 5

uM, 250 x4.6 mm). Predicted molecular masses were in each case verified using an

Applied Biosystems 4700 proteomics analyzer MALDI-TOF.

Circular Dichroism Spectroscopy

Far-ultraviolet (UV) spectra were obtained from 200 to 250 nm on an AVIV

spectropolarimeter as described (Vijay Pandyarajan, et al., 2014a) equipped with an

automated syringe-driven titration unit (Sreerama & Woody, 1993). Helix sensitive

wavelength 222 nm was used as a probe of protein denaturation by guanidine

105 hydrochloride using the AVIV-equipped automated syringe-driven titration unit.

Thermodynamic parameters were obtained by application of a two-state model as described. In brief, data were fit by nonlinear least squares to a two-state-model as shown in Equation 1

-DG0 -mx /RT ( H2O ) q +q e q(x) = A B -DG0 -mx /RT ( H2O ) 1+ e

(Eq. 1) where x is the concentration of guanidine hydrochloride, and A and B represent respective estimates of the baseline ellipticities of the protein in its unfolded and native states as extrapolated to a guanidine concentration of 0 (Sosnick, Fang, &

Shelton, 2000). Baseline values were approximated via pre- and post-transition lines

H2O H2O represented by equations A(x)A mAx and B(x)B mBx. Such simultaneous fitting avoids artifacts of linear plots of G versus concentration of denaturant (Pace & Shaw,

2000).

Receptor Binding Assays

Affinities for IR-A were measured by a competitive-displacement scintillation proximity assay as described (Pandyarajan, et al., 2016). This assay employed detergent solubilized holo-receptor with C-terminal streptavidin binding protein tags purified by sequential wheat-germ agglutinin (WGA) and Streptactin-affinity chromatography from detergent lysates of polyclonal stably transfected 293PEAK cell lines expressing each receptor. To obtain analog dissociation constants, competitive binding data were analyzed by non-linear regression by the method of Wang (Z. X. Wang, 1995), a model that provides an analytical solution for the binding of two ligands to a single receptor.

Rodent Assays

106

Male Lewis rats (mean body mass 300 g) were rendered diabetic by streptozotocin. To test the in vivo potency of insulin analogs relative to [OrnB29]-insulin, the analogs were made 10 µg per 100 µl in a formulation buffer (16 mg/ml glycerin, 1.6 mg/ml meta- cresol, 0.65 mg/ml phenol, and 3.8 mg/ml sodium phosphate (pH 7.4)) and injected intravenously into tail veins as described (Vijay Pandyarajan, et al., 2014a). WT insulin or analogs were each re-purified by HPLC, dried to powder, dissolved in Lilly diluent at the same maximum protein concentration, and re-quantitated by analytical C4 reverse phase HPLC to ensure uniformity of formulations; dilutions were made using the above buffer.

Rats were injected at time t0. Blood was obtained from the clipped tip of the tail at time t0 and every 10 min for the 1st h, every 20 min for the 2nd h, every 30 min for the

3rd h, and every hour thereafter to a final time of 5 h. The efficacy of WT insulin or analog to reduce the blood glucose concentration was calculated using the following: (a) the rate of change in blood glucose concentration over 240 minutes following initial injection, and (b) the integrated area between the X-axis and the curve representing fractional blood-glucose level with relation to initial blood-glucose concentration (area under the curve; AUC). Statistical significance was assessed using a Student’s t test.

Assessment of Fibril Formation

The physical stability of the analogs was assessed by their propensity to form fibrils.

To estimate lag times prior to onset of fibril formation, zinc-free WT insulin or analogs were made 60M in phosphate-buffered saline (pH 7.4) containing 0.01% sodium azide as an antimicrobial agent. The insulin solutions in triplicate were gently rocked at 37 °C in glass vials containing a liquid-air interface. Aliquots, taken at regular intervals, were

107 frozen to enable later analysis of thioflavin T fluorescence (Y. Yang, Petkova, et al.,

2010). For a given sample tube, the assay was terminated on the 2nd day of the appearance of cloudiness in the solution.

Assessment of Hexamer Stability

Visual absorption spectroscopy was used to probe the formation and disassembly of phenol-stabilized R6 Co-substituted insulin hexamers as first described by (Roy, et al.,

1989). WT insulin or analogs were made 0.6 mM in a buffer containing 50 mM Tris-HCl

(pH 7.4), 50 mM phenol, 0.2 mM CoCl2, and 1 mM NaSCN. Samples were incubated overnight at room temperature prior to the studies to ensure attaintment of conformational equilibrium. Spectra (450–700 nm) were obtained to monitor tetrahedral Co2+ coordination with its signature peak absorption band at 574 nm (D. T. Birnbaum, et al.,

1997). To determine the rate of Co2+ release from the hexamers, metal ion sequestration was initiated at 25 °C by addition of an aliquot of EDTA (50 mM at pH 7.4) to a final concentration of 2 mM. Attenuation of the 574-nm absorption band was monitored on a time scale of seconds to hours. Kinetic data were consistent with monoexponential decay.

Assessment of Insulin Self-assembly by Size Exclusion Chromatography

Oligomeric states of the insulin analogs were monitored by HPLC- size-exclusion chromatography (SEC) as described (Pandyarajan, et al., 2016). Insulin analogs were made 0.6 mM in a buffer consisting of 25 mM Tris-HCl (pH 7.4), 0.65 mg/ml phenol, 1.6 mg/ml meta-cresol, 16 mg/ml glycerol, and ZnCl2 at a ratio of 3 zinc ions per insulin hexamer. For zinc-free conditions, analogs were also made 0.6 mM in phosphate- buffered saline (pH 7.4). Protein samples (volume 10 µl) were applied through a Waters

717 autosampler onto a Zenix-C SEC-150 column (Sepax Technologies, Southborough,

108

MA) with a nominal fractionation range of 0.5–150 kDa. Proteins were fractionated at a flow rate of 1 ml/min using a Waters Binary HPLC system. Protein elution was monitored at 215 and 280 nm using a dual-lambda Waters 2487 absorbance detector. The mobile phase consisted of 10 mM Tris-HCl (pH 7.4) and 140 mM NaCl with or without

0.3 mM ZnCl2 and 50 mM cyclohexanol; the latter provided a non-aromatic replacement for the phenolic compounds often employed as R6-hexamer stabilizing- and antimicrobial agents in pharmaceutical formulations (Abedini & Raleigh; Brange, 2000;

Brange, Hansen, Havelund, & Melberg, 1987; U. Derewenda, et al., 1989). Data acquisition and processing utilized Waters HPLC Empower software. The column was calibrated for apparent molecular mass determination by fractionating standard proteins individually on the column as detailed in the figure legend.

In vivo insulin folding and secretion

HEK-293T cells were plated into 6 or 12-well plates 1 day before transfection. A total of 1–2 mg plasmid DNA was transfected using Lipofectamine (Invitrogen®). Cells were pulse-labeled with 35S-labeled amino acids 48 h after transfection and chased for the times indicated. A proteinase inhibitor mixture was added to cell lysates and chase media.

The samples were precleared with Zysorbin and immunoprecipitated with anti-insulin antibodies. Anti-insulin immunoprecipitates were boiled for 5 min in gel sample buffer

[1% SDS, 12% glycerol, and 0.0025% Serva Blue in 50 mM Tris (pH 6.8) with or without 100 mM DTT] and analyzed using tris-tricine-urea-SDS-PAGE under nonreducing or reducing conditions (M. Liu, Ramos-Castaneda, & Arvan, 2003).

BiP-Driven Luciferase Assay

109

Min6 cells were plated into 24-well plates 1 d before transfection. Using

Lipofectamine 2000 (Invitrogen), cells were co-transfected with pBiP-firefly-luciferase reporter plasmid (Tirasophon, Welihinda, & Kaufman, 1998), CMV-renilla-luciferase plasmid (Promega), and human wild type or mutant proinsulin at a DNA ratio of 1:2:5, respectively. At 48 h post-transfection, cell extracts were prepared for the dual-luciferase reporter assay (Promega) with BiP-luciferase normalized to Renilla luciferase activity

110

Chapter 3: Identification of a Minimal Prodomain in Insulin: A Weak Hydrogen

Bond is Critical for Proinsulin Folding, but Dispensible in the Mature Hormone

3.1 Chapter Summary

The dangers of proteotoxic aggregation and misfolding of proteins within cells necessitate efficient and specific folding. The sequences of proteins have evolved features that facilitate the folding process. A subset of such adaptations are structural features that are critical to the folding process of a nascent peptide, but are dispensable in the structure and function of the mature protein. Such features, part of the broader class of

“prodomains,” may vary in size and character constituting a full protein domain or a single amino-acid residue. In this chapter, we have identified a minimal prodomain that constitutes a single hydrogen bond between the two polypeptide chains of the metabolic hormone insulin, a long studied model of protein folding and evolution. Elimination of this hydrogen bond (formed between the amide proton of the C-terminal residue of the insulin A chain (AsnA21) and the the 23rd residue of the B chain (GlyB23)) by elimination of the AsnA21 prevents the oxidative folding of the precursor of insulin, proinsulin, in vitro and in transiently transfected mammalian cell lines. In spite of these deficits in folding, mature desA21 insulin analogs retain native-like structure, as determined by

NMR spectroscopy, as well as thermodynamic and physical stability. Although biological activity of desA21 insulin is impaired, this is thought to be the result of loss of contacts between the A21 residue and the insulin receptor rather than a consequence of the loss of the A21-B23 hydrogen bonds.

111

3.2 Introduction

The sequence of a globular protein contains critical information that determines its structure and function (C. M. 14685248 Dobson, 2003). Those components of protein sequence that are highly conserved are often determined to play a critical role in the function of the protein (Petsko, 2000). However, the conservation of other components of protein sequence has been shown to be the result of their role in the folding efficiency of the nascent polypeptide into the mature, natively folded protein (C. M. Dobson, 1999).

Such components, often referred to as “prodomains,” may be dispensable for the maintenance of the native structure of the mature protein and in its function (Bryan,

2002). In this study, we describe a prodomain in the peptide hormone insulin that comprises a single hydrogen bond (Chu, et al., 1987a). To our knowledge, this is the smallest prodomain that has been identified to date (Goldenberg & Creighton, 1994;

Jonathan S Weissman & Peter S Kim, 1992).

The modern view of protein folding, represents the process as a protein traversing a three-dimensional “funnel-like” landscape as it folds from a disordered peptide to its native conformation (C. M. Dobson & Karplus, 1999). The process is driven by a series of regional interactions in the nascent polypeptide that nucleate the folding process, which can propagate from the local to global level (J. Wang, Onuchic, & Wolynes, 1996;

Wolynes, 2005). On the whole, the protein rapidly approaches the absolute thermodynamic minimum, the native structure represented by the bottom of the funnel

(Zhuravlev, Materese, & Papoian, 2009). However, a series of local maxima and minima in the folding landscape, representing metastable folding intermediates, can impede the

112 progression of the folding process to the native state. These minima are commonly known as “kinetic traps (J. S. Weissman & P. S. 0001384063 Kim, 1992).”

Proteins have evolved structural motifs to prevent the formation of kinetic traps and to erect kinetic barriers that prevent the unfolding and misfolding of the proteins once they are folded (Weissman & Kim, 1991). Whereas mutation or elimination of motifs in the sequence of a protein leads to the destabilization of the native-state (M. Yu & King,

1988), the former class of motifs, known as prodomains, are entirely dispensible in the structure and function of the mature protein (Mitraki, Danner, King, & Seckler, 1993).

The peptide hormone insulin serves as a robust model of protein folding and misfolding (Q. Hua, 2010). Insulin is translated as prohoromone, proinsulin, that comprises an A domain, a B domain, and an intervening C domain that serves as a prodomain facilitation the formation of three disulfide bridges (between A7 and B7, A20 and B19, and A6 and A11) in the native molecule (Steiner, et al., 2006). A number of mutations to the proinsulin sequence that cause the misfolding of proinsulin leading to

ER stress in β cells. These mutations manifest clinically as the cause of mutant INS diabetes of the young, or in severe cases, permanent neonatal diabetes (Weiss, 2013b).

However, it is apparent that even the folding of native insulin may cause lethal ER stress in most human cell lines, and initiates some degree of ER stress in β-cells (Flamment,

Hajduch, Ferré, & Foufelle, 2012).

The precarious nature of the insulin doling has led to the conservation of several structural features (Weiss, 2012). The formation of the three native disulfide bridges represents the most significant kinetic barrier in the proinsulin folding process: unless it has degraded, mature insulin may readily re-fold into its native conformation after

113 denaturation as long as the disulfide bridges remain intact (Qiao, Min, Hua, Weiss, &

Feng, 2003; Weiss et al., 2000). For this reason, the oxidative folding of insulin in the ER remains the most significant event in the folding process that is catalyzed by such enzymes as protein disulfide isomerase and oxidoreductases (Khoo et al., 2011; Rajpal, et al., 2012). The C-terminal B-chain of insulin, comprising a β-turn spanning residues B20-

B23 and a triplet of aromatic residues spanning B24-B26, provides a kinetic barrier between the native fold of insulin and folding intermediates that may nucleate the aggregation of insulin into amyloid fibrils (Brange, et al., 1997). Moreover, this region of the molecule has been shown to play a role in the proper oxidative folding of proinsulin: several clinical mutations associated with MIDY and permanent neonatal diabetes have been identified in this region (Edghill, et al., 2008; M. Liu, Hodish, et al., 2010; Stoy, et al., 2007).

In this study, we propose a single weak hydrogen bond serves as a critical “lynchpin” in the oxidative folding process of the prohormone. This hydrogen bond is formed between the amide proton of AsnA21, the C-terminal residue of the A chain, and the carbonyl oxygen of GlyB23, the final residue in the B20-B23 β-turn (E. N. Baker, et al.,

1988). Our hypotheses regarding the importance of the A21-B23 hydrogen bond is based on the work of Katsoyannis and colleagues: this group was unable to synthesize a desA21 insulin analog by chain combination, but found that natively folded insulin could be produced by the amidation of the carboxyl group of CysA20, which restores the NH group that participates in the A21-B23 bond (Chu, et al., 1987a; Chu, Wang, Burke, Chanley, &

Katsoyannis, 1987b; J. Markussen, et al., 1988).

114

Figure 3-1. AsnA21-GlyB23 H-Bond within the Structure of Insulin (A) The primary sequence of insulin. The three canonical disulfide bridges are shown as black lines.

Residue AsnA21 is indicated with a red arrow and residue GlyB23 is indicated with a blue arrow. (B) The three dimensional structure of insulin. The B chain is shown as peach ribbon and the A chain is shown as periwinkle ribbon. Disulfides are represented as yellow sticks. The region of the molecule containing residues AsnA21 and GlyB23 is enclosed by the black box. The positions of AsnA21 and GlyB23 are respectively indicated by a red arrow and a blue arrow. The inset shows residues AsnA21 and GlyB23 in stick representation. The hydrogen bond between the amide-proton of the A21 residue and the carbonyl carbon of the B23 residue is indicated by a red dashed line. (C) Heteronuclear

B10 NMR spectrum of a monomeric insulin analog (Asp lispro, DKP) in H2O after 2s presaturation at a frequency downfield of protein resonances (red contours) overlaid with a spectrum of the same analog in presaturation at solvent frequency (blue contours). The peaks corresponding to the A21 amide proton are highlighted by the green box. The presence of red and blue contours indicates protection from the solvent by a hydrogen bond. Figure adapted from (Y. Yang, Petkova, et al., 2010) with permission of the authors.

115

116

Although residue A21 is not completely dispensable in native insulin, given the role of

Cα and carboxyl groups of residue A21 in insulin-IR complexation, we expect its removal to have minimal impact on the globular structure of insulin in its native state

(Chu, et al., 1987b). This hypothesis is based on the peripheral location and lack of intramolecular contacts apart from the hydrogen bond with B23 (E. N. Baker, et al.,

1988; Y. Yang, Petkova, et al., 2010). As such, the A21-B23 hydrogen bond may be identified as a “minimalist prodomain,” comprising only two atoms and a single non- covalent bond.

Results

In vitro proinsulin folding assays:

Wild-type proinsulin, des-A21 proinsulin, and A21-amide proinsulin were created as disordered peptides using native-chemical ligation (Dawson, Muir, Clark-Lewis, & Kent,

1994). Peptides were incubated in redox buffer in the presence of PDI for three hours and folding of each peptide was assessed by analytical rp-HPLC. Wild-type proinsulin displayed a reduction in retention time and eluted primarily as a single peak after three hours of incubation with PDI. Such a finding indicated the burial of hydrophobic surfaces during protein folding. The resulting peak was associated with folded proinsulin in its native state. A similar pattern was observed when t0 and t3 samples of A21-amide proinsulin were compared. Whereas the naïve des-A21 proinsulin peptide eluted as a single sharp peak, no single species could be identified after incubation suggesting des-

A21 proinsulin was unable to achieve its native conformation (Figure 3-2).

In vivo assessment of proinsulin folding and secretion:

117

The role of residue A21 in proinsulin processing in cells was assessed by comparing the folding efficiency and secretion of wild-type proinsulin, des-A21 proinsulin, and

GlyA21 proinsulin (a biological equivalent of A21-amide proinsulin). Analogs were expressed in HEK293T cells. Cells were pulsed with S35-labeled cysteine. Reducing gels run after chase and immunoprecipitation with anit-insulin antibodies were able to confirm the expression of all three insulin analogs. Whereas bands corresponding to wild-type proinsulin and GlyA21 proinsulin were visualized in a non-reducing gel, a band corresponding to natively-folded des-A21 proinsulin was absent. This indicated that the proinsulin variant had likely aggregated or misfolded into several alternative species

(Figure 3-3A). Survey of cell lysates and media 120 minutes after chasing with cold cysteine was used to determine the secretability of the proinsulin constructs. Whereas all three proinsulin constructs were detectable within cells before the chase, wild-type proinsunlin and GlyA21 proinsulin were detected primarily in the media after 120 minutes of chase. des-A21 proinsulin, on the other hand, was found predominantly within the cell, suggesting that the analog is unable to be secreted (Figure 3-3B). The localization of proinsulin constructs within pancreatic β cells was determined by expressing chimeric constructs of GFP and each proinsulin in rat β cell lines. Whereas wild-type and GlyA21 proinsulins were localized to puncta near the periphery of the cell, a localization corresponding to secretory granules, des-A21 proinsulin was appeared to be distributed diffusely within the cytoplasm. Des-A21 proinsulin was found to co-localize with protein disulfide isomerase, an enzyme expressed in the ER, through immunohistochemistry.

This co-localization was not observed in cells expressing wild-type and GlyA21 proinsulin. This result suggests that des-A21 proinsulin is sequestered in the ER, and is

118

Figure 3-2. Impact of foldability of A21-B23 hydrogen bond on in vitro proinsulin folding. Folding of proinsulin analogs created using native chemical ligation techniques was monitored by reverse-phase HPLC. HPLC traces corresponding to unfolded proinsulin are shown in the top half of panels A,B, and B. HPLC profiles of proinsulin after three hours of folding in redox buffer in the presence of protein disulfide isomerase is shown in the bottom half of panels A,B, and C.

119

120

unable to be packaged into secretory granules. This is likely the result of its inability to fold in the ER.

CD Studies:

The thermodynamic stability of des-[A21]-KP-amide, [GlyA21]-KP-amide, and KP- amide were assessed by circular dichroism (CD)-monitored guanidine titrations. The concentration-normalized CD-spectra of all three analogs closely matched that of native-

K- insulin indicating little change in the global secondary structure of the insulin hormone occurred as a result of (i) the introduction of an amide in place of the C-terminal carboxyl group of the insulin B-chain, (ii) the substitution of AsnA21 for glycine, or (iii) removal of the A21 residue. Guanidine denaturation studies indicated insignificant changes in thermodynamic stability resulting from the introduction of the C-terminal amide compared to the native KP (ΔGu of 2.8 ±0.1 kcal/mol and 2.9 ±0.1 kcal/mol, respectively). In contrast substitution of Asn21 for glycine or removal of the A21 residue in its entirety in the context of KP-amide insulin resulted in a substantial decrease in thermodynamic stability (ΔGu of 2.5 ±0.1 kcal/mol and 2.5 ±0.1 kcal/mol, respectively)

(Figure 3-4). Although these data suggest the favorability of the native asparagine at the

A21 position, the similar thermodynamic stabilities of GlyA21 and des-A21 analogs indicate that the increased stability of native KP-amide is a result of the effects of the side chain of the residue. That is, the A21-B23 hydrogen bond does not contribute to the thermodynamic stability of the mature insulin analog. For this reason, des-A21-KP- amide, which lacks the hydrogen bond, and GlyA21-KP-amide, in which the hydrogen bond remains intact, have identical thermodynamic stabilities.

121

Figure 3-3. Proinsulin folding experiments conducted in cells. (A) Gel migration patterns of proinsulin analogs in non-reducing (left) and reducing (right) conditions indicate incorrect disulfide pairing of disulfides in des-A21 proinsulin. (B) Pulse-chase assay conducted on cells expressing variant proinsulins using 35S cysteine. Lanes marked

“C” indicate samples collected from cell lysates, and lanes marked “M” denote samples collected from the media. Samples from cells and media were tested at 0 minutes post- chase and 120 minutes post-chase. (C) Whole cell imaging of primary rat beta cells expressing chimeric proinsulin-GFP constructs. Immunohistochemistry was employed to stain protein disulfide isomerase, an indicator of ER localization.

122

123

The consequences of the removal of the A21 residue on the physical stability of insulin was assessed by determining the fibrillation lag time of des-A21-KP-amide in relation to control analogs. Whereas wild-type human insulin displayed a fibrillation lag time

(confirmed by THT fluorescence) of 3.7±0.7 days, native KP-insulin displayed diminished lag time of 2.0± 0.1 days. Amidation of the C-terminal B chain of KP-insulin appeared to have a protective effect: native KP-amide exhibited a fibrillation lag time of

3.8± 0.3 days. In comparison, the fibrillation lag time of des-A21 KP-amide was diminished; the analog formed fibrils after only 2.0±0.1 days. However, GlyA21 KP-amide displayed a similarly diminished lag time of 2.3±0.2 days suggesting that the loss of the

A21-B23 hydrogen bond did not contribute to the reduced fibrillation lag time of the des-

A21 analog. Results are summarized in Table 3-1.

Cobalt-Coordinated Hexamer Studies- To probe the ability of des-A21to assemble into divalent metal ion-coordinated, phenol stabilized hexamers (termed R6), visible spectroscopy of the 574 nm absorption band of tetrahedrally coordinated cobalt was employed. The rate of hexamer dissociation was monitored by measuring absorbance at

574 nm after the addition of EDTA to a final concentration of 2 mM. Wavelength scans of insulin hexamers confirmed the ability of des-A21, GlyA21, and native OrnB29-amide insulin analogs to form R6 hexamers. Whereas the half-life of OrnB29–amide hexamers(450s ±44) was similar to that of native OrnB29 (454s ±20), des-A21-OrnB29- amide hexamers showed increased stability (half-life 884s ±45). GlyA21 OrnB29-amide formed hexamers that were less stable than native insulin (half-life 328s ±18). These results suggest the formation of the A21-B23 hydrogen bond may contribute to hexamer

124

Figure 3-4: Circular dichroism spectroscopy of amidated, monomeric insulin analogs. Left, wavelength scans of KP-amide (black), GlyA21 KP-amide (red), and des-

A21 KP-amide (blue). Right, CD monitored guanidine titration curves of KP-amide analogs. Curves were fit to a two-state thermodynamic unfolding equation to extrapolate free energies of unfolding (ΔGu). Results are summarized in Table 3-1.

125

126

Table 3-1. Stability of A21-Variant Insulin Analogs

a b Analog ΔGu Cmid M fibrillation lag (kcal mol-1) (M) (kcal mol-1 M-1) time (days) KP-amide 2.8 ± 0.1 5.0 ± 0.2 0.56 ± 0.03 3.8 ± 0.3 GlyA21 KP-amide 2.5 ± 0.1 4.4 ± 0.2 0.56 ± 0.02 2.3 ± 0.2 Des-A21 KP-amide 2.5 ± 0.1 5.0 ± 0.3 0.49 ± 0.03 2.0 ± 0.1 KP 2.9 ± 0.1 4.8 ± 0.2 0.60 ± 0.01 2.0 ± 0.1 WT 3.3 ± 0.1 4.9 ± 0.2 0.68 ± 0.01 3.6 ± 0.7 aThermodynamic parameters were inferred from CD-detected guanidine denaturation data by application of a two-state model. bThe m-value (slope Δ(G)/Δ(M)) correlates with extent of hydrophobic surfaces exposed on denaturation.

127

Figure 3-5. Hexamer Dissociation Assays of des-A21 and Control Analogs Visible- range spectroscopy was used to assess the stability of Co2+-coordinated, phenol-stabilized

2+ (R6) insulin hexamers. The signature absorbance band of tetrahedrally coordinated Co at 574 nm was used as a reporter of R6-hexamer formation. (A) Visible-range wavelength scans of A21 variants revealed the formation of R6 hexamers of all four analogs tested.

Small differences in amplitude are the result of differences in the T-R equilibrium of hexamers. (B) Hexamer dissociation rates were monitored as extinction of absorbance at

574 nm after the addition of an excess of EDTA to each sample. Dissociation curves are color-coded as in A. Half-lives of hexamers are summarized in Table 3-2.

128

129

2+ Table 3-2. Half-lives of Co -Substituted R6 Hexamers of Insulin Analogs

Analog t1/2 ± SD (min) OrnB29 454s ±20 OrnB29-amide 450s ±44 GlyA21,OrnB29-amide 884s ±45 Des-A21 OrnB29-amide 328s ±18

130 dissociation with some compensation to hexamer stability coming from the side chain of the asparagine residue (Figure 3- 5).

Fibrillation Experiments:

Biological Activity of des-A21 and Control Analogs: Affinity of analogs for solubilized insulin receptor (IR) was determined by competition assay against radioactive tracer. Results of receptor binding experiments have indicated that the affinity for des-

A21 lispro is approximately 1.5% of native lispro and 3% of lispro-amide (Figure 5).

Biological activity of insulin analogs was assessed by monitoring glucose clearance in streptozotocin-rendered diabetic rats after intravenous injection of analog of interest.

Both native KP-amide and GlyA21 KP-amide insulin analogs showed equivalent potencies and pharmacological profiles to KP control. des-A21-KP-amide showed markedly reduced potency compared to its native and GlyA21 counterparts (Figure 3-6).

NMR Spectroscopy:

Heteronuclear NMR was used to obtain a high-resolution view of the structures of a des-A21-insulin analog in relation to a native control. NMR experiments were conducted using an engineered monomeric (KP) insulin analog containing stabilizing substitutions of HisB10→ Asp and ThrA8 → Glu, which i) increased synthetic yields of labeled samples in yeast and ii) improved the quality of NMR spectra collected in 5% deuteroacetic acid

(required to prevent aggregation of the protein sample). The C-terminus of the B-chain of native and des-A21 analogs was amidated to facilitate digestion by carboxypeptidase-A.

Comparison of 1D 1H spectra of GluA8,AspB10-KP-amide and des-A21-GluA8,AspB10-

KP-amide revealed an overall similarity in the pattern of chemical shifts. Several peaks in the aromatic region of the native analog appeared to be better resolved than those of the

131

Figure 3-6. Biological Activity of des-A21-KP-amide. (A) Competitive displacement of radioactive tracer from solubilized IR by insulin analogs. Raw data (colored points) were fit to curves (corresponding colored lines) to determine Kd. Results are summarized in

Table 3-3. (B) Biological activity of analogs after IV injection in diabetic rats represented as glucose-lowering curves, color code as in A.

132

133

Table 3-3. In Vitro Affinities of A21-Variants for IR

Analog KD % Parent Analog (nM) KP 0.10 ± 0.02 - KP-amide 0.25 ± 0.04 100 Gly A21 KP-amide 0.59 ± 0.09 42 Des A21 KP-amide 7.11 ± 1.00 3.5

134 des-A21 analog (Figure 3-7A). 2D 13C-1H and 15N-1H HSQC spectra of both analogs were essentially identical (Figure 3-7B,C)). Residue GlyB20 exhibited the largest difference in chemical shift between the 15N-HSQC of native and des-A21 analog (Figure

3-7B). Such a change in shift may have been the result of increased solvent exposure due to the removal of residue Asn21 or due to small changes in the disposition of the B20-B23

β-turn.

3.4 Discussion.

The ubiquitous expression of insulin and its role in vertebrate metabolism imposed strict evolutionary constraints on its structure (J. M. Conlon, 2001). Such constraints affect not only the physiological function of insulin as a metabolic hormone and the stability of the mature protein but also the ability of proinsulin to achieve its native conformation (C. M. Dobson, 1999; Sikosek & Chan, 2014). Whereas the formation of the three canonical disulfide linkages of insulin represent critical kinetic barriers in the proinsulin folding process (Qiao, et al., 2003; Weiss, et al., 2000), the pre-oxidative folding of insulin has been shown to be critical to the formation of the critical A19-B20 disulfide bridge and subsequent folding process (Weiss, Nakagawa, et al., 2002).

Although the events that constitute the pre-oxidative folding process of proinsulin are unknown, structural components that contribute to the process have been identified.

This chapter has discussed the role of a hydrogen bond formed between the amide-proton of residue AsnA21 and the carbonyl carbon of GlyB23 in nascent proinsulin and in the mature insulin hormone. This hydrogen bond is dispensable in the structure of the mature insulin hormone: NMR spectroscopy has indicated poor protection of the AsnA21-amide proton, indicating that the A21-B23 bond is easily broken

135

Figure 3-7 NMR Spectroscopy of a Des-A21 Insulin Analog in Relation to that of its Native Counterpart. (A) 1D spectra of des-A21, GluA8, AspB10, KP-amide (top) and native GluA8, AspB10, KP-amide, (bottom). (B) 13C-1H HSQC of des-A21 (red contours) and native (green contours) analogs. The spectra of the two analogs were essentially identical. (C) 15N-1H HSQC of des-A21 and native analogs, color code as in B.

Tentatively assigned peaks are labeled. Residue GlyB20 exhibited the most significant difference in chemical shift between the two analogs.

136

137

(Y. Yang, Petkova, et al., 2010), and elimination of the bond by the removal of the A21 residue has little impact on the thermodynamic and physical stability of the hormone or on its structure in solution. However, the interaction between AsnA21 and GlyB23 has been shown to be critical to the native folding of proinsulin both in vitro and in mammalian cell lines. Indeed, the foldability of des-A21 proinsulin is rescued in vitro by the amidation of the carboxy group of CysA20.

Role of A21-B23 Hydrogen Bond in the Folding Pathway of Proinsulin

The critical kinetic barrier to the folding of proinsulin is the native pairing of its three canonical disulfide linkages. The successive formation of each linkage increases the thermodynamic stability of the native insulin structure, a process known as “landscape maturation (M. A. Weiss, 2009a).” The mispairing of cysteine residues (CysA6 and

CysB7/CysA7 and CysA11) creates a kinetic trap that prevents proinsulin from readily achieving its native conformation (Q. X. Hua, Jia, Frank, Phillips, & Weiss, 2002). The efficiency of the oxidative folding of proinsulin is maintained in pancreatic β-cells by the expression of ER chaperones ER-oxidoreductin-1β (ERO-1β) (Khoo, et al., 2011, and protein disulfide isomerase (PDI) {Rajpal, 2012 #17759).

Less apparent are the events that constitute the pre-oxidative folding process of insulin, which facilitates the native pairing of disulfide bridges in the absence of chaperones. It is this process that allows insulin to be synthesized by chain combination in vitro (Q. X. Hua et al., 2002). That des-A21 insulin analogs cannot be synthesized by chain combination suggests that the A21-B23 hydrogen bond is critical to the pre- oxidative folding of proinsulin. Furthermore, the proximity of both B23 and A21 to the

A20-B19 disulfide bridge imply that the interaction between the two residues may play in

138 a role in this initial disulfide pairing. Indeed, the coalescence of the hydrophobic core of insulin is thought to be the critical event that facilitates this initial pairing (Weiss,

Nakagawa, et al., 2002).

The A21-B23 hydrogen bond likely stabilizes the nascent core of the insulin molecule prior to disulfide formation. Whereas short linker domains are known to facilitate the coalescence of the structural motifs that are distant from one another in the primary sequence of a protein, the C-domain of proinsulin (35 amino-acid residues in length) is too long and unstructured to perform such a function (Eric Alm & David Baker, 1999).

Instead, the formation of the core of the insulin molecule is described by a binary collision model; this requires molecular recognition by the A- and B domains of proinsulin to initiate the formation of the globular core of the protein (Islam, Karplus, &

Weaver, 2002). The orientation of the B20-B24 region of the proinsulin B chain has been shown to be critical in process. This is evidenced by proinsulin misfolding associated with the clinical mutations PheB24→Ser (S. Shoelson, Haneda, et al., 1983), a mild mutation resulting from the putative disruption of packing within the core of insulin (M.

Liu, Hodish, et al., 2010), and GlyB23→Val, a severe mutation that disrupts the B20-B23

β-turn and prevents the C-terminal B chain of insulin from packing against the core of the molecule (Edghill, et al., 2008). The latter mutation further underscores the importance of the A21-B23 hydrogen bond: given that the integrity of the B20-B23 turn is critical to proinsulin folding, it may be hypothesized that the A21-B23 hydrogen bond stabilizes the

B20-B23 in the nascent fold of proinsulin. This bond may orient the aromatic side chain of residue PheB24 so that it may pack in its native position within the core of proinsulin prior to the formation of the A20-B19 disulfide bridge. The formation of the native core

139 may alter the overall structure of proinsulin causing the hydrogen bond to weaken in the way it is observed in mature insulin.

Role of A21-B23 Hydrogen Bond in Mature Insulin

Preliminary NMR analyses of the structure of a des-A21 insulin analog has revealed its largely nativelike structure suggesting that the A21-B23 is dispensable in the context of mature insulin. The forthcoming solution structure of des-A21, GluA8, AspB10 KP- amide and its native counterpart may further validate this finding. The thermodynamic stability of a des-A21 insulin analog, although diminished compared to the native

(AsnA21) analog, was equivalent to that of its equivalent GlyA21- insulin analog. This indicated that the A21-B23 hydrogen bond did not contribute to the thermodynamic stability of insulin structure. The decreased stability of GlyA21 and des-A21 insulin analogs in relation to the parent (KP-amide) insulin analog was likely the result of interactions involving the AsnA21 side chain; most likely, this interaction is the protection of the hydrophobic A20-B19 disulfide bridge by the polar side chain of AsnA21 (Prevost,

Wodak, Tidor, & Karplus, 1991). Unexpectedly, des-A21, OrnB29-amide insulin formed

A21 A21 more kinetically-stable R6 hexamers than its Gly or Asn counterparts. This finding was likely not the result of exposed hydrophobic surface favoring hexamerization, given the decreased stability of the GlyA21 hexamer. Instead, the loss of the A21-B23 bond may provide residue GlyB23 with additional conformational freedom, allowing the residue to form a stronger interaction across the dimer interface with ProB28; this interaction is critical for hexamer stability (D. T. Birnbaum, et al., 1997).

The removal of AsnA21 decreased the biological activity of insulin. This is consistent with previous reports of decreased biological activity associated with des-A21 insulin

140 analogs (Chu, et al., 1987a, 1987b). However, such decreases in activity were found to be the result of critical contacts between the α-carbon and carboxyl groups of AsnA21 and IR.

For this reason, GlyA21-KP-amide displayed native-like biological activity in diabetic rats.

The differences in the in vitro IR affinities of amidated insulin and native KP were unexpected. Such differences may be an artifact resulting from protein aggregation during the formulation of samples. However, that analogs with two- and six- fold decreases in receptor affinity having full potency in vivo is consistent with previous data.

Concluding Remarks

Prodomains are common features of proteins. Prodomains may broadly be divided into two classes. The first class comprises domains or regions within proteins that facilitate the folding of the polypeptide into its native conformation and are proteolytically excised in the mature protein. The second class comprises residues or regions of the protein that are critical for overcoming kinetic barriers to achieving the native fold of the protein but are dispensable in the context of the native state of the protein. Such prodomains are distinct from structural motifs in proteins that are required both in the folding process and to maintain the native structure of the mature protein (Bryan, 2002). Whereas a keystone residue such as insulin residue PheB24 falls into the latter category, the hydrogen bond between residues AsnA21 and GlyB23 falls into the former (Vijay Pandyarajan, et al.,

2014b).

A variety of prodomains have been identified. Among the most complex of these was the bacteriophage p22 tailspike protein, in which self-association of the protein subunits facilitates folding of the protein (Scott Betts & Jonathan King, 1999). Another such prodomain, found in the same protein, comprised a small group of residues that, when

141 mutated, resulted in a temperature-sensitive folding defect, but did not affect the stability of the mature protein (Sturtevant, Yu, Haase-Pettingell, & King, 1989). Prodomains constituting a single-residue have also been identified: among these is a catalytic cysteine residue in bovine pancreatic trypsin inhibitor that was required for the folding of the protein, but had no discernable function in the mature molecule (Creighton, Weissman, &

Kim, 1992). To our knowledge, the A21-B23 hydrogen bond, which is constituted by the interaction between two atoms, represents the minimal prodomain. Although remarkable, the presence of this domain in the structure of insulin, a molecule that has undergone over

500 million years of natural selection, is not surprising. Indeed, an invertebrate insulin- like molecule found in C. elegans was found to contain a fourth disulfide linkage between the A21 and B23 residues (Q. X. Hua et al., 2003). Such findings highlight the usefulness of insulin as a model molecule for protein structure and evolution and demonstrate the beauty of one of the most ubiquitously expressed and clinically utilized proteins.

3.5 Methods.

Preparation of Insulin Analogs

Analogs were prepared by trypsin-catalyzed semi-synthesis using an insulin fragment, des-octapeptide (B23-B30)-insulin, and modified octapeptides as described (Inouye, et al., 1981). The fragment was made by tryptic cleavage of human insulin and purified by reverse phase HPLC. Modified octapeptides were prepared by solid phase synthetic methods (Barany & Merrifield, 1980). The resulting 51-residue insulin analogs were purified by preparative reverse phase C4 HPLC (Higgins Analytical Inc., Proto 300 C4

,10 µM, 250 x 20 mm), and their purity was assessed by analytical reverse phase C4

142

HPLC (Higgins Analytical Inc., Proto 300 C4 5 uM, 250 x4.6 mm). Predicted molecular masses were in each case verified using an Applied Biosystems 4700 proteomics analyzer

MALDI-TOF.

Des-A21 analogs were produced from insulin analogs in which the C-terminus of the

B chain was amidated. Analogs were made 4 mg/mL in 6 M urea 0.1 M NH3CO3 and treated with 18µl Carboxypeptidase-A (Sigma Aldrich®) from a solution containing 1820 units/mL; a total of 33 units of enzyme were added. Reaction was gently rocked overnight at room temperature The resulting product was purified by rp-HPLC, as described. Identity was confirmed by MALDI-TOF and purity was confirmed by analytical rp-HPLC.

Circular Dichroism Spectroscopy

Far-ultraviolet (UV) spectra were obtained from 200 to 250 nm on an AVIV spectropolarimeter as described (Vijay Pandyarajan, et al., 2014a) equipped with an automated syringe-driven titration unit (Sreerama & Woody, 1993). Helix sensitive wavelength 222 nm was used as a probe of protein denaturation by guanidine hydrochloride using the AVIV-equipped automated syringe-driven titration unit.

Thermodynamic parameters were obtained by application of a two-state model as described. In brief, data were fit by nonlinear least squares to a two-state-model as shown in Equation 1

-DG0 -mx /RT ( H2O ) q +q e q(x) = A B -DG0 -mx /RT ( H2O ) 1+ e

Baseline values were approximated via pre- and post-transition lines represented by

H2O H2O equations A(x)A mAx and B(x)B mBx. Such simultaneous fitting avoids artifacts of linear plots of G versus concentration of denaturant (Pace & Shaw, 2000).

Receptor Binding Assays

Affinities for IR-B were measured by a competitive-displacement scintillation proximity assay as described (Pandyarajan, et al., 2016). This assay employed detergent solubilized holo-receptor with C-terminal streptavidin binding protein tags purified by sequential wheat-germ agglutinin (WGA) and Streptactin-affinity chromatography from detergent lysates of polyclonal stably transfected 293PEAK cell lines expressing each receptor. To obtain analog dissociation constants, competitive binding data were analyzed by non-linear regression by the method of Wang (Z. X. Wang, 1995), a model that provides an analytical solution for the binding of two ligands to a single receptor.

Rodent Assays

144 time t0 and every 10 min for the 1st h, every 20 min for the 2nd h, every 30 min for the

3rd h, and every hour thereafter to a final time of 5 h.

Assessment of Fibril Formation

The physical stability of the analogs was assessed by their propensity to form fibrils.

2010). For a given sample tube, the assay was terminated on the 2nd day of the appearance of cloudiness in the solution.

Assessment of Hexamer Stability

Visual absorption spectroscopy was used to probe the formation and disassembly of phenol-stabilized R6 Co-substituted insulin hexamers as first described by (Roy, et al.,

1989). WT insulin or analogs were made 0.6 mM in a buffer containing 50 mM Tris-HCl

145

Immunofluorescence of transiently transfected cells

INS1 cells were transfected using Lipofectamine 2000 (Invitrogen®) and at 48h fixed with 3.7% formalin in PBS (pH 7.4) for 20min, permeabilized with TBS containing 0.4%

TritonX-100, blocked with TBS containing 3% BSA and 0.2% Triton-X100, and then stained overnight (4oC) with primary mAb anti-human proinsulin (not cross-reacting with rodent proinsulin and not cross-reacting with human insulin) and rabbit anti-PDI diluted in TBS containing 3% BSA and 0.2% Tween. Thereafter, sections were rinsed and incubated with secondary antibodies conjugated to Alexa Fluor 488 or 568 (Invitrogen).

Slides were mounted with Prolong Gold with DAPI (Invitrogen) and imaged by epifluorescence in an Olympus FV500 confocal microscope with a X60 objective.

In vitro Proinsulin Folding Assays

WT, des-A21, and A21-amide proinsulin analogs were synthesized by one-pot native chemical ligation techniques as described (Luisier, Avital-Shmilovici, Weiss, & Kent,

2010). Folding experiments were performed on a small scale beginning from the linear unfolded polypeptides. The crude polypeptide was folded at room temperature at a concentration of 0.1 mg/mL in a buffer solution containing 10 mM Tris at pH 7.5, 10 mM

Gly, 1 mM EDTA, 1 mM GSH, 2 mM GSSG, 70 mM Gu·HCl and PDI in a

PDI/proinsulin molar ratio of 0.093 (Winter, Klappa, Freedman, Lilie, & Rudolph, 2002;

Winter, Lilie, et al., 2002). The folding solution was purged and sealed. Folding was assessed by HPLC after 3h, identities of products were confirmed by LC-MS.

NMR Spectroscopy

NMR spectra were acquired using Bruker AVANCE 700 MHz spectrometer equipped with a triple-resonance cryoprobe (Bruker Biospin Corp, Billerica, MA), in 5%

146 deuteroacetic acid (pD ~3.0 by direct meter reading) at 32° C for the one-dimensional and two-dimensional homo-nuclear 1H NMR and pH 7.4 and 25° C for two-dimensional

HSQC spectra of 13C,15N labeled samples as described (Q. X. Hua, Jia, & Weiss, 2011).

147

Chapter 4: Protein Evolution at the Edge of Non-Foldability: Impaired Biosynthesis of an Insulin Variant with Native Structure and Function

4.1 Chapter Summary

Proteins have evolved to be foldable, and yet determinants of foldability may be inapparent once the native structure is achieved. Here, we demonstrate that an invariant phenylalanine in human proinsulin (residue B24) is essential for its biosynthesis in a human secretory cell line. Whereas non-conservative mutations at B24 are known to induce toxic misfolding of the nascent polypeptide in a monogenic diabetes syndrome, our model studies focused on a conservative substitution: PheB24Tyr. Surprisingly, this substitution in proinsulin markedly impaired folding and secretion relative to wild-type.

Although endoreticular stress was induced, spectroscopic studies of the corresponding insulin analog revealed native-like structure with modest decrement in stability. The variant also exhibited native self-assembly as probed by cobalt absorption spectroscopy.

Packing of TyrB24 in the solution structure of an engineered insulin monomer was similar to that of the wild-type side chain, in each case adjoining the key B19-A20 disulfide bridge in the hydrophobic core. Although the analog’s affinity for the detergent- solubilized insulin receptor was decreased by 20-fold, biological testing in a diabetic rat model indicated essentially native function. Together, our findings suggest that exclusion of Tyr at this position in vertebrate insulins reflects a cryptic yet essential role of PheB24 in enabling efficient folding, trafficking and secretion. We envision that the para- hydroxyl group of TyrB24—although compatible with native structure and function— hinders pairing of cystine B19-A20 in an obligatory on-pathway intermediate. In vivo

148 this variant would presumably trigger β-cell dysfunction (leading to diabetes mellitus) and so is excluded by natural selection.

4.2 Introduction

Insulin is a small globular protein that regulates metabolic homeostasis in vertebrates

(G. Dodson & Steiner, 1998). Long a model for studies of protein structure (E. N. Baker, et al., 1988) and a mainstay of therapy for diabetes mellitus (DM) (Zaykov, et al., 2016), insulin is the two-chain product of a single-chain precursor, designated proinsulin

(Steiner et al., 1969). Expressed in pancreatic β-cells, this precursor has organization B-

C-A, wherein a connecting segment (C; 35 residues in human proinsulin) links the C- terminal residue of the B chain (30 residues; ThrB30 in human insulin) to the N-terminal residue of the A chain (21 residues; invariant GlyA1). Although proinsulin is refractory to crystallization, its structure in solution (as determined by heteronuclear NMR methods

(Y. Yang, Hua, et al., 2010)) contains a folded insulin domain and flexible connecting segment (Figure 4-1A). The present study has focused on an invariant Phe at position B24

(PheB24). This aromatic side chain packs within a non-polar crevice adjoining cystine

B19-A20 (Figure 4-1B, C), the critical first disulfide bridge in oxidative refolding (Q. X.

Hua, Chu, et al., 2002; Weiss, et al., 2000).

Insulin and insulin-like sequences form a metazoan superfamily, found both in vertebrates and invertebrates, including genetic model organisms Drosophila melanogaster (Varghese, Lim, & Cohen, 2010) and Caenorhabditis elegans (Q. X. Hua, et al., 2003) (Figure 4-2). This superfamily is defined by a distinctive pattern of cysteines and framework non-polar residues (Z. Y. Guo, Qiao, & Feng, 2008). Structures of invertebrate insulin-like proteins (Blumenthal, 2010; Chan, Cao, & Steiner, 1990) contain

149

Figure 4-1. Schematic of the role of PheB24 in proinsulin structure. A, schematic of the primary sequence of proinsulin. The C domain is shown as a black line, and the three canonical disulfide linkages are shown as yellow lines with circles representing the sulfur atoms. PheB24 is highlighted in red. B, the solution structure of proinsulin (PDB ID:

2KQP). The NMR ensemble is shown as gray lines, whereas the average structure is shown as colored ribbon. The A domain is colored green, the B domain is colored blue, and the C domain is colored orange. Disulfide linkages are shown as gold sticks with gold spheres representing the sulfur atoms. The side chain of PheB24 is shown as red sticks. C, a stereo view of PheB24 in its local environment in the solution structure of proinsulin. PheB24 (red sticks) packs against the hydrophobic core of insulin; the side chains of local residues are shown as sticks, with oxygen atoms colored red, with methyl groups represented as spheres. Disulfide linkages are shown in gold with spheres representing sulfur atoms. D, the side chain of PheB24 shown within the electrostatic potential surface of the surrounding residues within proinsulin.

150

151

Figure 4-2 Phylogenetic Tree of Vertebrate Evolution Various taxa within the vertebrate subphylum are displayed C. elegans and D. melanogaster are shown as outgroups. Appearance occurred during the Cambrian explosion, approximately 500-575 million years ago.

152

153

three β-helices whose orientation and packing are similar to that observed in mammalian insulins (J. M. Conlon, 2000; J. M. Conlon, 2001). Unlike invertebrate homologs, however, the α-helical domains of vertebrate insulins and insulin-like-growth factors

(IGFs) are extended by a C-terminal B-chain β-strand (B24-B28). This element contains a conserved triplet of aromatic residues—PheB24, PheB25 and TyrB26—whose contributions to structure and function have been extensively investigated (Mirmira, et al., 1991; Pandyarajan, et al., 2016; Vijay Pandyarajan, et al., 2014b). Whereas non- aromatic variation occurs at B26 and position B25 may be Phe or Tyr, PheB24 is remarkable for its absolute conservation in vertebrates (Figure 4-3) (Mirmira & Tager,

1989; Vijay Pandyarajan, et al., 2014b).

Why is PheB24 invariant? Interest in this question has been sharpened by the discovery of mutations at this site in association with DM (Edghill, et al., 2008; S. E. Shoelson, et al., 1984a). Although early studies focused on the relative affinities of B24 analogs for the insulin receptor (IR), recent studies have highlighted the importance of residue B24 in the biosynthesis of proinsulin (M. Liu, Haataja, et al., 2010). This pathway contains a complex series of steps (Qiao, Guo, & Feng, 2006; Qiao, et al., 2003) (Figure 4-4).

Diverse mutations in proinsulin can perturb one or more steps, leading to dysregulation of insulin biosynthesis (M. A. Weiss, 2009a). Of particular interest are heterozygous mutations in proinsulin leading to toxic misfolding and impaired secretion in trans of wild-type (WT) insulin (M. Liu, Hodish, et al., 2010; Weiss, 2013b). Aberrant aggregation triggers endoreticular (ER) stress, leading in turn to β-cell dysfunction and eventual death (Ming Liu, et al., 2015). First observed in the Akita mouse (mutation

CysA7Tyr in the ins1 gene) (Hodish et al., 2010; Izumi, et al., 2003), corresponding

154 human mutations were observed in patients with permanent neonata-onset DM (Nadja

Herbach et al., 2007).

Folding of proinsulin in the ER is coupled to native disulfide pairing. This process requires specialized machinery in the ER, including chaperonins and oxidoreductases

(Khoo, et al., 2011; Rajpal, et al., 2012). Two classes of clinical mutations impair disulfide pairing leading to aberrant inter-molecular cross-links. (i) Gain or loss of a cysteine (as respectively exemplified by PheB24Cys and CysA7Tyr; the latter recapitulating the Akita mutations) in either case results in an unpaired thiol. Many such mutations have been identified in the A- and B domains of proinsulin, in each case leading neonatal onset (M. Liu, Haataja, et al., 2010). (ii) Non-cysteine-related mutations are associated with variable ages of DM onset, ranging from neonatal (as exemplified by

GlyB8Ser (Colombo et al., 2008; Stoy, et al., 2007)) to early childhood (ArgB22Gln

(Stoy, et al., 2010)) and even in the third decade of life (PheB24Ser (S. E. Shoelson, et al., 1984b)). The later the onset, the less severe the presumed folding defect, resulting in milder ER stress (M. Liu, et al., 2007). Collectively, this syndrome is designated “mutant ins diabetes of the young” (MIDY) (M. Liu, Hodish, et al., 2010).

The present study has focused on conservative substitution PheB24Tyr. Like the codons encoding the clinical variants CysB24 (severe) and SerB24 (mild), the codon encoding TyrB24 may be obtained from the WT codon by a single nucleotide substitution—despite its exclusion from vertebrate insulin- and IGF sequences. We found that in a mammalian cell TyrB24-proinsulin encounters a folding defect intermediate between those conferred by CysB24 and SerB24. Despite this defect, a mature TyrB24- insulin analog (prepared by semi-synthesis (Inouye, et al., 1979)) exhibits native-like

155

Figure 4-3. Alignment of Vertebrate Insulin and IGF Sequences Alignment of the B chain (A) insulin and (B) IGF sequences of the B chain (or B domain) various taxa within the vertebrate subphylum. The B24 residue of insulin and the equivalent B23 residue of

IGF is highlighted in red.

156

157

Figure 4-4. The biosynthesis of insulin. Top, schematic of proinsulin trafficking within a cell and, bottom, an expanded schematic of proinsulin biosynthesis and transport within organelles. Steps are labeled as follows (a) translation in the rough endoplasmic reticulum (ER), (b) translocation into the ER with cleavage of a signal sequence (Okun &

Shields, 1992), (c) oxidative folding in the ER with the assistance of oxidoreductases to achieve native disulfide pairing (Khoo, et al., 2011), (d) quality control (associated with chaperone binding or ER-associated degradation) and (e) export to the Golgi apparatus,

(f) processing of the dibasic sites at the BC and CA junctions to yield the mature hormone and C peptide; (g) trafficking to glucose-regulated secretory vesicles; and (h) zinc- dependent assembly and (i) microcrystallization (Steiner, et al., 2009).

158

159 structure and function. Whereas stability is decreased, the conformation and environment of TyrB24 closely resembles the native packing of PheB24 as defined by NMR studies of an engineered insulin monomer (Q. X. Hua, Hu, et al., 1996). As a seeming paradox, a

TyrB24-insulin analog may be used to regulate metabolism in a rat model of DM. Our results indicate that proinsulin contains cryptic determinants of foldability that are inapparent in the mature hormone. We speculate that PheB24 functions in an oxidative folding intermediate to facilitate the pairing of CysB20 and CysA19.

4.3 Results

TyrB24 insulin analogs exhibited native biological activity.

The in vitro affinity of TyrB24, OrnB29-insulin for detergent-solubilized, lectin-purified

IR holoreceptor was determined in competition with a radiolabeled tracer. TyrB24, OrnB29- insulin displayed 4% and 6% affinity of WT and native OrnB29-insulin for IR isoforms A and B, respectively. Binding of the variant analog to IGF-1R was similarly proportional to control analogs. These results are consistent with previously published data on TyrB24 insulin analogs (Mirmira & Tager, 1989) (Figure 4-5A,B,C).

The in vitro affinities of insulin analogs often do not reflect their biological activity

(Ribel, et al., 1990). For this reason, the ability of TyrB24, OrnB29-insulin to lower the blood-glucose level of streptozotocin-renderd diabetic rats was tested. TyrB24, OrnB29- insulin exhibited reduced blood-glucose lowering capability after IV injection in relation to native OrnB29-insulin. However, a two-fold increase in dose of the analog produced a native-like pharmacodynamics profile, whereas a three-fold increase in dose lowered blood-glucose levels more than native OrnB29-insulin. These results suggested that TyrB24,

OrnB29-insulin is a functional variant although it has diminished biological activity

160

Figure 4-5. Biological activity of TyrB24, OrnB29-insulin in relation to control analogs.

A-C, competitive receptor-binding assays of WT-insulin or analogs for A, IR-isoform B,

B, IR-isoform A, C, IGF-1R. Data points are represented by colored symbols whereas corresponding lines indicate fit curves; the color code is provided in panel A. D, a dose- response experiment assessing the in vivo biological activity of TyrB24, OrnB29-insulin in relation to native OrnB29-insulin in diabetic rats.

161

162

Figure 4-6. Circular dichroism studies of TyrB24 insulin analogs in relation to controls. A. Far-UV CD spectra of WT-, OrnB29-, and TyrB24, OrnB29-insulin and, C, lispro and Tyr24-lispro, collected at 25°C. B, guanidine denaturation assays of WT-,

OrnB29-, and TyrB24, OrnB29-insulin and, D, monitored by ellipticity at 222 nm (color codes as in A and C, respectively).

163

164

Table 4-1. Thermodynamic stabilities of insulin analogs.

a b analog ΔGu Cmid m (kcal mol-1) (M) (kcal mol-1 M-1) Wild-type 3.2 ± 0.1c 5.2 ± 0.1 0.62 ± 0.02 OrnB29 3.2 ± 0.1 5.0 ± 0.1 0.64 ± 0.01 TyrB24 OrnB29 2.3 ± 0.1 4.7 ± 0.2 0.49 ± 0.03 Bovine 3.3 ± 0.1 4.6 ± 0.1 0.72 ± 0.02 Lispro 2.9 ± 0.1 4.9 ± 0.2 0.59 ± 0.01 TyrB24 Lispro 2.3 ± 0.1 4.9 ± 0.2 0.47 ± 0.02 a Parameters were inferred from CD-detected guanidine denaturation data by application of a two-state model; uncertainties represent fitting errors for a given data set. bThe m-value (slope Δ(G)/Δ(M)) correlates with extent of hydrophobic surfaces exposed on denaturation.

165

Figure 4-7. Circular Dichroism of Control Insulin Analogs CD wavelength scans

and (B) CD-monitored guanidine denaturation curves of of OrnB29 and TyrB24, OrnB29-

insulins at 37°C. Guanidine denaturation curves of bovine insulin and wild-type

human insulin are also included. Results are summarized in Table 4-2a. (C) CD

wavelength scans and (B) CD-monitored guanidine denaturation curves of AspB10-

lispro and AlaA6 AlaA11 AspB10 Lispro at 4°C. Results are summarized in table 4-2b.

166

167

Table 4-2a. Thermodynamic stabilities of insulin analogs at 37°C

Analog ΔGu Cmid m -1 (kcal (M) (kcal mol -1 -1 mol ) M ) Wild-type 2.5 ± 0.1 4.5 ± 0.2 0.57 ± 0.02 OrnB29 2.6 ± 0.1 4.6 ± 0.2 0.57 ± 0.02 TyrB24 OrnB29 1.7 ± 0.1 3.8 ± 0.3 0.45 ± 0.03 Bovine 2.4 ± 0.1 4.4 ± 0.1 0.54 ± 0.02

Table 4-2b. Thermodynamic stabilities of insulin analogs at 4°C

analog ΔGu Cmid m (kcal mol-1) (M) (kcal mol-1 M- 1) AspB10 Lispro 4.7 ± 0.1 6.2 ± 0.1 0.76 ± 0.01 AlaA6 AlaA11 1.8 ± 0.5 3.2 ± 0.6 0.57 ± 0.11 AspB10 Lispro

168

Figure 4-8. CD Spectrum of Diluted TyrB24, OrnB25 Insulin. Far-UV CD spectra of

WT-, OrnB29-, and TyrB24, OrnB29-insulin at 37⁰C after 10-fold dilution of the samples from Figure 4-7A (final concentration ~5µM). Dilution mitigates some of the spectral differences observed in Figure 4-7A.

169

170

(Figure 4-5D).

A monomeric TyrB24 insulin analog exhibited native structure with reduced stability.

Circular dichroism (CD) spectroscopy was utilized to assess the thermodynamic stability of mature TyrB24 insulin analogs. The CD spectrum of TyrB24 OrnB29 insulin displayed less α helical content than WT and native OrnB29 insulin at both 25°C and 37°C

(Figure 4-6A, Figure 4-7A; Table 5-1, 5-2). This was hypothesized to be the result of impaired dimerization of the TyrB24 variant: this is hypothesis is supported by the mitigation of the spectral difference when samples were diluted 10-fold (Figure 4-8).

Furthermore, the CD spectrum of TyrB24 introduced into an engineered monomeric insulin analog, insulin lispro, was identical to that of native lispro (Figure 4-7C).

Guanidine denaturation studies revealed decreased thermodynamic stabilities of TyrB24 insulin analogs in relation to control analogs. Whereas WT and native OrnB29 insulin had

B24 B29 similar free energies of unfolding (ΔGu), the ΔGu of Tyr , Orn -insulin was diminished by 0.9 kcal/mol at both 25°C and 37°C (Figure 4-6B). The difference in ΔGu between TyrB24-lispro and native lispro was less than that of TyrB24, OrnB29-insulin and

B29 B24 native Orn -insulin: Tyr -lispro exhibited a ΔGu of 2.3±0.1 kcal/mol compared to the

2.9±0.1 kcal/mol of native lispro (Figure 4-6D).

B24 To provide physiological context for the stability of Tyr insulin analogs, the ΔGu values of bovine insulin, a naturally occurring insulin variant with reduced physical stability, and AlaA6, AlaA11, AspB10-lispro, a partially folded insulin analog (Q. X. Hua, et al., 2001), were determined. Bovine insulin exhibited thermodynamic stability equivalent to that of WT-insulin at 25°C and modest decrease in stability at 37°C; this indicated that thermodynamic stability was not dispositive of physical degradation. AlaA6, AlaA11,

171

AspB10-lispro displayed markedly decreased thermodynamic stability in relation to

B20 Asp -lispro at 4°C (ΔΔGu= 1.8 kcal/mol) (Figure 4-7C,D). In spite of such a marked decrease in stability, AlaA6, AlaA11 proinsulins have been shown to undergo proteolytic processing and secretion in cultured cells (Gorr, Huang, Cowley, Kuliawat, & Arvan,

1999; M. Liu, et al., 2005); this indicates that decreased thermodynamic stability is not responsible for the decreased secretion of an insulin analog.

NMR Spectroscopy Revealed Native-like Solution Structure of TyrB24.

NMR spectroscopy was used to assess the impact of the TyrB24 mutation on the local and global structure of the insulin molecule in the context of the monomeric insulin analog AspB10 lispro (DKP). Comparison of the 1H spectra of native DKP and TyrB24-

DKP revealed a similar pattern of peaks within the aliphatic region. Differences observed between the two spectra may be attributed to the varying effects of the aromatic ring currents of residues PheB24 and TyrB24 causing different chemical shifts in a region with many overlapping peaks. Both native DKP and TyrB24-DKP displayed a greater degree of similarity in the aromatic region (Figure 4-9A).

2D proton NMR suggested the native packing of TyrB24 within the hydrophobic core of insulin. The signature NOEs between LeuB15 methyl protons and those of the B24 side chain were present in the spectra of both analogs (Figure 4-9B,C). Heteronuclear

NMR studies essentially identical 1H-13C HSQC spectra of the native and variant analogs

(Figure 4-9E). Whereas 1H-15N HSQC spectra also exhibited a high degree of overlap, residues within or near the B20-B23 β-turn of TyrB24-DKP (AsnA21, CysB19, and GluB21) exhibited small downfield shifts (~0.1 ppm) in relation to those of the native analog

(Figure 4-9F). The solution structure of TyrB24-DKP was indistinguishable from that of

172 native DKP and from a collection of x-ray crystal structures of WT insulin (Figure 4-10).

This observation extended to the atoms comprising the core of the insulin molecule and those at the C-terminal B chain. The local environment of the TyrB24 was similar to that of the native Phe; both residues exhibited a similar pattern of NOEs between side-chain protons and molecules comprising the core of insulin (Table 4-3)

The TyrB24 substitution weakens dimerization without perturbing the kinetic stability of the R6 hexamer.

The ability of TyrB24, OrnB29-insulin to form dimers was determined by size-exclusion chromatography (SEC). Analogs were made 0.6 mM in zinc-free solution and applied to a gel-filtration column. The elution profile of insulin lispro corresponded to monomeric insulin (calculated MW: 4.2 kDa). Although the elution peak of TyrB24, OrnB29-insulin contained a dimeric component, the analog was largely monomeric (4.7 kDa) in relation to both WT (6.8 kDa) or native Orn B29-insulin (6.6 kDa) (Figure 4-11).

To determine whether the effects of the TyrB24 mutation on the storage of insulin, the lifetime of zinc-coordinated, phenol stabilized (R6) hexamers formed by the variant was probed using Co2+-absorption spectroscopy. The characteristic absorbance band of

2+ B24 tetrahedrally coordinated Co at 574 nm (a reporter of R6 hexamer formation) of Tyr ,

OrnB29-insulin was essentially identical to that of native OrnB29-insulin and WT insulin

(Figure 4-12C). Native OrnB29-insulin exhibited the longest hexamer-dissociation time of

B24 B29 the analogs tested (t1/2 = 8.6 min). WT insulin and Tyr , Orn -insulin exhibited slightly more rapid dissociation (t1/2 = 7.7 and 7.6, respectively), which were within one standard deviation of the dissociation halflife of OrnB29-insulin. The difference in dissociation rates of TyrB24, OrnB29-insulin hexamers and those of OrnB29-insulin was

173

Figure 4-9 NMR studies of B24Tyr-DKP insulin analog. (A) 1D 1H NMR spectra of

B24Y-DKP insulin (top) and the corresponding parent analog DKP-insulin (bottom). (B)

Overlapped aromatic region of the two-dimensional TOCSY spectra of TyrB24-DKP insulin in black (B24Y is highlighted in red) and DKP-insulin in green. (C) Overlapped aromatic to aliphatic cross-peak region of NOESY spectra of B24Y-DKP-insulin and

DKP-insulin with same color coding as in C. (D) Overlapped 15N-HSQC spectra of

B24Y-DKP insulin (black) and DKP-insulin (green). (E) Overlapped 13C-HSQC spectra of B24Y-DKP insulin (black) and DKP-insulin (green). Spectra in D and E were collected using samples made with 13C and 15N enriched amino-acids at A1-A21 and B1-

B22 (additionally DKP-insulin sample had enriched B24F). Spectra in A-C were acquired at pD 7.6 (direct meter reading) at at 32 °C. TOCSY and NOESY spectra in B and C were acquired with a mixing time of 55ms and 200 ms respectively. 15N- and 13C-HSQC spectra in D and E were acquired at 25 °C and pH 7.4.

174

175

Figure 4-10 Solution Structure of a Monomeric TyrB24 Insulin Analog (A). 15 crystal structures of wild type insulin (PDB IDs: 1MSO, 4INS, 1APH, 1BPH, 1CPH, 1DPH,

1TRZ, 1TYL, 1TYM, 2INS, 1ZNI, 1LPH,1G7A) Aligned by the backbone atoms of the three canonical helices (A1-A8, A12-A19, B8-B19) of insulin. The A-chain is shown in light-gray, the B-chain in dark-gray. Side-chain of the B24 residue is shown as red sticks and that of B15 is shown as green sticks. The side chains of residues B25 and B26 are shown as gray sticks. The three disulfide bonds are shown as yellow sticks. (B) An

NMR ensemble of 20 structures of insulin lispro, aligned and color coded as in A. (C) An

NMR ensemble of 20 TyrB24,AspB10-lispro structures color aligned and color coded as in A and B.

176

177

Table 4-3. Comparison of observed NOEs involving PheB24 of WT (DKP-insulin) and the analog TyrB24.

Phe/TyrB24 Common WT only TyrB24 Proton analog only H CysA20-H & H, AsnA21- AsnA21-H - HN, LeuB15-H H CysA20-H, AsnA21-HN, TyrA19-H CysA20- H CysB19-H, LeuB15- H & H H LeuB15-H & H, TyrB16- CysA20-H ValB12- H & H, CysB19-H, GlyB23- H1,2 H1,2 H ValB12-H, LeuB15-H & - - H, TyrB16- H & H HN GlyB23- H & HN - CysA20-HN

178

Figure 4-11. Size-exclusion chromatography of WT-insulin and analogs. A SEC chromatogram of insulin analogs in zinc-free conditions. The void volume (V0, black arrow) was defined by thyroglobulin (MW 669 kDa). (B) Plot of log(molecular weight) vs elution ratio (Ve/V0) of molecular weight standards. Linear relationship between log[MW] to elution ratio (Ve /V0) is indicated by the red line with coefficient of

2 determination (R ) 0.996 and parameters log[molecular weight] = -1.71*(Ve /V0)+6.7012.

Elution times of molecular weight standards are indicated by blue squares (labeled by molecular weight). Identity of molecular-weight standards is as follows: 66 kDa, BSA; 45 kDa ovalbumin; 29 kDa, carbonic anhydrase, 17 kDa, myosin light chain; 12.4; cytochrome C, 6.5 IGF-I. Calculated MW are given in Table 1.

179

180

Table 4-4. Self-association properties of insulin analogs.

analog t1/2 hexamer calculated MW by dissociation SECa (min ± SD ) (kDa) Wild Type 7.7 ± 1.3 6.8 Lisprob 4.6 ± 0.3 4.2 OrnB29 c 8.6 ± 1.3 6.6 TyrB24 OrnB29 7.6 ± 0.3 4.7 a Proteins were made 0.6 mM in a buffer containing ZnCl2 at a ratio of 2 zinc ions per insulin hexamer and applied to SEC column as described in Methods. Masses were calculated from the plot in Fig. 4B.

181

small when compared to that of insulin lispro (t1/2 =4.6 min), a rapid-acting insulin analog (Figure 4-12D; Table 4-4).

The propensity of TyrB24 to form fibrils was not restrictive.

The fibrillation lag time of TyrB24, OrnB29-insulin was assessed to test whether the propensity of the mutant to form proteotoxic aggregates contributes to its evolutionary exclusion. TyrB24, OrnB29-insulin displayed a lag time of 2.1 ± 0.3h. Although diminished compared to WT insulin and native OrnB29 insulin (fibrillation lag times of 4.8 ±2.5h and

4.3±0.5h, respectively), TyrB24 exhibited longer fibrillation lag time than another naturally occurring insulin variant, bovine insulin (lag time 1.2±0.2h). This insulin established a lower bound for the physical stability of naturally occurring insulin sequences (Figure 4-13).

TyrB24-proinsulin exhibited impaired folding and secretion.

Pulse-chase assays were used to assess the foldability of TyrB24 proinsulin in relation to WT and CysB24 and SerB24 variants. Variants were expressed as recombinant proteins in HEK293T cells. After 1 hour of synthesis, S35-labeled proinsulin was immunoprecipitated and expression and secretion were determined using reducing and non-reducing SDS PAGE, respectively. The foldability statistic was calculated by comparing the intensity of the migration band corresponding to natively folded proinsulin on the non-reducing gel to that of the band on the reducing gel, which corresponded to total proinsulin.

Wild-type proinsulin displayed the highest foldability, whereas the CysB24 variant displayed a dramatically reduced fraction of native-folded proinsulin and an increased number of disulfide isomers, which are visualized as slow migrating bands in the non-

182

Figure 4-12. Hexamer dissociation of insulin analogs. A, ribbon representation of the R6 insulin hexamer. The insulin A chain is shown in dark gray, and the insulin B chain is shown in blue. Coordination of a Zn2+ ion (gray sphere) by HisB10 (blue sticks) is depicted. The R6 state of the insulin hexamer is stabilized by the allosteric binding of phenolic ligands (red sticks). B, an expanded view of the tetrahedral Zn2+-coordination

2+ site in R6 insulin hexamer (ball-and-stick model): the Zn ion is coordinated by three

HisB10 side chains and one chloride ion. C, Absorbance spectra of d-d bands in corresponding Co2+complex; the color code is indicated. D, Hexamer dissociation curves as monitored at 574 nm after addition of excess EDTA; the color code is as in B.

183

184

Figure 4-13. Dot plot of lag time to fibril formation of insulin analogs. Onset of fibrillation was determined by initial increase in Thioflavin-T fluorescence. The mean lag time is indicated by the black horizontal line.

185

186

Table 4-5 Fibrillation lag times of insulin analogs.

analog fibrillation lag time (h ± SD ) Wild-type 3.9 ± 1.1* Bovine 1.2 ± 0.2 OrnB29 4.3 ± 0.5 TyrB24 OrnB29 2.1 ± 0.3

*with the outlier included, values are 4.8 ± 2.7

187 reducing gel. Both SerB24 and TyrB24 variants exhibited a larger fraction of native- proinsulin than CysB24 variant, however, the variants displayed reduced foldability in relation to WT proinsulin. WT proinsulin underwent efficient secretion into the culture media 1h after synthesis. All B24 variants displayed significantly reduced secretability in relation to WT proinsulin: CysB24-proinsulin displayed the greatest reduction in secretion, whereas SerB24-proinsulin exhibited the least severe secretion defect of the three proinsulin variants. TyrB24-proinsulin was secreted at levels intermediate to SerB24- and

CysB24-proinsulin variants (Figure 4-14A,B).

In a separate set of experiments, ER-stress associated with the misfolding of proinsulin varaints was determined as function of BiP expression (Figure 4-14C).

HEK29T cells were co-transfected with constructs expressing variant proinsulin and luciferase driven by a BiP promoter. Luciferase levels were determined 24 hours after transfection. Whereas WT proinsulin initiated expression of the luciferase reporter at similar levels to the vehicle, CysB24-proinsulin initiated a two-fold increase in luciferase expression. Both SerB24- and TyrB24-proinsulin variants triggered a 1.5-fold increase in luciferase expression compared to WT-proinsulin; this luciferase level was less than that associated with CysB24-proinsulin to a statistically significant degree (Figure 4-14D).

The ability of B24 mutant proinsulins to interfere with the folding and secretion of

WT-proinsulin, a property referred to as trans-dominance, was assessed in HEK293T cells. WT proinsulin and B24 variants were co-expressed in cells with myc-tagged WT proinsulin. Secretion of 35S-labeled WT proinsulin into the culture media was assessed by

SDS PAGE following a one-hour chase. Whereas myc-tagged proinsulin was detected at lower levels in the media of cells expressing SerB24-proinsulin compared to those

188

Figure 4-14. Folding and secretion properties of proinsulin variants in a human cell line. A, a non-reducing SDS PAGE gel depicting migration bands corresponding to natively folded proinsulin (or variant) and non-native isomers after a pulse-chase assay.

B, secretion of proinsulin variantsform transfected cells into the culture media. The difference in %secretion between each variant is statistically significant (p < 0.05). C.

Cartoon representation of BiP activation in response to unfolded protein. Accumulation unfolded protein (red coil) within the ER leads to the dissociation of BiP (blue semicircle) from ATF6 (brown), IRE-1 (purple), and PERK (orange). Liberated BiP then binds to unfolded protein to prevent proteotoxic aggregation. Activated ATF6, IRE-1, and PERK initiate the unfolded protein response, which decreases translation of proteins within the cell and increases ER associated degradation (ERAD). The expression of ER- associated chaperones, including BiP, is upregulated by UPR. D, activation of the UPR by proinsulin variants as monitored by expression of a BiP-luciferase reporter. The asterisk (*) indicates statistically significant increase (p < 0.05) in luciferase reporter in relation to WT proinsulin. The octathorpe (#) indicates statistically significant reduction in BiP-luciferase expression in relation to CysB24 proinsulin.

189

190 expressing un-tagged WT-proinsulin, no myc-tagged proinsulin was detected in the media of cells expressing CysB24- and TyrB24-proinsulins (Figure 4-15). For this reason, CysB24 and TyrB24 mutations may be considered trans-dominant, whereas SerB24 has characteristics of a recessive mutation.

TyrB24- proinsulin was confined within the ER of β-cells.

The manifestation of the TyrB24 mutation at the cellular level was visualized in rat insulinoma cells (INS1) that natively express and secrete insulin. INS1 cells were transiently transfected with WT human proinsulin or TyrB24, SerB24, and CysB24 variants.

Immunostaining of cells for proinsulin indicated incomplete transfection of cells with proinsulin constructs: only a few cells per low power field expressed the construct. Cells expressing WT proinsulin, exhibited juxtanuclear staining consistent with Golgi localization. A small number of puncta were visualized beyond the juxtanuclear region; this suggested the presence of uncleaved proinsulin that had been packaged into secretory granules. This pattern of localization differs from that of cells expressing TyrB24, SerB24, and CysB24 variants, which exhibited diffuse localization within the cytoplasm; juxtanuclear localization was not visualized. This pattern of localization indicated the lack of trafficking of proinsulin variants into the Golgi apparatus (Figure 4-16).

This finding was strengthened by observation of the colocalization of proinsulin variants with calnexin, an ER-associated chaperone. WT proinsulin displayed modest colocalization with calnexin (immunostained red), indicating rapid egress from the ER. In contrast, all three variant proinsulins exhibited a high degree of colocalization with calnexin. This observation indicated that the proteins were sequestered within the ER, presumably as a consequence of the action of ER quality control.

191

Figure 4-15. Assessment of trans-dominance of proinsulin variants. A SDS-PAGE gel displaying the folding and secretion of C-myc-tagged WT proinsulin co-transfected with

WT proinsulin or B24 variants in HEK293T cells. The gel depicts the migration bands of

35S-labeled proteins in cell lysates at the beginning of the chase (leftmost lanes marked

“C”) and after 30 minutes of chase (middle lanes marked “C”). Labeled proteins secreted into the media after the chase were run in lanes marked “M.” Bands correspond to WT proinsulin or B24 variants, myc-tagged proinsulin as labeled. Lanes marked “Con” indicate untransfected cells. Bands corresponding to C-myc-tagged proinsulin are visible in the media of cells transfected with WT and SerB24- proinsulin, whereas no such band is visible in the media of cells transfected with CysB24 and TyrB24- variants, suggesting that these two variants are trans-dominant. The asterisk indicates that the lanes corresponding to TyrB24 proinsulin were cut and pasted from non-consecutive lanes in the same gel.

192

193

4.4 Discussion

The accumulation of unfolded and misfolded proteins within cells represents an evolutionary pitfall (C. M. Dobson, 1999). This concept is illustrated by the robust system of safeguards against protein misfolded within cells; among these are protein chaperones and ER quality-control systems (C. M. 14685248 Dobson, 2003; Koga,

Kaushik, & Cuervo, 2011). Beyond necessitating these explicit adaptations, toxic misfolding of proteins constrains the sequence of the proteins themselves. Indeed, the structure of a protein is determined not only by its function, but also by its ability to achieve and maintain its native fold (Ashenberg, Gong, & Bloom, 2013). Prodomains, which are highly conserved regions of a protein that are necessary for its proper folding, but dispensable in the context of the mature protein structure, exemplify such constraints

(Sturtevant, et al., 1989).

Insulin provides a robust model for the study of protein folding. The ubiquitous requirement for insulin signaling among vertebrates necessitated its availability and continuous synthesis. However, the biosynthetic pathway of the hormone functions at the edge of toxic misfolding (M. Liu, Hua, et al., 2010): even basal levels of insulin production has been shown to increase the expression of UPR-related proteins in β cells

(Lipson et al., 2006).

Analysis of disease phenotypes associated with mutations to the INS gene have led to the identification of a number of “hot spots” that are critical to the proper folding of insulin.

Our results demonstrate that PheB24, a residue invariant among vertebrate insulins, represents one such hot spot.

Native-like properties of TyrB24 insulin do not rationalize its evolutionary exclusion.

194

Figure 4-16. Intracellular distribution of human proinsulin wild type (WT), or mutations of human proinsulin PheB24 with Cys, Ser, or Tyr substitutions. INS1 cells expressing only rat proinsulins were transiently transfected to express the indicated human proinsulin constructs. By this procedure, only one to a few cells per low power field were actually transfected. All cells are immunoreactive for calnexin, which localizes to the ER, and transfected cells are immunoreactive with antibody that recognizes only human proinsulin. For the WT human proinsulin, the protein advances beyond the ER and can be seen to be concentrated in a juxtanuclear Golgi region rich in forming secretory granules; in some cells proinsulin distribution has migrated still further to a granule distribution subjacent to the plasma membrane. However, proinsulin CysB24,

SerB24, and TyrB24 demonstrate an ER distribution matching that of calnexin, indicating a defect in the intracellular proinsulin transport of these mis-sense mutants.

195

196

The structure of TyrB24 lispro was identical to that of the native molecule as determined by NMR spectroscopy. Moreover, the TyrB24 side-chain displayed native occupancy, packing against the core of the globular protein; determination of the NMR ensemble of TyrB24 lispro using restraints determined from native lispro produced a structure identical to that determined by NMR studies of TyrB24 lispro itself. In the context of the mature hormone, no local or global perturbations associated with TyrB24 that may affect the folding pathway of proinsulin were apparent.

CD studies revealed decreased stability of TyrB24 insulins in relation to control analogs. However, several factors mitigate the significance of these results with regard to insulin evolution. The qualitative differences between characteristics of the 200-250 nm spectrum of TyrB24, OrnB29-insulin and native OrnB29 insulin are likely the result of reduced dimerization of the TyrB24 analog. This hypothesis is supported by the partial mitigation of spectroscopic differences when samples were diluted 10 fold and by the similarity in spectra of monmeric lispro and TyrB24 lispro. The 0.9 kcal/mol decrease in thermodynamic stability of TyrB24, OrnB29-insulin does not sufficiently explain the folding and secretion defects associated with TyrB24 proinsulin: a two-disulfide insulin

A6 A11 B10 analog (Ala , Ala , Asp lispro) exhibited ΔGu of 1.7 kcal/mol at 4°C— 0.6 kcal/mol lower than that of TyrB24, OrnB29-insulin at 25°C—yet the analog exhibits nativelike folding and secretion characteristics (Gorr, et al., 1999; Q. X. Hua, et al., 2001).

TyrB24, OrnB29-insulin did not exhibit a significant increase in susceptibility proteotoxic misfolding. Whereas the fibrillation lag time of TyrB24, OrnB29-insulin was reduced relative to those of WT and native OrnB29-insulin, the mutant analog was more resistant to fibrillation than bovine insulin, which is naturally expressed in cattle and is

197 not known to form amyloid in vivo (Devlin, et al., 2006; Zimmerman, Moule, & Yip,

1974).

The difference in fibrillation lag times between TyrB24, OrnB29-insulin and its WT controls may be explained by its impaired ability to dimerize, which was determined by

SEC studies. Whereas WT and native OrnB29 insulin analogs are expected to be in a predominantly dimeric association state in the conditions used in the fibrillation assay,

TyrB24, OrnB29-insulin is predominantly monomeric. The dimerization of the two control analogs provides additional protection against proteotoxic misfolding that prolongs their fibrillation lag time (J. Dong, et al., 2003). The nativelike stability of the TyrB24, OrnB29 hexamer suggested that the dimerization defect of the variant—and its contribution to proteotoxic aggregation—is not likely to affect its storage within pancreatic β cells.

The biological activity of TyrB24, OrnB29 insulin is substantial, although it is diminished compared to WT or native OrnB29 controls. Whereas in vitro receptor binding assays indicated ~5% IR affinity of TyrB24, OrnB29 insulin in relation to native OrnB29 insulin, the analog displayed ~50% biological activity in relation to OrnB29 insulin in diabetic rats. This two-fold difference in biological activity is insignificant in the context of physiology; the reference ranges for insulinemia and insulin resistance (HOMA-IR) may vary 3-6 fold and up to 3 fold, respectively, in non-diabetic individuals (Esteghamati et al., 2010; Ikeda, SUEHIRO, NAKAMURA, KUMON, & HASHIMOTO, 2001). For this reason, TyrB24-insulin may be considered a biologically functional variant.

Foldability constrains divergence at B24.

Our studies demonstrated the impact of small structural perturbations at the B24 position on the folding and secretion of proinsulin. The substitution of the native PheB24

198 by Tyr, a residue that differs from the native Phe by a single para-hydroxyl group, impaired not only the foldability of the mutant proinsulin, but also its ability to be packaged into secretory vesicles. This result is particularly remarkable when compared to the folding properties of the clinical mutant SerB24 insulin (insulin Los Angeles).

Although the physical properties of Ser differ from those of Phe to a greater extent than those of Tyr, SerB24 proinsulin displayed similar foldability and increased secretability in relation to the TyrB24 variant. Moreover, unlike SerB24-proinsulin, TyrB24-proinsulin displayed trans-dominance, preventing WT proinsulin co-expressed with the variant from being secreted.

Immunostaining of INS1 cells expressing B24 variant proinsulins revealed the manifestations of the decreased foldability of B24 variants in cellular physiology. WT proinsulin exhibited a punctate, juxtanuclear pattern of localization. This pattern corresponds to localization to the Golgi apparatus. A small number of puncta found in the periphery of the cell. This pattern of corresponded to localization within insulin secretory granules, as determined by previous experiments where INS-1 cells were immunostained for C-peptide. This finding suggested that a small amount of intact proinsulin was packaged into insulin secretory granules and was representative of completion of the biosynethesis and trafficking of proinsulin.

This pattern contrasted with that of each of the three B24 mutants. Unlike WT proinsulin, CysB24, SerB24, and TyrB24 variants did not exhibit localization to juxtanuclear puncta. Instead, proinsulin analogs displayed diffuse localization that corresponded to sequestration within the ER as determined by co-localization with ER-chaperone calnexin. Previous studies have shown that clinical mutations such as CysA7→Tyr (Gorr,

199 et al., 1999; M. Liu, et al., 2007) which causes PNDM and ArgB22→Glu (Stoy et al.,

2017), which is associated with MODY, display similar patterns of proinsulin localization with marked dilation of the ER. This pattern may be attributed to the failure of misfolded variant proinsulins to exit the ER because of ER quality control enzymes: similar localization is observed in cells expressing WT proinsulin that lack critical ER chaperones (such as ER oxidoreductin-1α; Ero-1α) (Wright, et al., 2013) or secretory proteins necessary for egress from the ER (Fang et al., 2015).

The phenotypes of the CysB24 and SerB24 clinical mutant provides context for the severity of B24 mutations. Whereas CysB24 mutation causes PNDM, the SerB24 mutation manifests itself in patients as hyperinsulinemia; maturity onset diabetes also occurs with varying penetrance (S. Shoelson, Haneda, et al., 1983). Whereas the former feature is likely caused by the decreased biological potency of the mutant insulin analog, the latter feature is thought to be caused by inefficient folding of the mutant proinsulin causing ER stress leading to β cell dysfunction (M. Liu, Haataja, et al., 2010).

The putative phenotype of TyrB24-proinsulin is expected to result in more severe DM than its SerB24 counterpart, although less severe than that associated with CysB24. The impaired secretion and increased BiP translation levels associated with TyrB24 insulin suggested greater levels of unfolded or misfolded protein will accumulate in pancreatic β cells; this is expected to result in a more severe manifestation of diabetes with increased penetrance (M. Liu, Haataja, et al., 2010). Moreover, the ability of the TyrB24 mutant to prevent the proper folding of WT proinsulin in trans, a property that is not shared by

SerB24 proinsulin, suggested that TyrB24 may manifest as MIDY that occurs in the first decade of life or PNDM (Hodish, et al., 2010).

200

PheB24 provides an “aromatic anchor” in the proinsulin folding pathway.

The concept of a folding landscape is used to visualize the protein folding process.

Folding landscapes are funnel-shaped with the stem of the funnel representing the conformation of the protein corresponding to the thermodynamic minimum: this is typically the natively-folded state. A protein may take a number of paths from the rim to the base of the funnel. The steepness of the path corresponds to the speed at which particular steps in the folding process occur. “Peaks” and “valleys” within the funnel represent local thermodynamic maxima and minima, respectively. Maxima may impose kinetic barriers to protein folding whereas valleys may represent kinetic traps, or metastable folding intermediates that may impede the efficiency of folding.

In vitro and in vivo experiments of mutations to proinsulin and related proteins (such as IGF-1) have suggested several critical steps in the folding process of the hormone (Q.

X. Hua, Narhi, et al., 1996). The first of these steps is pre-oxidative folding: the A- and

B-domains of proinsulin, which are ordered in the folded prohormone, form nascent structure that coalesces into what becomes the core of the mature hormone (Q. X. Hua,

Chu, et al., 2002; Q. X. Hua, Mayer, et al., 2006). This step facilitates the formation of the A20-B19 disulfide bridge (the second step in the pathway) (Weiss, Nakagawa, et al.,

2002), which is the first major kinetic barrier to proinsulin folding. The formation of the

A7-B7 bridge and the A6-A11 bridge are respectively the third and fourth major steps in the folding pathway (Q. X. Hua, et al., 2001). The efficient oxidation of disulfide bridges is facilitated by the ER oxidoreductases. The formation of off pathway disulfide isomers may also occur during the formation of the latter two disulfide bridges; these isomers represent kinetic traps that are overcome with the action of chaperones such as protein

201 disulfide isomerase (PDI) (Q. X. Hua, et al., 1995; Rajpal, et al., 2012). The formation of each disulfide bridge increases the steepness of the protein folding landscape—a process known as landscape maturation. Both proinsulin and mature insulin readily refold into their native state after chemical denaturation provided the disulfide linkages remain intact.

Mutations that severely impact the foldability of insulin manifest as permanent neonatal diabetes. There are two classes of mutations that cause permanent majority. The first, comprising a majority of PNDM mutations (Stoy, et al., 2007) are mutation that either introduce ecotopic cysteine residues or mutate native cyteines to non-cyteine residues. The resulting unpaired cysteine interferes with disulfide formation. The second class of PNDM-associated mutations contains non-cysteine mutations to the INS-gene .

Such mutations affect structural motifs that are thought to destabilize on-pathway intermediates in the folding pathway of proinsulin. The severity of the phenotype of these mutations suggest that these mutations prevent the native pairing of cysteines (Q. X. Hua, et al., 2001).

The characterization of non-cysteine mutations causing PNDM has led to a deeper understanding of the proinsulin folding pathway. Efforts to synthesize insulin containing the PNDM-associated GlyB8→Ser mutation by chain-combination demonstrated that the substitution GlyB8 by an L-amino acid diminishes synthetic efficiency. It is hypothesized that the achiral glycine is required to properly orient the neighboring CysB7 residue for disulfide pairing. A similar effect on cysteine orientation may be expected to result from another PNDM mutation, LeuB6→Pro, whereas LeuB11→Pro is expected to destabilize the central helix of the proinsulin B domain and prevent native disulfide formation. Although

202

PNDM-mutations severely impact the foldability of proinsulin, their impact on the structure and function of the mature hormone may be variable. For example, SerB8 insulin, although less structurally stable in relation to WT, retains substantial biological activity. This highlights the importance of the evolutionary constraint of foldability in relation to those of biological activity.

The severity of the folding defect associated with the TyrB24 mutation is informative of its effects on the pre-oxidative folding of proinsulin. A large subset of non-cysteine mutations that are associated with PNDM are clustered at the residues that form the hydrophobic core of the mature insulin hormone (LeuB15, TyrA19, and GlyB23) (Stoy, et al.,

2017). For this reason, it is not surprising that the PheB24 residue, which packs against the hydrophobic core in mature insulin, has been selected to optimize foldability of proinsulin.

Along with previous studies, our data suggested that the integrity of the hydrophobic core of proinsulin is a critical step in the folding pathway of the prohormone. The severity of mutations to the region and their proximity to the A20-B19 disulfide bridge imply that nucleation of the core is required for the native pairing of the two cysteine residues. This putative step is likely critical in the folding pathway of proinsulin due to the separation of the ordered A- and B domains of proinsulin by a 35 residue disordered loop and the resulting 55 residue distance between the two domains. Such a loop necessitates a “binary collision” model of protein folding in which ordered regions of the

A- and B domains (in this case, the hydrophobic core residues) form metastable complexes to facilitate the maturation of the disulfide bridge (E. Alm & D. Baker, 1999).

The mechanism of the folding defect of TyrB24-proinsulin is likely distinct from that of

203

SerB24-proinsulin. Whereas smaller SerB24 side chain likely destabilizes the nascent core during the folding of proinsulin decreasing the efficiency of disulfide formation, the para-hydroxyl group of TyrB24 likely obstructs or interferes with the formation of either the A19-B20 disulfide bridge or the hydrophobic core of proinsulin itself. Although the position of TyrB24 in the context of mature insulin is identical to that of the native Phe, it is possible that the residue occupies a non-native position in which its para-hydroxyl group forms an obstructive interaction in the context of a less conformationally restricted folding intermediate.

Concluding Remarks

Diabetes-associated mutations in the INS gene have identified residues critical to the foldability of proinsulin. One such position is PheB24, invariant as Phe among vertebrate insulins and IGFs. Whereas CysB24 causes permanent neonatal-onset DM (Stoy, et al.,

2007) and is associated with an essentially complete block to the folding of proinsulin in cell culture (M. Liu, Haataja, et al., 2010), SerB24 was observed as a classical insulinopathy with onset in the third decade of life (S. Shoelson, Haneda, et al., 1983).

Although its perturbation to the efficiency of cellular biosynthesis and secretion in culture was mild, its expression triggered the UPR with induction of ER stress (M. Liu, Haataja, et al., 2010). These mutations representing the contrasting endpoints in the spectrum of clinical presentations, bookend the present study of TyrB24.

We chose to investigate the properties of TyrB24-insulin and the effects of the substitution on the biosynthesis of proinsulin because of an evolutionary paradox. Phe and Tyr are almost identical differing only by the absence (Phe) or presence (Tyr) or a para-hydroxyl group on corresponding aromatic rings. Indeed, in the evolution of many

204 globular proteins Phe and Tyr appear interchangeably in sequence alignments. Yet, in the vertebrate insulin/IGF family, Tyr has been excluded—a period of more than 500 million years.

To investigate possible reasons why PheB24 has been so maintained, we prepared

TyrB24 insulin analogs. Like the dog that did not bark in the night-time, these analogs exhibited native-like structure and function. Modest decrements were observed in stability and self-assembly of unclear physiologic significance. The propensity of these analogs to undergo toxic misfolding (as probed by fibrillation lag times) was unchanged.

In striking contrast to the above subtle findings, the cellular folding and secretion of

TyrB24-proinsulin was markedly perturbed. Quantitative analysis of such cell-based assays suggested that these perturbations were intermediate between those of CysB24 and

SerB24. Because a codon for Tyr can be reached from the native codon, we anticipate that a future MIDY mutation will be found presenting in childhood rather than in the neonatal period like CysB24 or in early adulthood like SerB24.

Our findings suggest that TyrB24 perturbs the formation of an on-pathway folding intermediate but is well tolerated once the native state is reached. The existence of such mutations was foreshadowed in prescient studies of a phage tail-spike protein by King and coworkers. Whereas the latter protein contains a multi-subunit β helix in which folding is coupled to self-assembly (Scott Betts & Jonathan King, 1999), proinsulin folds as a small globular minidomain (Qiao, et al., 2006; Qiao, et al., 2003). Its folding is coupled stepwise to formation of its three native disulfide bridges (Q. X. Hua, Chu, et al.,

2002; Q. X. Hua, et al., 2001). In the native state, PheB24 itself contacts CysB19, whose pairing with CysA20 is thought to represent an early step in folding. We envision that

205

PheB24 stabilizes a native-like folding nucleus that guides such pairing. It would be of future interest to investigate whether or how the para-hydroxyl group of TyrB24 perturbs the formation of this nucleus.

4.5 Methods

Materials.

Insulin was purchased from BioDel® (Danbury, CT). Bovine insulin was purchased from Sigma Aldrich® (St. Louis, MO). Reagents for peptide synthesis were as described

(Wan, Huang, Whittaker, & Weiss, 2008).

Preparation of Insulin Analogs.

Variant insulins were prepared by semisynthesis (Inouye, et al., 1979). In brief, synthetic peptides were coupled to a tryptic fragment of insulin (des-octapeptide [B23-

B30] insulin) in aqueous/organic solvent using trypsin as a catalyst. Following rp-HPLC purification, predicted molecular masses were confirmed by mass spectrometry (V.

Pandyarajan, et al., 2014).

Hexamer Disassembly Assays.

2+ Disassembly of phenol-stabilized (R6) Co -substituted insulin hexamers was monitored as described (Rahuel-Clermont, French, Kaarsholm, & Dunn, 1997). In brief,

WT insulin or variants were made 0.6 mM in buffer containing 50 mM Tris-HCl (pH

7.4), 50 mM Phenol, and 0.2 mM CoCl2 (Pandyarajan, et al., 2016) and incubated overnight at room temperature to attain conformational equilibrium. Spectra (400-750 nm) were obtained to monitor tetrahedral Co2+ coordination (Roy, et al., 1989) through its signature absorption band at 574 nm (Roy, et al., 1989). Co2+ sequestration initiated by addition of EDTA to a concentration of 2mM. Dissociation was probed via attenuation of

206

574 nm band (Roy, et al., 1989); data were fit to a monoexponential decay equation (D.

T. Birnbaum, et al., 1997).

Receptor-Binding Assays.

Analog affinities for detergent-solubilized IR-B, IR-A, or IGF-1R holoreceptor were measured by a competitive-displacement assay (Pandyarajan, et al., 2016). Successive dilutions of WT insulin or analogs were incubated overnight with WGA-SPA beads

(PerkinElmer Life Sciences®), receptor, and radio-labeled tracer before counting

(Pandyarajan, et al., 2016). To obtain dissociation constants, competitive binding data were analyzed by non-linear regression by the method of Wang (Z. X. Wang, 1995).

Rat Studies.

Male Lewis rats (mean mass ~300 g) were rendered diabetic by streptozotocin.

Effects of insulin analogs formulated in Lilly® buffer (Pandyarajan, et al., 2016) on blood-glucose concentration following intravenous injection were assessed in relation to

OrnB29 insulin (Pandyarajan, et al., 2016). Rats were anesthetized with isoflurane and insulin or analog was administered via tail vein injection.

Size-Exclusion Chromatography.

Analogs were made 0.6 mM in 10 mM Tris-HCl (pH 7.4), 1.6% glycerol (v/v), 0.3 mM ZnCl2, and 7 mM phenol (V. Pandyarajan, et al., 2014). Insulin samples (20 µl) were loaded on an Enrich® SEC70 column (10 mm x 300 mm with fractionation range 3-70 kDa); the mobile phase consisted of 10 mM Tris-HCl (pH 7.4), 140 mM NaCl, and

0.02% Sodium Azide. Elution times were monitored by absorbance at 280 nm. Molecular masses and void volume (V0) were inferred in reference to standard proteins (Pandyarajan, et al., 2016).

207

Circular Dichroism Spectroscopy.

CD spectra were acquired in 10 mM potassium phosphate (pH 7.4) and 50 mM

KCl (Vijay Pandyarajan, et al., 2014a). Free energies of unfolding (ΔGu) were inferred at

25 °C from two-state modeling of protein denaturation by guanidine-HCl (Vijay

Pandyarajan, et al., 2014a; Sosnick, et al., 2000).

Fibrillation Assays.

Insulin or analogs were made 60μM in 1xPBS with 0.02% sodium azide and 16μM thioflavin T. Samples were plated in a Costar® plate (250μl/well) and incubated at 37°C with continuous shaking at 400 RPM in a Biotech plate reader. THT fluoresence at 480 nm after excitation at 450 nm was assessed at 5 minute intervals. The time of initial THT fluorescence was recorded as fibrillation lag time.

NMR Spectroscopy

1H-NMR spectra were acquired at 700 MHz at pH 8.0 or pD 7.6 (direct meter reading) at 37° C (Q. X. Hua, et al., 2011). Chemical shifts of aromatic protons in residues B24-B24 (Phe-Phe-Tyr or Phe-Phe-Trp) were evaluated in relation to corresponding chemical shifts in respective octapeptides B23-B30, presumed to represent random-coil shifts.

NMR spectra were acquired using Bruker AVANCE 700 MHz spectrometer equipped with a triple-resonance cryoprobe (Bruker Biospin Corp, Billerica, MA), at pH 8.0 or pD

7.6 (direct meter reading) at 32° C for the one-dimensional and two-dimensional homonuclear 1H NMR and pH 7.4 and 25° C for two-dimensional HSQC spectra of 13C,15N labeled samples as described (Q. X. Hua, et al., 2011).

208

Distance-geometry/simulated annealing calculations (DG/SA) were performed using program DG-II (Havel & Snow, 1991); restrained molecular dynamics (RMD) were calculated using X-PLOR (Brunger, 1992). Nuclear overhauser enhancement (NOE)- related and dihedral-angle restraints were used for molecular modeling as described

(Nakagawa, et al., 1998).

In vivo insulin folding and secretion

The samples were precleared with Zysorbin and immunoprecipitated with anti-insulin antibodies. Anti-insulin immunoprecipitates were boiled for 5 min in gel sample buffer

BiP-Driven Luciferase Assay

Min6 cells were plated into 24-well plates 1 d before transfection. Using

Lipofectamine 2000 (Invitrogen), cells were co-transfected with pBiP-firefly-luciferase reporter plasmid (Tirasophon, et al., 1998), CMV-renilla-luciferase plasmid (Promega), and human wild type or mutant proinsulin at a DNA ratio of 1:2:5, respectively. At 48 h post-transfection, cell extracts were prepared for the dual-luciferase reporter assay

(Promega) with BiP-luciferase normalized to Renilla luciferase activity

Immunofluorescence of transiently transfected cells

209

INS1 cells were transfected using Lipofectamine 2000 (Invitrogen) and at 48h fixed with 3.7% formalin in PBS (pH 7.4) for 20min, permeabilized with TBS containing 0.4%

TritonX-100, blocked with TBS containing 3% BSA and 0.2% Triton-X100, and then stained overnight (4oC) with primary mAb anti-human proinsulin (not cross-reacting with rodent proinsulin and not cross-reacting with human insulin) and rabbit anti-calnexin diluted in TBS containing 3% BSA and 0.2% Tween. Thereafter, sections were rinsed and incubated with secondary antibodies conjugated to Alexa Fluor 488 or 568

(Invitrogen). Slides were mounted with Prolong Gold with DAPI (Invitrogen) and imaged by epifluorescence in an Olympus FV500 confocal microscope with a X60 objective.

210

Chapter 5: Structure-Based Stabilization of Insulin as a Therapeutic Protein

Assembly via Enhanced Aromatic-Aromatic Interactions

Adapted from:

Rege, N.K., Wickramasinghe, N.P., Tustan, A.N., Phillips, N.F.B., Yee, V.C., Ismail-

Beigi, F., Weiss, M.A. (2018) Structure-Based Stabilization of Insulin as a Therapeutic

Protein Assembly via Enhanced Aromatic-Aromatic Interactions. Journal of Biological

Chemistry, In Press

5.1 Chapter Summary

Key contributions to protein structure and stability are provided by weakly polar interactions, which arise from asymmetric electronic distributions within amino acids and peptide bonds. Of particular interest are aromatic side chains whose directional π systems commonly stabilize protein interiors and interfaces. Here, we consider aromatic-aromatic interactions within a model protein assembly: the dimer interface of insulin. Semi- classical simulations of aromatic-aromatic interactions at this interface suggested that substitution of residue TyrB26 by Trp would preserve native structure while enhancing dimerization (and hence hexamer stability). The crystal structure of a TrpB26-insulin

f analog (determined as a T3R 3 zinc hexamer at a resolution of 2.25 A) was observed to be essentially identical to that of wild-type insulin. Remarkably and yet in general accordance with theoretical expectations, spectroscopic studies demonstrated a 150-fold increase in the in vitro lifetime of the variant hexamer, a key pharmacokinetic parameter influencing design of long-acting formulations. Functional studies in diabetic rats indeed revealed prolonged action following subcutaneous injection. The potency of the TrpB26- modified analog was equal to or greater than an unmodified control. Thus exploiting a

211 general quantum-chemical feature of protein structure and stability, our results exemplify a mechanism-based approach to the optimization of a therapeutic protein assembly.

5.2 Introduction

Weakly polar interactions are ubiquitous among protein structures (S. K. Burley &

Petsko, 1988). Among such interactions, the relative packing of aromatic rings is of particular interest in relation to the organization of protein cores and subunit interfaces

(S. K. Burley & Petsko, 1985). Aromatic-aromatic interactions are governed by quantum-chemical properties, which underlie dispersion forces and give rise to asymmetric distribution of partial charges. Whereas aromatic stacking is prominent in nucleic-acid structures, pairs of aromatic side chains in proteins more often exhibit edge- to-face (ETF) contacts (Waters, 2002). Can such contacts be exploited in therapeutic protein engineering? Here, we have analyzed aromatic-aromatic interactions in the insulin hexamer (Blundell, Dodson, et al., 1972) as a basis for designing improved long- acting (basal) analogs. This class of analogs is central to the treatment of Type 1 and

Type 2 diabetes mellitus (Salsali & Nathan, 2006).

Classical crystal structures of insulin hexamers (M. J. Adams et al., 1969; G. Bentley,

Dodson, Dodson, Hodgkin, & Mercola, 1976) immediately suggested a pathway of assembly (Blundell, Dodson, et al., 1972). Pertinent to the mechanism of storage in the secretory granules of pancreatic β-cells (G. Dodson & Steiner, 1998), such assembly is also of pharmacologic importance (Brange & Langkjaer, 1997). Insulin assembly both protects the hormone from degradation in pharmaceutical formulations and modulates its pharmacokinetic properties (Brange, Owens, Kang, & Volund, 1990). Indeed, the first use of insulin analogs in diabetes therapy reflected efforts to destabilize the insulin

212 hexamer and thereby accelerate absorption of monomers and dimers from the subcutaneous (SQ) depot (M Berger et al., 1979; M. R. DeFelippis, Chance, & Frank,

2001)). Because it is more straightforward to introduce unfavorable substitutions than favorable ones, such engineering (corresponding to rapid-acting analogs) was more successful than complementary efforts to enhance the thermodynamic (and kinetic) stability of the insulin hexamer (Brange & Vølund, 1999). There are presently three rapid-acting insulin analogs in clinical use (J. P. Mayer, et al., 2007), but no basal products designed on the basis of enhanced hexamer assembly despite extensive efforts

(J. Markussen, et al., 2003). These difficulties were circumvented by alternative mechanisms of protracted action (acylation and pH-dependent SQ precipitation (J. P.

Mayer, et al., 2007)).

The dimer interface of insulin (repeated three times in the hexamer) contains a cluster of eight conserved aromatic rings (TyrB16, PheB24, PheB25, TyrB26, and their dimer-related mates in subunit D; Figure 5-1A,B,C). Of these, successive ETF contacts are formed by

B16-D26, B24-B26, B24-D24, and B26-D16; the B25 side chain is peripheral to this network. Whereas variation at these sites is in general constrained by the structure of the hormone-receptor interface (J.G. Menting, et al., 2014; Vijay Pandyarajan, et al., 2014b), our attention focused on the B26 side chain because of its functional tolerance to diverse substitutions (Pandyarajan, et al., 2016) and because of its partial exposure in the monomer, dimer, and hexamer (Blundell, Dodson, et al., 1972) (Figure 5-1C, D). We sought to investigate whether variant B16-D26 and B26-D16 ETF contacts across the dimer interface might in principle modulate—in either direction—the strength of these interactions. We hypothesized that enhanced ETF contacts at this interface might provide

213 the long-sought approach to stabilize the insulin hexamer and so improve the pharmacokinetc (PK) properties of basal formulations.

The present study had three parts. The first employed local modeling, using the standard CHARMM empirical energy function, to probe possible effects of a

TyrB26→Trp substitution on aromatic-aromatic interactions within the wild-type (WT) aromatic cluster. These molecular mechanics (MM) calculations suggested that substitution of TyrB26 by Trp could enhance dimer-related ETF contacts and yet otherwise preserve a native-like interface. We next prepared this analog to examine whether this substitution might indeed stabilize the insulin hexamer and retard its disassembly while preserving the biological activity of the monomeric hormone. The crystal structure of a

TrpB26-insulin analog (as a zinc-insulin hexamer) was essentially identical to that of WT insulin. Finally, we undertook studies in diabetic rats to obtain proof of principle that this approach could extend the duration of insulin action on SQ injection.

To our knowledge, our results represent the first exploitation of aromatic-aromatic interactions to enhance the physical and biological properties of a therapeutic protein.

Because standard MM calculations employ a simplified model of aromatic systems (i.e., approximating their quantum-mechanical (QM) properties via partial atomic charges

(Bernard R Brooks et al., 1983; Kramer, Gedeck, & Meuwly, 2012)), ab initio QM simulations of the aromatic cluster and their incorporation in QM/MM simulations (B. R.

Brooks et al., 2009) promise to establish a rigorous foundation for therapeutic protein design, including further optimization through incorporation of modified or non-standard amino acids

214

Figure 5-1 Structural overview of insulin. (A) Sequence of insulin with disulfide bridges in black. TyrB26 is highlighted in black; the present TyrB26→Trp substitution in orange; and modifications in pI-shifted clinical analog glargine in purple. Our semisynthetic pI-shifted analog contained Orn (green) instead of Lys or Arg. (B)

Structure of an insulin monomer; the A chain is shown in black and B chain in green.

TyrB26 is red whereas PheB24 and TyrB16 are blue (PDB: 4INS). (C) Structure of Zn-

B26 B24 B16 coordinated insulin hexamer (T6 state), a trimer of dimers; Tyr , Phe and Tyr are color-coded as in B. (D) Stereo view showing TyrB26 (sticks) in a cavity within insulin

f dimer (extracted from T3R 3 hexamer 1TRZ). (E) Corresponding stick model with residues labeled.

215

216

(El Hage, Pandyarajan, et al., 2016; Lieblich et al., 2017; Nakagawa & Tager, 1986).

The present results suggest that insulin’s conserved aromatic cluster can provide a natural laboratory for such foundational analysis and its therapeutic translation.

5.3 Results

Molecular mechanics calculations suggested that augmented aromatic-aromatic interactions were possible at the TrpB26 dimer interface.

MM simulations were employed to estimate the strength of aromatic-aromatic interactions of TrpB26 at the insulin dimer interface in relation to those of the native Tyr.

These calculations employed the CHARMM empirical energy function in which aromatic rings contain partial atomic charges, parametrized to mimic the electrostatic properties of the  system (Bernard R Brooks, et al., 1983; Kramer, et al., 2012). Working models of the variant dimer were obtained by local energy minimization (see Experimental

Procedures).

Energies of interaction between the eight aromatic residues at the insulin dimer interface (TyrB16, PheB24, PheB25, TyrB26, and their symmetry-related mates) were

B26 f calculated using local models in which Trp was substituted within WT T2, R2, and TR reference dimers ( extracted from PDB structures 4INS, 1ZNJ, and 1TRZ) (Ciszak &

Smith, 1994). The total interaction energy between B26 and the other aromatic residues at the TrpB26 interface, which was calculated using the full CHARMM potential energy function, was augmented by 2.0 and 0.8 kcal/mol relative to the minimized WT interface

f in the context of R2 and TR structures, respectively, and diminished by 1.5 kcal/mol in the context of the T2 structure. Results are summarized in Table 5-1.

217

A minimal model was utilized to further evaluate the potential impact of a TrpB26 substitution on aromatic-aromatic interactions at the dimer interface. To this end, a structural model of the dimer interface (extracted from a T6 insulin hexamer; PDB ID:

4INS) was first built containing residue B26 (TyrB26 or TrpB26) and its nearest aromatic neighbors (PheB24, PheD24, and TyrD16). Simulations predicted the orientation of the B26 ring corresponding to free-energy minima of electrostatic interactions between the aromatic residues (the partial-charge parametrization of aromatic residues in CHARMM is shown in Figure 5-2) (El Hage, Pandyarajan, et al., 2016). When substituted at position B26, Trp displayed improved electrostatic interactions with its three aromatic neighbors (relative to the WT Tyr) over a broad range of conformations (Figure 5-3).

This trend extended to conformations that are sterically permitted in the context of the

WT insulin hexamer.

TrpB26 analog exhibited markedly decreased hexamer dissociation rate.

The effect of the TrpB26 substitution on the lifetime of insulin hexamers under formulation conditions was assessed in the context of OrnB29-insulin, a structural equivalent of WT insulin that is amenable to production by trypsin-catalyzed semisynthesis (Riemen, Pon, 2+ & Carpenter, 1983). The lifetime of Co -substituted, phenol-stabilized (R6) hexamers of TrpB26, OrnB29-insulin was assessed at equilibrium in relation to native TyrB26, OrnB29- insulin. Optical absorbance spectra of these analogs (characteristic of Co2+ with tetrahedral coordination) were similar to WT (Figure 5-4A,B). Assessment of R6 dissociation rates (summarized in 5-2) revealed a 150-fold increase in hexamer half-life of the TrpB26 analog relative to its parent OrnB29-insulin (Fig. 5-4C,D). This increase is remarkable given that the difference between the half-lives of a rapidly

218

Figure 5-2. Comparison of Ab initio electrostatics and CHARMM parameters of

Tyr and Trp side chains. (A) Isosurface representation of electron density and molecular electrostatic potential (MEP) map of Tyr (left) and Trp (right). Electron density and

MEP were calculated using B3LYP method and 6-31G(d) basis set using Gaussian utility

Cubegen on Gaussian 09 (M.J. Frisch et al., 2009). The isosurface map was then generated using Jmol (Willighagen & Howard, 2007). (B) Ball-and-stick models of Tyr

(left) and Trp (right) side chains. Point charges of each atom as implemented in

CHARMM22 are indicated (B. R. Brooks, et al., 2009; Alexander D. MacKerell et al.,

2002).

219

220

Table 5-1a. Interaction energy between the B26 residue of an energy minimized model of TrpB26 insulin relative to native TyrB26 in the context of a TRf dimer.

Residue Interacting Residue Total Energy of Interaction (kcal/mol) 1TRZ TyrB26 1TRZ TrpB26 B26 PheB24 -0.344 -0.502 PheB25 -0.100 -0.113 TyrB16 -0.006 0.007 PheD24 -0.887 -1.465 PheD25 -0.029 -0.120 Tyr/TrpD26 0.000 -0.032 TyrD16 -1.254 -1.672 D26 PheB24 -0.828 -1.436 PheB25 0.005 -0.025 Tyr/TrpB26 0.000 -0.032 TyrB16 -1.594 -0.503 PheD24 -0.524 -0.699 PheD25 -0.036 -0.084 TyrD16 0.006 -0.005 Total -6.986 -7.811

221

Table 5-1b. Local energy-minimized model of TrpB26 insulin relative to native TyrB26 where polar atoms of Tyrosine and Tryptophan rings are replaced with nonpolar synthetic residues.

Residue Interacting Residue Total Energy of Interaction (kcal/mol) 1TRZ TyoB26 1TRZ TroB26 B26 PheB24 -0.358 -0.503 PheB25 -0.090 -0.138 TyrB16 0.006 0.009 PheD24 -0.882 -1.501 PheD25 -0.054 -0.095 TyrD26 0.000 0.001 TyrD16 -1.416 -1.527 D26 PheB24 -0.821 -1.478 PheB25 0.005 -0.039 TyrB26 0.000 -0.001 TyrB16 -0.938 -0.209 PheD24 -0.566 -0.708 PheD25 -0.049 -0.061 TyrD16 0.006 0.003 -6.531 -7.438 a Tabulation of non-bonded interaction energy at the insulin dimer interface in the energy-minimized naïve model of TyrB26→Trp displays interactions most improved by the substitution. b This table demonstrates the impact of π-π interactions on non-bonded interaction energy across the insulin dimer interface. Polar oxygen and nitrogen atoms are replaced, in silico, with non-polar synthetic atoms eliminating the confounding effects of polar interactions involving those atoms on free-energy calculations.

222

Table 5-1c. Interaction energy between the B26 residue of a local energy-minimized

B26 B26 model of Trp insulin relative to native Tyr in the context of an R2 dimer.

Residue Interacting Residue Total Energy of Interaction (kcal/mol) 1ZNJ TyrB26 1ZNJ TrpB26 B26 PheB24 -0.501 -0.571 PheB25 -0.142 -0.146 TyrB16 0.008 0.008 PheD24 -1.032 -1.308 PheD25 -0.002 -0.027 Tyr/TrpD26 -0.011 -0.003 TyrD16 -1.569 -1.996 D26 PheB24 -0.738 -1.283 PheB25 0.004 -0.034 Tyr/TrpB26 -0.011 -0.003 TyrB16 -0.613 -1.326 PheD24 -0.683 -0.635 PheD25 -0.120 -0.145 TyrD16 0.004 0.008 Total -5.406 -7.461

223

Table 5-1d. Interaction energy between the B26 residue of a local energy-minimized

B26 B26 model of Trp insulin relative to native Tyr in the context of an T2 dimer.

Residue Interacting Residue Total Energy of Interaction (kcal/mol) 4INS TyrB26 4INS TrpB26 B26 PheB24 -0.397 -0.549 PheB25 -0.027 -0.058 TyrB16 0.003 0.006 PheD24 -1.405 -1.181 PheD25 -0.008 -0.017 Tyr/TrpD26 -0.008 -0.054 TyrD16 -1.643 -1.868 D26 PheB24 -1.242 -0.679 PheB25 -0.016 -0.022 Tyr/TrpB26 -0.016 -0.054 TyrB16 -2.710 -1.261 PheD24 -0.365 -0.466 PheD25 -0.049 -0.140 TyrD16 0.004 0.006 Total -7.879 -6.337

224

Figure 5-3. Molecular simulations of aromatic interactions in the insulin dimer. (A)

Aromatic-aromatic interactions across insulin's dimer interface involve PheB24, TyrB16,

D24 D26 D26 Phe (sticks) and either Tyr (left) or Trp (right). Residues were extracted from T6 structure 4INS. (B) Contour maps depicting empirical interaction energies between B26

(Tyr on left and Trp on right) at varying χ1 and χ2 angles and the other three residues shown in A. The orientation of TyrB26 in WT crystal structure is indicated by a green “x”; orientation of TrpB26 in naive model is indicated by a green asterisk.

225

226

Figure 5-4. Hexamer dissociation of TrpB26 analog. (A) Tetrahedral Zn2+-coordination

B10 site in R6 insulin hexamer (ball-and-stick model): three His side chains and one chloride ion. (B) Absorbance spectra of d-d bands in corresponding Co2+complex; the color code is indicated. (C) Hexamer dissociation curves as monitored at 574 nm after addition of excess EDTA; the color code is as in B. The lifetime of the TrpB26, OrnB29- insulin hexamer was markedly prolonged (asterisk). (D) Dissociation of TrpB26, OrnB29 hexamer from 0-8000 sec in relation to that of parent OrnB29-insulin (black arrow). Half- lives are given in Table 5-2.

227

228

Table 5-2. Self-association properties of insulin analogs.

a analog t1/2 hexamer dissociation calculated MW by SEC (min ± SD ) (kDa) Wild Type 7.7 (± 1.3) 9.7 Lisprob 4.6 (± 0.3) 5.1 OrnB29 c 8.2 (± 0.8) 8.2 B26 B29 3 Trp , Orn 1.2(± 0.3) x 10 28.0, 4.0

a Proteins were made 0.6 mM in a buffer containing ZnCl2 at a ratio of 2 zinc ions per insulin hexamer and applied to SEC column as described in Methods. Masses were calculated from the plot in Fig. 4B. b “Lispro” describes insulin analogs containing ProB28→Lys and LysB29→Pro substitutions. These substitutions impair dimerization (D. T. Birnbaum, et al., 1997; Brems, et al., 1992). c Use of Orn simplified trypsin-catalyzed semisynthesis (V. Pandyarajan, et al., 2014).

229

Figure 5-5 Size-exclusion chromatography of TrpB26 hexamer. (A) SEC chromatogram of insulin analogs in presence of zinc and phenol. The void volume (V0, black arrow) was defined by thyroglobulin (MW 669 kDa). (B) Plot of log(molecular weight) vs elution ratio (Ve/V0) of molecular weight standards. Linear relationship between log[MW] to elution ratio (Ve /V0) is indicated by the red line with coefficient of

2 determination (R ) 0.996 and parameters log[molecular weight] = -1.71*(Ve /V0)+6.7012.

Elution times of molecular weight standards are indicated by blue squares (labeled by molecular weight). Identity of molecular-weight standards is as follows: 66 kDa, BSA; 45 kDa ovalbumin; 20 kDa, carbonic anhydrase, 17 kDa, myosin light chain; 12.4; cytochrome C, 6.5 IGF-I. Calculated MW are given in Table 5-2.

230

231

Figure 5-6. Size exclusion chromatography (SEC). Elution times of insulin in monomeric and hexameric states in SEC provide context for species identified in Figure

4. (A) SEC profile of monomeric lispro (Zn2+-and phenol-free). The molecular weight

(MW; or molecular mass) of the species was 3.1 kDa (calculated as in Fig. 4). (B) WT insulin formulated with 0.3 mM ZnCl2 and phenol was run in a mobile-phase containing

50 mM cyclohexanol and 0.3 mM ZnCl2. The sample eluted as a hexamer (Calculated molecular mass 48 kDa) with dissociation intermediates constituting the “tail” of the peak. (C) Calibration plot of SEC column with mobile phase used in panel B. Linear fit

(red line) of log(MW) to Ve/V0 of MW standards (blue squares). The equation of the line

2 is log(MW) = -1.79 (Ve/V0) +6.83 (R = 0.986). The following standards were used for calibration: thyroglobulin (669 kDa, V0), ovalbumin (45 kDa), carbonic anhydrase (29 kDa), elastase (26 kDa), ribonuclease A (13.7 kDa), cytochrome C (12.3 kDa) and synthetic peptides (3.6 and 1.2 kDa).

232

233 dissociating analog in clinical use (lispro1 (D. T. Birnbaum, et al., 1997; Brems, et al.,

1992; Roy, et al., 1989)) and WT is <2-fold (Table 5-2).

TrpB26 imposed kinetic barriers to the dissociation of hexamers into monomers.

R6 dissociation kinetics were further examined by size-exclusion chromatography

(SEC). The TrpB26 analog (formulated in the presence of phenol and zinc ions) was injected onto an SEC column using a zinc- and phenol-free mobile phase. Subsequent dissociation of the R6 hexamers was monitored in the chromatograms (Figure 5-5A, Table

5-2). Absence of a void-volume signal (V0) indicated that none of the proteins formed large non-specific aggregates. Whereas WT and OrnB29-insulin eluted as a broad peak representing an association state intermediate between monomer and dimer (9.7 and 8.2 kDa respectively) and whereas insulin lispro eluted essentially as a monomer (5.1 kDa),

TrpB26, OrnB29-insulin eluted in two distinct peaks. The larger peak corresponded to a trimeric or tetrameric association state (MW 28 kDa), and the smaller corresponded to a monomeric state (4 kDa; Figure 5-5A,B; see Figure 5-6)for reference chromatograms of monomer and hexamer). These findings suggest that oligomers intermediate between hexamer and dimer may delay dissociation of the TrpB26 analog; in particular, the absence of broad elution tails implies that TrpB26 imposes significant barriers to rapid dissociation.

TrpB26 analog retained native biological activity.

The in vitro affinity of the TrpB26 analog for the lectin-purified insulin receptor (IR; isoform B holoreceptor) was determined to be 0.14(±0.02) nM, approximately 50% that of OrnB29-insulin (0.07(±0.02) nM) and WT insulin (0.08(±0.03) nM) (Figure 5-7A). The potencies of these analogs, evaluated by intravenous (IV) injection in diabetic rats, were nonetheless indistinguishable from WT insulin (Figure 5-7B).

234

TrpB26 analog displayed a zinc-dependent delay in onset of biological activity.

Hexamer assembly delays absorption of insulin from its SQ injection site (M Berger, et al., 1979; M. R. DeFelippis, et al., 2001). To assess the onset and duration of TrpB26,

OrnB29-insulin relative to OrnB29-insulin, the pharmacodynamics (PD) profile of these proteins (made 0.15 mg/ml, corresponding to a monomer concentration of 27 µM and a putative hexamer concentration of 4.5 µM) were evaluated as zinc-free solutions or as pre-assembled phenol-stabilized R6 hexamers in the presence of excess zinc ions (0.30 mM ZnCl2; 70 zinc ions per hexamer). A zinc-dependent delay in onset of activity was observed on SQ injection of TrpB26, OrnB29-insulin but not on injection of OrnB29-insulin or WT insulin (Fig. 5-7C; See Figure 5-8 for WT results). Whereas the latter PD profiles exhibited a nadir at ca. 120 min irrespective of zinc-ion concentration, the PD profile of

TrpB26, OrnB29-insulin occurred at (i) 120 min in the absence of zinc ions and (ii) 150 min in the presence of 0.30 mM zinc ions with corresponding delays in rate of fall over the first 30 min (Figure 5-7). Together with the

B26 above in vitro results, these findings suggest that the prolonged lifetime of the Trp R6 hexamer (as inferred from the above kinetic studies of the Co2+ -substituted hexamer) are responsible for the inferred zinc-dependent delay in SQ absorption.

TrpB26 protracted the PD profile of a model pI-shifted analog.

Insulin analogs with isoelectric points (pI) shifted to neutral pH generally exhibit prolonged activity due to precipitation in the SQ depot (Gillies, et al., 2000). To determine whether TrpB26 might further prolong the activity of such analogs, this substitution was introduced into a GlyA21, OrnB29, OrnB31, OrnB32-insulin. This “glargine- like” framework was designed to recapitulate the pI shift of glargine with greater ease of

235

Figure 5-7 Pharmacology of TrpB26 analog. (A) Receptor-binding affinities (isoform

B). The affinity of TrpB26, OrnB29-insulin was reduced by twofold relative to OrnB29- insulin (respective Kd estimates 0.14(±0.03) and 0.07(±0.02) nM). Color code is inset. (B)

Time-course of [blood glucose] following IV injection in rats (N=15); color code as in A.

(C) Time-course of [blood glucose] following SQ injection in absence or presence of 0.3 mM ZnCl2 (N=18). (D) Histogram summarizing rate of fall of [blood-glucose] over first

30 min in panel C, black bars indicate S.D. (E) Time-course of [blood glucose] following

SQ injection of pI-shifted analogs: GlyA21, OrnB29, OrnB31, Orn32-insulin and its

TrpB26 derivative (N=6; color code in panel).

236

237

Figure 5-8. Effects of Zn2+ on PD profile of WT insulin. (A) Time course of [blood glucose] after SQ injection of WT insulin formulated in absence of Zn2+ (blue; N=8) or in presence of 0.3 mM ZnCl2 (black; N=9). (B) Normalized curves from panel A.

238

239

2 semisynthesis . The proteins (formulated at 0.6 mM with 0.3 mM ZnCl2, corresponding to 3 zinc ions per hexamer) were each injected SQ in diabetic rats.

The pI-shifted parent analog displayed peak activity at ca. 120 min with blood-glucose levels returning to baseline after about 360 min. By contrast, its TrpB26 derivative displayed a prolonged PD profile: peak activity occurred 180 min with slow return to baseline >800 min (Fig. 5-7E; See Figure 5-9 for IV potency). Such a marked delay in peak activity was not observed in the parent glargine-like analog or the TrpB26 derivative when administered in the absence of zinc (Figure 5-10). These results suggest that TrpB26 may favorably be incorporated into current basal analogs as a complementary mechanism of prolonged SQ absorption.

Crystal structure of TrpB26 analog demonstrated native-like dimer interface.

The crystal structure of TrpB26, OrnB29-insulin was determined as a zinc-coordinated hexamer in the presence of phenol to a resolution of 2.25 Å. Diffraction and refinement statistics are provided in Table 5-3. The asymmetric unit constituted a “TRf” dimer3

(Chothia, Lesk, Dodson, & Hodgkin, 1983; G. G. Dodson, et al., 1993). The overall structures of the T- and Rf protomers were essentially identical to those of WT insulin

(Figure 5-11) with respective RMSDs 1.11 ± 0.30 and 1.36 ± 0.30. Additional RMSDs are given in Tables 5-4 and 5-5. Side-chain packing near the B26 position was largely unperturbed. In both protomers the TrpB26 indole group was oriented with its six-member ring packing against conserved core residues IleA2, ValA3, and ValB12 (Figure 5-12); the

f B26 indole NH group is exposed to solvent in the TR dimer and T3R 3 hexamer. The Trp side chain in both R- and T protomers also displayed dihedral angles within the range of the native Tyr in WT crystal structures (Tables 5-6 and 5-7, respectively) with a slight

240

Figure 5-9. In vivo potency of pI-shifted insulin analogs. (A) Time course of [glucose] after IV injection of parent pI-shifted insulin analog (blue; N=6) and its TrpB26 derivative

(red; N=6). (B) Bar graph showing area over curve (AOC) of curves from panel A. Black bars indicate S.E.M. The TrpB26 derivative displayed 82 ± 6% potency relative to the parent analog but is a complete agonist on injection of higher doses.

241

242

Figure 5-10 Effects of Zn2+ on the PK/PD Profile of pI-shifted insulin analogs. (A)

Time course of [blood glucose] after SQ injection of zinc-free parent pI-shifted insulin analog (blue; N=6) and TrpB26 derivative (red; N=6). Normalized data are shown in B.

(C-D) Corresponding plots of analogs administered after formulation in presence of 0.3 mM ZnCl2, color-coded as above (N=6).

243

244

Table 5-3. Data collection and refinement statistics pertaining to the crystal structure of TrpB26, OrnB29-insulin.

Data Collection Source SSRL BL7-1 wavelength (Å) 0.9795 space group R3 Unit-cell dimensions a = 79.7552 Å, b = 79.7552 Å, c = 37.61 Å, α = 90°, β = 90°, γ = 120° resolution (Å) 33.04-2.25 Rmerge (%) 3.698 (10.28) I/σI 45.78 (8.99) completeness (%) 99.5 (99.8) redundancy 10.1 (9.6) no. of reflections/used in refinement 4273/4226 Refinement resolution (Å) 33.04-2.25 R-factor/RFree (%) 16.8/24.9 r.m.s.d. bond length (Å) 0.036 r.m.s.d. bond angle (°) .971 median B-factors (Å2) chain A 28 chain B 22 chain C 27 chain D 18 Ramachandran plot (%) most favored region 100 additionally allowed regions 0 generously allowed regions 0 disallowed regions 0

245

Figure 5-11. Comparison of the structure of TrpB26, OrnB29-insulin to a collection of WT insulin structures. (A) Rf-state protomer of TrpB26, OrnB29-insulin. The A chain is shown in red, and the B chain in blue (B1-B8) or green (B9-B30). (B) Rf-state protomer of TrpB26, OrnB29-insulin (sticks, color coded as above) in relation to extensive set of crystal structures of insulin and insulin analogs (PDB entries: 1BEN, 1G7A,

1RWE, 1EV3, 1EV6, 1MPJ, 1TRZ, 1TYL, 1MPJ, 1ZEG, 1ZNJ and 1ZNI; gray sticks).

Structures are aligned with respect to the main-chain atoms of residues A1-A21 and B3-

B28. RMSD values are given in Table 5-4. (C-D) Corresponding representations of the

T-state protomer of TrpB26, OrnB29-insulin, color code as above. PDB entries used for alignment are as follows: 1APH, 1DPH, 1BEN, 1MPJ, 1TRZ, 1TYL, 1TYM, 1RWE,

1G7A, 1ZNI, 2INS and 4INS. RMSD values are given in Table 5-5.

246

247

Figure 5-12 Crystal structure of TrpB26, OrnB29-insulin. (A) Electron density of

TrpB26 in T-state protomer showing surrounding density in TRf asymmetric unit (contour level 2.0 Å). (B) Stick model corresponding to map in A; TrpB26 is orange. (C) Surfaces of residues surrounding TrpB26 (sticks) as indicated above. (D-F) Corresponding map and models of TrpB26 in the R-state protomer.

248

249

Table 5-4. RMSD values of main-chain and side-chains of the T-state protomer of

B26 B29 the crystal structures of Trp , Orn -insulin and selected WT T6 and T3R3 hexamers

PDB ID Main-chain RMSD side-chain RMSD (Å) (Å) 1RWE T (T3R3) 0.302 1.18 1BEN T (T3R3) 0.289 0.67 1TRZ T (T3R3) 0.316 1.09 1TYL T (T3R3) 0.319 1.38 1TYM T (T3R3) 0.295 1.07 1ZNI T (T3R3) 0.283 0.68 1MPJ T (T3R3) 0.382 0.84 3MTH T (T3R3) 0.318 0.94 1LPH T (KP T3R3) 0.486 0.95 1G7A T1 0.307 0.75 a (rhombohedral T3R3) T2 0.334 1.16 4INS T1 (2ZN T6) 1.127 1.94 T2 0.686 1.49 b 1APH (cubic T2) 0.620 1.00 1BPH (cubic T2) 0.653 1.12 1CPH (cubic T2) 0.667 1.32 1DPH (cubic T2) 0.659 1.21 Average 0.47 ± 0.11 1.11 ± 0.30 a Rhombohedral crystals contain two unique T-state protomers in the asymmetrical unit that are related by 3-fold symmetry to the two other TR dimers comprising an T3R3 hexamer. b Cubic T2 crystals contain two T-state protomers forming an insulin dimer. They are zinc-free crystals.

250

Table 5-5. Comparison of χ1 and χ2 dihedral-angle values of the T protomer of the

B26 B29 Trp , Orn -insulin crystal structure to the B26 side chains of WT T6 or T3R3 insulin crystal structuresa

PDB ID chain χ1 χ2 TrpB26, OrnB29- B 167.131 -108.457 insulin 1TRZ B -179.531 78.679b 4INS B 170.768 66.198 D 175.602 -113.722 2INS B 170.867 63.894 D 170.708 -111.824 1ZNI D 178.670 67.486 1BEN B 178.543 67.578 D 172.629 60.183 3MTH D 178.321 70.378 a B26 Tabulation of χ1 and χ2 angles demonstrates the native-like orientation of the Trp ring of the T protomomer of the TrpB26, OrnB29-insulin crystal structure. b Due to the symmetric structure of the Tyr side chain, χ2 and χ2-180° are equivalent. For example, 78.679° is equivalent to -101.321°.

251

Table 5-6. Comparison of χ1 and χ2 dihedral angle values of the R protomomer of

B26 B29 f Trp , Orn -insulin crystal structure to the B26 side chains of WT R6 or R insulin crystal structuresa

PDB ID chain χ1 χ2 WB26 D -179.108 -102.751 1TRZ D 179.520 81.505b 1ZNJ B -173.898 82.727 D 168.865 -90.454 F -176.674 69.837 H 176.540 74.677 J -173.848 80.139 L 175.389 72.440 B -178.950 77.319 1BEN D -162.232 84.877 1MPJ B -177.306 -89.456 3MTH B -173.822 -87.442 a B26 Tabulation of χ1 and χ2 angles demonstrates the native-like orientation of the Trp ring of the R protomomer of the TrpB26, OrnB29-insulin crystal structure. b Due to the symmetric structure of the Tyr side chain, χ2 and χ2-180° are equivalent.

252 deviation in the positioning of the peptide backbone to accommodate the larger indole side chain.

Spectroscopic probes revealed native-like structure and thermodynamic stability of

TrpB26 analogs in solution.

The native-like crystal structure of TrpB26, OrnB29-insulin is in accordance with its unperturbed circular dichroism (CD) spectrum and thermodynamic stability under monomeric conditions (Figure 5-13A). Free energies of unfolding (ΔGu 3.3 ±

0.1) kcal/mole at 25 °C as inferred from two-state modeling of chemical denaturation (V.

Pandyarajan, et al., 2014)) were indistinguishable due to small and compensating changes in transition midpoint and slope (m value) (V. Pandyarajan, et al., 2014; Sosnick, et al.,

2000) (Fig. 5-12B, Table 5-8). Further evidence that the crystal structure extends tothe monomer in solution was provided by 2D 1H-NMR studies of TrpB26 substituted within an engineered insulin monomer (lispro (Holleman & Hoekstra, 1997)). Whereas the spectrum of lispro (at pD 7.6 and 37 oC) exhibits sharp resonances for each aromatic spin system (Figure 5-14A), as expected for a monomeric analog (Holleman & Hoekstra,

1997), the spectrum of its TrpB26 derivative exhibits broadening of resonances at the dimer interface (B16, B24-B26). The latter spin systems can be observed on TOCSY spectrum (Fig. 8C) but not in the corresponding DQF-COSY spectrum due to antiphase cancellation. Like the aromatic ring TyrB26 in spectra of insulin lispro

(Figure 5-14A,B), the indole ring exhibited regiospecific nonlocal nuclear Overhauser enhancements (NOEs) from its six-member moiety to the methyl resonances of

ValB12 and IleA2 (Figure 5-14B,C,D).

253

Table 5-7. Thermodynamic stabilities of insulin analogs.

a b Analog ΔGu Cmid m (kcal mol-1) (M) (kcal mol-1 M-1) Wild-type 3.4 ± 0.1c 5.0 ± 0.1 0.68 ± 0.02 OrnB29 3.3 ± 0.1 4.9 ± 0.1 0.67 ± 0.01 TrpB26, OrnB29 3.3 ± 0.1 5.1 ± 0.2 0.64 ± 0.03 a Parameters were inferred from CD-detected guanidine denaturation data by application of a two-state model; uncertainties represent fitting errors for a given data set. b The m-value (slope Δ(G)/Δ(M)) correlates with surface area exposed on denaturation. c Analysis of replicates of TrpB26, OrnB29-insulin, parent OrnB29-insulin, and WT samples indicated that experimental standard errors were equal to or less than the above fitting errors: ±0.1 kcal mol-1 (ΔGu), ±0.1 M (Cmid), and ±0.01 kcal mol-1 M-1 (m).

254

Figure 5-13 Thermodynamic Stability of TrpB26, OrnB29-insulin. (A) CD spectra of

TrpB26, OrnB29-insulin, OrnB29-insulin, and WT insulin. (B) Guanidine denaturation assays of insulin analogs monitored by ellipticity at 222 nm; color code as in panel A. Stabilities are given in Table 5-7.

255

256

The pattern of secondary shifts in the variant is similar to that in the parent monomer.

In particular, the aromatic 1H-NMR resonances of TrpB26 (red cross peaks in Figure 5-

14C) exhibit upfield features (relative to Trp in the isolated B23-B30 octapeptide; dashed lines) similar to those of TyrB26 in the parent spectrum (purple cross peaks in Figure 5-

14A versus dotted lines) (Q. X. Hua, et al., 2011). Dilution of the TrpB26 sample partially mitigated resonance broadening but preserved these trends in dispersion. Indole-specific

NOEs indicated that the side chain assumes one predominant and asymmetric conformation within a native-like crevice between A- and B-chain α-helices (Table 5-8).

B26 Because Tyr undergoes rapid ring rotation about the Cβ-Cγ bond axis ("ring flips"), analogous side-chain specific NOEs (inferred in prior studies from molecular modeling) cannot be observed directly (Table 5-9).

MM calculations suggested improved aromatic-aromatic interactions within the variant crystal structure.

The contribution of aromatic-aromatic interactions involving TrpB26 to the stability of

f the variant dimer interface of the T3R hexamer was evaluated through calculation of non- bonded interaction energies among aromatic residues B16, B24, B25, and B26 in the TRf dimer. These calculations, which employed the variant crystal structure, were in overall accordance with expectations based on our initial local MM-based modeling (above). In particular, based on aromatic-aromatic interactions alone, the TrpB26, OrnB29 dimer displayed an increase in interaction energy of 1.4 kcal/mol relative to WT TRf reference structure 1TRZ; the results of these calculations are given in Table 5-10a,b. Although the standard CHARMM empirical energy function, when applied to analyze either crystallographic and MM-minimized models of TrpB26 insulin, suggested that the

257

Figure 5-14. Homonuclear 2D NMR of a TrpB26 analog. 2D NMR studies of insulin analogs indicate similar TyrB26- and TrpB26 environments. (A, B) Spectra of parent monomer insulin lispro (LysB28, ProB29-insulin): A, aromatic region of TOCSY spectrum with TyrB26 cross peaks (magenta) shown relative to Tyr spin system in free octapeptide

GFFYTKPT (dotted lines); and B, region of NOESY spectrum showing contacts between aromatic protons (vertical axis, ω2) and methyl groups (horizontal axis, ω1). (C, D)

Spectra of TrpB26 analog of insulin lispro: C, aromatic TOCSY spectrum highlighting TrpB26 cross peaks (red) relative to Trp spin system in free octapeptide

GFFWTKPT (dashed lines) and D, region of NOESY spectrum corresponding to B. B26-

A3 related NOEs are shown in red. Cross-peak assignments: (a) γ-CH3 Val , (b) γ-CH3

B12 A2 B15 A3 B12 Val , (c) γ-CH2, γ-CH3 Ile , (d) δ-CH3 Leu , (e) γ-CH3 Val , (f) γ-CH3 Val , (g) γ-

A2 A2 B15 CH2, γ-CH3 Ile , (h) δ-CH3 Ile , and (i) δ-CH3 Leu . TOCSY mixing times in spectra

A and C were 55 ms; NOESY mixing times in B and D were 150 ms.

258

259

Table 5-8. Comparison of predicted proton-proton distances involving TrpB26 from

TrpB26, OrnB29-insulin crystal structure and TrpB26-associated NOEs in TrpB26 lispro

Trp Residue Proton Predicted NOE

Proton Distancea (Å) Strengthb A2 H-ε3 Ile H-δ1 3.82 w H-ζ3 ValB12 H-γ1c 6.93 s H-ζ3 ValB12 H-γ2 4.61 s H-η2 ValB12 H-γ1 7.61 m H-η2 ValB12 H-γ2 4.68 m B12 H-ζ2 Val H-γ1 7.38 w H-ζ2 ValB12 H-γ2 4.40 w H-ε3 ValB12 H-γ1 8.85 w H-ε3 ValB12 H-γ2 4.24 w H-ε3 LeuB15 H-δ1 3.24 w H-ε3 LeuB15 H-δ2 6.04 w

H-ζ3 LeuB15 H-δ1 4.74 w H-ζ3 LeuB15 H-δ2 6.85 w a Predicted distances were obtained from the AB (T-state) protomer of the TrpB26, OrnB29-insulin crystal structure. b NOEs were categorized as strong (s, <4.0 Å), medium (m, 4.4-5.5 Å), and weak (w, >5.5 Å). c Predicted proton-proton distances are provided for protons at equivalent carbon positions although associated NOEs are difficult to differentiate.

260

Table 5-9. Comparison of predicted proton-proton distances involving TyrB26 from

WT crystal structure and TyrB26-associated NOEs from lispro NMR spectrum.

Tyr Residue Proto Predicted NOE Proton n Distancea (Å) Strengthb H-ε IleA2 H-δ1 5.17 w H-ε ValA3 H-γ1c 6.82 w H-ε ValA3 H-γ2 4.05 w A3 H-δ Val H-γ1 4.12 w H-δ ValA3 H-γ2 6.25 w B11 H-ε Leu H-β 4.61 w H-ε LeuB11 H-γ 3.13 w H-ε LeuB11 H-δ1 5.42 w H-ε LeuB11 H-δ2 3.67 w H-δ LeuB11 H-δ1 5.75 w B11 H-δ Leu H-δ2 5.24 w H-ε ValB12 H-γ1 7.21 w H-ε ValB12 H-γ2 4.72 w H-ε LeuB15 H-δ1 5.12 w H-ε LeuB15 H-δ2 7.80 w B15 H-δ Leu H-δ1 3.14 m H-δ LeuB15 H-δ2 6.14 w H-δ PheB24 H-ε 4.55 w B27 H-δ Thr H-γ2 6.88 w a Predicted distances were obtained from the AB (T-state) protomer of WT structure 1TRZ. b NOEs were categorized as strong (s, <4.0 Å), medium (m, 4.4-5.5 Å) or weak (w, >5.5 Å). c Predicted proton-proton distances are provided for protons at equivalent carbon positions although associated NOEs are difficult to differentiate.

261 electrostatic properties of the Trp side chain were the primary contributors to the increased stability of the dimer, this physical interpretation may reflect the limitations of the partial-charge representation (El Hage, Bereau, Jakobsen, & Meuwly, 2016; El Hage,

Pandyarajan, et al., 2016). Indeed, preliminary ab initio QM simulations of a minimal model (consisting of two aromatic rings in vacuo) predict that enhanced Van der Waals interactions may also make a significant contribution (Figure 5-15) (see Discussion).

5.4 Discussion

The physical origins of protein stability and recognition define a foundational problem in biochemistry (Sikosek & Chan, 2014) with central application to molecular pharmacology (Baron & McCammon, 2013). The zinc-insulin hexamer provides a favorable system for structure-based design due to its long history of crystallographic investigation (Blundell et al., 1971). Indeed, the hexamer’s rigidity, as interrogated by

NMR spectroscopy (X. Chang, Jorgensen, Bardrum, & Led, 1997), renders the overall structure robust to diverse amino-acid substitutions (Jacoby, et al., 1996 ; Keller,

Clausen, Josefsen, & Led, 2001), even those that destabilize the dimer interface4 (Brange, et al., 1988; G. G. Dodson, et al., 1993). This rigid framework has often enabled analysis of discrete interactions without complications due to the long-range transmission of conformational change (Chothia, et al., 1983; G. G. Dodson, et al., 1993).

Our studies, stimulated by the seminal recognition of aromatic-aromatic interactions by Burley and Petsko more than 30 years ago (S. K. Burley & Petsko, 1985), focused on the classical dimer interface, a basic building block of the hexamer (G. G. Dodson, et al.,

1993). Long appreciated as "a thing of beauty" (Blundell, Dodson, et al., 1972; Weiss &

Lawrence, 2018), this interface contains eight aromatic residues, six of which engaged in

262

Figure 5-15. Ab initio calculations of energy of interaction between pairs of isolated aromatic molecules. Phenol-phenol (top), phenol-benzene (middle), and phenol-indole

(bottom). The phenol-indole pair was determined to form the most stable ETF interaction as a result of Van der Waals forces. Interaction energies were calculated using MP2 method and aug-cc-pVDZ basis set using Gaussian 09 (M. J. Frisch et al., 2009).

263

264 a successive set of aromatic-aromatic interactions. Quantum-chemical simulations of model systems have suggested that nearest-neighbor interactions predominate even in the presence of multiple rings (Sinnokrot & Sherrill, 2006). Pairwise dissection of insulin’s dimer interface has highlighted the potential opportunity to enhance its stability through substitution of TyrB26 by Trp. Whereas our crystallographic analysis verified that this substitution preserves native architecture, a TrpB26 insulin analog exhibited a dramatic increase in hexamer lifetime in vitro. Results of animal testing demonstrated native intrinsic potency (i.e., on IV bolus injection) but with prolonged activity on SQ injection, presumably due to delayed dissociation of the variant zinc hexamers in the SQ depot.

Protein engineering of insulin analogs is constrained by the complexity of insulin’s

“conformational lifecycle”: from oxidative folding intermediates and self-assembly in the pancreatic β-cell (G. Dodson & Steiner, 1998) to adoption of an active, “open” conformation on receptor binding (Figure 5-16A,B) (J.G. Menting, et al., 2014). Specific residues may play distinct roles at each stage. In particular, because interfaces within the insulin hexamer overlap the hormone’s receptor-binding surface—essentially invariant among vertebrates (J. M. Conlon, 2001)—modifications often impair activity (Brange &

Vølund, 1999). A given WT residue may represent a compromise among competing structural tasks.

A recent survey of 18 substitutions at position B26 demonstrated that Tyr is suboptimal with respect to IR-binding affinity but enhances self-assembly relative to more active alternatives (Ala, Ser or Glu) (Pandyarajan, et al., 2016). The latter side chains destabilize the “closed” dimer interface but are favorable at the solvated B26- related edge of the open hormone-receptor interface (Figure 5-16C).

265

Figure 5-16 Binding surface of TyrB26 on an IR fragment. (A) model of WT insulin (in classical T-state) overlaid on structure of insulin bound to an IR fragment (PDB entry

4OGA). The L1 domain and part of CR domain are shown in powder blue whereas αCT is shown in purple. PheB24 and TyrB26 are respectively shown as gray and red sticks. The

B chain of IR-bound insulin is shown in dark gray (B6-B19) or black (B20-B27); the green tube indicates classical location within overlay of residues B20-B30 (green arrow), thereby highlighting steric clash of B26-B30 with αCT. Insertion of the insulin B20-B27 segment between L1 and αCT was associated with a small rotation of the B20-B23 β-turn and changes in main-chain dihedral angles flanking B24. (B) Stick representation of B- chain residues B20-B27 packed between αCT and the L1 β2 strand. Color code in insulin segment: carbon atoms (green), nitrogen (blue) and oxygen (red). Residues B8-B19 are shown as a black ribbon, and the A chain is shown as a yellow ribbon. Key contact surfaces of αCT with B24-B26 are highlighted in magenta and of L1 with B24-B26 are highlighted in cyan. (C) Stereo view of environment of TyrB26 within its binding site.

Neighboring side chains in L1 and αCT are as labeled. This figure was adapted from

(Pandyarajan, et al., 2016) with permission of the authors.

266

267

Protective self-assembly is of key pharmacologic importance.

Insulin self-assembly protects the hormone from degradation and toxic misfolding in pancreatic β-cells5 (G. Dodson & Steiner, 1998; Jekl, et al., 2011) and in pharmaceutical formulations (Brange & Havelund, 1984). Because the zinc insulin hexamer exhibits delayed SQ absorption relative to monomers and dimers, mutational destabilization of the hexamer (M Berger, et al., 1979; Brange, et al., 1990) led to development of rapid-acting insulin analogs (J. P. Mayer, et al., 2007). Efforts to stabilize the insulin hexamer—as a converse strategy to obtain protracted action—were less successful (Brange & Vølund,

1999). Current long-acting insulin analogs rely instead on higher-order self-association of hexamers within the SQ depot (Gillies, et al., 2000) and binding acylated monomer to albumin as a circulating depot (I. Jonassen, et al., 2012; Peter Kurtzhals, et al., 1995).

The problem of how to improve a protein interface is in general more subtle than its opposite, for destabilizing substitutions abound at conserved interfaces whereas stabilizing substitutions can be rare (Brange & Vølund, 1999). Structure-based candidate substitutions may encounter entropy-enthalpy (EEC) compensation (Duane T Birnbaum, et al., 1996; Chodera & Mobley, 2013) or cause unintended biological perturbations (Q.

X. Hua, Nakagawa, et al., 2006). The challenge posed by insulin is magnified by the structural elegance of its self-assembly (as emphasized by Hodgkin and colleagues in a classic review) (Blundell et al., 1972). Diverse structure-based strategies were previously undertaken with only limited success. Alternate strategies previously used to create basal insulin analogs are summarized in Table 5-11. One approach focused on the general nonpolar character of dimer interface: additional hydrophobic substitutions were introduced in an effort to enhance this feature (J. M. Markussen, et al., 1987). Such

268 designs were not successful and also impaired biological activity, although analogs were identified whose sparing solubility slowed SQ absorption (J. M. Markussen, et al., 1987).

A second approach exploited the classical TR transition among insulin hexamers

(Michael A Weiss, 2009). At the pivot point of this allosteric transition, an invariant glycine (GlyB8) was substituted by Ser in an effort to stabilize the more stable R-state hexamer (J. Markussen et al., 1987). This analog was unstable as a monomer (Q. X. Hua,

Nakagawa, et al., 2006) and exhibited reduced activity (J. Markussen, et al., 1987). Yet another approach sought to stabilize the hexamer by relieving electrostatic repulsion created by the internal clustering of six acidic side chains (GluB13): their isosteric substitution by Gln indeed promoted assembly of zinc-free hexamers but impaired biological activity (G. A. Bentley, et al., 1992). Overlap between the self-assembly surfaces of insulin and its receptor-binding surfaces thus compounded the optimization problem.

The present study revisited the architecture of the insulin hexamer in light of recent insights into the hormone’s receptor-binding surface (Pandyarajan, et al., 2016).

Prominent roles are played by a quartet of aromatic residues in the C-terminal B-chain β- strand: TyrB16, PheB24, PheB25 and TyrB26 (V. Pandyarajan, et al., 2014). These four residues—and the clustering of eight dimer-related side chains—have long been the focus of structure-activity relationships (Mirmira, et al., 1991). Photo-activatable aromatic probes (para-azido-Phe) at any of these sites exhibited efficient cross-linking to the IR

(Huang, et al., 2004; Xu, Hu, Chu, Huang, et al., 2004). The co-crystal structure of an insulin monomer bound to a fragment of the IR ectodomain (the μIR model) revealed distinct binding sites at the surface of the L1 domain of the IR β-subunit (B16, B24 and

269

B26) or its αCT element (B24 and B25). PheB24 packs within a nonpolar pocket near aromatic residues in L1 (residues Leu37, Phe39) and αCT (residue Phe714); one wall of this pocket is defined by the aliphatic side chains of LeuB15, CysA20, and CysB19 as in free insulin (J.G. Menting, et al., 2014). This environment differs in detail from those in the insulin dimer but exhibits analogous general features. Although aromaticity is not required to fill the B24-binding pocket (V. Pandyarajan, et al., 2014), its specific size and shape constrain potential substitutions. TyrB16 lies at the periphery of L1 (near Phe39 and

Lys40). Its substitution by Ala or other aromatic residues preserves activity (Hu, Burke, &

Katsoyannis, 1993). PheB25 occupies a cleft between αCT residues (Val715, Pro716, Arg717, and Pro718) that can only accommodate trigonal γ-carbons (Nakagawa & Tager, 1986).

B25-related aromatic-aromatic interactions are limited due to the peripheral location of this residue.

We chose to focus on position B26 because of its broad functional tolerance of diverse substitutions (Pandyarajan, et al., 2016). As illustrated above (Figure 5-

15C), TyrB26 binds at the solvated periphery of the μIR interface. Indeed, substitution of small, polar, or charged amino acids, (such as Ala, Ser, or Glu) enhances receptor affinity

—but at the price of impaired self-assembly and decreased thermodynamic stability with heightened susceptibility to physical degradation (J.G. Menting, et al., 2014). These findings highlighted the evolutionary importance of the native dimer interface and dual role of TyrB26. We thus hypothesized that an aromatic substitution at the B26 position might enhance self-assembly without loss of biological activity.

The classical structure of the insulin dimer motivated study of successive aromatic- aromatic interactions as a physical mechanism of stability

270

Table 5-10a. Energy of non-bonded interaction of TyrB26 with local aromatic residuesa

Protomer Residue Interaction Energy

(kcal/mol) B26 Tyr B24F -0.323 B25F -0.111 B16Y 0.007 D24F -1.010

D25F -0.014 D26Y 0.000 D16Y -1.310 TyrD26 B24F -1.063 B25F 0.001 B26Y 0.000 B16Y -1.391 D24F -0.443 D25F -0.033 D16Y 0.006 Total -7.252

a As calculated from WT structure 1TRZ.

271

Table 5-10b. Energy of non-bonded interaction of TrpB26 with neighboring aromatic residuesa

Protomer Residue Interaction Energy

(kcal/mol) B26 Trp B24F -0.375 B25F -0.162 B16Y 0.009 D24F -1.266

D25F -0.023 D26Y 0.015 D16Y -1.411 D26 Trp B24F -1.517 B25F -0.013

B26Y 0.015 B16Y -2.073 D24F -0.733 D25F -0.073 D16Y 0.007 Total -8.670 a Table (calculated from TrpB26 OrnB29 structure) show specific aromatic- aromatic aromatic interactions improved in the TR dimer of TrpB26 OrnB29-insulin in relation to WT insulin.

272

Table 5-11. Summary of previous designs of basal insulin analogsa

Modification Mechanism Drawbacks Human Proinsulin C-peptide slows absorption, Safety concerns regarding prolongs activity (Galloway hyperproinsulinemia and et al., 1992) cardiovascular side effects Not a truly basal insulin (Galloway, et al., 1992)

Neutral Protamine Hagedorn Forms crystals with partner Consistent dosing difficult / Neutral Protamine Lispro molecule protamine (Lauritzen, Pramming, Gale, (Hagedorn, et al., 1936) Deckert, & Binder, 1982) Intermediate acting (Owens & Bolli, 2008) Can cause allergic reactions (Stewart, Mcsweeney, Kellett, Faxon, & Ryan, 1984) GlnB13 Forms zinc-free hexamers Dramatically reduced (G. A. Bentley, et al., 1992) biological activity (J. M. Markussen, et al., 1987) LysB27 Positive charge of stabilizes Effective only in complement packing of multiple hexamers with other modifications (J. (J. Markussen, et al., 1987) Markussen, et al., 1987) Shifts pI of insulin GlnB13 GlnA17 LysB30 Iso-electric precipitation Dramatically reduced Stabilization of Hexamer biological activity (J. Markussen, et al., 1988) GlyA21,ArgB27,ThrB30-NH2 Iso-electric precipitation in Macrophage invasion and subcutaneous depot, GlyA21 inflammation at injection site prevents chemical (Peter Kurtzhals, et al., 1995; degradation in acidic J Markussen, et al., 1996) formulations (J. Markussen, et al., 1988) GlyA21 ArgB30ArgB31 Iso-electric precipitation in Reports of association with subcutaneous depot, GlyA21 neoplasms (Suissa et al., prevents chemical 2011) degradation in acidic PK/PD profile less formulations predictable than other basal analogs (Yamamoto et al., 2016) Unknown effect of enzymatic degradation products (Sommerfeld et al., 2010; Suissa, et al., 2011) Myristic acid-coupled LysB29 Slows absorption from SubQ Reduced potency (Peter 273

depot, binds to albumin Kurtzhals, et al., 1995) (Peter Kurtzhals, et al., 1995) Hexadodecandioic acid- Slows absorption from SubQ Reduced potency, has not yet coupled LysB29 depot, forms multihexameric been used long-term complexes, binds to albumin clinically(Gough, et al., (Gough, et al., 2013; I. 2013) Jonassen, et al., 2012)

274

(M. J. Adams, et al., 1969; E. N. Baker, et al., 1988). The increased stability of ETF aromatic-aromatic interactions involving Trp over those involving Tyr contributes to the increased stability of the TrpB26 hexamer. The larger size of the delocalized π orbital of

Trp in relation to that of Tyr causes a stronger negative charge to accumulate on the

“face” of the indole ring. For this reason, hydrogen atoms surrounding the aromatic rings of local residues form stronger electrostatic interactions with the face of the indole ring of

Trp than with the phenol ring of Tyr (Guvench & Brooks, 2005; Samanta, Pal, &

Chakrabarti, 1999). Aromatic pairs involving Trp residues are less common than those involving Tyr or Phe (S. K. Burley & Petsko, 1988). However, the strength of aromatic- aromatic interactions involving Trp is evidenced by functional importance of Trp-based interactions; an example is provided by a Trp-Tyr aromatic “lock” that stabilizes the active conformation of the ghrelin receptor (B. Holst et al., 2010). Indeed, comparison of electrostatic interactions of TrpB26 and TyrB26 within the minimized model of the

B16/B24/B26 aromatic network of insulin revealed that interactions involving Trp were favored over those involving Tyr across a broad variety of orientations of the respective aromatic rings. Extension of the aromatic "lock" metaphor (introduced by Holst, B. et al. to describe conformational “trapping” in GPCR structure (B. Holst, et al., 2010)) to the insulin hexamer highlights the kinetic effect of the B26 substitution on the rate of hexamer disassembly (as probed by the Co2+-EDTA sequestration assay), which was more dramatic than effects on equilibrium association (as probed by SEC). It would be of future interest to measure activation energies for disassembly. Insight into the structural origins of the prominent TrpB26-associated kinetic lock may be provided by activated molecular-dynamics simulations of hexamer disassembly.

275

TrpB26 side chain exhibited an orientation similar to native Tyr.

The TrpB26 side chain in the crystal structure of TrpB26, OrnB29-insulin displayed an orientation similar to that of the native Tyr. A slight main-chain shift in the B24-B28 β- strand (0.2 [±0.03] Å) was sufficient to enable native-like packing of the larger indole ring against the core of a protomer (Figure 5-17). In both T and Rf subunits, the six- membered component of the indole side chain was oriented towards IleA2, ValA3, and

ValB12. Based on the classical 3.5-6.5 inter-centroid distance, residue TrpB26 (of the T protomer) showed potential interactions with residue TyrD16 and PheD24. An interplanar angle of 78° between TrpB26 and TyrD16 was indicative of classical aromatic-aromatic packing within proteins (S. K. Burley & Petsko, 1985), which generally range from 50-

90°. TrpB26 and PheD24 displayed an interplanar angle of 36°, however; this orientation is less common. Similarly, in the Rf protomer TrpD26 packed near TyrB16 and PheB24. The interplanar angle between TrpD26 and TyrB16 was 62° whereas that between TrpD26 and

PheB24 was 48°. Previous studies have suggested that Trp residues may form aromatic- aromatic interactions over longer distances than those formed by Tyr-Phe pairs

(Bhattacharyya, Samanta, & Chakrabarti, 2002). Thus, the two intra-chain aromatic pairs, TrpB26/PheB24 and TrpD26/PheD24 (respectively separated by 7.8 Å and 7.1 Å) may interact more efficiently than the corresponding Tyr/Phe pairs in WT insulin (ring geometries are summarized in Table 5-12a,b).

CHARMM calculations of aromatic-aromatic interactions across the dimer interface of

TrpB26, OrnB29-insulin revealed 1.4 kcal/mol increased interaction energy. Residue-by- residue analysis of each component of the aromatic network indicated that the increased strength of interaction across the dimer interface was the result of the interactions

276

Figure 5-17. Comparison of B26 crevice within TRf dimers of TrpB26, OrnB29-insulin to WT (1TRZ). (A) Stereo view of TyrB26 (sticks) of 1TRZ T-state protomer within an electrostatic potential surface (generated using APBS plug-in to Pymol® (N. A. Baker,

Sept, Joseph, Holst, & McCammon, 2001)) formed by the surrounding residues (Fig. 1D,

E). Positively-charged surfaces are represented in blue, negatively charged surfaces in red, and neutral surfaces in white. (B, C) Two possible orientations of the TrpB26 that retain χ1 and χ2 angles of the WT structure shown in panel A. Due to the asymmetric structure of the Trp side chain, two possible χ2 angles of correspond in principle to the native Tyr. However, TrpB26 encounters a steric clash with residues in the core of insulin in the orientation shown in C. (D) TrpB26 of the Rf-state protomer from the TrpB26,

OrnB29-insulin crystal structure depicted within an electrostatic potential surface formed by surrounding residues. (E) Depiction of TrpB26 from panel D within the B26 crevice from panel A (WT). The TrpB26 side chain does not encounter a steric clash. (F)

Alignment of the naïve model of TrpB26 from panel C (green sticks) to the TrpB26, OrnB29- insulin structure (orange sticks). Residues are depicted within the WT crevice (panels A,

B, C, and E). The steric clash predicted in the naïve model is mitigated in the TrpB26,

OrnB29-insulin structure by (a) a local shift (0.2 Å) in the backbone of the C-terminal B-

B26 chain and (b) a slight difference in the χ1 torsion angle of Trp .

277

278

involving TrpD26 (see Table 5-10b). This result suggests that TyrB26→Trp may only display stabilizing properties when in an R-state protomer. If so, the T6 hexamer formed by TrpB26 insulin would be expected to have dissociation kinetics similar to WT insulin whereas the corresponding R6 hexamer may be expected to have markedly increased stability. The R→T transition, which is rapid in WT insulin on release of phenol, may represent the kinetic barrier responsible for the meta-stable association state observed in

SEC experiments.

The packing of TrpB26/D26 within respective cores of T- and Rf crystallographic protomers is similar in each case to the WT TyrB26/D26 and oriented such that the indole's nonpolar six-membered portion projects more deeply into a crevice between A- and B-chains than does its proximal heterocycle. Our 1H-NMR studies of TrpB26 within an engineered monomer (Holleman & Hoekstra, 1997) provided evidence that this overall conformation does not require self-assembly. Analogous partial burial of TyrB26 and TrpB26 in respective protein structures would in itself be expected to augment the variant's stability due to enhanced solvation free energy6 (L. C. Lee, Chou,

Chen, Lee, & Shaw, 2009) (i.e., as predicted by water-octanol transfer studies of free Tyr and Trp (L. C. Lee, et al., 2009); Table 5-13). Guanidine-denaturation nonetheless indicated that their stabilities are indistinguishable7.

We speculate that the predicted residue-specific differences in solvation free energy are attenuated by differences in protein dynamics leading to EEC (Glidden, Yang, et al.,

2017). Although the two B26 side chains each exhibit upfield secondary 1H-NMR shifts and analogous inter-residue NOEs, it is possible that the substitution is associated with local or non-local differences in protein dynamics. It would be of future interest to

279 investigate dynamic features by amide proton 1H-2H exchange and heteronuclear NMR relaxation methods (Glidden, Yang, et al., 2017). Because TrpB26 promotes partial dimerization of insulin lispro under NMR conditions (as indicated by concentration- dependent 1H-NMR resonance broadening), such studies may require use of an alternative monomeric template.

Given the ubiquity of EEC as a confounding general aspect of protein design

(Frederick, Marlow, Valentine, & Wand, 2007), the profound effects of TrpB26 on the properties of the insulin hexamer seem all the more remarkable. We envision that EEC is circumvented in this case by the rigidity of the insulin hexamer (including the interlocked aromatic residues at its dimer interfaces) with efficient burial of the WT and variant B26 side chains (J. Dong, et al., 2003). With similar internal structures and external solvation properties, the variant hexamer would gain (relative to WT) two advantages from the

Tyr → Trp substitution: (i) greater B26-related solvation transfer free energy (Wimley &

White, 1996) (Table 5-14) augmented by (ii) an uncompensated enthalpic advantage arising from more favorable ETF aromatic interactions as next discussed.

Molecular mechanics calculations rationalized physical and pharmacologic properties.

The structural rigidity of the R6 insulin hexamer (Hassiepen, Federwisch, Mulders, &

Wollmer, 1999; Jacoby, et al., 1996) motivated local MM-based modeling to assess the interactions contributing to the stability of the TrpB26-insulin hexamers. Analysis insulin oligomers by Raman spectroscopy revealed dampened conformational fluctuations of the

R6 hexamer in relation to lower-order oligomers and T6 hexamers (J. Dong, et al., 2003).

Moreover, the thermodynamic stability of the assembly was evidenced by the lack of

280

Table 5-12a. Intercentroid distances and angles between TyrB26 and local aromatic residuesa

B26 D26 Residue r Dih.A r Dih.A B16 13.6 76 6.0 52 B24 7.3 85 5.9 87 B25 10.9 34 10.6 70 B26 - - 12.0 77 D16 5.8 58 13.1 89 D24 5.8 38 7.5 78 D25 12.6 38 9.5 45 D26 12.0 77 - -

Table 5-12b. Intercentroid distances and angles between TyrB26 and local aromatic residuesa

B26 D26 Residue r Dih.A r Dih.A B16 12.8 90 5.5 20 B24 7.8 79 5.9 25 B25 9.7 37 13.0 44 B26 - - 12.2 73 D16 5.6 51 13.7 66 D24 6.6 46 7.0 89 D25 10.9 73 11.1 10

D26 12.2 73 - - a Tables S11a and S11b (respectively calculated from WT structure 1TRZ and the present crystal structure of the TrpB26 analog) indicate the character of aromatic- aromatic interactions involving Tyr or TrpB26 at the insulin dimer interface. Whereas some interactions are classical ETF interactions, others show some deviation in ring-to-ring dihedral angle (Dih.A).

281

Table 5-13a. Thermodynamic stabilization associated with the B26 side chain during folding of insulin analogsa

b Side ΔSA ΔGu Side Chain ΔSA ΔGu Chain (kcal/mol) (kcal/mol) TyrB26 68.9% 0.49 a TrpB26 69.3% 1.4

TyrD26 78.6% 0.56 TrpD26 70.1% 1.5 Mean ΔΔGu: 0.9 kcal/mol (not observed)

Table 5-13b. Thermodynamic stabilization associated with the B26 side chain during dimerization of insulin analogsa

Side ΔSA ΔGu Side Chain ΔSA ΔGu Chain (kcal/mol) (kcal/mol) TyrB26 25.9% 0.18 TrpB26 26.6% 0.55 D26 D26 Tyr 78.6% 0.56 Trp 26.5% 0.55 Mean ΔΔGu: 0.37 + 0.42 = 0.8 kcal/mol/dimer interface a The relative thermodynamic stabilities of TyrB26 and TrpB26 insulin monomers calculated using hydrophobic transfer free energies contrasted with those determined by guanidine-denaturation assays. However, NMR data suggests dimerization may be favored in TrpB26 lispro to a greater extent than in native lispro. b Tabulated values of respective H2O/octanol transfer free energies are -0.71 kcal/mol (Tyr) and -2.09 kcal/mol (Trp) as described (Wimley & White, 1996). This calculation pertains only to changes in exposure of residue B26 and does not consider secondary changes in exposure of neighboring side chains.

282 conformational changes visualized by NMR spectroscopy over a temperature range of

10-80 °C (Jacoby, et al., 1996). The resistance of the R6 hexamer to structural perturbation has also been shown in the context of mutant insulin analogs: native-like x- ray crystal structures have been reported of R6 containing a broad range of substitutions

(Table 5-15) (U Derewenda, DEREWENDA, DODSON, & BRANGE, 1987; M. Liu, et al., 2009; Vijay Pandyarajan, et al., 2014b). Even substitutions that were shown to destabilize the dimer interface of insulin, such as the substitution of PheB24 by the non- aromatic cyclohexylalanine (Cha), were shown to have little impact on the global

B26 structure of the R6 hexamer. For this reason, the Tyr →Trp substitution was expected to affect only the local structure of the insulin hexamer. Thus, the effects of the mutation were amenable to initial analysis in simplified (eight-ring) models of the dimer interface of insulin.

Energy minimization of a local model of TrpB26-insulin (i.e., as substituted into a WT

f insulin dimer extracted from a representative crystal structure of a T3R 3 zinc hexamer) yielded a native-like framework with enhanced nearest-neighbor B26-related aromatic- aromatic interactions. The partial-charge (monopole) model of the aromatic rings in the

CHARMM empirical energy function—parametrized in accordance with ab initio simulations (M. J. Frisch, et al., 2009)—predicted an increase of 0.8 kcal/mol. Although this calculation could in principle have been confounded by transmitted conformational perturbations and did not consider potential changes in conformational entropy or solvation, its conservative features were verified by x-ray crystallography. That the structure of TrpB26, OrnB29-insulin is essentially identical to WT suggests that local

283

Table 5-14. Native-like crystal structures of R6 hexamers formed by mutant insulin.

PDB ID or Reference Modification 1EV6 WT Human Insulin 1ZNJ Porcine insulin: AlaB30 at chain terminus 4E7V Bovine Insulin: AlaA8, ValA10, AlaB30 at surfaces 3GKY HisA8, ValA16 at surface (A8) or in core (A16) 4P65 ChaB24 at dimer interface 3ROV D-AlaB20, D-AlaB23, lispro within -turn 5HRQ cis-hydroxyprolineB28 at edge of dimer interface 5HPU trans-hydroxyprolineB28 at edge of dimer interface 5URU DihydroxyprolineB28 at edge of dimer interface 5UQA 4-fluoro-prolineB28 at edge of dimer interface 2WS6 N-methyl-TyrB26 at dimer interface and inter-chain crevice 2WS7 ProB26 at dimer interface 3ZS2 TyrB25, N-methyl-PheB26, lispro at dimer interface 3ZQR N-methyl-PheB25 at dimer interface 5EMS 3-iodotyrosineB26, NorleucineB29 at dimer interface 1QIY TyrB5 at inter-chain crevice 1ZEG AspB28 at edge of dimer interface 3ZU1 Des-B30, Nε-ω-carboxyheptadecanoyl- LysB29 Derewenda, 1987 (U ValB12 Ile at dimer interface Derewenda, et al., 1987; M. Liu, et al., 2009; Vijay Pandyarajan, et al., 2014b)

284 properties of TrpB26 directly underlie the observed increase in hexamer stability and lifetime.

The general asymmetry of the variant TRf dimer in the crystallographic hexamer was associated with differences in the details of corresponding aromatic-aromatic interactions across the dimer interface. Although CHARMM calculations predicted an increase of a

1.4 kcal/mol in interaction energies (0.7 kcal/mole per protomer) in accordance with our initial modeling, residue-by-residue decomposition ascribed this increase primarily to

TrpD26 (in the Rf protomer) and not to TrpB26 (in the T protomer). It is formally possible that TrpB26 is only stabilizing in an R state, but additional studies would be required to resolve this issue. Our cobalt-EDTA sequestration studies focused on the R6 state as the preferred storage vehicle in a pharmaceutical formulation. On SQ injection, rapid diffusion of phenolic ligands from the depot leads to a TR transition. That TrpB26 was found to delay subsequent absorption into the blood stream suggests that this substitution also enhances the kinetic stability and lifetime of the T6 hexamer.

A seeeming paradox is posed by the evolutionary exclusion of Trp at position B28 of vertebrate insulins despite the evident compatibility of this aromatic side chain with native structure and function. We speculate that this exclusion reflects the biological importance of the rapid disassembly of zinc insulin hexamers on their secretion by pancreatic β-cells. Whereas the enhanced thermodynamic and kinetic stabilities of

TrpB26-insulin hexamers would seem favorable for storage within secretory granules (as within pharmaceutical formulations) (G. Dodson & Steiner, 1998), delayed disassembly of the variant hexamers in the portal circulation would be predicted to reduce the hormone's bioavailability on first pass through the liver, as insulin dimers and hexamers

285 cannot bind to the IR (S. E. Shoelson, Lu, Parlautan, Lynch, & Weiss, 1992). Such delayed disassembly might also decrease the delivery of free zinc ions to the liver, recently predicted to constitute a regulatory signal in its own right (for review, see

(O’Halloran, Kebede, Philips, & Attie, 2013)).

The resulting impairment in hormonal regulation of hepatic metabolism (and accompanying systemic hyperinsulinemia (M Dodson Michael et al., 2000)) could in principle have imposed a selective disadvantage in the course of vertebrate evolution.

Although to our knowledge, such a kinetics-based mechanism has not been observed in vivo, the converse—abnormally rapid clearance of insulin—has been found on release from zinc-deficient secretory granules; presumably zinc-free insulin oligomers more rapidly dissociate on dilution in the portal circulation and so are more efficiently cleared than WT zinc insulin hexamers (Tamaki et al., 2013). Similarly, the exclusion of Trp at position B26 of vertebrate IGFs may reflect a disadvantageous competition between self- assembly and binding to IGF-binding proteins, which are critical to the integrated physiology of IGF function (for review, see (Forbes, McCarthy, & Norton, 2012)). Such functional complexity may impose hidden constraints on the evolution (and so divergence) of protein sequences. The conserved “aromatic triplet” of vertebrate insulins and IGFs provides a natural laboratory to uncover such evolutionary constraints (Weiss &

Lawrence, 2018). These considerations further suggest that structural features of a protein pertinent to its endogenous function may in general be distinct from those biophysical properties that may re-engineered to optimize molecular pharmacology.

Comparison of TrpB26-insulin and a corresponding iodo-Tyr analog.

286

Design of TrpB26-insulin was in part motivated by prior studies of an analog containing

3-iodo-Tyr at the B26 (El Hage, Pandyarajan, et al., 2016; Kim A Heidenreich, Yip,

Frank, & Olefsky, 1985; Vijay Pandyarajan, et al., 2014b). The latter analog (“3I-Y- insulin”) exhibited a variety of favorable properties: increased affinity for the IR (Kim A

Heidenreich, et al., 1985), increased thermodynamic stability and augmented resistance to fibrillation (Vijay Pandyarajan, et al., 2014b). Moreover, when formulated as an R6 hexamer, this analog exhibited a decreased dissociation rate (Vijay Pandyarajan, et al.,

2014b). Because modified amino acids raise the cost and complexity of protein manufacture, we wondered whether a natural amino acid might mimic, at least in part, the structure of 3I-Y-insulin and so confer favorable pharmacologic properties with conventional manufacturing.

This translational goal motivated our initial MM-based local modeling of TrpB26 insulin. Analysis of the crystal structure of 3I-Y insulin, determined as an R6 zinc hexamer (El Hage, Pandyarajan, et al., 2016), revealed that the large, nonpolar iodo- substituent packed within the core of the insulin protomer with preservation of a native- like dimer interface. Molecular-dynamics simulations, undertaken with a multipolar electrostatic model of the modified aromatic ring (El Hage, Pandyarajan, et al., 2016), rationalized this conformation: the iodine atom efficiently filled a cryptic packing defect in WT insulin, lined by the conserved side chains of IleA2, ValA3, and TyrA19. The enhanced packing efficiency of the modified insulin and novel network of halogen- specific electrostatic interactions (“weakly polar” interactions (S. K. Burley & Petsko,

1988)) appear to underlie the analog’s increased thermodynamic stability and resistance to fibrillation. Whereas the standard partial-charge model of aromatic systems failed to

287 account for the conformation of iodo-Tyr observed in the crystal structure, a multipolar electrostatic model rationalized thermodynamic stabilization of the dimer interface by this halogen “anchor” (El Hage, Pandyarajan, et al., 2016). Subtle changes in the geometry of aromatic-aromatic interactions were observed both in the simulations and in the crystal structure. Although activated MD simulations were not undertaken to probe the process of dimer dissociation, we presume that the above mechanisms of ground-state stabilization—enhanced core packing efficiency and a halogen-specific weakly polar network—also underlie the increased barrier to dissociation as indicated by the prolonged lifetime of the variant R6 hexamer (El Hage, Pandyarajan, et al., 2016).

Although the profound QM effects of halo-aromatic substitutions, including weakly polar interactions and “halogen bonding” (Scholfield, Zanden, Carter, & Ho, 2013), cannot be recapitulated by natural amino acids, it seemed possible that enhanced core packing efficiency might be achieved by analogy to 3-I-Y-insulin. Indeed, the steric profile of the asymmetric indole side chain of Trp is similar in size to 3-iodo-Tyr.

Accordingly, we imagined that the offset six-membered portion of the bicyclic indole ring might pack within the core of insulin in a manner similar to the iodine atom. In this intuitive picture the extended π-system of TrpB26 was envisioned to interact with neighboring residues to recapitulate, at least in part, the favorable electrostatic properties of iodo-TyrB26 (El Hage, Bereau, et al., 2016; El Hage, Pandyarajan, et al., 2016).

The above line of reasoning led to the present set of studies. In accordance with our original intuition and local MM-based modeling, the crystal structures of 3I-Y-insulin and TrpB26, OrnB29-insulin exhibited similar features. Nevertheless, salient differences between the quantum-chemical properties of iodo-Tyr and Trp might make the similar

288 structures of these B26 analogs a fortuitous outcome of distinct mechanisms. Whereas effects of 3-iodo-Tyr on aromatic-aromatic interactions at the insulin dimer interface are subtle, presumably reflecting indirect inductive effects of iodo-substitutent (El Hage,

Pandyarajan, et al., 2016), substitution of Tyr by Trp introduces a larger aromatic surface at this interface. It is possible that this feature underlies the more marked impact of

TrpB26 on insulin oligomerization relative to 3-iodo-Tyr.

Aromatic-aromatic interactions exemplify limitations of classical models.

The quantum origins of aromatic-aromatic interactions are complex, and so classical electrostatic models (such as partial-charge model of Tyr in the standard CHARMM empirical energy function) can be incomplete (Marshall, 2013). Although QM calculations more fully capture this complexity, a trade-off is encountered between rigor and computational feasibility, especially in complex systems such as proteins (A. D.

MacKerell et al., 1998), but even in ab-initio simulations of benzene-benzene interactions

(Sinnokrot & Sherrill, 2006). Standard MM and MD methods thus employ parametrized force fields to approximate quantum-chemical interactions (B. R. Brooks, et al., 2009).

Although parameters (e.g., partial charges assigned to an aromatic ring) have been chosen to provide reasonable protein models, such use of “monopolar” electrostatics neglects the polarizability of aromatic systems and so omits dispersion forces (Alexander D.

MacKerell, et al., 2002). Parametrized classical model may thus mischaracterize the strength or directionality of intermolecular interactions, particularly those involving more than one aromatic group or an aromatic group and a charged moiety. Although the pioneering studies of Burley and Petsko suggested that the partial-charge model of aromatic rings in proteins was sufficient for characterizing aromatic-aromatic interactions

289

(S. Burley & Petsko, 1986b), more recent work has highlighted its limitations with respect to delineating underlying physical mechanisms (Kamerlin, Haranczyk, &

Warshel, 2008; Marshall, 2013). The limits of parametrized classical force fields are of particular importance when applications are sought in nonstandard systems (such as unnatural protein mutagenesis (Řeha et al., 2002)) for which the parameters were not intended. Examples are provided by halogen-modified aromatic systems (widely employed in medicinal chemistry) (Steward & Chamberlin, 1998), which are associated with marked changes in quantum-chemical properties (Dougherty, 2000). Formal

QM/MM simulations of proteins may nonetheless be circumvented through force fields incorporating multipolar electrostatic models of aromatic-aromatic networks (Marshall,

2013).

The correspondence between our initial local MM-based modeling and our experimental findings, however striking, may be coincidental (van Gunsteren et al.,

2018): rigorous elucidation of the physics of the variant aromatic cluster at insulin’s dimer interface may require application of free-energy MD-based simulations with explicit inclusion of water molecules (“molecular alchemy” (Aleksandrov, Thompson, &

Simonson, 2010)). This approach may also provide insights into whether or how EEC may be circumvented. Nevertheless, standard MM calculations can guide initial biochemical analysis of protein structure as a guide for protein engineering. In the present study such initial modeling reinforced our structural intuition by highlighting the plausibility that substitution of TyrB26 by Trp might preserve a native-like interface and in fact enhance its weakly polar properties. Because the rigid hexameric framework provided a favorable context for local modeling of amino-acid substitutions at subunit

290 interfaces and yet it is the monomer that functions as a hormone in the bloodstream, it would be of future interest to apply more sophisticated computational techniques to simulate the structure and dynamics of TrpB26 insulin as a monomer in solution. Such predictions may in principle be tested through biophysical studies of an engineered insulin monomer containing TrpB26. Although this substitution is favorable in the context of the hexamer (and so of potential pharmacologic benefit), design of a monomeric NMR model will need to overcome the confounding effects of TrpB26, as defined in this experimental context, to promote self-association.

Concluding Remarks

To our knowledge, this study represents the first exploitation of aromatic-aromatic interactions to enhance the biophysical properties of a therapeutic protein (Kannan &

Vishveshwara, 2000). Our approach may be broadly applicable in protein engineering

(as ETF interactions are ubiquitous) and generalizable to non-standard aromatic moieties.

The latter would be likely to require QM/MM methods rather than classical force fields parametrized with partial charges (Řeha, et al., 2002). Overall effects of such substitutions on protein stability and self-assembly will require an integrated analysis of solvation free energies (A. D. MacKerell, et al., 1998), changes in protein dynamics

(Szyperski, Luginbuhl, Otting, Guntert, & Wüthrich, 1993) and potential EEC (Freire,

2009). Three- and four-dimensional heteronucelar NMR experiments would be expected to provide higher-resolution information regarding the local and non-local interactions of the optimized aromatic system. In the present application, analyses of 15N relaxation and

1H-2H amide-proton exchange are expected to improve understanding of the molecular dynamics of the TrpB26-modified aromatic cluster and so provide a more rigorous

291 biophysical context for its enhanced self-association properties (Glidden, Yang, et al.,

2017; Szyperski, et al., 1993).

The present application to insulin demonstrates a direct relationship between stabilization of the insulin hexamer and prolonged activity of a basal analog. Continuous and flat 24-hour insulin activity ("peakless" basal formulations) is of clinical interest in the treatment of Type 1 and Type 2 diabetes mellitus to reduce the risk of hypoglycemia

(especially at night) at a given level of glycemic control (P. E. Cryer, 2002). Because this mechanism is unrelated to present strategies to achieve protracted action, we envision that TrpB26-related enhancement of dimer-specific aromatic-aromatic interactions could favorably be introduced into current basal insulin formulations. In the future a combination of orthogonal molecular strategies might enable development of a once-a- week basal insulin therapy analogous to that of GLP-1 agonists (Christensen & Knop,

2010). Optimization of weakly polar interactions may thus assume a central place in the toolkit of molecular pharmacology.

5.5 Methods

Materials

Insulin was purchased from BioDel® (Danbury, CT). Insulin glargine was obtained from Lantus® (Sanofi-Aventis, Paris, FR). Reagents for peptide synthesis were as described (Wan, et al., 2008).

Preparation of Insulin Analogs

Variant insulins were prepared by semisynthesis (Inouye, et al., 1979). In brief, synthetic peptides were coupled to a tryptic fragment of insulin (des-octapeptide [B23-

B30] insulin) in aqueous/organic solvent using trypsin as a catalyst. Following rp-HPLC

292 purification, predicted molecular masses were confirmed by mass spectrometry (V.

Pandyarajan, et al., 2014).

Hexamer Disassembly Assays

2+ Disassembly of phenol-stabilized (R6) Co -substituted insulin hexamers was monitored as described (Rahuel-Clermont, et al., 1997). In brief, WT insulin or variants were made 0.6 mM in buffer containing 50 mM Tris-HCl (pH 7.4), 50 mM phenol, and

0.2 mM CoCl2 (Pandyarajan, et al., 2016) and incubated overnight at room temperature to attain conformational equilibrium. Spectra (400-750 nm) were obtained to monitor tetrahedral Co2+ coordination (Roy, et al., 1989) through its signature absorption band at

574 nm (Roy, et al., 1989). Co2+ sequestration initiated by addition of EDTA to a concentration of 2 mM. Dissociation was probed via attenuation of 574 nm band (Roy, et al., 1989); data were fit to a monoexponential decay equation (D. T. Birnbaum, et al.,

1997).

Protein Crystallography

Crystals were obtained by hanging-drop vapor diffusion at room temperature in the presence of a 1:1.7 ratio of Zn2+ to protein monomer and a 3.5:1 ratio of phenol to protein monomer in Tris-HCl. Diffraction was observed using synchrotron radiation at a wavelength of 0.9795 Å at the Stanford Synchrotron Radiation Light Source (Beamline

BL7-1; Stanford, CA); crystals were flash frozen to 100 K. Structure determination was carried out using molecular replacement using CCP4 (Winn et al., 2011) and Phenix structure-determination suites (P. D. Adams et al., 2010). The resulting structure was validated using PDB Redo server (Joosten, Womack, Vriend, & Bricogne, 2009). The lattice contained one TRf dimer per asymmetric unit. The main-chain conformations of

293 the 97 residues in the refined model of the TRf dimer in the asymmetric unit (excluding 2

Thr, 1 Orn, and 1 Phe residues) each resided in a most favored Ramachandran region.

Receptor-Binding Assays

Analog affinities for detergent-solubilized IR-B holoreceptor were measured by a competitive-displacement assay (Pandyarajan, et al., 2016). Successive dilutions of WT insulin or analogs were incubated overnight with WGA-SPA beads (PerkinElmer Life

Sciences®), receptor, and radio-labeled tracer before counting (Pandyarajan, et al., 2016).

To obtain dissociation constants, competitive binding data were analyzed by non-linear regression by the method of Wang (Z. X. Wang, 1995).

Rat Studies

Male Lewis rats (mean mass ~300 g) were rendered diabetic by streptozotocin.

Effects of insulin analogs formulated in Lilly® buffer (Pandyarajan, et al., 2016) on blood-glucose concentration following SQ injection were assessed in relation to WT- or

OrnB29-insulin (Pandyarajan, et al., 2016). OrnB29 insulin and TrpB26, OrnB29 insulin were made in the above buffer. GlyA21, OrnB29, OrnB31, OrnB32, TrpB26 and GlyA21, OrnB29,

OrnB31, OrnB32 were dissolved in dilute HCl (pH 4) containing meta-cresol, glycerol, and a 1:2 ratio of ZnCl2: insulin monomer. Rats were injected SQ with 3.44 nmoles of insulin or insulin analogs (~12-13.7 nmoles).

Animals used in this study were housed in the Association for Assessment and

Accreditation of Laboratory Animal Care (AAALAC)-accredited facilities of Case

Western Reserve University (CWRU) School of Medicine. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) Office at CWRU, which provided Standard Operating Procedures and reference materials for animal use (in

294 accordance with the NIH Guide for the Care and Use of Laboratory Animals). The animal health program for all laboratory animals was directed by the CWRU Animal

Resource Center. Animal care and use was further monitored for Training and

Compliance issues by Veterinary Services.

Size-Exclusion Chromatography

70 kDa); the mobile phase consisted of 10 mM Tris-HCl (pH 7.4), 140 mM NaCl, and

0.02% sodium azide. Elution times were monitored by absorbance at 280 nm. Molecular masses and void volume (V0) were inferred in reference to standard proteins

(Pandyarajan, et al., 2016).

Spectroscopy

CD spectra were acquired in 10 mM potassium phosphate (pH 7.4) and 50 mM KCl

(V. Pandyarajan, et al., 2014). Free energies of unfolding (ΔGu) were inferred at 25 °C from two-state modeling of protein denaturation by guanidine-HCl (V. Pandyarajan, et al., 2014; Sosnick, et al., 2000). 1H-NMR spectra were acquired at 700 MHz at pH 8.0 or pD 7.6 (direct meter reading) at 37 °C (Q. X. Hua, et al., 2011). Chemical shifts of aromatic protons in residues B24-B24 (Phe-Phe-Tyr or Phe-Phe-Trp) were evaluated in relation to corresponding chemical shifts in respective octapeptides B23-B30, presumed to represent random-coil shifts.

Molecular Mechanics Calculations

295

Calculations were performed using CHARMM (kindly provided by Prof. M. Karplus).

Its standard empirical energy function was employed (in whose development aromatic rings were parametrized by partial atomic charges in accordance with ab initio QM simulations (B. R. Brooks, et al., 2009)). Representative WT insulin dimers were obtained from PDB entries 4INS, 1ZNJ and 1TRZ. These structures and corresponding

TrpB26 homology models were subjected to local energy minimization (100 steps of

Steepest Descent followed by Adopted Basis Newton-Raphson with gradient tolerance tolg 0.0008/10000 steps). Minimizations were halted either at 1000 steps or when the above tolerance was reached. Changes in conformation were allowed only to eight side chains (TyrB16, PheB24, PheB25, TyrB26 (or TrpB26), and their dimer-related mates; the remaining atoms in the respective dimers were fixed. Total interaction energies and respective electrostatic components were obtained between the side chain of residue B26 and the neighboring three aromatic side chains (PheB24 and dimer-related TyrD16 and

PheD24). Following this survey of crystallographic dimers, such energies were further evaluated in a simplified molecular model that contained only the side chains of residues

TyrB26 (or TrpB26) and the same neighboring three aromatic residues as extracted from PDB entry 4INS; this yielded the electrostatic interaction energy map shown in

Figure 5-3, in which B26 χ1 and χ2 dihedral angles were systematically varied without energy minimization.

Quantum Mechanics Calculations

Electron density and Molecular Electrostatic Potential (MEP) of Tyr and Trp side chains were calculated using B3LYP and 6-31G(d) basis set using Gaussian utility

296

Cubegen in Gaussian09 (M. J. Frisch, et al., 2009). The isosurface map was generated using Jmol (Willighagen & Howard, 2007).

Ab initio energies of interaction between pairs of isolated aromatic rings were determined by calculating interaction energies using the MP2 method with aug-cc-pVDZ basis set in Gaussian 09 (M. J. Frisch, et al., 2009).

5.6 Endnotes

1 Insulin lispro, containing substitutions ProB28→Lys and LysB29→Pro, exhibited an attenuated 574 nm band(L. C. Lee, et al., 2009); this is associated with a greater

f proportion of spectroscopically attenuated T3R 3 hexamer and invisible T6 hexamer

(Figure 5-2).

2 This analog simulated pI-shifted insulin glargine, which contains two additional arginine residues at the C-terminal B chain (B31 and B32) and an AsnA21 substitution

(See Fig. 5-1A) (D. T. Birnbaum, et al., 1997; L. C. Lee, et al., 2009 ). Use of Orn permitted semisynthesis.

3 One protomer in the (dimeric) asymmetric unit adopted the classical “T” conformation (Figure 5-1B); the other adopted in the frayed-R state (Chothia, et al., 1983;

G. G. Dodson, et al., 1993) in which residues B3-B8 extend the B chain α-helix.

4 An exception is provided by a substitution at the primary Zn-binding site

(HisB10→Asp), which converts the hexamer into a novel dodecamer with Zn- coordination mediated by HisB5 (Riemen, et al., 1983). The structure of each protomer is native-like.

5 An experiment of nature provided evidence for the relationship in vivo between zinc coordination and protection from toxic insulin misfolding: in the rodent species Octogon

297 degus, substitution of HisB10 by Asn is associated with senile amyloidosis of the islets due to insulin fibrillation(Glidden, Yang, et al., 2017), resulting in DM (Jekl, et al., 2011).

|6 The identification of a destabilizing Trp→Tyr mutation in E. coli thioesterase-I demonstrated, in breach, the improved cavity-filling properties of Trp relative to Tyr (J.

Dong, et al., 2003).

7 That TrpB26, OrnB29-insulin exhibited a greater exposure of nonpolar surfaces on denaturation in guanidine-HCl than did OrnB29-insulin was suggested by a change in m values in two-state modeling (Table 2).

298

Chapter 6: 4SS-Insulin

6.1 Chapter Summary

The shelf-life of insulin and its analogs is limited, in part, by its propensity to form amyloid fibrils in formulation. The development of fibrillation-resistant insulin analogs has been a long-sought goal of the insulin-research community. Two independently discovered modifications to the insulin hormone have been shown to confer resistance to amyloid fibrillation. One such modification is the introduction of a foreshortened C- domain (4-8 residues in length) that spans the C-terminus of the insulin B chain and the N terminus of the insulin A chain and prevents insulin from adopting the elongated conformation necessary for the nucleation of amyloid fibrils. Analogs containing this modification are referred to as single-chain insulins (SCIs) The second modification is the introduction of a fourth, interchain, disulfide linkage between residues A10 and B4.

This ectopic cystine anchors the N-terminal region of the insulin B chain to the core of the insulin hormone and thus prevents the elongated prefibrillar conformation of insulin.

This class of analog is called “4SS-insulin.” In this chapter, I describe further characterization of 4SS insulin and the introduction of the 4SS modification in the context of an SCI analog. The resulting class of molecules, termed 4SS-SCIs, exhibit ultrastability, displaying little to no changes in secondary structure at temperatures exceeding 98°C. Such analogs have pharmacological properties that differ significantly from those of either parent analog and show potential therapeutic application as fabricated polymer-insulin blends.

299

6.2 Introduction

Insulin is used ubiquitously in the treatment of T1DM and in a subset of patients with

T2DM. In spite of evolutionary adaptations that prevent proteotoxic aggregation and misfolding of insulin in during storage in the secretory granules of pancreatic β-cells, insulin forms amyloid fibrils in pharmaceutical formulations (Brange, et al., 1997;

Brange & Langkjaer, 1997). This property limits the shelf-life of insulin products and limits their availability in areas of the world that lack reliable refrigeration systems

(Alberti, 1994). This limitation is particularly prominent for rapid-acting analogs whose diminished self-association capability reduces the protective effects of self-assembly. The development of insulin analogs with increased stability that are resistant to amyloid fibril formation has been a major goal of the insulin research field (Weiss, 2013a). At present, two classes of such insulin analogs have been described.

The first class of fibrillation-resistant insulin analogs comprises single chain insulins

(SCIs) (Q. X. Hua, et al., 2008). The design of SCI analogs was based on the structure of single-chain precursor to mature insulin, proinsulin, and single-chain precursors used in the production of recombinant insulin. Proinsulin contains a 35 amino-acid C-domain that connects the C-terminus of the insuin B chain to the N-terminus of its A chain. Proinsulin retains native-like biological activity and exhibits high affinity for the molecular target of insulin, the insulin receptor (IR). Although proinsulin is capable of forming amyloid fibrils, its fibrillation lag time is prolonged in relation to native insulin (Huang, et al.,

2005). In contrast, single-chain precursors of recombinant insulin, such as porcine insulin precursor (PIP) or mini-proinsulin (MPI), that contained foreshortened C-domains of length 2 amino acids and -1 amino acids (B29-A1 peptide bond), are resistant to

300 fibrillation and exhibit increased thermodynamic stability to native insulin, but are biologically inert (Q. X. Hua et al., 1998; Q. X. Hua, et al., 2008; Kobayashi, Sasaoka,

Sugibayashi, Iwanishi, & Shigeta, 1989). These findings led to the discovery of a “sweet spot” C-domain length of 4-8 amino acids (Figure 6-1A,B), which successfully dampened the conformational fluctuations of the insulin and prevented it from adopting the elongated conformation that nucleates the assembly of amyloid fibrils, yet was provided enough flexibility to allow insulin to adopt its receptor-binding conformation. The resulting class of analogs displayed a complete resistance to fibril formation, increased chemical and thermodynamic stability and quantitatively native-like biological potency

(Glidden, Yang, et al., 2017; Q. X. Hua, et al., 2008).

The second class of fibrillation-resistant insulin analogs are called “4SS-insulins.”

Such analogs contain an engineered fourth disulfide linkage between residues A10 and

B4 that recapitulates the hydrogen bond formed between A11 and B4 in the native structure of insulin (Figure 6-1C) (Tine N Vinther, et al., 2015; T. N. Vinther, et al.,

2013). The ectopic disulfide stabilizes the three-dimensional structure of insulin by preventing the N-terminal B chain from separating from the inter-helical segment of the

A chain and thus “traps” the insulin molecule in a native-like globular conformation, which naturally excludes formation of the cross-β structure associated with amyloid fibrils. The introduction of the fourth disulfide linkage has also been determine to make the insulin molecule more compact by burying hydrophobic residues within the globular core of the hormone that may be transiently exposed in native insulin. 4SS insulin analogs have also been shown to have increased biological activity in relation to their native-like, “3SS” counterparts (Figure 6-1A) (T. N. Vinther, et al., 2013).

301

The orthongonal nature of the stabilizing properties of both the foreshortened C- domain of SCIs and the A10-B4 disulfide linkage of 4SS insulin analogs motivated the incorporation of both modifications into a single insulin analog (Figure 6-1D). Such an analog was expected to be “ultrastable” and display resistance to fibrillation as well as thermal and chemical denaturation. The utility of such analogs in relation to parent SCIs or 4SS-insulins extends beyond their ability to stabilize insulin in traditional formulations. In particular, the increased thermal and physical stability of “4SS-SCIs” allows the analogs to overcome a critical barrier that has prevented the incorporation of insulin into slow-release polymeric matrices, which require high temperatures for fabrication (S.-E. Kim et al., 2014).

This chapter describes the development of tree distinct “4SS-SCIs,” designated 4SS-

SCI-1, -2, and -3, that are respectively based on previously characterized long-acting, biphasic, and rapid-acting SCI analogs. In addition to confirming the thermodynamic stability of 4SS-SCIs, our studies revealed the unique pharmacodynamic profiles of such analogs in relation to those of parent 4SS and SCI analogs (Glidden, Aldabbagh, et al.,

2017; Glidden, Yang, et al., 2017). Such findings may be indicative of novel insulin-IR binding kinetics or unique receptor-ligand interactions involving 4SS-SCIs. We further extended our studies to demonstrate the unique utility of ultra-stable insulin analogs by fabricating 4SS-SCIs into PLGA-based polymeric matrices and confirming their retained biological activity.

302

Figure 6-1 Structures of Wild-Type Insulin and Stabilized Analogs. (A) Ribbon representation of wild-type human insulin (PDB:4INS). The A chain is shown in blue and the B chain is shown in orange. Disulfide bridges are shown as yellow sticks. (B) Ribbon representation of an SCI. Color code as in A; the foreshortened C domain is represented as a black loop and marked by a black asterisk (PDB: 2LWZ). (C) Ribbon representation of 4SS-insulin, color code as above. The A10-B4 disulfide bridge is shown as yellow sticks and indicated by the red asterisk. (D) A model of the expected structure of a 4SS-

SCI analog. The A domain is shown as red ribbon, the N-terminal loop and central α- helix of the B domain are shown as orange ribbon and the C-terminal β-strand is shown in cyan. The c domain is shown as a gray loop. The engineered disulfide linkage is shown in yellow ball-and-stick representation and highlighted by the yellow asterisk.

303

304

6.3 Results

Two-Chain 4SS-insulin Exhibits Increased Thermodynamic Stability

. Circular Dichroism (CD) spectroscopy was used to determine the effects of the additional disulfide bridge on the thermodynamic stability of two-chain 4SS-insulin, a property that is independent of the fibrillation lag time of the analog. Qualitative differences in the far-UV CD spectrum of two-chain 4SS-OrnB29-insulin suggested increased α-helical content in relation to native OrnB29-insulin: both analogs exhibited similar mean residue ellipticity at 222 nm, the region of the CD spectrum that corresponds most closely with α-helix, yet native OrnB29-insulin displayed a blue-shifted

208 nm peak with greater amplitude than that of the 4SS analog (Figure 6-2A). This finding suggested that at equilibrium a greater number of α-helices are intact in 4SS- insulins than in their 3SS counterparts. Such spectral differences were also observed in des-[B23-B30]-octapeptide insulin (DOI), produced by the trypsinization of 3SS- or 4SS- insulin analogs. Whereas tryptic removal of the C-terminal B chain of insulin analogs appears to destabilize the secondary structure of both native and 4SS-insulins by removing aromatic anchor residues B24 and B26 and thus exposing the hydrophobic core of the protein, 4SS-DOI retains greater α-helical content than its 3SS counterpart (Figure

6-2C).

CD-monitored guanidine denaturation assays were utilized to assess the thermodynamic stabilities of full length 4SS-insulin and 4SS-DOI in relation to 3SS counterparts. OrnB29-insulin and native DOI displayed thermodynamic stabilities of 3.4 ±

0.1 kcal/mol and 0.9 ± 0.2 kcal/mol, respectively. Neither the stability of 4SS-OrnB29- insulin nor that of 4SS-DOI could be determined: no post-translational baseline baseline

305 was observed in the denaturation curves of both 4SS samples, indicating that neither full- length 4SS-OrnB29-insulin, nor 4SS-DOI underwent complete denaturation at 7.82 M guanidine hydrochloride. Comparison of these curves with previously reported stabilities of insulin analogs suggested that both full-length 4SS-OrnB29 insulin and 4SS-DOI had thermodynamic stabilities exceeding 5.5 kcal/mol (Figure 6-2B,D).

Conformational Restriction of the Core of 4SS-Insulin Analogs Confers

Resistance to Amyloid Fibril Formation

4SS-insulin and SCIs achieve fibrillation resistance through independent mechanisms.

Whereas SCIs prevent the C-terminal B chain of insulin from separating from the hydrophobic core of insulin, 4SS-insulin analogs anchor the N-terminal B chain to the periphery of the globular core of the insulin protein. Fibrillation lag times of DOI produced from WT insulin and 4SS-insulin analogs were assessed to demonstrate that the fibrillation resistance of 4SS-insulin analogs does not depend upon the conformation of their C-terminal B chain.

Fibrillation lag times were determined at room temperature because of the rapid fibrillation of WT insulin (lag time < 4 hours) at 37ºC. Three 4SS-DOIs were tested.The first contained no modifications other than the ectopic disulfide bridge. The other two

DOIs contained additional stabilizing substitutions: one of these contained a substitution

ThrA8 by Glu, which has been shown to stabilize the N-terminal helix of the insulin

(Weiss, Hua, et al., 2002), whereas the other contained a substitution of HisB10 by Asp

(Burke, Schwartz, & Katsoyannis, 1984; Slieker et al., 1997), which has been shown to stabilize the B9-B19 α-helix. Under these conditions, WT DOI formed fibrils within 4 days. No 4SS-DOIs formed fibrils after 100 days of agitation, as confirmed by THT

306

Figure 6-2 Thermodynamic Stability of 4SS Analogs(A) Far UV CD-wavelength spectra of 4SS-OrnB29 and native OrnB29-insulin and (C) spectra of a tryptic fragment,

DOI, of wild-type insulin and of 4SS-insulin. (B,D) CD-monitored guanidine denaturation assays of analogs.

307

308 fluorescence. These results suggested that 4SS-DOI is resistant to amyloid fibril formation.

4SS-insulin Analogs Exhibit Unique Pharmacodynamic Profiles Unrelated to IR-

Affinity

The affinity of 4SS-insulin analogs for IR was reported to be nearly 2-fold that of WT insulin. The affinities of three 4SS-insulin analogs were assessed to reproduce these results. As expected, 4SS-lispro1 exhibited a nearly 2-fold increased IR-affinity (0.04 ±

0.01 nM) in relation to native lispro (0.07 ± 0.02 nM). 4SS-AspB10-lispro displayed an affinity of 0.02 nM, whereas 4SS-GluA8-lispro exhibited an affinity of 0.05 nM (results are summarized in Table 2). These findings indicate that the AspB10 and GluA8 substitutions affect the IR-affinity of 4SS-insulin analogs in a manner proportional to that observed in native (3SS) insulin analogs: AspB10 and GluA8 insulins have been reported to have ~180% and 80% receptor affinity in relation to WT insulin, respectively (Burke, et al., 1984; Slieker, et al., 1997; Weiss, Hua, et al., 2002) (Table 6-1).

The in vitro affinity of insulin for IR does not reflect its biological activity. The biological activity of 4SS-insulin analogs was assessed by IV injection in streptozotocin- rendered diabetic rats (Figure 6-3). Native 4SS-GluB29-insulin2, 4SS-AspB10-GluB29- insulin, 4SS-GluA8-GluB29-insulin were tested against native GluB29-insulin. All three 4SS analogs displayed increased potency in relation to native GluB29-insulin. Moreover, each of the three 4SS analogs displayed an anomalous duration of biological activity that was visualized as a “tail” of the blood-glucose-clearance curve. Although 4SS-AspB10-insulin displayed greater potency than the other two 4SS analogs, the presence of the “tail” of the glucose-clearance curve was ubiquitous among 4SS insulins.

309

Table 6-1 In Vitro Affinities of 4SS Analogs for IR.

analog Kd IR-B (nM± SEM ) Lispro 0.07 ± 0.02* 4SS-Lispro 0.04 ± 0.01 4SS-AspB10-Lispro 0.02 ± 0.01 4SS-GluA8-Lispro 0.05 ± 0.01

310

Design of 4SS-SCI Analogs

The increased potency of 4SS-insulin analogs and the orthogonal nature of their stability in relation to that of SCIs motivated the creation of 4SS-SCIs. The 4SS modification (IleA10→Cys and GlnB4→Cys substitutions) was introduced in the context of three SCIs. The first, SCI-1, was designed as a pI-shifted long-acting insulin analog, whereas SCI-2 and SCI-3 are biphasic and rapid-acting analogs, respectively. In addition to foreshortened linker domains, these SCI analogs contained amino-acid substitutions that have been shown to stabilize the secondary structure of insulin. Design of the analogs is summarized in Table 6-2.

4SS-SCIs Retain α-Helical Structure at 95°C

The thermal stability of 4SS-SCIs was assessed using CD spectroscopy. Far-UV CD spectra of 4SS-SCI-1 and 4SS-SCI-3 were collected at pH 3.0 and pH 7.4, respectively at

15°C and 95°C, the rapid acting insulin analog, lispro, was used as a native control. At both pH 3 and pH 7.4, insulin lispro displayed a primarily α-helical secondary structure at

15°C. At 95°C, a blue shift and loss attenuation of 222 nm band was observed at both pH values indicating the presence of a largely unfolded, “random coil” protein (Figure 6-

4B,D). In contrast, both 4SS-SCI-1 and 4SS-SCI-3 displayed largely α-helical structure at

3 both 15 and 95°C. A 1 θ x 10 attenuation of mean residue ellipticity [θ]222 band was observed in the spectra of both 4SS-SCIs at 95°C. This difference is likely the result of temperature-based changes in the CD spectrum that are unrelated to protein denaturation, or local unfolding of regions of less-stable helical regions of insulin such as the N- terminal α-helix of the insulin A-chain (Weiss, 1990) (Figure 6-4A,C).

311

Figure 6-3 Activity of 4SS-GluB29 Analogs Glucose-lowering curves after IV injection of equal doses of 4SS and control analogs in diabetic rats (n=6). All 4SS analogs exhibited increased potency and prolonged duration of activity in relation to native

GluB29-insulin. 4SS- AspB10, GluB29-insulin displayed increased activity in relation to

GluA8 and native 4SS analogs.

312

313

Table 6-2 Summary of Modifications in 4SS Analogs. analog Linker Substitutions Rationale 4SS 2-chain none CysA10, CysB4 increased stability 4SS-SCI-1 EEGSRRSR CysA10, CysB4 increased stability ArgA8, ArgB29, pI-shifted + GlyA21 (prolonged action) 4SS-SCI-2 EEGPRR CysA10, CysB4 increased stability HisA8, GluA14 pI-shifted — AspB28, ProB29 monomeric association 4SS-SCI-3 EEGPRR CysA10, CysB4 increased stability HisA8, GluA14 pI-shifted — GluB29 avoid AspB28

314

The inadequacy of chemical- and temperature-based denaturation techniques to quantify the thermodynamic stability of 4SS-SCIs motivated use of NMR-monitored amide-proton exchange for this purpose. The rates of exchange of 1H amide protons of

2 the monomeric 4SS-SCI-2 with H protons (from D2O solvent) were used to determine the frequency of exposure of various regions of the insulin protein. These rates were then correlated to the frequency of unfolding using techniques developed by Vugumeyster and colleagues (Vugmeyster, Kuhlman, & Raleigh, 1998). Using this technique, the thermodynamic stability of 4SS-SCI-2 was determined to be 6.4 kcal/mol compared to

5.4 kcal/mol for the parent SCI-2 analog (Figure 6-5).

4SS-SCIs Exhibited Unique Pharmacodynamic Profiles

The biological activity of 4SS-SCI analogs was assessed in diabetic rats. IV injection of 4SS-SCIs revealed native-like potency. Unexpectedly, the “tailing” effect on the glucose clearance curves associated with two-chain 4SS-insulins was not observed in

4SS-SCI-1 (Figure 6-6A,B). However, 4SS-SCI-2 displayed biphasic activity similar to that of its parent SCI analog.

4SS-SCIs Retain Potency After Exposure to High Temperatures

The ability of 4SS-SCIs to remain functional after exposure to high temperatures was assessed using heating experiments. 4SS-SCIs and parent SCIs were heated at 98°C for two hours and injected intravenously in diabetic rats. Whereas the parent 4SS-SCI-2 displayed a 50% reduction in potency after heating, the biological activity of 4SS-SCI-2 was unaffected by prolonged exposure to high temperatures. Parent SCI-1 exhibited a foreshortened pharmacodynamics profile after heating: a decrease blood-glucose

315

Figure 6-4 Circular Dichroism 4SS-SCIs. (A) Far UV-CD spectra of 4SS-SCI-1 at 25 and 95°C, collected at pH 3.0 and (B) corresponding spectra of insulin lispro. (C) Far

UV-CD spectra of 4SS-SCI31 at 25 and 95°C, collected at pH 7.4 and (B) corresponding spectra of insulin lispro. Both 4SS-SCI-1 and 4SS-SCI-3 exhibited minimal differences in secondary structure between 25 and 95°C.

316

317

Figure 6-5. NMR-Monitored Amide-Proton Exchange of 4SS-SCI-2 NMR-monitored amide-proton exchange was used quantify the stability of the monomeric 4SS-SCI-2. 15N- labeled samples were taken up in D2O and attenuation of amide-proton peaks by exchange of 1H for spectroscopically quiescent 2H was monitored. The rates of disappearance of peaks corresponding to residues within the core of insulin (reporting global stability) were fit to monoexponential decay curves to determine the free energy of unfolding (ΔGu).

318

319

Figure 6-6. In Vivo Potency of 4SS-SCIs. (A) Time course of glucose lowering in diabetic rats after IV injection of equal doses of insulin lispro, 4SS-insulin, SCI-2, and

4SS-SCI-1. 4SS-SCI-1 did not exhibit the prolonged biological activity observed in two- chain 4SS analogs (n=5). (B) Time course of glucose lowering in diabetic rats after IV injection of equal doses of insulin lispro, SCI-2, and 4SS-SCI-2. Both analogs exhibited biphasic biological activity.

320

321 clearance was observed in the late phase of activity. In contrast, the pharmacodynamic profile of 4SS-SCI-1 was unaffected by heating (Figure 6-7).

4SS-SCIs Retain Potency After Fabrication into PLGA Matrices

The ability of 4SS-SCI-1 to retain biological activity after fabrication in Poly-lactic- co-glycolic acid (PLGA)-based polymer matrices was tested in diabetic rats. 4SS-SCI-1 and control analogs were extruded into PLGA matrices (S.-E. Kim, Harker, De Leon,

Advincula, & Pokorski, 2015; S.-E. Kim, et al., 2014). Insulin was extracted from the cylindrical PLGA-insulin co-polymer products by incubation in 0.1% trifluoroacetic acid

(TFA). After complete dissolution of the polymer, insulin remaining the solution was quantified and injected into diabetic rats. 4SS-SCI-1 retained potency after fabrication and extraction from the PLGA matrix. In contrast, two chain 4SS-lispro, and parent SCI-

1 displayed marked decreases in biological activity (Figure 6-8).

Extraction of Insulin from PLGA Matrices in Vitro

To improve the extraction of insulin from PLGA-matrices, 5% polyethylene glycol

(PEG) was introduced into the polymer matrix (P. Lee, Towslee, Maia, & Pokorski,

2015). Matrices fabricated with 4SS-SCI-1 were incubated in 1xPBS and the amount of insulin extracted into solution was determined by optical density at 280 nm. Findings confirmed a steady increase in insulin concentration in solution over 11 days with total insulin concentration plateauing between 11 and 15 days of incubation (Figure 6-9A).

Although such a result suggested the feasibility of such co-polymers as basal insulin delivery systems, SQ implantation of PLGA-4SS-SCI-1 co-polymers did not have any effect on blood-glucose concentrations in diabetic rats (Figure 6-9B).

322

Figure 6-7. Activity of SCI Analogs After Incubation at High Temperatures.

Biological activity of unheated 4SS-SCI and parent SCI analogs and that of analogs incubated at 95°C for 2h.

323

324

Figure 6-8. Activity of Insulin Analogs after Extraction from PLGA Matrices. (A)

4SS-SCI-1 and controls were extruded into a PLGA-based polymeric matrix shown in a vial in the photograph. Insulin was extracted from the matrix after soaking the cylindrical polymer in 0.1% TFA overnight. Extracted 4SS-lispro (B), SCI-1 (C), and 4SS-SCI-1 (D) were injected SQ into diabetic rats to assess biological activity.

325

326

6.4 Discussion

The development of ultra-stable insulin analogs represents a novel avenue for the treatment of T1DM and T2DM: the technology overcomes the limitation of the physical stability of insulin to allow its use in slow-release polymeric matrices (Brange, Hallund,

& Sorensen, 1992; Brange, Havelund, & Hougaard, 1992; Brange & Langkjaer, 1992).

Although the findings discussed in this chapter demonstrate that the application of 4SS-

SCIs in such fabrications is feasible, the technology will require significant fine-tuning.

Nevertheless, 4SS-SCI analogs are of considerable interest from a structure-function perspective: the unique pharmacokinetic profiles of 4SS-SCIs in relation to their parent analogs provide a scientifically rich platform for the study of insulin-IR complexation, kinetics, and signal transduction.

4SS-Insulin and SCIs are Stabilized by Independent Mechanisms

The differences in in the designs of SCIs and 4SS insulin analogs suggested that both classes of insulin analogs achieve resistance to amyloid formation through different mechanisms. The fibrillation resistance of 4SS-DOIs confirmed this hypothesis. The foreshortened C-domain of SCIs prevent the local unfolding of the C-terminal B chain of insulin. This prevents the B chain from adopting the elongated conformation required for fibrillation (Q. X. Hua, et al., 2008). A break in the polypeptide chain of SCIs at the C- terminal B chain would render such analogs susceptible to fibrillation. In contrast, 4SS- insulins do not form fibrils even after the complete removal of residues B23-B30. The engineered disulfide linkage of 4SS-insulins anchors the N-terminal B chain to the central

A chain, thus preventing insulin from adopting conformations associated with fibril formation. Whereas the foreshortened C-domain of SCIs has variable effects on the

327

Figure 6-9. In Vitro and In Vivo Elution of 4SS-SCI-3 from PLGA Matrix (A) In vitro elution of 4SS-SCI-3 from a cylindrical PLGA-PEG copolymer after incubation in

1xPBS. The total concentration of insulin in solution was measured as a function of optical density readings at 2800 nm. (B) BGL measured by a continuous glucose monitor

(CGM) in diabetic rats implanted with 4SS-SCI-3 or PLGA-matrix vehicle. Aberrant hypoglycemic readings were observed early in the course of the experiment, which were found to be spurious. No evidence of elution of insulin was observed.

328

329 thermodynamic stability of insulin, the A10-B4 disulfide linkage has a marked effect on the stability of the globular core of insulin as evidenced by the incomplete unfolding of

4SS-DOI at high concentrations of denaturant.

The combination of the orthogonal stabilizing mechanisms utilized by 4SS-insulins and SCIs was able to produce an ultra-stable class of insulin analogs. 4SS-SCIs exhibited remarkable stability at temperatures exceeding 90°C, and they were able to withstand the physical stresses associated with extrusion into polymer matrices. Unlike two-chain 4-

SS- or parent SCIs, 4SS-SCIs displayed native-like activity after extrusion and recovery from PLGA polymer matrices. This finding suggests the utility of such analogs beyond their use in polymer-based delivery. 4SS-SCIs may be capable of withstanding freeze- thaw cycles, allowing them to be shipped and stored as frozen solutions or aerosolized as inhalable insulin analogs (Bhatnagar, Bogner, & Pikal, 2007; Nyambura, Kellaway, &

Taylor, 2009; Ungaro et al., 2009). Continuing work will focus on the exploration and fine-tuning of the clinical application of 4SS-SCIs.

4SS-SCIs Display Novel Pharmacokinetic Profiles

Perhaps the most intriguing aspect of 4SS-SCI analogs (particularly 4SS-SCI-1 and

4SS-SCI-3) is their native-like PD profiles that lack the prolonged duration of insulin activity observed in two-chain 4SS insulin analogs. The non-additive behavior of 4SS and SCI modifications on the PD profiles of insulin analogs highlights the complexity of insulin IR-interaction, signaling, and physiology. That the protracted action of 4SS- insulin is observed after IV injection suggests that insulin-IR interaction, rather than SQ absorption, is responsible for the protracted action of the analog. One explanation for the observed differences in biological activity of the analogs is that 4SS-insulins have high

330

“on” rates and “off” rates of IR binding, a process that is governed by a conformational change in IR directing insulin to its secondary binding site (P. De Meyts, 2015; Knudsen,

De Meyts, & Kiselyov, 2011). Since a majority of insulin is cleared by receptor-mediated pathways, fast-on/fast-off insulin analogs are expected to be cleared more slowly than native analogs resulting in prolonged biological activity (Di Guglielmo et al., 1998;

Genuth, 1972). The introduction of a C-domain in 4SS-SCIs may be hypothesized to proportionately decrease the on and off rates of insulin-IR complexation leading to native-like clearance of the analog. This hypothesis will be tested with future experiments that measure the off rates of insulin analogs using whole-cell receptor-binding assays

(Knudsen, et al., 2011) and measuring the clearance of such analogs by ELISA in rats.

A competing hypothesis suggests that the differences in topography between 4SS, SCI, and 4SS-SCI analogs lead to a differential activation of insulin-receptor substrates (IRSs) by IR upon insulin complexation. The protracted activity of two-chain 4SS insulin analogs would be expected to be the result of activation of slower-decaying signal transduction intermediates. “Biased signaling” has been famously described in the context of GPCRs (Damian, Martin, Mesnier, Pin, & Banères, 2006; Wisler, Xiao,

Thomsen, & Lefkowitz, 2014), and the well-documented mitogenicity of AspB10-insulin suggests that IR-signaling bias may be achieved between the metabolic and mitogenic pathways (Gallagher et al., 2013; Sciacca et al., 2010). However, no such bias nor the basis for such a bias has been reported within the metabolic pathway of IR (P. De Meyts,

2000).

Concluding Remarks

331

The development of 4SS-SCIs represents the use of protein engineering to improve upon evolution. The A10-B4 disulfide bridge of 4SS insulin analogs recapitulates a hydrogen bond between residues A11 and B4 that stabilizes the native conformation of insulin. Although the stabilization of proteins through the increased frequency of disulfides is a phenomenon observed in thermophilic bacteria (Ladenstein & Antranikian,

1998), such adaptations remain difficult to access without significant selective pressure.

In the case of insulin, mutations introducing ectopic cysteines almost always lead to β- cell destruction through insulinopathy (Stoy, et al., 2017; Stoy, et al., 2010). Similarly, the foreshortening of the C domain of proinsulin would require a deletion of ~30 amino acids of the proinsulin C-domain as well as elimination of the dibasic sites used for prohormone convertase cleavage. Furthermore, the structure of native insulin is adequate for preventing proteotoxic misfolding or degradation during storage in secretory granules

(Emdin, Dodson, Cutfield, & Cutfield, 1980). For this reason, there is little selective pressure favoring ultra-stable insulin variants. Yet, findings of previous studies of the structural evolution of insulin and of the molecular evolution of proteins provided critical insight into developing a technology that has tremendous potential for therapeutic application.

The biological activity of 4SS-SCIs in relation to that native insulin and parent analogs provides a platform for further fundamental studies on insulin-IR complexation and physiology. It is apparent from the proportionate binding affinities of native 4SS-insulin,

4SS-AspB10-insulin, and 4SS-GluA8 insulin that 4SS insulin analogs occupy the similar binding site as native insulin (Hansen, et al., 2011; Weiss, Hua, et al., 2002). Yet, each of these analogs exhibit different PD profiles from native 3SS insulin analogs and 4SS-

332

SCIs. Future studies may these analogs to study the structural determinants of insulin-IR complexation and the effects of the structure of insulin on the conformational changes in

IR. Such studies may reveal the ability of IR to differentially activate IRS isoforms to modulate the duration and strength of the consequent signal transduction pathways

(Saltiel & Pessin, 2002).

6.5 Methods

Preparation of Insulin Analogs

4SS-SCIs and parent analogs were expressed in Pichia pastoris that secrete a precursor of the mature protein into growth media. Media were filtered and prepared for hydrophobic interaction chromatography: media were prepared with 1M (NH4)2SO4 and mixed with butyl-derivatized sepharose resin. Protein was liberated from resin with distilled, deionized water and acidified with TFA for reverse-phase HPLC purification.

Rp-HPLC was carried out with a mobile phase consisting of 0.1% TFA as an aqueous solvent and 0.1% TFA in acetonitrile as an organic elution solvent; the stationary phase consisted of C4-derivatized silica resin. Purified products were lyophilized and enzymatically converted to their mature form with endoLys-C-derivatized resin in a

2mg/mL solution containing 1M urea at pH 8.4. Conversion was carried out at room temperature overnight and quenched by acidification with TFA. Product was purified by

RP-HPLC using a preparative C8-derivatized silica resin stationary phase and the mobile phase described for initial purification (above). Purity of product was assessed by RP-

HPLC using an analytical C4-column. Identity was confirmed by MALD-TOF mass spectrometry.

333

Two-chain 4SS analogs were prepared by semisynthesis. A tryptic fragment of 4SS insulin, des-[B23-B30]-octapeptide insulin was created by trypsinization of a single-chain precursor, miniproinsulin, containing the 4SS modification. MPI was expressed and purified from yeast using the protocol described above. MPI was trypsinized in 0.1 mM ammonium bicarbonate (pH 8.0) containing 1 M urea at 2.5 mg/ml with 10% trypsin. The

DOI was purified by RP-HPLC as described above.

Trypsin-catalyzed semi-synthesis using DOI (4SS or WT) and modified octapeptides as described (Inouye, et al., 1981). Modified octapeptides were prepared by solid phase synthetic methods (Barany & Merrifield, 1980). The resulting 51-residue insulin analogs were purified by preparative reverse phase C4 HPLC (Higgins Analytical Inc., Proto 300

C4 ,10 uM, 250 x 20 mm), and their purity was assessed by analytical reverse phase C4

HPLC (Higgins Analytical Inc., Proto 300 C4 5 uM, 250 x4.6 mm). Predicted molecular masses were in each case verified using an Applied Biosystems 4700 proteomics analyzer

MALDI-TOF.

Circular Dichroism Spectroscopy

Analogs were made 40-60μM in KPi buffer (50 mM KCl, 10 mM KPi, pH 7.4) or 30% acetic acid (pH 3.0). Far-ultraviolet (UV) spectra were obtained from 200 to 250 nm at

25°C and 95°C on an AVIV spectropolarimeter as described (Vijay Pandyarajan, et al.,

2014a). Resulting spectra were normalized against buffer blanks, concentration corrected, and processed to reflect per-residue ellipticity.

Polymer Fabrication and In Vitro Extration

SCIs were processed within PLGA thermal melts at 95 oC to form cylinders (1 mm radius X 1 cm length) (Makadia & Siegel, 2011). The polymer blends containing insulin

334 analogs were incubated in 0.1% TFA at room temperature until complete dissolution of the polymer matrix. Insulin released was quantified and injected SQ in diabetic rats.

In vitro Elution Experiments under Physiological Conditions

The polymer blends were incubated in PBS (at skin pH 5.5 (Lambers, Piessens,

Bloem, Pronk, & Finkel, 2006)) wherein 50-50 PLGA degrades in ~1 month at 37 oC.

Elution of insulin was evaluated as a function of absorbance at 280 nm.

Receptor Binding Assays

Rodent Assays

335 phase HPLC to ensure uniformity of formulations; dilutions were made using the above buffer. Rats were injected at time t0. Blood was obtained from the clipped tip of the tail at time t0 and every 10 min for the 1st h, every 20 min for the 2nd h, every 30 min for the

3rd h, and every hour thereafter to a final time of 5 h.

In polymer implantation experiments, rats were anesthetized with isofluorane and a 1- inch polymer cylinder was implanted subcutaneously. BGL was recorded by a continuous glucose-monitoring system over the course of 12 days.

6.6 Endnotes

1 Insulin lispro is a monomeric insulin analog that contains a transposition of residues

ProB28 and LysB29

2 The LysB29→ Glu substitution is found in insulin glulisine. This substitution does not affect the biological activity of insulin.t

336

Chapter 7: Towards a Glucose-Responsive Insulin Analog

7.1 Chapter Summary

The maintenance of adequate glycemic control in patients with insulin-dependent forms of DM is a major clinical challenge. The major sequelae of DM are caused by chronic hyperglycemia, whereas the repeated episodes of mild hypoglycemia have been associated with dementia and cognitive deficit. Severe acute episodes of both conditions can result in death. For this reason, insulin dosing regimens are tightly controlled and require the use of long-, rapid-, and often intermediate-acting insulin analogs. Such regimens would be simplified by the development of glucose-responsive insulin (GRI) systems. The term “GRI” refers to a broad category of insulin analogs or delivery systems that provide glucose-lowering insulin activity at levels proportional to the glycemic state of the patient.

Whereas the glucose-responsive insulin analogs that have been described in current literature rely co-formulation with non-insulin agents, sequestration within nanoparticles or polymeric assemblies, or delivery through a system containing a continuous glucose monitor and insulin pump, this chapter describes the early stages of the development of an GRI that contains an internal glucose-sensing motif. Although this study remains in its early stages, preliminary results demonstrate the promise of this design and underscore its scientific and translational richness.

7.2 Summary of previous GRI Research

Adapted from

337

Rege, N.K., Phillips, N.F.B, and Weiss, M.A. (2017) Development of glucose- responsive 'smart'insulin systems. Current Opinion in Endocrinology, Obesity, and

Metabolism, 24(4) 267-278

Introduction:

Insulin replacement therapy is integral to the treatment of Type I diabetes (T1DM) and often required in the treatment of Type II diabetes (T2DM). These distinct diseases are respectively caused by an absolute or relative lack of insulin (S. Kahn, 2003). Insulin initiates the uptake of glucose within peripheral tissues by binding to its molecular target, the insulin receptor (IR) (J.G. Menting, et al., 2014). This receptor tyrosine kinase in turn activates a complex set of intracellular signaling processes (White & Kahn, 1994) leading to translocation of glucose transporter GLUT4 to the cell surface (Saltiel & Pessin, 2002).

Regulated secretion of insulin by pancreatic β-cells in response to increased interstitial glucose levels ordinarily enables the blood-glucose level (BGL) to be maintained within a narrow range, which in healthy individuals lies between 80 and 120 mg/dl (4.4-5.6 mM) with fasting values between 80 and 100 mg/dl (Petznick, 2011).

Unlike pancreatic secretion of insulin, standard pharmacological administration of insulin is not regulated by an endogenous feedback mechanism. Even with carefully designed dosing regimens and use of modern insulin products (i.e., engineered basal and rapid-acting insulin analogs) and even in the context of strict dietary and lifestyle adherence by patients, patients with diabetes often experience periods of hyperglycemia or hypoglycemia (M. D. Campbell et al., 2015). The chronic health risks associated with elevated mean glycemia (micro- and macrovascular disease, renal disease, retinopathy, and neuropathy) are insidious relative to the dramatic presentation of acute metabolic

338 decompensation (diabetic ketoacidosis in T1DM and hyperglycemic coma in T2DM) or the immediate signs and symptoms of hypoglycemia (adrenergic and neuroglycopenic effects leading to anxiety, tremulousness, and altered mental status leading in severe cases to coma) (Beregszàszi et al., 1997; P. E. Cryer, Davis, S.N., Shannon, H, 2003).

Further, such acute events can trigger adverse cardiovascular events in the setting of preexisting vascular disease (Snell-Bergeon & Wadwa, 2012), and repeated severe hypoglycemic episodes can cause cumulative damage to the brain with cognitive decline

(Philip E Cryer, 1997). For these reasons avoidance of hypoglycemia often limits individual hemoglobin A1c goals (an integrated indicator of mean glycemia over three months) and the rigor of current insulin treatment regimens.

A long-standing goal has been development of insulin delivery systems that would co- optimize treatment of hyperglycemia and prevention of hypoglycemia. Such systems would, like endogenous pancreatic b-cells, provide insulin action proportionate to the glycemic state of the patient. Accordingly, “smart insulin” systems (broadly designated glucose-responsive insulin; GRI) have attracted the attention of the research community for four decades (Brownlee & Cerami, 1979; Zaykov, et al., 2016). Such systems are defined as a class of insulin-delivery devices or insulin formulations that provide insulin activity commensurate with the metabolic needs of the patient. Due to the clinical considerations summarized above, the development of GRI technology has been a strategic priority of the National Institute of Diabetes and Digestive and Kidney Diseases

(NIDDK) (NIDDK, 2011) and the Juvenile Diabetes Research Fund (JDRF) (Insel,

Deecher, & Brewer, 2012).

339

Our definition of GRI is based on the conceptual goal of glucose-regulated delivery rather than any particular mechanical or molecular embodiment. Because this goal is shared by closed-loop delivery systems (i.e., the “artificial pancreas” as based on CGM- coupled insulin pumps (Halvorson, Carpenter, Kaiserman, & Kaufman, 2007)) and because such systems are near regulatory approval, they provide a starting point for our discussion. This perspective may be unconventionally, but highlights clinical goals as our ultimate objective. In the coming decade closed-loop systems are likely to provide the gold standard with respect to which the efficacy and safety of “smart” molecular strategies will be evaluated.

With this in mind, GRIs may be broadly classified as follows: (i) algorithm-based mechanical GRI systems (Halvorson, et al., 2007) in accordance with the above definition; (ii) polymer-based systems in which insulin is encapsulated within a glucose- responsive polymeric matrix-based vesicle or hydrogel (Ravaine, Ancla, & Catargi,

2008); or (iii) molecular GRI analog systems, which involve the introduction of a glucose-sensitive motif to the insulin molecule or its formulation, in either case conferring glucose-responsive changes to the bioavailability or activity of the hormone

(Thomas Hoeg-Jensen et al., 2005).

Mechanical GRIs

Mechanical GRI systems have three components: (i) a CGM providing real-time measurement of the interstitial glucose concentration; (ii) an insulin pump capable of receiving data from the CGM and (iii) a computer-encoded algorithm that predicts the appropriate dose of insulin for subcutaneous (SQ) injection (E. Renard, 2002), (E.

Renard, Costalat, & Bringer, 2002). Optimization of such systems has led to efforts to

340 enhance the accuracy and precision of CGMs and the robustness of the control algorithms. Because changes in interstitial glucose concentration lag by ~20 minutes behind changes in BGL, because the insulin analog (once injected) may require 20-40 min for absorption into the blood stream and because the hormone (once in the blood steam) may exert biological effects lasting 3-4 hours, a critical feature of modern algorithms is prediction of future trends in BGL. The robustness of such predictive algorithms is likely to be enhanced by the development of “ultra-fast” pump insulin analog formulations that minimize any delay in SQ absorption and in the duration of target-cell signaling, once the insulin receptor is engaged (Breton et al., 2012).

Advances in engineering over the past decade have led to the creation of compact devices in which CGMs are integrated within algorithm-controlled insulin pumps capable of precise adjustments in SQ delivery rate (Ly et al., 2015). Further, through analysis of clinical data and implementation of rapid signal-processing techniques, hierarchical algorithms that optimize glycemic control while prioritizing patient safety have been developed (Breton, et al., 2012; Doyle, Huyett, Lee, Zisser, & Dassau, 2014; Pinsker et al., 2016). These algorithms include such parameters as insulin potency, time since previous bolus dose, rate of increase of BGL, heart rate, and temperature to create closed- loop insulin delivery systems designed to maintain euglycemia during a variety of conditions: extent of exercise or activities of daily life (DeBoer et al., 2016; Jacobs et al.,

2016; Miller, Nimri, Atlas, Grunberg, & Phillip, 2011), fluctuations in ambient temperature (Turksoy et al., 2017), changes in diet (H. Lee & Bequette, 2009; H. Lee,

Buckingham, Wilson, & Bequette, 2009), variation in insulin sensitivity (Steil, Clark,

Kanderian, & Rebrin, 2005), and amount of insulin already on-board (Ellingsen et al.,

341

2009). To more effectively prevent or treat hypoglycemia, more complex pumps have been developed to independently provide either insulin or its counter-regulatory hormone glucagon (Bakhtiani et al., 2014; Haidar, Legault, Matteau-Pelletier, et al., 2015; Haidar,

Legault, Messier, et al., 2015; Jacobs, et al., 2016). This dual-hormone algorithm triggers

SQ injection of a stabilized glucagon formulation based on trends in CGM readings predicting a hypoglycemic excursion. Such a bihormonal AP has been shown to maintain

BGL with the targeted range more efficiently (i.e., with fewer hyperglycemic excursions and reduced time in hypoglycemia) than conventional CGM-coupled insulin pumps

(Blauw, Keith-Hynes, Koops, & DeVries, 2016; El-Khatib et al., 2016).

Mechanical GRIs may be further streamlined by optimization of each of its components. CGMs may be improved, for example, through ultra-miniaturization of sensors and improvements in their accuracy (Graf et al., 2016; Ribet, Stemme, & Roxhed,

2017; Thabit & Hovorka, 2016). Development of novel insulin analogs with more rapid onset of activity and shorter duration of action (relative to current insulin analog products

Humalog®, Novolog®, and Apidra® (Becker, et al., 2005; Lougheed et al., 1997; Mudaliar et al., 1999)) could improve closed-loop systems by providing more predicable pharmacodynamic profiles and hence more accurate calculation of insulin doses.

Examples of more rapid formulations are provided by Novolog-based FIAsp (a reformulation of insulin aspart to contain excepients that promote capillary absorption) or addition of the degradative enzyme hyaluronidase to wild-type insulin (rHuPH20;

Halozyme, Inc.) (Tim Heise, Hövelmann, et al., 2017; Tim Heise, Pieber, Danne,

Erichsen, & Haahr, 2017; Vaughn et al., 2009); the latter approach is no longer being pursued.

342

Intraperitoneal (IP) delivery of insulin may also offer advantages within closed-loop systems. IP delivery leads to ultra-rapid absorption and, like native pancreatic secretion, provides first-pass treatment of the liver. These features have led to the development of intraperitoneal infusion (IPI) devices (David S Schade & Eaton, 1980; D. S. Schade,

Eaton, Friedman, & Spencer, 1980) as implanted and refillable pumps (Irsigler et al.,

1981). Despite their theoretical advantages and long-term reservoirs (up to three months), regulatory approval of IPIs in the United States has been withheld due to their susceptibility to catheter occlusion (E. Renard et al., 1996). Such occlusions are in part inflammatory, provoked by formation of immunogenic and pro-inflammatory amyloid fibrils (Bally, Thabit, & Hovorka, 2017; Jeandidier et al., 1995). In addition, fibrillation of insulin within the pump reservoir or catheter alters the flow properties of the solution and inactivates the hormone. Occlusive events limited the feasibility of routine IPI delivery in an otherwise encouraging clinical trial at U.S. Veterans Administration hospitals (Saudek et al., 1996). Nonetheless, long-term experience with intraperitoneal insulin pumps has been documented in Montpellier, FR with use of a special surfactant- stabilized U-400 insulin formation (Insuplant®; Sanofi); periodic in-hospital alkaline washes of the implanted reservoir are required to remove any fibrils or amyloidogenic seeds (E. Renard, et al., 1996).

The challenge of insulin degradation in pumps, whether external (Cobelli et al., 2012) or implanted (Bally, et al., 2017), may be addressed by development of ultra-stable, fibrillation-resistant insulin analogs. Two classes of insulin analogs have been shown to be refractory to fibrillation: single-chain insulins (SCIs) containing a foreshortened C domain (Fig. 1A) (Q. X. Hua, et al., 2008) and two-chain insulin analogs containing an

343 engineered non-canonical disulfide bridge between the A and B chains (T. N. Vinther, et al., 2013) (Fig. 1B). In each case the altered topology (connectivity) of the polypeptide is thought to be incompatible with regular cross-b assembly, the general structural mechanism underlying formation of amyloid (C. M. 14685248 Dobson, 2003; Tycko,

2011; Tycko & Ishii, 2003). Although such analogs have not been tested clinically, insertion of a non-canonical cystine is associated in rat studies with anomalously prolonged duration of activity on intravenous bolus injection (T. N. Vinther, et al., 2013); this would be an unfavorable feature of a pump insulin.

Polymer and Matrix-Based GRIs

Polymer- or matrix-based GRI systems rely on sequestration of native or derivatized insulin molecules within a matrix suitable for SQ injection. The matrix is designed to sense ambient glucose concentrations and release a proportional amount of insulin for systemic absorption. Three classes of glucose-sensitive motifs permit glucose-dependent insulin sequestration: (i) glucose-binding proteins (GBPs), a class that includes lectins like concanavalin A (ConA), (ii ) glucose oxidase (GoD), an enzyme that catalyzes oxidation of glucose to gluconic acid with release of a proton (hence lowering the pH), and (iii) phenylboronic acid (PBA; Figure. 7-2), which forms reversible ester linkages with diol-containing molecules, including glucose itself (Table 7-1).

Brownlee and Cerami created the first model GRI system: glycosylated insulin complexed with ConA (Brownlee & Cerami, 1979; Zaykov, et al., 2016). This complex was intended to sequester insulin in the SQ space during normoglycemia and release the hormone during hyperglycemia via competition with ambient glucose molecules.

Although the strategy was successful in vitro, competitive setpoint was above typical

344

Figure 7-1 Ribbon representation of structure of ultra-stable insulin analogs. (A)

Single chain insulin (SCI) contains the native A-domain (dark gray) and B-domain (light gray) and the three native disulfide linkages (yellow) of insulin. In addition, SCI analogs contain a foreshortened “C-domain” (6-8 amino acids in length) (orange) that connects the C-terminal strand of the B-domain (green) to the N-terminus of the A-domain. This domain dampens the conformational fluctuations of the molecule thus increasing its thermodynamic stability and rendering it resistant to fibrillation (Q. X. Hua, et al., 2008).

(PDB ID: 2LWZ) (B) 4SS-insulin is a two-chain insulin analog (A-chain, dark gray, B- chain, light gray) that contains an engineered fourth disulfide linkage (red asterisk) between the A- and B-chains in addition to its three native disulfide linkages (yellow).

The analog was reported to be resistant to fibrillation and have increased temperature stability (T. N. Vinther, et al., 2013). (PDB ID: 4EFX)

345

346

hyperglycemic concentrations (Brownlee & Cerami, 1983). Later studies suggested that the immunogenicity and mitogenicity of ConA would impair clinical feasibility

(Ballerstadt, Evans, McNichols, & Gowda, 2006; Coutinho, Larsson, Grönvik, &

Andersson, 1979).

Advances in materials science have allowed development of polymer-based GRI systems. In such systems insulin is encapsulated within polymer matrices as a “smart” SQ depot. Polymeric vesicles (polymerosomes) and micro- or nanoparticles encapsulate insulin in a controlled SQ environment that may mitigate foreign-body reactions or immune responses (Fig. 3A) (Frost, 2007; Lopes, Santos, Barata, Oliveira, & Lopes,

2013) Such nanosystems are more amenable to fine tuning than is ConA complexation

(Figure 7-4A).

Polymer-based GRIs can employ a variety of encapsulation chemistries. These include poly-ethylene glycol, poly-N-vinyl-pyrrolidone, and succinyl-amidophenyl- glucopyranoside (Kitano, Koyama, Kataoka, Okano, & Sakurai, 1992; Y. M. Lee, Kim,

& Cho, 1996; Podual, Doyle, & Peppas, 2004; Ravaine, et al., 2008) or designed bio- molecules such as modified peptides or lipids (Anirudhan, Nair, & Nair, 2016; X. Li et al., 2017). Whereas the matrices are impermeable to insulin during normoglycemic or hypoglycemic conditions, their permeability may increase as a result of structural changes that cause swelling or increased water solubility of the polymer in

347

Table 7-1 Summary of GRI Technologies

Motif and Examples Advantages Disadvantages Implementation Description

Glucose Binding wheat germ -Naturally -Non- -Pre-bound Proteins agglutinin (WGA) occurring (Weis physiological complexes with & Drickamer, glucose insulin (Brownlee Bind glucose and concanavalin A 1996) affinity(Drickamer & Cerami, 1983) oligosaccharides (ConA) -Scalable & Taylor, 1993) GLUT1 - Immobilized in Manufacture(Paul - hydrogels ová, Tichá, Immunogenicity( Entlicher, Koštíř, Ballerstadt, et al., -Used to coat & Kocourek, particles 1970) 2006) -ConA mitogenic - Structural -Structurally molecule in robust (Paulová, et polymeric al., 1970) matrices

Glucose-Oxidase Glucose Oxidase- - Available -Peroxide - Coupled to 1 (Wilson & Turner, byproduct is toxic insulin Catalyzes 1992) oxidation of Glucose Oxidase- -Active at low - Co-formulated glucose to 2 - High glucose glucose with insulin gluconic acid, specificity concentrations glargine produces H2O2 as (Janssen & a byproduct Ruelius, 1968) -pH changes - Encapsulated in difficult to reverse pH, hypoxia, and - Easily coupled to rapidly peroxide-sensitive other molecules(Ngo & polymer matrices Lenhoff, 1983) (Leathem & Brooks, 1987) - Stable(Coulthard et al., 1945) - High turnover (Janssen & Ruelius, 1968) Boronic Acid Boronic acid, - Small molecule -Interaction not -Coupled to Phenylboronic specific to glucose insulin for co- Forms reversible - Easily used as Acid, 3-Fluoro-4- (Deshayes et al., injection with esters with cis- derivative Carboxy 2013) polyols diols Phenylboronic (Friedman & Acid, Amino- Pizer, 1975) (Z. - Limited - PBA polyol co- Phenylboronic Guo, Shin, & translational derivative insulin Yoon, 2012) Acid function -Coupled to - Forms - Can degrade insulin via fatty- reversible,

348 covalent bonds acyl linkers (Yan, 2004) (Springsteen & -Co-polymerized Wang, 2002) with diol- containing - Glucose affinity molecules in in physiological polymeric range(Thomas matrices Hoeg-Jensen, et al., 2005) - PBA chemical properties can be modified using substituent groups (Nicholls & Paul, 2004; Yilmaz et al., 2015) (Matsumoto et al., 2012) -Sensitive to pH and glucose levels(Y. Wang, Zhang, Han, Cheng, & Li, 2012) - No toxicity(W. Yang, Gao, X., and Wang, B., 2003)

349

Figure 7-2 Schematic of phenylboronic acid-monosaccharide complexation.

Phenylboronic acid (PBA) comprises an aryl ring containing a boronic acid substituent group. When the boronic acid moiety is ionized to phenylboronate, the group may form reversible ester bonds with cis-diols, including those found on monosaccharides

(Springsteen & Wang, 2002). The pKa of unmodified PBA is 8.0 and it has an affinity for glucose of approximately 10 mM. Both of these properties may be modulated by introducing electron-withdrawing substituents to the aryl ring (red circle marked with

“X”) (Z. Guo, et al., 2012). A variety of groups, including nitro-, fluoro-, carboxyl, and sulfone groups have been shown to lower pKa of PBA to below physiological pH (7.4), affinities of these derivatives for glucose range from 8-14 mM (Thomas Hoeg-Jensen, et al., 2005).

350

351 response to an increase in interstitial glucose concentration (Figure 7-3B). To this end, molecular glucose sensors (GBPs, GoD, and PBA) have been incorporated into such polymer scaffolds.

Several strategies have been tested to incorporate glucose sensors into polymer-based

GRIs. These include insulin-encapsulating particles composed of matrices co-derivatized with PBA or other boronic acids (Jin et al., 2009; H. Wang, Yi, Yu, & Zhou, 2017; Y.

Wang, Huang, Sun, Gao, & Chai, 2017), dot-immobilized carbon or glucose ester-based crosslinks (Y. Dong et al., 2016). Immobilized GBPs have also been exploited in conjunction with glucose-modified polymers. The latter strategy underlies SmartInsulin® as developed by SmartCells, Inc. (C. Wang et al., 2017; Zion, Tsang, & Ying, 2003).

GoD has also been encapsulated in matrices that are chemically sensitive to H2O2, hypoxia, or decreases in local pH. Unlike PBA and GPB-based technologies, polymeric matrices in GoD-based systems isomerize (X. Li, et al., 2017) or are chemically modified by redox reactions (Anirudhan, et al., 2016) that occur as a consequence of GoD activity, resulting in increased water-permeability of the encapsulating polymer (Figure 7- 3C).

GoD-based polymerosomes were recently employed by Yu and colleagues (2017) (J.

Yu et al., 2017) to encapsulate insulin in self-assembling polymersomes composed of poly(ethylene glycol) (PEG) and polystyrene (as modified with 2-nitroimmidazole through a thioether linker). In this elegant scheme, local hypoxia induced by GoD activity promoted reduction of 2-nitroimmidazole into the more hydrophilic 2-aminoimmidazole by SQ reductases, causing the encapsulating polymer to swell and become more water- permeable. The thioester linker served a dual purpose: to scavenge H2O2 (a byproduct of

GoD activity) and in so doing, through conversion of the thioester to a sulfone, further

352

Figure 7-3: Cartoon representation of polymer-based GRIs. (A) Insulin is encapsulated within a polymeric matrix in large implants, transdermal patches, or micro- or nanoparticles. The native or derivatized insulin analog is typically sequestered within a water-containing cavity within the hydrophobic, water impermeable polymer matrix.

Such matrices are compact during hypo- or euglycemia, but swell during hypoglycemia to release sequestered insulin (B). Several different methods have been used to create glucose-responsive polymers (C). (Top panels, C) PBA (green hexagon) and glucose have been incorporated into polymeric scaffolds to maintain the compactness of the matrix. This interaction is competitively disfavored as ambient glucose concentrations rise, causing the polymer to swell (top panels, C). (Middle panels, C) Immobilized or co- encapsulated GBPs (red shapes) have been used as agents that stabilize the compactness of polymeric matrices in a glucose-dependent fashion. (Bottom panels, C) Matrices that are sensitive to the byproducts of co-encapsulated or immobilized GoD (red polygon), gluconic acid and H2O2, along with the resulting decrease in pH and local hypoxia, have been developed. Oftentimes, polymeric matrices will undergo conformational changes resulting from changes in the protonation state of components (blue circles, red lines) of the scaffold or from reactions that are catalyzed by the byproducts of GoD .

353

354 increase the permeability of the matrix. After in vitro confirmation of the mechanism of action, the polymersomes were integrated with cross-linked hyaluronic acid microneedle arrays for painless transcutaneous delivery (Xie, Li, & Yu, 2015). This system was able to lower blood-glucose levels in diabetic mice within 60 minutes and maintain adequate glycemic control for approximately five hours. Critically, the system did not induce hypoglycemia in mice pre-treated with insulin.

Challenges faced by polymer-based GRI technologies include limited particle stability

(Shi et al., 2016) and suboptimal response rates: too slow in responding either to an increase or decrease in GBL (respectively leading to hyperglycemic or hypoglycemic excursions). Addressing these limitations has often yielded paradoxical results: sensitizing matrices to hyperglycemia, for example, can limit their ability to attenuate insulin release at low glucose levels. The latter problem can be exacerbated by matrix degradation, raising the risk of a severe hypoglycemic episode. (Gu et al., 2013; Yesilyurt et al., 2016)

Several groups have addressed the above challenges by creating GRI systems with multiple and complementary molecular mechanisms of glucose sensing and response. An example is provided by the combination of GoD, pH-sensitive matrices, and PBA modifications of the matrix or encapsulated insulin itself (J. Li et al., 2017; Y. Wang, et al., 2012). Such a nanotechnology may respond rapidly to GoD activity and also retard the release of insulin during normoglycemia through its interactions with diols in the polymer matrix (Kikuchi et al., 1996; Matsumoto, Yoshida, & Kataoka, 2004; Yesilyurt, et al., 2016). Variation in particle size and permeable surface area (L. Sun, Zhang, Zheng,

Wu, & Li, 2013) and addition of multiple layers of polymeric matrix (Shi, et al., 2016)

355 have also been shown to modulate glucose-response rates. In several micro- and nanogel systems insulin has been encapsulated in largely impermeable polymer matrices that contain only small patches (or “pores”) of glucose-responsive polymeric material. This strategy seeks to limit the maximal rate of glucose release (thus mitigating risk of extreme hypoglycemia) while maintaining a rapid response to hyperglycemia (Gu, et al.,

2013).

An alternative strategy for polymer-based GRI systems exploits the self-regulatory capability of cadaveric pancreatic β-cells or those cultured from pluripotent stem cells

(SC-β) to create a glucose-responsive system (Hirshberg, 2007). A major barrier to β-cell implantation is posed by the foreign-body response (FBR) (Robertson, 2004).

Encapsulation in polymeric matrices has long been investigated as a means to isolate implanted cells from such tissue reactions at the injection site (Lim & Sun, 1980).

Whereas several studies have shown the ability of polymer-encapsulated SC-βs to provide glycemic control in immunosuppressed diabetic animals (Lum et al., 1991;

Schneider et al., 2005), the first example of a functional polymer-encapsulated SC-β GRI in an immunocompetent animal was shown by Vegas and colleagues in 2016 (Vegas et al., 2016). SC-βs were encapsulated in 1.5 mm microspheres composed of azole- modified alginate (triazolethiomorpholine dioxide; TMTD). This demonstrated that in diabetic mice such particles maintained euglycemia with no hypoglycemic episodes for

150 days. Further, the particles elicited only limited immune activation and FBR as assessed by analysis of immunological molecules adsorbed to the surface of extracted particles and of fibrosis-associated biomolecular markers (M. Kolb et al., 2002). The large size of the implanted particles (relative to other polymer-based GRI devices) and

356 the derivatization of the alginate matrix with imidazole were identified as key innovations

(Vegas, et al., 2016). These modifications have not yet been shown to mitigate immune response in humans, however, potentially limiting clinical applicability (Elliott et al.,

2005; Vegas, et al., 2016).

Molecular GRIs

Molecular GRIs represent a novel class of molecules in which the insulin molecule or its formulation confers glucose-dependent activity or bioavailability. Present candidate technologies rely on sequestration of active insulin hormone within the SQ space or within the bloodstream (as inactive complexes) with enhanced release or activation only during hyperglycemia. Although such strategies remain in the early stages of development, the potential convenience and cost-effectiveness of molecular GRIs make them an attractive target for future analog design.

An early strategy employed an insulin-GoD fusion molecule (Ito & Imanishi, 1994).

A cysteine-based linkage was broken as the enzyme oxidized glucose. Although proof of principle was obtained in vitro, the low Km of GoD for glucose led to liberation of the hormone under hypoglycemic as well as hyperglycemic conditions (Ito & Imanishi,

1994). In addition, the release of hydrogen peroxide by GoD (as a byproduct of glucose oxidation) could damage surrounding tissues via generation of reactive oxygen species

(ROS). In an unrelated approach, GoD-dependent changes in ambient pH were exploited by Kashyap and colleagues to modulate the solubility (and hence bioavailability) of insulin glargine, the active component of Lantus (Hilgenfeld, Seipke, Berchtold, & D.R.

Owens, 2014). Because this basal analog is insoluble at neutral pH but soluble under acidic conditions, GoD-mediated acidification (via production of gluconic acid) was

357 predicted to increase the bioavailability of the insulin analog (Gillies, et al., 2000) (Fig.

4B). Although this mechanism-based approach showed promise in vitro and demonstrated some improvement in glycemic control in pilot animal studies, no comprehensive report has been published (Kasyhap, 2010).

PBA was first described in the context of a GRI system by Hoeg-Jensen et al.

(Thomas Hoeg-Jensen, et al., 2005). The group demonstrated that derivatives of PBA could be coupled to insulin without affecting its biological activity, in turn enabling the analog to bind diol-containing sequestering agents (Figure 7-4C). The investigators further demonstrated that insulin analogs co-derivatized with PBA and polyols could form glucose-dependent multimeric complexes in vitro (Figure 7-4D) (T. Hoeg-Jensen,

Havelund, & Markussen, 2008), (T. Hoeg-Jensen et al., 2008). This form of glucose responsiveness has been engineered within insulin degludec, a basal acylated insulin (I.

Jonassen, et al., 2012). To our knowledge, no in vivo results have been described.

Recent use of a PBA-based GRI was described by the Langer-Anderson groups at MIT

(Chou et al., 2015). This study employed an acylated insulin analog (detemir, the active compotent of Levemir®; Novo-Nordisk). Insulin detemir contains a myristic acid coupled to a unique lysine (LysB29), which mediates binding to serum albumin and so provides a long-lived circulating depot. The MIT groups sought to derivatize the acyl tag with PBA so that its affinity for albumin would be glucose-responsive (Figure 7-3E; (Chou, et al.,

2015). Although this goal was not achieved in vitro, several of the candidate GRI analogs nonetheless exhibited glucose-responsive biological activity in a peritoneal glucose- infusion assay in mice. Molecular mechanisms underlying these provocative findings were not defined.

358

Figure 7-4 Cartoon representation of derivatization of the insulin molecule in GRI technology. The insulin hormone comprises two polypeptide chains that are designated

“A” and “B” and are connected by two disulfide linkages spanning residues A7 and B7 and A20 and B19. Brownlee, et al created a molecular GRI by coupling mono- and disaccharides to the N-termini of one or both polypeptide chains of insulin (red). These analogs were bound to ConA (red star) before administration. ConA was expected to sequester the insulin in the SQ space during euglycemic conditions (left, A) and allow it to liberated by competitive binding of ambient glucose (blue hexagons) during hyperglycemia (right, A (Brownlee & Cerami, 1979), (Brownlee & Cerami, 1983). B)

Kashyap and colleagues created a GRI system using the clinical analog insulin glargine.

Glargine contains an addition of two arginine residues at the C-terminus of its B-chain and a substitution of AsnA21 for glycine (Kasyhap, 2010). These modifications shift the isoelectric point of the protein to near physiological conditions causing it to form precipitates after SQ injection. In this GRI system glargine was co-injected with GoD

(red polygon), which was expected to lower the local pH by oxidizing glucose to gluconic acid at a rate proportional to the glycemic state of the patient increasing the solubility and, hence, the bioavailability of glargine. (C-E) A number of groups have modified the ε-amino group of LysB29 with PBA derivatives (green hexagons) to create

GRI systems. (C) Hoeg-Jensen and colleagues directly coupled LysB29 with PBA derivatives and demonstrated the ability of the analog to couple to diol-containing polymer carriers (orange rectangles with red circles) in a glucose-dependent fashion

(Thomas Hoeg-Jensen, et al., 2005). (D) The same group also derivatized residue B29 with a molecule containing a PBA and a polyol group (black lines with red circles) and

359 demonstrated the ability of the analog to form multi-hexameric complexes in vitro that could dissociate in a glucose-dependent fashion (T. Hoeg-Jensen, Havelund, Markussen, et al., 2008). (C) Most recently, Chou and colleagues created a GRI that contained a PBA derivative coupled via a fatty acyl linker (black, jagged line with green hexagon) to

LysB29, it was hypothesized that this analog would bind to albumin (blue oval) under normoglycemic conditions and become liberated during hyperglycemia as free glucose complexed with the PBA molecule and decreased the affinity of the analog for albumin

(Chou, et al., 2015). Although this analog did show glucose-responsiveness in a mouse model, glucose-dependent association or dissociation from albumin was not confirmed.

360

361

A new avenue for molecular GRI design may be opened by recent crystallographic studies of insulin bound to fragments of the IR ectodomain (J. G. Menting, et al., 2013;

J.G. Menting, et al., 2014). Such studies have revealed a major change in the conformation of insulin on receptor binding (Figure 7- 5). It may be possible to exploit this mechanism to design a glucose-dependent conformational switch such that binding of the modified insulin to the IR is impaired under conditions of hypoglycemia. This notion represents an active area of research.

Clinical Significance and Conclusions:

The development of safe and effective GRIs promises to enhance the health and quality of life of patients with T1DM and of patients with T2DM refractory to oral therapy. Strict glycemic control has been shown to retard or prevent microvascular complications in T1DM (DCCT, 1993b) and is likely to be beneficial in early stages of

T2DM; (Duckworth et al., 2009; Patel et al., 2008; UKPDS, 1998). Aggressive insulin regimens nonetheless increase the acute and long-term risks of hypoglycemia (Control &

Group, 1997; Gerstein et al., 2008; Group, 1991). Of particular concern as patients with

T1D live longer with current standards of medical management, recurring hypoglycemic episodes are associated with cognitive decline (P. E. Cryer, Davis, S.N., Shannon, H,

2003).

In summary GRI technology offers the hope to provide intensive glycemic control without increased risk of hypoglycemia and with less burden on patients (Beregszàszi, et al., 1997). Whereas CGM-pump-based GRIs presently employ the most mature component technologies, recent innovations in matrix-based and molecular GRIs are

362

Figure 7-5 Structural representation of insulin IR binding. (Original Figure) (A) The ectodomain of insulin receptor (IR) (anterior monomer of homodimer shown in color), the molecular target of insulin. The L1 and αCT (purple) domains of IR comprise the insulin binding site. (B) In its “closed” or storage conformation, insulin cannot engage its binding site due to a steric clash between αCT (purple) and the C-terminal B-chain of insulin (brown). (C) To engage its primary binding site, insulin must undergo a conformational change in which the C-terminal B-chain detaches from the hydrophobic core of the globular protein via unfolding of a β-turn spanning residues B20-B23 (red box in B and C). (D) The detachment of the C-terminal B-chain from the hydrophobic core of insulin allows the molecule to intercalate between the L1 and αCT domains and occupy its binding site.

363

364 elegant and suggest promising routes toward functional replacement of the pancreatic β- cell. We anticipate continuing progress in the coming years.

7.2 Design of Glucose-Responsive Insulin Analogs Containing Internal Glucose-

Sensing Motifs

This chapter describes the initial studies in an interdisciplinary, collaborative project to develop a molecular GRI with an internal glucose-sensing motif. The design of this novel class of insulin analogs is unique in that it does not rely on co-formulation with an external agent, such as GOD, and potentially does not require sequestration in an artificial or naturally-occurring reservoir in order to have glucose-responsive activity.

The design of a molecular GRI that functions in the absence of a separate pharmacological entity relies on the identification of a “switch” that controls insulin activity. Such a structural switch in insulin was identified by the landmark co-crystal structure of insulin bound to a fragment of IR containing the L1 and CR domains and a soluble αCT peptide (collectively referred to as μIR). The crystal structure revealed the requirement for the separation of the C-terminal B chain of insulin from its hydrophobic core via the unfolding of the B20-B23 β-turn (J.G. Menting, et al., 2014). The inability of the C-terminal B chain to separate from the globular core of insulin prevents the exposure of residues (including B24, B25, B26, A2, and A3), which form critical interactions at the insulin-IR interface, resting in the inability of insulin to engage IR (see Figure 7-5, above) (J. G. Menting, et al., 2013). This concept is illustrated by single chain insulin precursors miniproinsulin (MPI) and porcine insulin precursor (PIP), in which the C- terminus of the insulin B chain is connected by a peptide bond either directly or via a two

365 amino-acid linker to the N-terminus of the A chain. Such a linkage prevents the C- terminal B chain of insulin from separating from the core of the hormone causing both

MPI and PIP to be biologically inert (Q. X. Hua, et al., 1998; Q. X. Hua, et al., 2008;

Kobayashi, et al., 1989).

Our design aimed to exploit the separation of the C-terminal B chain of insulin from its globular core through the creation of glucose-dependent MPI- or PIP-like linkage. 3- fluoro-4-carboxyphenylboronic acid (PBA-F) was selected as the glucose-sensing motif in our design due to its physiologically relevant affinity for glucose (Kd = 10 mM)

(Thomas Hoeg-Jensen, et al., 2005), its chemical stability, and its lack of toxicity (Chou, et al., 2015). Our design envisioned the creation of a glucose-dependent linkage between a PBA-F molecule coupled to the N-terminus of the insulin A chain and a small molecule containing a cis-diol coupled to a residue near the C-terminus of insulin B chain. During hypo- or euglycemia, the tether was expected to remain intact, as the proximity of the

PBA-F and diol overcame entropic barriers to the formation of ester linkages between the two motifs. However, in hyperglycemic conditions, the PBA-F moiety was expected to favor binding to ambient glucose molecules, thus liberating the C-terminal B chain and allowing it to adopt its receptor-binding conformation.

The formation of an intramolecular PBA-diol linkage was expected to have a number of consequences on the biophysical properties of the GRI. First, the intramolecular nature of the linkage was expected to provide reversible activation of the GRI. That is, once broken, the PBA-diol linkage was could be expected to re-associate once ambient glucose concentrations returned to euglycemic levels. Second, GRIs were expected to have glucose-dependent hexamer dissociation rates: the formation of the PBA-diol linkage was

366 expected to dampen conformational fluctuations of the C-terminal B chain of insulin, which comprises a majority of the dimer interface of insulin (D. T. Birnbaum, et al.,

1997). Finally, the PBA-diol linkage is expected to protect the GRI from amyloid fibril formation by preventing the C-terminal insulin B chain from detaching from the hormone’s globular core, a step that is required for the nucleation of amyloids (evidenced by the resistance of MPI and PIP to fibril formation) (Q. X. Hua, et al., 2008; Huang, et al., 2005).

7-3 Results

Development of Candidate GRIs

Design of candidate GRIs was rationalized by the structure of insulin in its storage

(closed) and receptor-bound (open) conformation. To develop a broad range of analogs that could be screened for glucose-responsive activity, several components of the GRI design were varied as follows. (i) Candidate analogs would contain PBA-F coupled to either the N-terminus of the insulin A chain or the N-termini of both the A and B chains.

(ii) The position of the diol-containing moiety at the C-terminal B chain would be varied.

Diols at position B27-B30 and the extended positions B30 and B31 would be tested. (iii)

The character of the diol-containing moiety would be varied. Diol containing molecules would include monosaccharides, disaccharides, linear diols (such as gluconic acid), and aromatic diols such as gallic acid and the non-standard amino acid 3,4- dihydroxyphenylalanine (DOPA). (iv) The “scaffold” used to couple the diol-containing- and PBA moieties to the insulin protein will also be varied. PBA-F molecules may be coupled to the N-termini of the insulin A- and B chains by acyl linkers ranging from 4 to

12 carbons in length. The use of longer linkers is expected to confer a basal PK/PD

367 profile to the GRI analog (via binding to circulating albumin) (I. Jonassen, et al., 2012) and potentially allow inter-hexameric PBA-diol complexation (T. Hoeg-Jensen,

Havelund, Markussen, et al., 2008). The scaffolding of the diol containing molecule will be varied: diol-containing small molecules will be coupled to the side chains of various amino acids include D/L-lysine, ornithine, diaminobutyric acid, cysteine, threonine, and serine. The non-standard amino acids, L- and D-DOPA, which contain an intrinsic planar diol, will also be tested. A summary of the design scheme is provided in Figure 7-6.

Creation of a Prototype GRIs

The prototype set of GRIs contained PBA-F moieties coupled to the N-terminus of the insulin A chain or the N-termini of both the A- and B chains. The des-[B230-B30]- octapeptide insulin (DOI) precursor for synthesis of the latter class of GRIs was readily created by the treatment of a tryptic fragment of insulin with an N-hydroxysuccinamide ester of PBA-F, which resulted in the coupling of PBA-F to both N-termini. However, selectively coupling the N-terminus of the insulin A chain proved to be more challenging.

To overcome this challenge, creation of A-chain-coupled DOI utilized “TK insulin,” an insulin molecule that contains N-terminal extension of the B-chain. Treatment of TK insulin with PBA-F-NHS ester produced an insulin molecule with PBA-F coupled to both the N-terminus of the insulin A-chain and the N-terminus of the TK extension.

Trypsinization of this product produced DOI with PBA-F coupled to N-terminus of the A chain and no modification of the B chain (Figure 7-7). Both single- and double-PBA-F- coupled DOI products were purified by HPLC. Full-length insulin analogs were created by trypsin-catalyzed semisynthesis: the prototype set of analogs contained O-linked monosaccharides or disaccharides at threonine residues B27 and B30.

368

Figure 7-6. Schematic of GRI Design The polypeptide chains of insulin are shown as blue rectangles and the canonical disulfide bridges are shown as black lines. In our design, PBA-F molecules will be coupled to the N-terminus of the insulin A chain

(indicated by the red triangle) via acyl linkers of varying length (shown in the red box).

Diol-containing molecules will be introduced at various positions along the insulin B chain (a monosaccharide is shown as a yellow hexagon with red circles indicating hydroxylgroups; alternate positions for diol placement are indicated by the green asterisks). The character and coupling scheme of the diol will be varied to optimize orientation, affinity for PBA-F, and length of the linkage (green box).

369

370

Figure 7-7. Scheme for PBA-F Coupling to the N-terminus of the Insulin A chain

The “TK” single chain precursor was used to achieve high-yield A-chain-specific PBA-F coupling. (a) The TK-SCI was digested with Endolys-C enzyme removing the linker domain of the SCI. This exposed the N-terminus of the A chain for PBA-F coupling.(b).

The resulting two-chain insulin was treated PBA-F NHS ester resulting in an insulin analog with PBA-F coupled to the N-terminus of the insulin A chain and to the N- terminus of the TK extension on the insulin B chain (c) The double-coupled product was digested with trypsin, which removes the TK extension and C-terminal B chain of insulin producing the tryptic fragment des-[B23-B30]-octapeptide insulin (DOI), with PBA-F coupled to the N-terminus of its A chain. (d) the DOI could then be synthesized into a diol-containing full-length insulin analog through trypsin-catalyzed semisynthesis. Diol- containing peptides are synthezied by solid-state protein synthesis.

371

372

Although several candidate GRI molecules were synthesized, only those analogs containing mannose at the B30 position could be produced in large quantities. This was largely the result of the dissociation of the O-linked mono- or disaccharide from the threonine side chain. Nevertheless, mannose-B30 analogs served as reliable prototypes for our GRI design.

The biological activity B30-mannose analogs with PBA-F coupled to one, both, or neither of the polypeptide chains of insulin was tested by IV injection in diabetic rats.

Without modification with PBA-F, Mannose-B30 insulin retained native biological activity, whereas analogs modified by PBA-F at the N-terminal A chain exhibited a

~15% decrease in potency (Figure 7-8A). Analogs modified by PBA-F at the N-termini of both polypeptide chains displayed minimal biological activity (Figure 7-8B). The minimal activity of double-PBA-F-coupled insulin lacking a C-terminal diol suggested that it was the modifications of both N-termini that caused the decrease in activity. Single coupled (1xPBA-F) analogs with and without mannose at the B30 position displayed a delated onset of biological activity accompanied by an increased duration of action after

SQ injection in diabetic rats (Figure 7-8C). This finding suggested that the PBA-F moiety was decreasing the rate of absorption of insulin from the SQ depot possibly due to interactions with diol-containing molecules in the interstitial space. A dose-response experiment of SQ injection of the prototype GRI revealed that the initial rate of fall in

BGL could be overcome by increasing the dosage of the PBA-F-coupled analog (Figure

7-8D). A dose of 40μg of 1xPBA-F Mannose-30 insulin produced a rate of fall equal to that of 10μg of insulin lispro (KP) for the first 50 minutes after injection. Surprisingly, although the initial rate of fall after SQ injection of 60μg of the candidate GRI was

373 slightly greater than that after injection of 40μg, the total activity associated with the two doses appeared to be similar. This may be the result of more rapid degradation of the

PBA-coupled analog in the interstitial space or bloodstream, or it may simply be the result of a the small sample size used in the experiment.

The hypothetical complexation of the O-linked mannose at position B30 with the N-

2+ terminal PBA-F was tested by measuring the dissociation rate of Co -substituted R6

(phenol stabilized) insulin hexamers. Single-chain insulin analogs form more kinetically stable hexamers as a result of the conformational dampening of the C-terminal B chain, which comprises a majority of the insulin dimer interface. For this reason, if a PBA-F- mannose linkage were to form, it would be expected that the R6 hexamers formed by

1xPBA-F, mannose B30 insulin would exhibit decreased dissociation rates in the absence of glucose than in its presence. Whereas unmodified KP-insulin, and mannose-B30-KP

(unmodified by PBA-F) displayed similar hexamer dissociation rates in the absence of glucose as they did in the presence of 25mM glucose, a nearly two-fold decrease in the hexamer dissociation rate of 1xPBA-F, B30-mannose-lispro in the absence of glucose in relation to that observed in the presence of 25 mM glucose (Figure 7-9). This suggested that a B30-A1 linkage was formed in the absence of glucose. A similar trend was observed for a 2xPBA-F, mannose-B30, which displayed more rapid hexamer dissociation in the presence of glucose than in its absence. Remarkably, in both the presence and absence of glucose, hexamer dissociation, as reported be attenuation of Co2+ signal at 574 nm, of the double-coupled insulin analog was not consistent with monoexponential decay. This suggested the potential formation of inter-hexamer interactions through the formation of PBA dimers through B-chain-coupled PBA groups of

374

Figure 7-8. Activity of Prototype GRI Candidate Analogs in Diabetic Rats (A) IV potency of candidate GRI analog in diabetic rats in relation to native KP (lispro), and glycosylated control. The prototype GRI exhibited approximately 20% decrease in activity in relation to controls. (B) IV potency of a prototype GRI with PBA-F coupled to the termini of both polypeptide chains. This analog displayed markedly reduced biological activity in relation to native KP. (C) Pharmacodynamic profile of 1xPBA-F,

Mannose B30-insulin, 1xPBA-F KP, and KP-insulin after SQ injection. Analogs with

PBA-F coupled to the N-terminus of the A chain display a delay in the onset and subsequent prolongation of biological activity. (D) SQ dose-response experiment of

1xPBA-F, Mannose B30 in diabetic rats. A 40μg dose of the prototype GRI appeared to have the same initial rate of fall as 10μg KP. A 60μg dose of the GRI appeared to have a similar profile to that of the 40μg dose.

375

376

neighboring hexamers. The disruption of such interactions in the presence of glucose resulted in the hexamer dissociation curve of the double-coupled analog displaying characteristics more closely resembling monoexponential decay. This hypothesis was further strengthened by a similar pattern of hexamer dissociation of a 2xPBA-F insulin analog lacking a diol group at its C-terminal B chain. That the putative PBA-PBA interaction was formed by the B-chain-coupled PBA-F group was indicated by similar hexamer dissociation rates of a 1xPBA-F insulin analog lacking a B-chain diol modification. The hexamers formed by this analog exhibited monoexponential decay in both the presence and absence of glucose (results are summarized in Table 7-2).

Glucose Responsiveness of Prototype GRI in Vivo

The glucose responsiveness of 1xPBA-F, mannose-B30 lispro was tested in a non- diabetic rat model. After a dose-response experiment in diabetic rats, a dose of 40 μg of the GRI candidate, determined to be approximately equivalent to 10 μg of naïve lispro was injected SQ in diabetic rats. Although results of this experiment suggested a potential reduction in biological activity of the GRI candidate in euglycemic, non-diabetic rats, the results were largely inconclusive.

Positional Screening of Diol-Containing Molecules

The lack of chemical stability, expense, and cumbersome nature of the synthesis of glycosylated peptides motivated the use of novel diol-containing molecules in the development of GRIs. To this end, the non-standard amino acid, dihydroxyphenylalanine

(DOPA), was used to scan positions along the C-terminal B chain to determine the optimal diol placement for PBA-diol complexation. The use of DOPA offered several distinct advantages over that of monosaccharides. As an amino acid, DOPA could be

377

Figure 7-9 Glucose-Dependent Hexamer Dissociation Rates of Prototype GRIs

378

Dissociation rates of phenol-stabilized metal ion-coordinated insulin hexamers were determined using visible range spectroscopy of Co2+-substituted insulin hexamers.

Dissociation rates were measured in absence of glucose or in the presence of 25mM glucose. Results are summarized in Table 7-2.

379

380

Table 7-2. Glucose-Dependent Hexamer Dissociation Rate of GRI Candidates. analog t1/2 hexamer dissociation (min ± Error ) No Glucose 25 mM Glucose KP (lispro) 1.4 ± 1.5 ± B30 Mannose KP 1.1 ± 1.2 ± B30 Mannose OrnB29 1.2 ± 1.1 ± 1xPBA-F B30 Mannose KP 5.6 ± 3.2 ± 2xPBA-F B30 Mannose OrnB29 37.4 ± 13.3 ± 2xPBA-F KP 78.2 ± 36.0 ± 1xPBA-F Orn B29 3.4 2.9 ± 0.3

381

Figure 7-10. Activity of Candidate GRI in Non-Diabetic Rats The biological activity of 1xPBA-F Mannose B30 insulin was tested in non-diabetic rats. A 40μg dose, equivalent to 10μg KP, was injected SQ. The biological activity of this dose exceeded that of the standard 3μg dose of KP. No conclusions regarding glucose-responsive activity could be drawn from these data.

382

383 incorporated into octapetpide during its chemical synthesis eliminating cumbersome coupling steps. Additionally, unlike monosaccharides, which may adopt a number of conformations to minimize steric clashes with their local environment, the cis-diol of

DOPA is restricted to planar geometry. This property eliminates a confounding factor that may affect PBA-F-diol complexation. Finally, DOPA has been shown to bind PBA variants in vitro with a physiologically-relevant affinity of 12 mM (Z. Guo, et al., 2012).

The glucose-responsiveness of 1xPBA-F insulin analogs containing DOPA at the B28,

B29, or B30 positions was tested in diabetic rats. These analogs contained an additional substitution of ThrA8 by His to improve IR affinity (Weiss, Hua, et al., 2002). After calibration in diabetic rats, equivalent dosages of GRI candidates and unmodified lispro were injected SQ in non-diabetic rats. Results suggested that 1xPBA-F, DOPAB29-lispro displayed no glucose-dependent activity, whereas 1xPBA-F, DOPAB30-lispro displayed an intermediate level of glucose responsiveness (as measured by attenuation of biological activity in non-diabetic rats). 1xPBA-F, DOPAB28-lispro displayed a significant degree of glucose-responsive insulin activity (Figure 7-11).

The B28 position was selected as the optimal position for diol placement based on the results of DOPA-scanning experiments. In addition to 1xPBA-F, DOPAB28-lispro, analogs containing dihydroxybutyric acid (DHBA) or gallic acid coupled to the ε-amino group of LysB28 were tested for glucose-responsiveness in rats. Of this subset of GRI candidates, 1xPBA-F LysB28-gallic acid-lispro displayed the greatest degree of attenuation of its activity in non-diabetic rats in relation to that in diabetic rats, indicating glucose-responsive activity. The analog containing DHBA coupled to LysB28 displayed little to no evidence of glucose-responsiveness (Figure 7-12). of

384

Figure 7-11 Activity of Positional DOPA Variants in Diabetic and Non-Diabetic Rats. The SQ pharmacodynamic profiles of 1xPBA-F insulin analogs with DOPA at positions B28, B29, and B30 were tested (A,C, and E, respectively) at 15μg and 30μg doses in relation to native KP (10 μg).(B,D,F) The SQ biological activities of B28, B29, and B30 DOPA analogs, respectively, in non-diabetic rats compared to a 3μg dose of KP, which was expected to be less than the equivalent dosage of PBA-F-coupled analogs used.

385

386

Figure 7-12 Activity of GRI Candidates with Aromatic Diols At LysB28 in Diabetic and Non-Diabetic Rats The pharmcodynamic profiles of 1xPBA-F analogs containing

(A) dihydroxybutyric acid (DHBA) or (C) gallic acid coupled to the ε-amino group of

LysB29 after SQ injection in diabetic rats. (B, D) Biological activity of the GRI candidate analogs after SQ injection in non-diabetic rats. Once again, the more than the equivalent dosage of GRI candidates in relation to 3μg KP was administered.

387

388

Figure 7-13. Comparison of Activity of GRIs in Diabetic and Non-Diabetic Rats

Comparison of the biological activity (calculated as area over the glucose-lowering curve; AOC) of GRI candidates with glucose-responsive activity in diabetic (left) and non-diabetic (right) rats. Data from diabetic rats are normalized against 10μg KP, and those from non-diabetic rats are normalized against 3μg KP. Black bars indicate standard deviation.

389

390

Such preliminary experiments motivated further investigation 1xPBA-F LysB28-gallic acid-lispro (B28 gallic acid-insulin). A complete set of dose-response experiments were performed in diabetic rats comparing the biological activity of B28 gallic acid-insulin to native lispro. The results of these experiments revealed an “S-shaped” dose-response curve B28 gallic acid-insulin (Figure 7-14A). The dose-response curve of B28 gallic acid- insulin indicated the existence of a “threshold” dosage for biological activity. This phenomenon may be explained by the binding of PBA-F to diols in the interstitial space, which prevents insulin from being absorbed into the bloodstream. Higher doses of insulin overcome this effect by saturating the diols at the SQ injection site thus causing a greater amount of insulin to enter systemic circulation. Although this finding somewhat confutes the evidence of glucose-response, dose-response experiments conducted in non-diabetic rats provided some evidence of glucose-dependent biological activity: higher doses injected in non-diabetic rats displayed diminished biological activity in non-diabetic rats in relation to that in diabetic rats (Figure 7-14B). Preliminary studies have utilized glucose-clamps (DeFronzo, Tobin, & Andres, 1979) to circumvent the experimental artifacts of dose-dependent pharmacokinetics. Non-diabetic rats were injected with equal doses of native lispro or B28 gallic acid-insulin after being clamped for 2 hours at BGLs of 100 mg/dl (corresponding to euglycemia) or 200 mg/dl (corresponding to mild hyperglycemia). Pilot experiments gallicB28-insulin displayed attenuated biological activity, as measured by increased glucose-infusion rate (GIR) after injection, in rats clamped at 100 mg/dl than those clamped at 200 mg/dl. Data were normalized against native lispro to verify relative activities.

391

Figure 7-14 Dose-Dependent Biological Activity of B28 Gallic Acid-Insulin (A) Dose- dependent biological activity of SQ injected B28 Gallic acid insulin analog in diabetic rats determined by calculating the area over the glucose-clearance curve (AOC). The blue arrow indicates the dosage equivalent to 10μg KP. (B) Corresponding graph of experiments in non-diabetic rats. The red arrow indicates the dosage equivalent to 4μg

KP in non-diabetic rats.

392

393

In spite of the encouraging in vivo evidence, biophysical assays have been unable to confirm the formation of a linkage between PBA-F and gallic acid. The far-UV CD spectra and thermodynamic stabilities (as determined by CD-monitored guanidine- denaturation assays) of gallicB28-insulin were identical in the absence or presence of 25 mM glucose.This indicated that the secondary structure was not affected by the formation of a hypothetical linkage between A1 and B28, not did such a linkage manifest itself in the denaturation curve of the analog. B28 Gallic acid-insulin was not able to form stable

R6 hexamers under formulation conditions. For this reason, comparison of hexamer dissociation rates in the absence and presence of glucose could not be used to assess

PBA-diol complexation. Finally, NMR spectra collected in the presence of 25 mM glucose and in its absence were identical. Such findings collectively suggest that a PBA- diol linkage does not form in the context of B28 gallic acid-insulin.

7-4 Discussion

Future work will focus on the creation of GRI analogs containing a linear polyol, gluconic acid, at the C-terminal B chain. The presence of multiple diols within gluconic acid which are each sterically accessible to PBA, along with increased in vitro affinity of

PBA for linear diols in relation to monosaccharides suggests a stronger linkage may be formed between the A and B chains of insulin analogs containing such a modification.

PBA-coupled analogs with gluconic acid coupled to positions B28, B29, and B30 will be tested for PBA-diol complexation in vitro, prior to validation of glucose-responsiveness in rats (Springsteen & Wang, 2002). Hexamer dissociation assays, which have previously provided evidence of a PBA-diol linkage, will be used as an initial screen of analogs.

Circular dichroism and NMR spectroscopy may be used to either further confirm the

394 presence of such a linkage or test analogs that are not capable of forming stable hexamers.

It remains possible that the self-association of GRI analogs is necessary for glucose- dependent activity. The rate of the stochastic dissociation of the C-terminal B chain from the core of the insulin hormone occurs on the order of 100 of nanoseconds (Papaioannou,

Kuyucak, & Kuncic, 2015). The time scale required for both PBA and diol-containing groups to sample a conformation at which they may complex may require additional time. For this reason, the dampening of conformational fluctuations provided by dimerization and hexamerization (the latter dissociating on the order of several minutes) may be necessary to overcome the kinetic barriers preventing a PBA-diol linkage from forming (J. Dong, et al., 2003). This may indicate that the glucose-dependent activity of our candidate analogs may not be reversible as initially hypothesized.

Current experimental results indicate that no true GRI has been developed using our design. However, a number of critical findings will inform the future direction of this study. First, we determined that insulin analogs with PBA coupled to the N-termini of both polypeptide chains have minimal biological activity, limiting their use as clinical analogs. Second, results of hexamer-dissociation assays on prototype GRIs confirmed the feasibility of the formation of a glucose-dependent intramolecular linkage between PBA and diol-containing moieties. Finally, SQ injections of PBA-coupled analogs revealed a delay in absorption from the SQ depot that presumably results from interaction between the PBA-moiety and diols in the interstitial space. This delay manifests itself in the dose- response relationship of PBA-coupled insulin analogs as a “lag phase” in insulin activity

395 at low doses: a threshold dose must be exceeded before significant insulin activity can be observed.

7-5 Methods.

Activation of PBA-F

4-carboxy-3-fluorophenylboronic acid, pinacol (PBA-F Pinacol, MW: 266.1) was made 250 mM in ice-chilled ethyl acetate. A small excess (1:1.2 molar ratio) of

Dicyclohexyl carboiimide (DCC, MW: 206.33) and N-hydroxysuccinimide were added to this solution and stirred overnight at room temperature. Reaction was centrifuged at

13k RPM for 5 minutes to pellet dicyhlohexylcarbourea byproduct. The resulting supernatant was collected and solvent was removed ex vacuo. The resulting solid product

(PBA-F-NHS ester) was dissolved in acetone and recrystallized from hexane to remove excess dicyclohexylcarbourea.

Production of PBA-F Coupled Precursors for Semisynthesis

DOI or two-chain TK-insulin was dissolved in 0.1M Sodium Carbonate pH 7.6 at a concentration of 10mg/200μl, a 1:1 volume/volume ratio of a saturated solution of PBA-

F NHS ester in acetonitrile were added to the mixture. The reaction was gently agitated at room temperature overnight. Products were purified by rp-HPLC. TK insulin was trypsinized in 0.1 mM NH4CO3 (pH 8.0) and 1mM urea with 10% wt/wt trypsin overnight at room temperature. The products of trypsinization were purified by rp-HPLC.

Identities of products were confirmed by MALDI-TOF mass spectrometry, and their purity was confirmed by analytical rp-HPLC.\

Semisynthesis of Insulin

396

Insulin analogs were prepared by trypsin-catalyzed semi-synthesis using an insulin fragment, des-octapeptide (B23-B30)-insulin, and modified octapeptides as described

(Inouye, et al., 1981). The fragment was made by tryptic cleavage of human insulin and purified by reverse phase HPLC. Modified octapeptides were prepared by solid phase synthetic methods (Barany & Merrifield, 1980). The resulting insulin analogs were purified by preparative reverse phase C4 HPLC (Higgins Analytical Inc., Proto 300 C4

,10 uM, 250 x 20 mm), and their purity was assessed by analytical reverse phase C4

HPLC (Higgins Analytical Inc., Proto 300 C4 5 uM, 250 x4.6 mm). Predicted molecular masses were in each case verified using an Applied Biosystems 4700 proteomics analyzer

MALDI-TOF.

Glucose-Dependent Hexamer Dissociation Assays

Visual absorption spectroscopy was used to probe the formation and disassembly of phenol-stabilized R6 Co-substituted insulin hexamers in the presence and absence of glucose as first described by (Roy, et al., 1989). WT insulin or analogs were made 0.6 mM in a buffer containing 50 mM Tris-HCl (pH 7.4), 50 mM phenol, 0.2 mM CoCl2, and

1 mM NaSCN, ±25mM D-glucose. Samples were incubated overnight at room temperature prior to the studies to ensure attaintment of conformational equilibrium.

Spectra (450–700 nm) were obtained to monitor tetrahedral Co2+ coordination with its signature peak absorption band at 574 nm (D. T. Birnbaum, et al., 1997). To determine the rate of Co2+ release from the hexamers, metal ion sequestration was initiated at 25 °C by addition of an aliquot of EDTA (50 mM at pH 7.4) to a final concentration of 2 mM.

Attenuation of the 574-nm absorption band was monitored on a time scale of seconds to hours. Kinetic data were consistent with monoexponential decay.

397

Rodent Assays

Male Lewis rats (mean body mass 300 g) were rendered diabetic by streptozotocin for experiments in diabetic rats. To test the in vivo potency of insulin analogs relative to

[OrnB29]-insulin, the analogs were made 10 µg per 100 µl in a formulation buffer (16 mg/ml glycerin, 1.6 mg/ml meta-cresol, 0.65 mg/ml phenol, and 3.8 mg/ml sodium phosphate (pH 7.4)) and injected intravenously into tail veins as described (Vijay

Pandyarajan, et al., 2014a). Control insulins or candidate GRI analogs were each re- purified by HPLC, dried to powder, dissolved in a glycerol-free Lilly diluent containing re-quantitated by analytical C4 reverse phase HPLC to ensure uniformity of formulations; dilutions were made using the above buffer. For SQ experiments, a ratio of 3:1 molar ratio of ZnCl2 to insulin was added to the formulation.

Rats were injected at time t0. Blood was obtained from the clipped tip of the tail at time t0 and every 10 min for the 1st h, every 20 min for the 2nd h, every 30 min for the

3rd h, and every hour thereafter to a final time of 5 h. The efficacy of insulin analogs to was calculated using the following: (a) the rate of change in blood glucose concentration over 240 minutes following initial injection, and (b) the integrated area between the glucose-lowering curve of the insulin analog and a curve generated by the injection of insulin-free diluent into diabetic or non-diabetic rats (area over the curve; AOC).

398

Chapter 8: Continuing Work, Future Directions, and Conclusions

8.1 Summary of Dissertation

This dissertation describes studies on the insulin molecule from the perspective of both fundamental protein biochemistry and therapeutic translation. The bridging of these two aspects celebrates an approach to the study of insulin structure that has been taken since the discovery of the hormone nearly 100 years ago (Das & Shah, 2011). In spite of the long history of insulin research, the concepts and innovations discussed in this dissertation demonstrate how study of the insulin molecule continues to uncover novel fundamental concepts and improve current therapeutic technologies.

Chapters 2, 3, and 4 of this thesis explore the hidden constraints of foldability and proteotoxicity on the structure of the insulin hormone. The region of the insulin molecule spanning residues B20 and B23 has been identified as being critical to the native folding of the hormone. Although this region is also known to facilitate the binding of insulin to the classical site of IR (J.G. Menting, et al., 2014; Mirmira, et al., 1991), studies discussed in this dissertation indicate that it is the impact of mutations to this region on the pre-oxidative folding of insulin that have led to its broad conservation. Indeed,

Chapter 4 demonstrates how the addition of a single para-hydroxyl group to PheB24 prevents the hormone from efficiently folding into its native conformation. As demonstrated in Chapter 2, the allosteric nature of insulin-IR complexation allows GlyB24 insulin to bind IR with high affinity with non-standard residue occupancy; yet, the diminished folding efficiency and stability of the variant have led to its strict evolutionary exclusion. That both the aromatic TyrB24 and the sidechain-less GlyB24 variants impair the foldability of proinsulin highlights the precarious nature of the insulin folding pathway:

399 formation of the A19-B20 disulfide bridge may be frustrated by the obstruction of disulfide bond formation (by the para-hydroxy group of TyrB24) or by failure of the C- terminal B-chain to achieve its native occupancy (as may be hypothesized to occur in

GlyB24 proinsulin). Such hypotheses are further evidenced by the requirement for the integrity of the hydrogen bond between A21 and B23 (Chapter 4), which likely plays a critical role in orienting residue B24 into a favorable occupancy.

Studies on the fundamental aspects of insulin biochemistry have informed translational efforts. Dissection of the dimer interface of insulin revealed the critical components that contribute to the stability of insulin hexamers, among them the aromatic-aromatic interactions between residues TyrB16, PheB24, and TyrB26 (El Hage, Pandyarajan, et al.,

2016). Such studies, along with mutational analyses that revealed the tolerance of the biological activity of insulin to substitutions at the B26 position, motivated the substitution of TyrB26 by Trp (Pandyarajan, et al., 2016). Indeed, this modification stabilized the insulin hexamer by strengthening aromatic-aromatic interactions at dimer interface producing an insulin analog with a basal PK/PD profile (Chapter 5). Structural knowledge about the dynamics of the C-terminal B chain of insulin, which was gained through mutational analyses and crystallographic studies (J.G. Menting, et al., 2014;

Mirmira, et al., 1991), revealed the existence of a conformational switch in insulin that could be exploited in the development of GRI analogs (Chapter 7). Finally, decades-long studies of amyloid structure and of the structural factors contributing to insulin fibrillation led to the development of SCI and 4SS technologies (C. M. 14685248

Dobson, 2003; Jimenez, et al., 2002; Jimenez, et al., 2001), which are combined in the context of ultra-stable 4SS-SCIs (Chapter 6).

400

8.2 Continuing Work and Future Directions

8.2.1 Unraveling Conformational Requirements for Proinsulin Folding

Findings discussed in Chapters 2, 3, and 4 provided a basis for hypotheses regarding the conformation of on-pathway proinsulin folding intermediates: folding of proinsulin requires efficient packing of the C-terminal B chain against the core of insulin (Weiss,

Nakagawa, et al., 2002), yet only amino acid side chains that do not interfere with disulfide formation are tolerated. A deeper understanding of the proinsulin folding pathway may be gained by examining the steric limitations imposed on the B24 residue by the nascent core of the prohormone. This may be achieved by introducing non- standard amino acid residues at the B24 position (using native chemical ligation techniques) and examining their effect on the in vitro folding efficiency of proinsulin

(Steward & Chamberlin, 1998). For example, the foldability of proinsulin variants containing para-chloro-Phe or para-methyl-Phe at the B24 position can be tested in vitro to determine whether the diminished foldability of TyrB24-proinsulin is a consequence of steric clashes involving its para-hydroxyl group or its electrostatics. Similarly, the foldability of analogs with ortho- or meta- hydroxyl (or methyl) phenylalanine at the B24 position would indicate whether the unfavorable interaction formed by TyrB24 is direction specific.

Preliminary results indicating diminished foldability of LeuB24-proinsulin suggest that the planar geometry of PheB24 may be a requirement for the efficient folding of proinsulin. This hypothesis may be tested by substituting PheB24 by naturally-occurring aliphatic residues Val and Ile and testing the folding efficiency of the variant proinsulins in vivo and in vitro. Non-standard mutagenesis may be employed to test additional

401 geometries: norleucine, alloisoleucine, or cycleohexylalanine (Cha) may be introduced at the B24 position for in vitro assessment of foldability (Vijay Pandyarajan, et al., 2014b).

A more abstract problem is the identification of the unfavorable conformations of proinsulin folding intermediates that are prevented by the A21-B23 hydrogen bond and that may be, in part, recapitulated in GlyB24 proinsulin variants. Such a problem may be resolved by studying des-A21 mutations in the context of SCIs with foreshortened C- domains. Preliminary findings have indicated that a des-A21 SCI (parent SCI-2 described in Chapter 6), is capable of re-folding at pH 10 (Figure 8-1). Such a finding suggests that the dampening effect of the six-residue C-domain of SCI-2 on the C-terminal B domain may extend to the preoxidative folding intermediate of proinsulin. The degree of

“dampening” required for des-A21 SCI folding may be determined by testing the foldability of des-A21 SCIs with varying C-domain lengths (eg. 6-12 amino acids). Once the maximum length of the C-domain that permits the folding of a des-A21 SCI and the minimum length that renders the SCI unfoldable has been identified NMR spectroscopy and MD simulations of the two analogs of interest be compared to identify those conformations that are accessible only to the unfoldable analog (Glidden, Yang, et al.,

2017). These conformations may be identified as those that disfavor native folding of proinsulin. The use of residual dipolar coupling may be particularly useful in studying the aberrant dynamics of the C-terminal B chain (domain) of SCIs. These fluctuations have been modeled to occur on a 100 microsecond timescale. Although the use of such a model is not without limitations—the degree to which the dynamics of the mature SCI correlate to those in the folding intermediate remains unknown—previous mutational analyses indicate a native-like structure of the insulin core prior to disulfide formation.

402

Figure 8-1. Redox Re-Folding of des-A21 SCI-1RP-HPLC traces of native SCI-2

(right) and des-A21-SCI-2 (top) in their native, folded state, (middle) after denaturation in 10 mM DTT and 6M Guanidine Hydrochloride, and (bottom) after refolding for 16h in redox buffer containing 100 mM NaCl and 10 mM glutathione pH 10. Both analogs were able to re-fold into their native conformations.

403

404

8.2.2 Insight into Structural Ancestors of Insulin: Allostery of Insulin-IR

Complexation and Oligomerization

The “register shift” phenomenon that confers native-like biological activity to GlyB24- insulin has revealed that insulin may adopt non-standard occupancies to bind with high affinity to IR. This finding suggests that is possible for insulin to fold (or adopt a native- like receptor-bound conformation) during the receptor-binding process. Such a phenomenon has been observed in the context of other proteins. For example, intrinsically unfolded peptides adopt secondary structure upon engaging a binding partner

(Fink, 2005). Studies from within our own laboratory group demonstrated that mutations that mutant HMG boxes of the transcription factor SRY that lack significant tertiary structure regain native-like conformation and stability upon binding to their target DNA sequence (Racca et al., 2014).

The assess the extent of the register shift phenomenon, insulin analogs containing insertions of one or two glycine residues between residues B23 and B24 were created.

Whereas the analog containing one glycine insertion, named GG-OrnB29 insulin, retained native-like biological (Figure 8-2A) activity and IR-affinity (Table 8-1), GGG-OrnB29 insulin (containing an insertion of two glycine residues) displayed native IR-affinity and near native-like biological activity in diabetic rats. These findings suggest that high- affinity insulin-IR binding is possible even in the absence of an ordered secondary structural motif, such as a β-turn, directing residues B24-B26 onto their binding surfaces on IR. This is especially true of GGG-OrnB29 insulin, in which the C-terminal B chain is hypothesized to be a disordered peptide tethered by a short linker to the insulin molecule.

Testing of the biological activity of insulin analogs with additional (3+) glycine

405 insertions will be useful in identifying the extent of conformational freedom that is tolerated in insulin IR binding.

Findings regarding the biological activity of glycine-inserted analogs, which are locally disordered at the C-terminal B chain, motivated assessment of the IR-affinity of a globally disordered insulin analog containing a TyrB16→Pro substitution. Although ProB16 insulin lacked discernable secondary structure as assessed by CD spectroscopy, the insulin analog retained 1% IR-affinity- higher than that of analogs such as ChaB25,

OrnB29-insulin. A ProB16 analog containing ThrA8→His and HisB10→Asp substitutions, which increase IR affinity but had little effect on the secondary structure of ProB16 insulin retained 15% affinity of native insulin for IR (Table 8-1). The biological activity of the analog was also substantial in diabetic rats (Figure 8-2B). This finding presented a preliminary validation of the hypothesis that insulin is capable of folding during the IR- binding process.

Continuing works seeks to extend this concept to insulin-IR binding with the goal of establishing the role of insulin’s modern structure in its molecular evolution. The ubiquity of insulin signaling among vertebrates (J. M. Conlon, 2001), its close relationship with signaling systems associated with growth (the intrinsic mitogenic pathway of IR and those of IGF-I and IGF-II) (Steiner & Chan, 1988) provides a rich physiological setting for the evolution of specificity of insulin-IR binding. Moreover, the constraints of proteotoxic misfolding provide an alternative mechanism constraining the molecular evolution of the hormone. Given the well-known influences of both physiology and foldability on insulin structure (Cohen & Dillin, 2008; Cohen et al., 2010; M. Liu, et al., 2007), it may be hypothesized that the globular structure of insulin has evolved

406 entirely to maintain the stability of the natively folded molecule and that insulin-IR complexation requires only a few critical contacts.

This hypothesis may be validated by further characterization of the binding of disordered insulin analogs to IR and associated receptors. Whereas in vitro IR-binding experiments have tested the affinity of ProB16 and Gly-insertion analogs, measuring the affinity of these analogs for IGF-1R, a related receptor, can be used to establish their specificity. Further mutational analysis of ProB16 or other disordered insulin analogs (such as the recently discovered, unpublished clinical mutant, ProB15-insulin) may be used to identify the minimal binding surface required for high-affinity IR complexation. Such efforts would not only provide insight into the evolutionary origins of insulin but may inform the design of insulin-mimetic peptides or small molecules—a long-standing goal of the pharmacology community.

Assessment of the stability of R6 hexamers formed by Gly-insertion analogs provided additional insight into the evolutionary origins of the insulin core. Both GG-OrnB29 insulin and GGG-OrnB29 insulin formed more stable hexamers than GlyB24, OrnB29 insulin. In fact, GG-OrnB29 insulin exhibited a diminished hexamer-dissociation rate compared to that of native OrnB29 insulin (Figure 8-2, Table 8-2). Such a finding suggested that the C-terminal B chain of insulin is directed towards its natural position, with residues B24 and B26 occupying their native crevices in the globular core, by the formation of the R6 hexamer. The similarity between the B24 binding pocket of L1

(visualized in the μIR co-crystal structure) and B24-related crevice in the insulin molecule may suggest a common evolutionary root (J.G. Menting, et al., 2014; Vijay

Pandyarajan, et al., 2014b). That is, the structure of the core of the insulin hormone may

407

Figure 8-2 Biological Activity of Destabilized Insulin Analogs Glucose-lowering curves of diabetic rats after IV injection of 10μg doses of native OrnB29-insulin, GG-

OrnB29-insulin, and GGG-OrnB29 insulin (n=6). (B) Glucose-lowering curves of diabetic rats after IV injection of 200μg of ProB16-DesDi (defined below), 25μg of HisA8, AspB10,

ProB16-DesDi, 10μg native DesDi, and vehicle (diluent). Whereas 25μg of HisA8, AspB10,

ProB16-DesDi was equipotent to 10μg native DesDi, a 200μg dose of ProB16-DesDi was unable to match the biological activity of 10μg of the control analog (n=5).

408

409

Table 8-1. In vitro IR-Affinities of Destabilized Insulin Analogs

analog Kd IR-B (nM±SEM ) Lispro 0.07±0.02* ProB16 –desDia 7.05±1.01 A8 B10 B16 His , Asp , Pro -desDi 0.31±0.05 OrnB29 0.08±0.02 GG-OrnB29 0.10±0.04 GGG-OrnB29 0.04±0.02 a desDi refers to a LysB28 analog in which residues B29 and B30 are deleted

410

have evolved to replicate the B24 binding pocket on IR to properly accommodate the residue during storage of the insulin hormone.

8.2.3 Characterization of Zinc-Insulin Hexamers Using TrpB26

The insulin hexamer is an efficient and intricate protein assembly. The evolutionary optimization from the packing of the C-terminal B chain against the core of insulin to the subtle—yet critical—interactions that that form across the interface is remarkable. It is the presence of such subtle details that that led Dorothy Hodgkin and colleagues to describe the insulin hexamer as “a thing of beauty” in their classic review(Blundell,

Dodson, et al., 1972; Weiss & Lawrence, 2018). Yet, even since these remarks were published, novel forms of the insulin hexamer, such as the phenol-stabilized R6 hexamer and (Brader, Kaarsholm, Lee, & Dunn, 1991; U. Derewenda, et al., 1989) zinc- independent hexamers formed by GlnB13-insulin have been discovered (G. A. Bentley, et al., 1992). Even today, the stability of the zinc-insulin hexamer may be further improved, as exemplified by the development of TrpB26 insulin (Chapter 5). Future work may accomplish this goal through further dissection of observations regarding the insulin dimer interface.

Apart from studies of the biophysical properties of TrpB26-insulin, a general

B29 observation regarding a slight decrease in the R6-hexamer dissociation rate of Orn - insulin in relation to that of wild-type human insulin motivated examination of the B29 residue in hexamer stability. We hypothesize that the ε-amino group of LysB29 forms transient interactions that stabilize the R6 insulin hexamer, and that decreasing the distance between the Cα of LysB29 and its amino group decreases the probability of such an interaction forming. This would explain the decreased stability of R6 hexamers formed

411 by OrnB29 insulin, in which the amino group and Cα of the B29 residue are separated by three rather than four carbon atoms. We propose that the transient nature of such an interaction has made it difficult to characterize (E. N. Baker, et al., 1988; Nakagawa &

Tager, 1987). Indeed, the side chain of LysB29 is poorly resolved in x-ray crystal structures and is inferred to be disordered in solution from NMR studies.

The hypothesis was tested by measuring the dissociation rates of R6 insulin hexamers formed by WT-insulin, OrnB29 insulin, and insulin analogs containing diaminobutyric acid (Dab) and diamnioproprionic acid (Dap) at the B29 position. The latter two amino acids are similar instructure to Lys, but contain alkyl chains of two and one carbon between Cα and the side-chain amino group. Preliminary results revealed that, indeed,

R6-insulin hexamers grew progressively less stable as the length of the side chain of the

B29 residue decreased (Figure 8-4).

The presence of putative interactions involving the side chain of LysB29 may be validated by MD simulations which may be able to successfully model an interaction occurring on a rapid timescale (El Hage, Pandyarajan, et al., 2016). Further characterization of the interaction may also be undertaken. Insulin analogs with varying charges and polarities at the B29 position may be surveyed for hexamer stability. Among the amino-acid residues that may be substituted at the B29 position are norleucine (non-polar), glutamic acid

(acidic), aspartic acid (acidic), asparagine (polar, uncharged), and glutamine (polar, uncharged). The potential identification of such novel interactions within the insulin hexamer may motivate the re-examination of regions of the assembly that were thought to be well characterized.

412

Figure 8-3. Hexamer Dissociation Rates of Gly-Inserted Analogs R6 hexamer- dissociation curves of native OrnB29, GlyB24OrnB29, GG-OrnB29, and GGG-OrnB29-insulin analogs monitored as a function of sequestration of tetrahedrally coordinated Co2+

(reported be absorbance at 574 nm).

413

414

Table 8-2. Half-Lives of R6 Insulin Hexamers of Gly-Inserted Insulin Analogs analog t1/2 hexamer dissociation (min ± SD ) OrnB29 8.2 ± 0.8 GlyB24, OrnB29 1.2 ± 0.2

Gly-GlyB24, OrnB29 11.0 ± 1.5 B24 B29 Gly-Gly-Gly , Orn 2.5 ± 0.3

415

TrpB26 insulin provides a robust platform for such an approach. The studies described in Chapter 5 discuss the improved aromatic-aromatic interactions associated with TrpB26 at the dimer interface of insulin. However, this chapter also raised questions regarding the packing of TrpB26 within the core of insulin. Although TrpB26 is well-tolerated within the globular core of the hormone, TrpB26 OrnB29 insulin does not exhibit increased thermodynamic stability, as has been observed for insulin analogs containing 3- iodotyrosine (a residue similar in size to Trp) at the B26 position. Such variant analogs may be used to study the character of the B26 packing crevice within the insulin monomer as it relates to self-association.

The properties of TrpB26 that allow its accommodation by the B26 crevice and its impact on hexamerization were assessed by testing the hexamer dissociation rates of insulin analogs containing diverse aromatic amino acid substitutions at the B26 position.

Analogs were made in the context of OrnB29 insulin containing a B chain that was C- terminally extended by Glu residues at the B31 and B32 positions. This modification was introduced to maintain the solubility of the peptide comprising the C-terminal B chain during semisynthesis. Such analogs were termed POTEE-insulins. In total, four POTEE analogs were created: native TyrB26-POTEE, PheB26-POTEE, pentafluorophenylalanineB26-POTEE (5F-PheB26-POTEE), and 2-naphthylalanineB26-

POTEE (NalB26-POTEE).

Whereas TrpB26-POTEE displayed the longest hexamer lifetime, that of both PheB26-

POTEE and 5F-PheB26-POTEE was increased in relation to native POTEE (Figure 8-5A,

Table 8-4). This finding suggests that the hydrophobicity of the B26 residue likely has a stabilizing effect on the

416 lifetime of the R6 insulin hexamer. In spite of weaker aromatic interactions between Phe-

Tyr pairs in relation to Tyr-Tyr or Trp-Tyr pairs, PheB26-POTEE hexamers were more stable than those of native TyrB26-POTEE. Following the same trend, 5F-Phe, a residue more hydrophobic than Tyr or Phe, confers increased hexamer lifetime beyond that of

Tyr and Phe at the B26 position. It is worth noting that the distribution of partial charges around the aromatic ring of 5F-Phe caused by the fluorine substituents may result in novel ring packing at the dimer interface of insulin. However, such a possibility will require confirmation by x-ray crystallography. That none of the three Phe-based B26 variants were able to form hexamers as kinetically stable as those of TrpB26 emphasizes the importance of aromatic-aromatic interactions at the dimer interface in stabilizing the

TrpB26 hexamer. Perhaps most significantly, NalB26-POTEE hexamers displayed the most rapid dissociation rate likely due to steric clashes involving the large aromatic side chain of Nal with residues in the core of the insulin monomer. This finding highlights that it is not only the size of the aromatic ring of TrpB26 but also its packing efficiency within the core of the insulin molecule that contributes to the stability of its R6 hexamers.

The thermodynamic consequences of the packing efficiency of the B26 side chain within the hydrophobic core of insulin was assessed using CD-monitored guanidine denaturation assays of the POTEE analogs. The stability of NalB26 POTEE could not be assessed due to the absorption of the Nal side chain at 222 nm. As may be inferred, both PheB26 and

5F-PheB26 variants displayed increased stability in relation to native POTEE, presumably because of the increased hydrophobicity of these side chains in relation to the native Tyr

(Table 8-5). Unexpectedly, the stability of TrpB26-POTEE was augmented by 0.2 kcal/mol in relation to native POTEE. This finding suggests that the character of the

417

Figure 8-5 Self-Association Properties of B26-Variant Insulin Analogs (A) R6 hexamer-dissociation curves of B26-variant insulin analogs monitored as a function of sequestration of tetrahedrally coordinated Co2+ (reported be absorbance at 574 nm) (half- lives are summarized in Table 8-4). (B) SEC profile of zinc-free formulations of TrpB26,

OrnB29-insulin and native OrnB29-insulin. An association state intermediate to a hexamer and a dimer was visualized in the chromatograph of TrpB26, OrnB29-insulin. (C) SEC profile of zinc-free formulations of AspB10, TrpB26, OrnB29-insulin and native AspB10,

OrnB29-insulin. The intermediate association state was not visualized in the chromatograms of either AspB10 sample.

418

419

Table 8-4. Hexamer Dissociation Half-Lives of Non-Standard B26 Variants analog t1/2 hexamer dissociation (min ± SEM ) OrnB29, GluB31, GluB32 4.9 ± 0.2 PheB26, OrnB29, GluB31, GluB32 11.1 ± 0.3 5F-PheB26, OrnB29, GluB31, GluB32 54.8 ± 0.4 B26 B29 B31 B32 2 Trp , Orn , Glu , Glu (10.2 ± 1)x10 NalB26, OrnB29, GluB31, GluB32 1.4 ± 0.1

420

Table 8-5. Thermodynamic Stabilities of Non-Standard B26 Variants

a b analog ΔGu Cmid m (kcal mol-1) (M) (kcal mol-1 M-1) OrnB29, GluB31, GluB32 3.2 ± 0.1c 4.9 ± 0.65 ± 0.02 PheB26, OrnB29, GluB31, GluB32 3.8 ± 0.1 5.3 ± 0.72 ± 0.01 5F-PheB26, OrnB29, GluB31, GluB32 3.9 ± 0.1 5.4 ± 0.72 ± 0.02 TrpB26, OrnB29, GluB31, GluB32 3.5 ± 0.1 5.7 ± 0.61 ± 0.02 a Parameters were inferred from CD-detected guanidine denaturation data by application of a two-state model; uncertainties represent fitting errors for a given data set. bThe m-value (slope Δ(G)/Δ(M)) correlates with extent of hydrophobic surfaces exposed on denaturation.

421

C-terminal B chain (eg. the presence of the Glu-Glu extension) may correct the defects that prevent TrpB26 OrnB29 insulin from exhibiting increased stability. If such a defect involves increased exposure of the hydrophobic Trp side chain in the context of the insulin monomer, as was hypothesized in Chapter 5, the Glu residues at B31 and B32 may have a shielding effect, protecting the hydrophobic surface from solvent exposure.

These findings collectively reveal the importance of both the aromaticity and the geometry of the B26 side chain in the formation of stable zinc-insulin hexamers. Even large, aromatic residues such as Nal are unable to form stable hexamers as a result of their presumed inability to occupy the B26 crevice in the insulin monomer. The dimensions of the B26 crevice may be probed in future studies by the introduction of Trp residues methylated at various positions around the indole ring at the B26 position. Those residues in which the methyl group encounters a steric clash will exhibit diminished hexamer lifetimes and thermodynamic stabilities. In this way, the rigidity and space surrounding the various vertices of the TrpB26 ring may be identified.

The novel self-association properties of the TrpB26 insulin analogs may also be further studied. SEC profiles of zinc- and phenol-free formulations of TrpB26, OrnB29-insulin analogs revealed the presence of an oligomer intermediate in size to a hexamer and a dimer (Figure 8-5B). This oligomer corresponded to the dissociation intermediate observed in SEC studies of TrpB26, OrnB29-insulin formulated with zinc and phenol

(discussed in Chapter 5). This intermediate was not observed in chromatograms of

AspB10, TrpB26, OrnB29-insulin, suggesting that its formation was mediated by the native

HisB10 side chain (Figure 8-5C) (Nourse & Jeffrey, 1998). Future studies using SEC combined with multi-angle light scattering (MALS) may be conducted to determine the

422 exact molecular mass of this intermediate (Attri, et al., 2010b; Attri & Minton, 2005).

The residues contributing to the stability of this novel intermediate may be studied using

NMR spectroscopy. Comparison of heteronuclear-NMR spectra of isotopically labeled

TrpB26, OrnB29 insulin and AspB10, TrpB26, OrnB29-insulin may be used to differentiate interactions unique to the novel oligomer from those of the dimer.

In spite of its native biological activity and stability, TrpB26 is not a naturally-occurring variant among known vertebrate insulin sequences. Our current hypothesis states that the increased stability of TrpB26 insulin hexamers would limit bioavailability of the hormone to the liver after its secretion as microcrystals into the portal vein (Tamaki, et al., 2013).

This hypothesis is further supported by the formation of large zinc-free oligomers by

TrpB26 insulin analogs. In vivo studies may be utilized to test this hypothesis. One such experiment can determine the amount of TrpB26 or native insulin (formulated in the presence of zinc and phenol) consumed after injection into the portal veins of ex vivo, perfused rodent livers. A majority of native insulin is expected to be consumed by IR- expressing cells in the liver. For this reason, a low concentration of insulin would be detected exiting the liver through the hepatic vein. In contrast, the stable oligomers formed by TrpB26-insulin would be expected to prevent the hormone from binding to IR in the liver resulting in a greater amount of the insulin exiting through the hepatic vein

(M Dodson Michael, et al., 2000). Such a prediction is reinforced by preliminary experiments in which TrpB26, OrnB29- and native OrnB29-insulins formulated at 0.6 mM in the presence of zinc and phenol were injected intravenously in diabetic rats (Figure 8-

6A). Whereas OrnB29-insulin displayed native potency, TrpB26, OrnB29-insulin exhibited reduced potency. It may be hypothesized a large proportion of the hormone was renally

423 cleared before it could successfully engage IR. The physiological consequences of reduction of hepatic insulin signaling in relation to that in peripheral tissues may be studied using a genetically engineered animal model, such as zebrafish. The metabolic characteristics of transgenic zebrafish expressing TrpB26-insulin (introduced in the framework of native zebrafish insulin) may be compared to those of unmodified fish to identify any detrimental effects associated with the TrpB26 variant.

Aside from their evolutionary and biophysical significance, TrpB26-insulin analogs are likely to be viable therapeutic analogs. The mechanism by which the TrpB26 modification protracts the PD profile of insulin is orthogonal to those utilized in current basal analogs

(detemir, degludec, and glargine). For this reason, the introduction of TrpB26 insulin into the framework of any of the three current clinical analogs is expected to further protract their duration of action (Gillies, et al., 2000; Gough, et al., 2013; J Markussen, et al.,

1996). This hypothesis was demonstrated, in part, in preliminary experiments in which

TrpB26 was introduced in the context of a pI-shifted insulin analog containing a

GluB13→Gln substitution (which stabilizes the insulin hexamer) (G. A. Bentley, et al.,

1992). This analog also contained a substitution of ThrA8 by Gln to rescue the decrease in biological activity associated with GlnB13 insulin (Weiss, Hua, et al., 2002). In relation to the TyrB26 analog, the TrpB26, GlnB13 insulin analog displayed a marked decrease in rate of fall of blood glucose level after SQ injection in diabetic rats. The analog also appeared to have a duration of action that was predicted by extrapolation to last over 20 hours

(Figure 8-6B). Future studies will follow up on these results and introduce TrpB26 in the framework of the current clinical analogs and utilize continuous glucose monitoring to study the protracted PD profiles of the novel analogs. Cell-based assays that test the

424

Figure 8-6. PD Profiles of High-Concentration (u100) TrpB26 Analogs in Diabetic Rats (A) Glucose-lowering curves of diabetic rats after IV injection of equal doses of

TrpB26, OrnB29 and native OrnB29-insulin formulated at u100 (n=8). (B) Glucose-lowering curves of diabetic rats after SQ injection of equal doses of TrpB26, GlnA8, GlnB13, OrnB29, native GlnA8, GlnB13, OrnB29-insulin, and insulin glargine formulated at u100. Both GlnB13 analogs analogs were amidated at the C-terminus of the B chain to shift the pI of the analog toward neutrality (n=8).

425

426 mitogenicity and toxicity of such analogs will also be utilized to test the clinical viability of such insulin analogs (Glidden, Aldabbagh, et al., 2017).

8.2.4 The Molecular Basis of Insulin Action and Design of Analogs of the Future

Current clinical insulin analogs exhibit rapid- or basal action as a result of modifications that modulate their PK profiles. Whereas rapid-acting insulin analogs are absorbed rapidly form the subcutaneous injection site (as a result of their rapidly dissociating oligomers or small molecules promoting absorption in their formulations), basal insulins form depots in the depots at the injection site (by isoelectric precipitation or the formation of multi-hexameric complexes) or in the bloodstream (via albumin binding) (Gillies, et al., 2000; Gough, et al., 2013). Yet, the intrinsic biological activity of analogs in current use is essentially identical to that of native insulin, save for differences in potency associated with diminished IR-affinity. This contrasts with a number of analogs that have been described in literature, which have shown novel biological properties independent of bioavailability. These properties may include prolongation or foreshortening of activity in animals after IV injection (as observed in 4SS-insulins and SCI-2, discussed in Chapter

6), tissue selectivity, or IR-isoform selectivity (Glidden, Aldabbagh, et al., 2017; Glidden,

Yang, et al., 2017; T. N. Vinther, et al., 2013). The identification of these insulin analogs raises important questions about the current understanding of the biological activity as it relates to insulin-IR complexation, activation of IR-associated signal-transduction pathways, and the physiological response of tissues to insulin signaling(P. De Meyts,

2015).

A long-standing goal in insulin research has been the development of IR-isoform- selective insulin analogs (T. Glendorf et al., 2011; S. G. Vienberg, et al., 2011). As

427 discussed in Chapter 1, the two isoforms of insulin, A and B, are differentiated by the presence of a 12 amino acid sequence (IR-B) or its absence (IR-A) in the insert domain of

IR. The tissue distribution and expression of the two isoforms varies in mammals (Sara G

Vienberg et al., 2011); in humans, IR-A is predominantly expressed in the brain, lymphocytes, and in fetal and neonatal tissues. Skeletal muscle and adipose tissue also display a greater proportion of of IR-A (~35%) in relation to IR-B than other tissues. The

B isoform is predominant in skeletal muscle, liver, adopose tissues, and in pancreatic β cells themselves. Possibly because of its distribution, IR-B is associated with metabolic signaling whereas IR-A, which is also commonly expressed by cancer cells, is associated with growth and mitogenicity (Belfiore, Frasca, Pandini, Sciacca, & Vigneri, 2009). For this reason, the development of insulin analogs that selectively bind IR-B may be beneficial to prevent exacerbation of malignancies, and IR-A-specific analogs may be beneficial in preventing macrovascular complications of insulin therapy.

The design of isoform-selective insulin analogs requires modulation of interactions between the alternatively spliced region (exon 11) of IR and insulin in its IR-bound conformation (Seino & Bell, 1989). Conveniently, exon 11, which spans residues 717-

728 in IR-B, was revealed to overlap with the classical binding site of insulin on the αCT domain of IR. This provided a structural basis for analog design (J.G. Menting, et al.,

2014). One such design suggested that modifying the C-domain of SCIs may be modified to favor IR-A binding. This design was informed by the requirement for αCT to traverse the loop formed by the C-domain, C-terminal B domain, and core of SCIs, in order for the molecule to successfully engage IR (Figure 8-7A). The increased size of the region of

IR-B spanning the αCT domain suggested that the binding of SCIs to this receptor form

428 may be sterically disfavored. Indeed, previous studies have indicated that SCIs, including

SCI-2 and SCI-3, preferentially bind IR-A over IR-B (Glidden, Aldabbagh, et al., 2017).

Future designs may introduce modifications to the C-domain of SCIs to increase the selectivity of the analog. This may be accomplished by the introduction of bulkier residues (such as Trp or Leu) to the linker domain that increase steric clashes with the exon-11-coded region of IR-B (Tine Glendorf et al., 2011).

The development of an IR-B selective insulin analog is somewhat more challenging.

Future efforts may focus on the B25 position of insulin as a site for modification to create such analogs; the μIR-insulin co-crystal structure revealed that the binding surface PheB25 included residues 715, 716, 717, and 718 of IR-A (Figure 8-7B). The inclusion of exon

11 in IR-B is expected to cause IR-A residues Pro717 and Ser718 away from the B25 binding surface putting in their place Arg717 and Lys718 (J.G. Menting, et al., 2014). A negatively-charged amino acid at the B25 position may be hypothesized to form favorable interactions at its binding site in the context of IR-B and to disfavor intercalation with a the relatively non-polar binding surface of IR-A. Indeed, the IR-B selectivity of a low affinity AspB25 analog has been reported (Tine Glendorf, et al., 2011).

Our preliminary studies have indicated that the B25 position may be amenable to modification: although high-affinity IR binding requires an aromatic amino acid at the

B25 position, B25-substiuted insulin analogs retain native thermodynamic stability.

These findings suggest that analogs containing non-standard aminoacids such as 4- carboxyphenylalanine, para-fluoro-phenylalanine, or DOPA at the B25 position may be

IR-B selective.

429

The study of the biological activity of insulin and its analogs has revealed the complexity of insulin-IR complexation and the subsequent activation of signal transduction pathways. The identification of insulin analogs that produce qualitatively different pharmacodynamic profiles that cannot be explained by their bioavailability or IR-affinity suggests that kinetics of insulin-IR complexation or differential activation of signal transduction pathways may play a major role in the in vivo activity of insulin. This concept is exemplified by 4SS-insulin, 4SS-SCIs, and SCI-2. 4SS-insulin analogs display prolonged duration of biological activity after IV injection in rats. This feature was not observed in the PD profiles of insulin analogs with similar or equivalent potencies to

4SS-insulin, such AspB10 or HisA8 insulins, suggesting that IR-affinity alone does not adequately explain its presence. The addition of a C-domain to 4SS insulin eliminates this tailing effect, as observed in the PD profile of 4SS-SCI-1. In contrast, SCI-2 displays biphasic activity after IV administration in rats. The PD profile of the analog is apparently unaffected by the introduction of the A10-B4 disulfide linkage of 4SS insulin.

The three novel insulin analogs may be used as a platform for characterization of the determinants of the biological activity of insulin. It is hypothesized that 4SS-insulin displays protracted biological activity due to “fast-on/fast-off IR-binding kinetics.

Because a vast majority of insulin is cleared through receptor-mediated pathways (M

Dodson Michael, et al., 2000), an insulin analog that binds IR long enough to activate associated signal transduction pathways, but dissociates before it may be internalized and degraded would display prolonged biological activity as a result of its delayed clearance

(Di Guglielmo, et al., 1998). It is further hypothesized that the addition of the C-domain delays dissociation of the 4SS-SCI from IR by interacting with αCT of IR resulting in

430

Figure 8-7. Binding of SCIs and WT Insulin in the Insulin-μIR Co-Crystal Structure

(A) MD simulation of the binding of an SCI with a six amino-acid C-domain to μIR. The

αCT domain of IR (purple ribbon) must traverse the loop formed by the C-domain of the

SCI (red loop), the C-terminal B chain of insulin (brown loop), and the core of the SCI molecule, composed of the A chain (yellow ribbon) and the central helix of the B chain

(black ribbon). This allows insulin to intercalate between αCT and the L1 domain (blue loops and white ribbon) (B) The binding cleft of PheB25 was somewhat poorly resolved in the insulin. The side chain of residue B25 (green sticks) was visualized to interact with

αCT residues 715, 716, 717, and 718 (identified using IR-A numbering scheme; figure adapted from (J.G. Menting, et al., 2014) with permission of the authors..

431

432 native-like clearance and biological activity. Future studies may test such a hypothesis by measuring the clearance rates of two-chain 4SS insulin analogs and 4SS-SCIs and measuring their on- and off rates of IR binding using surface plasmon resonance (SPR)

(Myszka, 1997) or cell-based assays(Knudsen, et al., 2011).

An alternative hypothesis is two-chain 4SS-insulins, 4SS-SCIs, and SCI-2 form distinct interactions with IR that result in biased activation of signal transduction pathways associated with IR. Such biased signaling has not been studied in detail in the context of

IR, yet, its plausibility has been confirmed by reports regarding AspB10-insulin, which disproportionately activates the mitogenic signaling pathways associated with IR and induces rapid endocytosis of the insulin-IR complex (Gallagher, et al., 2013; Sciacca, et al., 2010). Validation of this hypothesis requires detailed analysis of the phosphorylation state of IR substrates including IRS (insulin-receptor substrate) proteins, Gab-1, Shc, and

Cbl (Saltiel & Pessin, 2002). This may be accomplished by the use of Western blots of cells treated with insulin analogs conducted over a time course to detect phosphorylation of IR-associated signaling proteins. The consequences of biased signaling on cell physiology may be monitored by live cell imaging of cells expressing fluorescent IR constructs (that report conformational changes upon insulin complexation and internalization of the complex) or immunostaining fixed cells to assess Glut4 membrane localization, the major metabolic consequence of IR signaling.

The identification of discrepancies in the biological activity of insulin analogs and their affinity for IR in vitro further emphasizes the complexity of the interaction between the two molecules. That the in vitro-IR affinity of insulin analogs does not correlate with their potency in animal models has been well established. Such reports identify receptor-

433 mediated clearance and the internalization of IR affecting the availability of the receptor and its ability to activate signaling pathways as compensatory mechanisms (Ribel, et al.,

1990). In contrast to these findings, we have identified insulin analogs that exhibit diminished IR-affinity and yet display increased in vivo potency in rats, and analogs that exhibit decreased potency in rats in spite of native IR affinity. This phenomenon is not adequately explained by the clearance of the insulin analog or internalization of the insulin-IR complex. This set of analogs comprises insulins that are methylated at the 2, 3, or 4 position of Phe24. Whereas 2-methyl-PheB24 displayed the highest in vitro IR affinity of the three analogs (Table 8-6), it exhibited reduced potency in relation to 3- and 4- methly PheB24 analogs (Figure 8-8). Conversely, 3- and 4-methyl analogs exhibited lower in vitro affinity to native insulin controls, but respectively exhibited equal and greater potency in rats.

Present knowledge of the biological activity of insulin does not provide an adequate explanation for these phenomena. That 2-methyl-PheB24 insulin displays native-like affinity for IR but diminished biological activity suggests that the analog is a partial agonist. No partial agonists of IR have currently been identified; this is thought to be a result of conformational change in IR that guides insulin from binding site 1 to binding site 2, enabling high-affinity binding, simultaneously activating downstream signaling pathways (Knudsen et al., 2012). Further characterization of the biological activity 2- methyl-PheB24 insulin analogs will be required to determine whether it is a true partial agonist. However, such a property is confirmed, future studies may structurally model how such an analog is able to bind to IR with high affinity (i.e. bind to site 2) without fully activating the tyrosine kinase domain.

434

The discrepancy between the in vitro and in vivo characteristics of 4-methyl-PheB24 insulin is more difficult to rationalize, yet it is possible that this analog is tissue selective.

Whereas the solubilized IR holoreceptor used in receptor binding assays is expressed from stably transfected cell lines, which produce essentially identical IR proteins, IR expressed in mammalian tissues may display tissue-specific glycosylation (particularly noted in the brain (K. A. Heidenreich, Zahniser, Berhanu, Brandenburg, & Olefsky,

1983)) or signaling substrates (Belfiore et al., 2017). For this reason, an insulin analog that is compatible with a specific glycosylation pattern or that produces an insulin-IR complex that activates a tissue-specific signaling pathway, may display discrepant in vivo and in vitro potency (Ye, et al., 2017). This hypothesis may be validated through phosphoproteomic analyses of tissues extracted from animals injected with the putative tissue-selective analogs.

The discovery of an insulin analog that is tissue specific as a result of its biased IR affinity or signaling properties would represent a remarkable breakthrough. Insulin analogs that are capable of initiating the uptake of glucose by skeletal muscle and the liver without initiating fatty acid synthesis in adipose tissue are expected to mitigate the weight gain associated with insulin therapy. Current efforts to create hepatospecific insulin analogs have involved the conjugation of insulin to a molecule that is sequestered in the liver such as mannose, albumin, or polyethylene glycol (PEG). Although the albumin-conjugated insulin albulin and PEGylated insulins did exhibit hepatospecific properties in rodents, such analogs were also shown to cause hepatotoxicity and, in the case of insulin albumin, cachexia and renal damage (Duttaroy et al., 2005; Moore et al.,

2014; Thibaudeau et al., 2005).

435

Table 8-6 In vitro IR-Affinity of B24-Methylated Analogs

analog Kd IR-B (nM±SEM ) OrnB29 0.08±0.02 2-methyl-PheB24, OrnB29 0.07±0.01

3-methyl-PheB24, OrnB29 0.45±0.06 B24 B29 4-methyl-Phe , Orn 0.35±0.05

436

Figure 8-8 Biological Activity of B24-Methylated Analogs Glucose lowering curves in diabetic rats after injection of equal doses of PheB24-methylated insulin analogs or native

OrnB29-insulin (n=12).

437

438

Collectively, the in vitro/in vivo discrepancies and aberrant PD profiles of insulin analogs highlight a scientifically challenging set of problems associated with characterizing insulin-IR complexation. The current level of understanding regarding this process is derived from experiments conducted a low resolution using techniques electron microscopy and fluorescence-based assays. The characterization of current insulin analogs and the design of analogs of the future will require an understanding of the finer details of conformational changes involved in insulin-IR binding. The insulin-μIR crystal structure represents a remarkable breakthrough in characterizing the binding surfaces of the classical insulin binding site on IR. Subsequent MD simulations of both the μIR complex and of the holoreceptor have likewise provided hypothesis regarding conformational fluctuations associated with insulin binding that may be experimentally validated by mutational analysis. Further characterization of IR may be conducted through photocrosslinking experiments targeting site 2, which remains unmapped and yet appears to be the defining feature in modulating the biological activity of insulin (P. De

Meyts, 2015). Finally, use of structural techniques such as cryo-EM tomography, which have undergone significant advances in the past decade, may be used to develop deeper structural knowledge of the insulin holoreceptor, which has proven difficult to crystallize

(Croll, et al., 2016).

8.2 Concluding Remarks

This dissertation has described the characterization of insulin from the perspective of fundamental biochemistry and its application to the treatment of diabetes mellitus. The findings discussed in chapters 2, 3, and 4 collectively propose a novel view of the evolution of insulin structure revealing that the foldability of insulin is more sensitive to

439 mutation than is its ability to bind IR. These chapters have provided insight into the folding pathway of proinsulin, in particular, the importance of occupancy of the C- terminal B chain in the nascent structure of the prohormone. Such findings provide a novel view of protein structure that is likely applicable to a broad variety of proteins.

Future studies of insulin evolution, which represents the divergence of metabolic and growth-related signaling in vertebrates, is of considerable clinical interest, since it may offer deeper insights into the evolutionary roots of diseases such as T2DM.

The engineered insulin analogs described in chapters 5, 6, and 7, are examples of how interrogation of protein structure from a fundamental perspective may be used to treat disease. Previous mutational screening of the B26 position of insulin revealed its amenability to modification and its contribution to the stability of insulin oligomers.

Likewise, chiral mutagenesis at the B20 and B23 positions, as well as the study of clinical mutations at the B23 were key studies that informed the design of SCIs described in Chapter 6 and GRIs described in Chapter 7. Still, the characterization of these therapeutic entities raised questions about fundamental biochemistry. TrpB26 insulin analogs provide a platform through which the B26 crevice at the insulin dimer interface could be further studied, whereas the novel PD profiles of two-chain 4SS-insulins and

4SS-SCIs raise questions regarding the structural determinants of IR-associated signaling.

In this way, the work discussed in this thesis demonstrates how the study of insulin truly embodies the “bench-to-bedside” principle.

440

Bibliography

Abedini, A., & Raleigh, D. P. (2005). The role of His-18 in amyloid formation by human islet amyloid polypeptide. Biochemistry, 44, 16284-16291.

Adams, M. J., Blundell, T. L., Dodson, E. J., Dodson, G. G., Vijayan, M., Baker, E. N., . . . Sheet, S. (1969). Structure of rhombohedral 2 zinc insulin crystals. Nature, 224(5218), 491-495.

Adams, P. D., Afonine, P. V., Bunkóczi, G., Chen, V. B., Davis, I. W., Echols, N., . . . Grosse-Kunstleve, R. W. (2010). PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. Sect. D-Biol. Crystallogr., 66(2), 213-221.

Ahmed, A. M. (2002). History of diabetes mellitus. Saudi medical journal, 23(4), 373- 378.

Alam, P., Siddiqi, K., Chturvedi, S. K., & Khan, R. H. (2017). Protein aggregation: from background to inhibition strategies. International journal of biological macromolecules, 103, 208-219.

Alberti, K. (1994). Insulin dependent diabetes mellitus: a lethal disease in the developing world. BMJ: British Medical Journal, 309(6957), 754.

Aleksandrov, A., Thompson, D., & Simonson, T. (2010). Alchemical free energy simulations for biological complexes: powerful but temperamental…. Journal of Molecular Recognition, 23(2), 117-127.

Alm, E., & Baker, D. (1999). Prediction of protein-folding mechanisms from free-energy landscapes derived from native structures. Proceedings of the National Academy of Sciences, 96(20), 11305-11310.

Alm, E., & Baker, D. (1999). Prediction of protein-folding mechanisms from free-energy landscapes derived from native structures. Proc. Natl. Acad. Sci. U. S. A., 96, 11305-11310. doi: PMC18029

Anirudhan, T., Nair, A. S., & Nair, S. S. (2016). Enzyme coated beta-cyclodextrin for effective adsorption and glucose-responsive closed-loop insulin delivery. International journal of biological macromolecules, 91, 818-827.

Ashenberg, O., Gong, L. I., & Bloom, J. D. (2013). Mutational effects on stability are largely conserved during protein evolution. Proceedings of the National Academy of Sciences, 110(52), 21071-21076.

Association, A. D. (2016). Acute and Chronic Complications. Diabetes, 65(Supplement 1), A541-A556.

441

Attri, A. K., Fernandez, C., & Minton, A. P. (2010a). pH-dependent self-association of zinc-free insulin characterized by concentration-gradient static light scattering. Biophys. Chem., 148(1-3), 28-33. doi: 10.1016/j.bpc.2010.02.002

Attri, A. K., Fernandez, C., & Minton, A. P. (2010b). Self-association of Zn-insulin at neutral pH: investigation by concentration gradient--static and dynamic light scattering. Biophys. Chem., 148(1-3), 23-27. doi: 10.1016/j.bpc.2010.02.001

Attri, A. K., & Minton, A. P. (2005). New methods for measuring macromolecular interactions in solution via static light scattering: basic methodology and application to nonassociating and self-associating proteins. Anal. Biochem., 337(1), 103-110. doi: 10.1016/j.ab.2004.09.045

Baker, E. N., Blundell, T. L., Cutfield, J. F., Cutfield, S. M., Dodson, E. J., Dodson, G. G., . . . Reynolds, C. D. (1988). The structure of 2Zn pig insulin crystals at 1.5 Å resolution. Philos. Trans. R. Soc. Lond. B Biol. Sci., 319, 369-456.

Baker, N. A., Sept, D., Joseph, S., Holst, M. J., & McCammon, J. A. (2001). Electrostatics of nanosystems: application to microtubules and the ribosome. Proceedings of the National Academy of Sciences, 98(18), 10037-10041.

Bakhtiani, P. A., Caputo, N., Castle, J. R., El Youssef, J., Carroll, J. M., David, L. L., . . . Ward, W. K. (2014). A novel, stable, aqueous glucagon formulation using ferulic acid as an excipient. Journal of diabetes science and technology, 9(1), 17-23.

Ballerstadt, R., Evans, C., McNichols, R., & Gowda, A. (2006). Concanavalin A for in vivo glucose sensing: a biotoxicity review. Biosens. Bioelectron., 22(2), 275-284. doi: 10.1016/j.bios.2006.01.008

Bally, L., Thabit, H., & Hovorka, R. (2017). Finding the right route for insulin delivery– an overview of implantable pump therapy. Expert opinion on drug delivery(just- accepted).

Barany, G., & Merrifield, R. B. (1980). In E. Gross & J. Meienhofer (Eds.), The Peptides (pp. 273-284). New York: Academic Press.

Baron, R., & McCammon, J. A. (2013). Molecular Recognition and Ligand Association. Annual review of physical chemistry, 64(1), 151-175. doi: 10.1146/annurev- physchem-040412-110047

Becker, R. H., Frick, A. D., Burger, F., Potgieter, J. H., & Scholtz, H. (2005). Insulin glulisine, a new rapid-acting insulin analogue, displays a rapid time-action profile in obese non-diabetic subjects. Exp. Clin. Endocrinol. Diabetes, 113, 435-443.

442

Belfiore, A., Frasca, F., Pandini, G., Sciacca, L., & Vigneri, R. (2009). Insulin receptor isoforms and insulin receptor/insulin-like growth factor receptor hybrids in physiology and disease. Endocr. Rev., 30, 586-623.

Belfiore, A., Malaguarnera, R., Vella, V., Lawrence, M. C., Sciacca, L., Frasca, F., . . . Vigneri, R. (2017). Insulin Receptor Isoforms in Physiology and Disease: An Updated View. Endocr Rev, 38(5), 379-431. doi: 10.1210/er.2017-00073

Bentley, G., Dodson, E., Dodson, G., Hodgkin, D., & Mercola, D. (1976). Structure of insulin in 4-zinc insulin. Nature, 261, 166-168.

Bentley, G. A., Brange, J., Derewenda, Z., Dodson, E. J., Dodson, G. G., Markussen, J., . . . Xiao, B. (1992). Role of B13 Glu in insulin assembly. The hexamer structure of recombinant mutant (B13 Glu®Gln) insulin. J. Mol. Biol., 228, 1163-1176.

Beregszàszi, M., Tubiana-Rufi, N., Benali, K., Noël, M., Bloch, J., & Czernichow, P. (1997). Nocturnal hypoglycemia in children and adolescents with insulin- dependent diabetes mellitus: prevalence and risk factors. The Journal of pediatrics, 131(1), 27-33.

Berger, M., Cuppers, H. J., Hegner, H., Jorgens, V., & Berchtold, P. (1982). Absorption kinetics and biologic effects of subcutaneously injected insulin preparations. Diabetes Care, 5(2), 77-91.

Berger, M., Halban, P. A., Girardier, L., Seydoux, J., Offord, R., & Renold, A. (1979). Absorption kinetics of subcutaneously injected insulin. Diabetologia, 17(2), 97- 99.

Betts, S., & King, J. (1999). There's a right way and a wrong way: in vivo and in vitro folding, misfolding and subunit assembly of the P22 tailspike. Structure, 7, R131- R139.

Betts, S., & King, J. (1999). There's a right way and a wrong way: in vivo and in vitro folding, misfolding and subunit assembly of the P22 tailspike. Structure, 7(6), R131-R139.

Bhatnagar, B. S., Bogner, R. H., & Pikal, M. J. (2007). Protein stability during freezing: separation of stresses and mechanisms of protein stabilization. Pharm Dev Technol, 12(5), 505-523. doi: 783461155 [pii]10.1080/10837450701481157

Bhattacharyya, R., Samanta, U., & Chakrabarti, P. (2002). Aromatic-aromatic interactions in and around a-helices. Protein Eng., 15(2), 91-100.

Birnbaum, D. T., Dodd, S. W., Saxberg, B. E., Varshavsky, A. D., & Beals, J. M. (1996). Hierarchical modeling of phenolic ligand binding to 2Zn− insulin hexamers. Biochemistry, 35(17), 5366-5378.

443

Birnbaum, D. T., Kilcomons, M. A., DeFelippis, M. R., & Beals, J. M. (1997). Assembly and dissociation of human insulin and LysB28ProB29-insulin hexamers: a comparison study. Pharm. Res., 14, 25-36.

Blauw, H., Keith-Hynes, P., Koops, R., & DeVries, J. H. (2016). A review of safety and design requirements of the artificial pancreas. Annals of Biomedical Engineering, 44(11), 3158-3172.

Bliss, M. (Ed.). (1982). The discovery of insulin (Vol. 16). Chicago, Illinois: University of Chicago Press.

Bluestone, J. A., Herold, K., & Eisenbarth, G. (2010). Genetics, pathogenesis and clinical interventions in type 1 diabetes. Nature, 464(7293), 1293.

Blumenthal, S. (2010). From insulin and insulin-like activity to the insulin superfamily of growth-promoting peptides: a 20th-century odyssey. Perspect. Biol. Med., 53(4), 491-508. doi: 10.1353/pbm.2010.0001

Blundell, T. L., Cutfield, J. F., Cutfield, S. M., Dodson, E. J., Dodson, G. G., Hodgkin, D. C., & Mercola, D. A. (1972). Three-dimensional atomic structure of insulin and its relationship to activity. Diabetes, 21, 492-505.

Blundell, T. L., Cutfield, J. F., Cutfield, S. M., Dodson, E. J., Dodson, G. G., Hodgkin, D. C., . . . Vijayan, M. (1971). Atomic positions in rhombohedral 2-zinc insulin crystals. Nature, 231, 506-511.

Blundell, T. L., Dodson, G. G., Hodgkin, D. C., & Mercola, D. A. (1972). Insulin: the structure in the crystal and its reflection in chemistry and biology. Adv. Protein Chem., 26, 279-402.

Boland, B. B., Rhodes, C. J., & Grimsby, J. S. (2017). The dynamic plasticity of insulin production in β-cells. Molecular metabolism, 6(9), 958-973.

Brader, M. L., Kaarsholm, N. C., Lee, R. W., & Dunn, M. F. (1991). Characterization of the R-state insulin hexamer and its derivatives. The hexamer is stabilized by heterotropic ligand binding interactions. Biochemistry, 30(27), 6636-6645.

Brange, J. (2000). Physical stability of proteins. In S. Frokjaer & L. Hovgaard (Eds.), Pharmaceutical formulation development of peptides and proteins (1st ed., pp. 89-109). Philadelphia, PA, USA: Taylor & Francis.

Brange, J., Andersen, L., Laursen, E. D., Meyn, G., & Rasmussen, E. (1997). Toward understanding insulin fibrillation. J. Pharm. Sci., 86(5), 517-525. doi: 10.1021/js960297s

444

Brange, J., Hallund, O., & Sorensen, E. (1992). Chemical stability of insulin. 5. Isolation, characterization and identification of insulin transformation products. Acta Pharm. Nord., 4, 223-232.

Brange, J., Hansen, J. F., Havelund, S., & Melberg, S. G. (1987). Studies of insulin fibrillation process. In P. Brunetti & W. K. Waldhäusl (Eds.), Advanced models for the therapy of insulin-dependent diabeses (pp. 85-90). New York: Raven Press.

Brange, J., & Havelund, S. (1984). United States Patent No. United States Patent and Trademark Office patent full-text database: U. S. P. a. T. Office.

Brange, J., Havelund, S., & Hougaard, P. (1992). Chemical stability of insulin. 2. Formation of higher molecular weight transformation products during storage of pharmaceutical preparations. Pharm. Res., 9, 727-734.

Brange, J., & Langkjaer, L. (1992). Chemical stability of insulin. 3. Influence of excipients, formulation, and pH. Acta Pharm. Nord., 4, 149-158.

Brange, J., & Langkjaer, L. (1997). Insulin formulation and delivery. Pharm. Biotechnol., 10, 343-409.

Brange, J., Owens, D. R., Kang, S., & Volund, A. (1990). Monomeric insulins and their experimental and clinical implications. Diabetes Care, 13, 923-954.

Brange, J., Ribel, U., Hansen, J. F., Dodson, G., Hansen, M. T., Havelund, S., . . . Snel, L. (1988). Monomeric insulins obtained by protein engineering and their medical implications. Nature, 333, 679-682.

Brange, J., & Vølund, A. (1999). Insulin analogs with improved pharmacokinetic profiles. Adv. Drug Deliv. Rev., 35, 307-335.

Brems, D. N., Alter, L. A., Beckage, M. J., Chance, R. E., DiMarchi, R. D., Green, L. K., . . . Frank, B. H. (1992). Altering the association properties of insulin by amino acid replacement. Protein Eng., 5, 527-533.

Breton, M., Farret, A., Bruttomesso, D., Anderson, S., Magni, L., Patek, S., . . . Kovatchev, B. (2012). Fully integrated artificial pancreas in type 1 diabetes: modular closed-loop glucose control maintains near normoglycemia. Diabetes, 61(9), 2230-2237. doi: 10.2337/db11-1445

Brooks, B. R., Brooks, C. L., Mackerell, A. D., Nilsson, L., Petrella, R. J., Roux, B., . . . Karplus, M. (2009). CHARMM: The biomolecular simulation program. J. Comput. Chem., 30(10), 1545-1614. doi: 10.1002/jcc.21287

445

Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan, S. a., & Karplus, M. (1983). CHARMM: a program for macromolecular energy, minimization, and dynamics calculations. Journal of computational chemistry, 4(2), 187-217.

Brown, R. J., & Rother, K. I. (2008). Effects of beta‐cell rest on beta‐cell function: a review of clinical and preclinical data. Pediatric diabetes, 9(3pt2), 14-22.

Brownlee, M., & Cerami, A. (1979). A glucose-controlled insulin-delivery system: semisynthetic insulin bound to lectin. Science, 206(4423), 1190-1191.

Brownlee, M., & Cerami, A. (1983). Glycosylated insulin complexed to Concanavalin A. Biochemical basis for a closed-loop insulin delivery system. Diabetes, 32(6), 499- 504.

Brunger, A. T. (1992). X-PLOR (Version 3.1) a system for x-ray crystallography and NMR (3.1 ed.). New Haven, CT: Yale University Press.

Bryan, P. N. (2002). Prodomains and protein folding catalysis. Chemical reviews, 102(12), 4805-4816.

Bucciantini, M., Giannoni, E., Chiti, F., Baroni, F., Formigli, L., Zurdo, J., . . . Stefani, M. (2002). Inherent toxicity of aggregates implies a common mechanism for protein misfolding diseases. Nature, 416, 507-511.

Burke, G. T., Schwartz, G., & Katsoyannis, P. G. (1984). Nature of the B10 amino acid residue. Requirements for high biological activity of insulin. Int. J. Pept. Protein Res., 23, 394-401.

Burley, S., & Petsko, G. (1986a). Amino-aromatic interactions in proteins. Febs Letters, 203(2), 139-143.

Burley, S., & Petsko, G. (1986b). Dimerization energetics of benzene and aromatic amino acid side chains. Journal of the American Chemical Society, 108(25), 7995-8001.

Burley, S. K., & Petsko, G. A. (1985). Aromatic-aromatic interaction: a mechanism of protein structure stabilization. Science, 229, 23-28.

Burley, S. K., & Petsko, G. A. (1988). Weakly polar interaction in proteins. Adv. Protein Chem., 39, 125-189.

Campbell, M. D., Walker, M., Bracken, R. M., Turner, D., Stevenson, E. J., Gonzalez, J. T., . . . West, D. J. (2015). Insulin therapy and dietary adjustments to normalize glycemia and prevent nocturnal hypoglycemia after evening exercise in type 1 diabetes: a randomized controlled trial. BMJ Open Diabetes Research and Care, 3(1), e000085.

446

Campbell, R. K., Campbell, L. K., & White, J. R. (1996). Insulin lispro: its role in the treatment of diabetes mellitus. Ann. Pharmacother., 30, 1263-1271.

Carpenter, M. C., & Wilcox, D. E. (2014). Thermodynamics of formation of the insulin hexamer: metal-stabilized proton-coupled assembly of quaternary structure. Biochemistry, 53(8), 1296-1301. doi: 10.1021/bi4016567

Chan, S. J., Cao, Q. P., & Steiner, D. F. (1990). Evolution of the insulin superfamily: cloning of a hybrid insulin/insulin- like growth factor cDNA from Amphioxus. Proc. Natl. Acad. Sci. U. S. A., 87, 9319-9323. doi: PMC55156

Chan, S. J., & Steiner, D. F. (2000). Insulin Through the Ages: Phylogeny of a Growth Promoting and Metabolic Regulatory Hormone. Integrative and Comparative Biology, 40(2), 213-222.

Chang, B. S., Kendrick, B. S., & Carpenter, J. F. (1996). Surface-induced denaturation of proteins during freezing and its inhibition by surfactants. J Pharm Sci, 85(12), 1325-1330. doi: 10.1021/js960080y10.1021/js960080y [pii]

Chang, X., Jorgensen, A. M., Bardrum, P., & Led, J. J. (1997). Solution structures of the R6 human insulin hexamer. Biochemistry, 36, 9409-9422.

Chap, Z., Ishida, T., Chou, J., Hartley, C. J., Entman, M. L., Brandenburg, D., . . . Field, J. B. (1987). First-pass hepatic extraction and metabolic effects of insulin and insulin analogues. American Journal of Physiology-Endocrinology and Metabolism, 252(2), E209-E217.

Chiti, F., & Dobson, C. M. (2006). Protein misfolding, functional amyloid, and human disease. Annu. Rev. Biochem., 75, 333-366.

Chodera, J. D., & Mobley, D. L. (2013). Entropy-enthalpy compensation: role and ramifications in biomolecular ligand recognition and design. Annu Rev Biophys, 42, 121-142. doi: 10.1146/annurev-biophys-083012-130318

Chothia, C., Lesk, A. M., Dodson, G. G., & Hodgkin, D. C. (1983). Transmission of conformational change in insulin. Nature, 302, 500-505.

Chou, D. H.-C., Webber, M. J., Tang, B. C., Lin, A. B., Thapa, L. S., Deng, D., . . . Anderson, D. G. (2015). Glucose-responsive insulin activity by covalent modification with aliphatic phenylboronic acid conjugates. Proc. Natl. Acad. Sci. USA, 112(8), 2401-2406. doi: 10.1073/pnas.1424684112

Christensen, M., & Knop, F. K. (2010). Once-weekly GLP-1 agonists: How do they differ from exenatide and liraglutide? Current diabetes reports, 10(2), 124-132.

447

Chu, Y. C., Wang, R. Y., Burke, G. T., Chanley, J. D., & Katsoyannis, P. G. (1987a). Possible involvement of the A20-A21 peptide bond in the biological activity of insulin. 1. [21-Desasparagine,20-cysteinamide-A]insulin and [21- desasparagine,20-cysteine isopropylamide-A]insulin. Biochemistry, 26, 6966- 6971.

Chu, Y. C., Wang, R. Y., Burke, G. T., Chanley, J. D., & Katsoyannis, P. G. (1987b). Possible involvement of the A20-A21 peptide bond in the expression of the biological activity of insulin. 3. [21-Desasparagine,20-cysteine ethylamide- A]insulin and [21-desasparagine,20-cysteine 2,2,2-trifluoroethylamide-A]insulin. Biochemistry, 26, 6975-6979.

Ciszak, E., Beals, J. M., Frank, B. H., Baker, J. C., Carter, N. D., & Smith, G. D. (1995). Role of C-terminal B-chain residues in insulin assembly: the structure of hexameric LysB28ProB29-human insulin. Structure (Lond.), 3, 615-622.

Ciszak, E., & Smith, G. D. (1994). Crystallographic evidence for dual coordination around zinc in the T3R3 human insulin hexamer. Biochemistry, 33, 1512-1517.

Cobelli, C., Renard, E., Kovatchev, B. P., Keith-Hynes, P., Ben Brahim, N., Place, J., . . . Avogaro, A. (2012). Pilot studies of wearable outpatient artificial pancreas in type 1 diabetes. Diabetes Care, 35(9), e65-67. doi: 10.2337/dc12-0660

Cohen, E., & Dillin, A. (2008). The insulin paradox: aging, proteotoxicity and neurodegeneration. Nat. Rev. Neurosci., 9(10), 759-767. doi: 10.1038/nrn2474

Cohen, E., Du, D., Joyce, D., Kapernick, E. A., Volovik, Y., Kelly, J. W., & Dillin, A. (2010). Temporal requirements of insulin/IGF-1 signaling for proteotoxicity protection. Aging Cell, 9, 126-134. doi: PMC3026833

Colombo, C., Porzio, O., Liu, M., Massa, O., Vasta, M., Salardi, S., . . . Early Onset Diabetes Study Group of the Italian Society of Pediatric Endocrinology and Diabetes. (2008). Seven mutations in the human insulin gene linked to permanent neonatal/infancy-onset diabetes mellitus. J. Clin. Invest., 118, 2148-2156. doi: PMC2350430

Conlon, J. M. (2000). Molecular evolution of insulin in non-mammalian vertebrates. Amer. Zool., 40, 200-212.

Conlon, J. M. (2001). Evolution of the insulin molecule: insights into structure-activity and phylogenetic relationships. Peptides, 22, 1183-1193.

Control, D., & Group, C. T. R. (1997). Hypoglycemia in the diabetes control and complications trial. Diabetes, 46(2), 271-286.

448

Cosmatos, A., Okada, Y., & Katsoyannis, P. G. (1976). Synthesis of two biologically active insulin analogs with modifications at the N-terminal and N-and C-terminal amino acid residues. Biochemistry, 15(18), 4076-4082.

Coulthard, C., Michaelis, R., Short, W., Sykes, G., Skrimshire, G., Standfast, A., . . . Raistrick, H. (1945). Notatin: an anti-bacterial glucose-aerodehydrogenase from Penicillium notatum Westling and Penicillium resticulosum sp. nov. Biochemical Journal, 39(1), 24.

Coutinho, A., Larsson, E. L., Grönvik, K. O., & Andersson, J. (1979). Studies on T lymphocyte activation II. The target cells for concanavalin A‐induced growth factors. European journal of immunology, 9(8), 587-592.

Creighton, T. E., Weissman, J. S., & Kim, P. S. (1992). The disulfide folding pathway of BPTI. Science, 256(5053), 111-114.

Croll, T. I., Smith, B. J., Margetts, M. B., Whittaker, J., Weiss, M. A., Ward, C. W., & Lawrence, M. C. (2016). Higher-Resolution Structure of the Human Insulin Receptor Ectodomain: Multi-Modal Inclusion of the Insert Domain. Structure, 24(3), 469-476. doi: 10.1016/j.str.2015.12.014

Cryer, P. E. (1997). Hypoglycemia: pathophysiology, diagnosis, and treatment: Oxford University Press, USA.

Cryer, P. E. (2002). Hypoglycaemia: the limiting factor in the glycaemic management of type I and type II diabetes. Diabetologia, 45. doi: 10.1007/s00125-002-0822-9

Cryer, P. E., Davis, S.N., Shannon, H. (2003). Hypoglycemia in Diabetes. Diabetes Care, 26, 1902-1912.

Damian, M., Martin, A., Mesnier, D., Pin, J. P., & Banères, J. L. (2006). Asymmetric conformational changes in a GPCR dimer controlled by G‐proteins. The EMBO Journal, 25(24), 5693-5702.

Das, A., & Shah, S. (2011). History of diabetes: from ants to analogs. J Assoc Physicians India, 59(Suppl)), 6-7.

Dawson, P. E., Muir, T. W., Clark-Lewis, I., & Kent, S. B. H. (1994). Synthesis of proteins by native chemical ligation. Science, 266, 776-779.

DCCT. (1993a). The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N. Engl. J. Med., 329, 977-986.

DCCT. (1993b). The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus.

449

The Diabetes Control and Complications Trial Research Group. N. Engl. J. Med., 329(14), 977-986. doi: 10.1056/nejm199309303291401

De Meyts, P. (2000). The Insulin Receptor and Its Signal Transduction Network.

De Meyts, P. (2002). Insulin and insulin-like growth factors: the paradox of signaling specificity. Growth Horm. IGF Res., 12, 81-83.

De Meyts, P. (2004). Insulin and its receptor: structure, function and evolution. Bioessays, 26, 1351-1362.

De Meyts, P. (2015). Insulin/receptor binding: the last piece of the puzzle? What recent progress on the structure of the insulin/receptor complex tells us (or not) about negative cooperativity and activation. Bioessays, 37(4), 389-397. doi: 10.1002/bies.201400190

De Meyts, P., Roth, J., Neville, J., Gavin, J. R., & Lesniak, M. A. (1973). Insulin ineractions with its receptors: experimental evidence for negative cooperativity. Biochem. Biophys. Res. Commun., 55, 154-161.

DeBoer, M. D., Cherñavvsky, D. R., Topchyan, K., Kovatchev, B. P., Francis, G. L., & Breton, M. D. (2016). Heart rate informed artificial pancreas system enhances glycemic control during exercise in adolescents with T1D. Pediatric diabetes.

DeFelippis, M. R., Chance, R. E., & Frank, B. H. (2001). Insulin self-association and the relationship to pharmacokinetics and pharmacodynamics. Crit. Rev. Ther. Drug Carrier Syst., 18, 201-264.

DeFelippis, M. R., DiMarchi, R. D., & Ng, K. (2007). Pre-mixes of GLP-1 and basal insulin: Google Patents.

DeFronzo, R., Tobin, J., & Andres, R. (1979). Glucose clamp technique: a method for quantifying insulin secretion and resistance. [Journal Article]. Am. J. Physiol. Gastrointest. Liver Physiol., 237(3), G214-G223.

Derewenda, U., Derewenda, Z., Dodson, E. J., Dodson, G. G., Reynolds, C. D., Smith, G. D., . . . Swenson, D. (1989). Phenol stabilizes more helix in a new symmetrical zinc insulin hexamer. Nature, 338, 594-596. doi: 10.1038/338594a0

Derewenda, U., DEREWENDA, Z., DODSON, G., & BRANGE, J. (1987). THE CRYSTAL-STRUCTURE OF THE B-12 ILE HUMAN INSULIN PREPARED BY SITE-DIRECTED MUTAGENESIS. Paper presented at the PROTEIN ENGINEERING.

Deshayes, S., Cabral, H., Ishii, T., Miura, Y., Kobayashi, S., Yamashita, T., . . . Kataoka, K. (2013). Phenylboronic acid-installed polymeric micelles for targeting

450

sialylated epitopes in solid tumors. Journal of the American Chemical Society, 135(41), 15501-15507.

Devlin, G. L., Knowles, T. P., Squires, A., McCammon, M. G., Gras, S. L., Nilsson, M. R., . . . MacPhee, C. E. (2006). The component polypeptide chains of bovine insulin nucleate or inhibit aggregation of the parent protein in a conformation- dependent manner. J. Mol. Biol., 360, 497-509.

Di Guglielmo, G. M., Drake, P. G., Baass, P. C., Authier, F., Posner, B. I., & Bergeron, J. J. (1998). Insulin receptor internalization and signalling. Mol Cell Biochem, 182(1-2), 59-63.

Dill, K. A., & Chan, H. S. (1997). From Levinthal to pathways to funnels. Nat. Struct. Biol., 4, 10-19.

Dobson, C. M. (1999). Protein misfolding, evolution and disease. Trends Biochem. Sci., 24, 329-332.

Dobson, C. M. (2001). Protein folding and its links with human disease Biochem. Soc. Symp. (Vol. 68, pp. 1-26). London: The Biochemical Society.

Dobson, C. M. (2003). Protein folding and disease: a view from the first Horizon Symposium. Nat. Rev. Drug Discov., 2, 154-160.

Dobson, C. M. (2003). Protein folding and misfolding. Nature, 426, 884-890.

Dobson, C. M., & Karplus, M. (1999). The fundamentals of protein folding: bringing together theory and experiment. Curr. Opin. Struct. Biol., 9, 92-101.

Dodson, G., & Steiner, D. (1998). The role of assembly in insulin's biosynthesis. Curr. Opin. Struct. Biol., 8, 189-194.

Dodson, G. G., Dodson, E. J., Turkenburg, J. P., & Bing, X. (1993). Molecular recognition in insulin assembly. Biochem. Soc. Trans., 21, 609-614.

Dong, J., Wan, Z., Popov, M., Carey, P. R., & Weiss, M. A. (2003). Insulin assembly damps conformational fluctuations: Raman analysis of amide I line widths in native states and fibrils. J. Mol. Biol., 330, 431-442.

Dong, Y., Wang, W., Veiseh, O., Appel, E. A., Xue, K., Webber, M. J., . . . Langer, R. (2016). Injectable and Glucose-Responsive Hydrogels Based on Boronic Acid– Glucose Complexation. Langmuir, 32(34), 8743-8747.

Dougherty, D. A. (2000). Unnatural amino acids as probes of protein structure and function. Current opinion in chemical biology, 4(6), 645-652.

451

Doyle, F. J., Huyett, L. M., Lee, J. B., Zisser, H. C., & Dassau, E. (2014). Closed-loop artificial pancreas systems: engineering the algorithms. Diabetes Care, 37(5), 1191-1197.

Drickamer, K., & Taylor, M. E. (1993). Biology of animal lectins. Annual review of cell biology, 9(1), 237-264.

Du, X., & Tang, J. G. (1998). Hydroxyl group of insulin A19Tyr is essential for receptor binding: studies on (A19Phe)insulin. Biochem. Mol. Biol. Int., 45, 255-260.

Duckworth, W., Abraira, C., Moritz, T., Reda, D., Emanuele, N., Reaven, P. D., . . . Huang, G. D. (2009). Glucose control and vascular complications in veterans with type 2 diabetes. N. Engl. J. Med., 360(2), 129-139. doi: 10.1056/NEJMoa0808431

Duttaroy, A., Kanakaraj, P., Osborn, B. L., Schneider, H., Pickeral, O. K., Chen, C., . . . Schulingkamp, R. (2005). Development of a long-acting insulin analog using albumin fusion technology. Diabetes, 54(1), 251-258.

Edghill, E. L., Flanagan, S. E., Patch, A. M., Boustred, C., Parrish, A., Shields, B., . . . Ellard, S. (2008). Insulin mutation screening in 1044 patients with diabetes: mutations in the INS gene are a common cause of neonatal diabetes but a rare cause of diabetes diagnosed in childhood or adulthood. Diabetes, 57, 1034-1042.

Edwards, M. S. (2009). Nutrition and behavior of degus (Octodon degus). Vet. Clin. North Am. Exot. Anim. Pract., 12(2), 237-253, viii. doi: 10.1016/j.cvex.2009.01.003

Eisenberg, D., & McLachlan, A. D. (1986). Solvation energy in protein folding and binding. Nature, 319, 199-203.

El-Khatib, F. H., Balliro, C., Hillard, M. A., Magyar, K. L., Ekhlaspour, L., Sinha, M., . . . Thompson, M. J. (2016). Home use of a bihormonal bionic pancreas versus insulin pump therapy in adults with type 1 diabetes: a multicentre randomised crossover trial. The Lancet.

El Hage, K., Bereau, T., Jakobsen, S., & Meuwly, M. (2016). Impact of Quadrupolar Electrostatics on Atoms Adjacent to the Sigma-Hole in Condensed-Phase Simulations. J. Chem. Theory Comp. doi: 10.1002/jcc.23786

El Hage, K., Pandyarajan, V., Phillips, N. B., Smith, B. J., Menting, J. G., Whittaker, J., . . . Weiss, M. A. (2016). Extending Halogen-based Medicinal Chemistry to Proteins: IODO-INSULIN AS A CASE STUDY. J Biol Chem, 291(53), 27023- 27041. doi: 10.1074/jbc.M116.761015

Ellingsen, C., Dassau, E., Zisser, H., Grosman, B., Percival, M. W., Jovanovič, L., & Doyle III, F. J. (2009). Safety constraints in an artificial pancreatic β cell: an

452

implementation of model predictive control with insulin on board. Journal of diabetes science and technology, 3(3), 536-544.

Elliott, R., Escobar, L., Tan, P., Garkavenko, O., Calafiore, R., Basta, P., . . . Bambra, C. (2005). Intraperitoneal alginate-encapsulated neonatal porcine islets in a placebo-controlled study with 16 diabetic cynomolgus primates. Paper presented at the Transplantation proceedings.

Emdin, S. O., Dodson, G. G., Cutfield, J. M., & Cutfield, S. M. (1980). Role of zinc in insulin biosynthesis. Some possible zinc-insulin interactions in the pancreatic B- cell. Diabetologia, 19, 174-182.

Esteghamati, A., Ashraf, H., Khalilzadeh, O., Zandieh, A., Nakhjavani, M., Rashidi, A., . . . Asgari, F. (2010). Optimal cut-off of homeostasis model assessment of insulin resistance (HOMA-IR) for the diagnosis of metabolic syndrome: third national surveillance of risk factors of non-communicable diseases in Iran (SuRFNCD- 2007). Nutrition & metabolism, 7(1), 26.

Fajans, S. S., & Bell, G. I. (2011). MODY: history, genetics, pathophysiology, and clinical decision making. Diabetes Care, 34(8), 1878-1884.

Fang, J., Liu, M., Zhang, X., Sakamoto, T., Taatjes, D. J., Jena, B. P., . . . Kowluru, A. (2015). COPII-dependent ER export: a critical component of insulin biogenesis and β-cell ER homeostasis. Molecular Endocrinology, 29(8), 1156-1169.

Fink, A. L. (2005). Natively unfolded proteins. Curr. Opin. Struct. Biol., 15, 35-41.

Fitzpatrick, A. W., Debelouchina, G. T., Bayro, M. J., Clare, D. K., Caporini, M. A., Bajaj, V. S., . . . Müller, S. A. (2013). Atomic structure and hierarchical assembly of a cross-β amyloid fibril. Proceedings of the National Academy of Sciences, 110(14), 5468-5473.

Flamment, M., Hajduch, E., Ferré, P., & Foufelle, F. (2012). New insights into ER stress- induced insulin resistance. Trends in Endocrinology & Metabolism, 23(8), 381- 390.

Forbes, B., McCarthy, P., & Norton, R. (2012). Insulin-like growth factor binding proteins: a structural perspective. Frontiers in endocrinology, 3, 38.

Frayling, T. M. (2007). Genome–wide association studies provide new insights into type 2 diabetes aetiology. Nature Reviews Genetics, 8(9), 657.

Frederick, K. K., Marlow, M. S., Valentine, K. G., & Wand, A. J. (2007). Conformational entropy in molecular recognition by proteins. Nature, 448(7151), 325.

453

Freire, E. (2009). A Thermodynamic Approach to the Affinity Optimization of Drug Candidates. Chemical Biology & Drug Design, 74(5), 468-472. doi: 10.1111/j.1747-0285.2009.00880.x

Friedman, S., & Pizer, R. (1975). Mechanism of the complexation of phenylboronic acid with oxalic acid. Reaction which requires ligand donor atom protonation. Journal of the American Chemical Society, 97(21), 6059-6062.

Frisch, M. J., Trucks, G. W., Schlegal, H. B., Scuseria, G. E., Robb, M. A., Cheeseman, J. R., . . . Fox, D. J. (2009). Gaussian09 Revision D.01. Wallingford, CT: Gaussian Inc.

Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria, G. E., Robb, M. A., Cheeseman, J. R., . . . Fox, D. J. (2009). Gaussian 09, Revision A.02 (Version Revision A.02).

Frost, G. I. (2007). Recombinant human hyaluronidase (rHuPH20): an enabling platform for subcutaneous drug and fluid administration. Expert opinion on drug delivery, 4(4), 427-440.

Gallagher, E. J., Alikhani, N., Tobin-Hess, A., Blank, J., Buffin, N. J., Zelenko, Z., . . . LeRoith, D. (2013). Insulin receptor phosphorylation by endogenous insulin or the insulin analog AspB10 promotes mammary tumor growth independent of the IGF-I receptor. Diabetes, 62(10), 3553-3560. doi: 10.2337/db13-0249

Galloway, J. A., Hooper, S. A., Spradlin, C. T., Howey, D. C., Frank, B. H., Bowsher, R. R., & Anderson, J. H. (1992). Biosynthetic human proinsulin. Review of chemistry, in vitro and in vivo receptor binding, animal and human pharmacology studies, and clinical trial experience. Diabetes Care, 15, 666-692.

Garg, S., Ampudia-Blasco, F. J., & Pfohl, M. (2010). Rapid-acting insulin analogues in basal-bolus regimens in type 1 diabetes mellitus. Endocr. Pract., 16, 486-505.

Garg, S. K., Ellis, S. L., & Ulrich, H. (2005). Insulin glulisine: a new rapid-acting insulin analogue for the treatment of diabetes. Expert Opin. Pharmacother., 6, 643-651.

Genuth, S. M. (1972). Metabolic clearance of insulin in man. Diabetes, 21(10), 1003- 1012.

Gerich, J. E., Charles, M. A., & Grodsky, G. M. (1976). Regulation of pancreatic insulin and glucagon secretion. Annual review of physiology, 38(1), 353-388.

Gerstein, H. C., Miller, M. E., Byington, R. P., Goff, D. C., Jr., Bigger, J. T., Buse, J. B., . . . Friedewald, W. T. (2008). Effects of intensive glucose lowering in type 2 diabetes. N. Engl. J. Med., 358(24), 2545-2559. doi: 10.1056/NEJMoa0802743

454

Gibson, T. J., & Murphy, R. M. (2006). Inhibition of insulin fibrillogenesis with targeted peptides. Protein Sci., 15, 1133-1141. doi: PMC2242521

Gillies, P. S., Figgitt, D. P., & Lamb, H. M. (2000). Insulin glargine. Drugs, 59(2), 253- 260.

Glendorf, T., Sorensen, A. R., Nishimura, E., Pettersson, I., & Kjeldsen, T. (2008). Importance of the solvent-exposed residues of the insulin B chain α-helix for receptor binding. Biochemistry, 47, 4743-4751.

Glendorf, T., Stidsen, C. E., Norrman, M., Nishimura, E., Sorensen, A. R., & Kjeldsen, T. (2011). Engineering of insulin receptor isoform-selective insulin analogues. PLoS One, 6(5), e20288. doi: 10.1371/journal.pone.0020288

Glendorf, T., Stidsen, C. E., Norrman, M., Nishimura, E., Sørensen, A. R., & Kjeldsen, T. (2011). Engineering of insulin receptor isoform-selective insulin analogues. PloS one, 6(5), e20288.

Glidden, M. D., Aldabbagh, K., Phillips, N. B., Carr, K., Whittaker, J., Yee, V., . . . Weiss, M. A. (2017). Crystal Structure and Function of an Ultra-Stable Single- Chain Insulin Analog. J Biol Chem.

Glidden, M. D., Yang, Y., Phillips, N. B., Carr, K., Smith, N. A., Wickramasinghe, N., . . . Weiss, M. A. (2017). Solution structure and dynamics of a single-chain insulin analog. J Biol Chem.

Goeddel, D. V., Kleid, D. G., Bolivar, F., Heyneker, H. L., Yansura, D. G., Crea, R., . . . Riggs, A. D. (1979). Expression in Escherichia coli of chemically synthesized genes for human insulin. Proceedings of the National Academy of Sciences, 76(1), 106-110.

Goldenberg, D. P., & Creighton, T. E. (1994). Phage tailspike protein. A fishy tale of protein folding. Curr. Biol., 4, 1026-1029.

Gorr, S. U., Huang, X. F., Cowley, D. J., Kuliawat, R., & Arvan, P. (1999). Disruption of disulfide bonds exhibits differential effects on trafficking of regulated secretory proteins. Am. J. Physiol., 277, C121-C131.

Gough, S., Harris, S., Woo, V., & Davies, M. (2013). Insulin degludec: overview of a novel ultra long‐acting basal insulin. Diabetes, Obesity and Metabolism, 15(4), 301-309.

Graf, A., McAuley, S. A., Sims, C., Ulloa, J., Jenkins, A. J., Voskanyan, G., & O’Neal, D. N. (2016). Moving Toward a Unified Platform for Insulin Delivery and Sensing of Inputs Relevant to an Artificial Pancreas. Journal of diabetes science and technology, 1932296816682762.

455

Greenfield, J. R., & Campbell, L. V. (2004). Insulin resistance and obesity. Clinics in dermatology, 22(4), 289-295.

Grönke, S., Clarke, D. F., Broughton, S., Andrews, T. D., & Partridge, L. (2010). Molecular evolution and functional characterization of Drosophila insulin-like peptides. PLoS Genet., 6(2), e1000857. doi: 10.1371/journal.pgen.1000857

Group, D. R. (1991). Epidemiology of severe hypoglycemia in the diabetes control and complications trial. The American journal of medicine, 90(4), 450-459.

Gruppuso, P. A., Gorden, P., Kahn, C. R., Cornblath, M., Zeller, W. P., & Schwartz, R. (1984). Familial hyperproinsulinemia due to a proposed defect in conversion of proinsulin to insulin. N. Engl. J. Med., 311, 629-634.

Gu, Z., Dang, T. T., Ma, M., Tang, B. C., Cheng, H., Jiang, S., . . . Anderson, D. G. (2013). Glucose-responsive microgels integrated with enzyme nanocapsules for closed-loop insulin delivery. ACS Nano, 7(8), 6758-6766.

Gubitosi-Klug, R. A., Braffett, B. H., White, N. H., Sherwin, R. S., Lachin, J. M., Tamborlane, W. V., . . . Group, C. R. (2017). Risk of severe hypoglycemia in type 1 diabetes over 30 years of follow-up in the DCCT/EDIC study. Diabetes Care, 40(8), 1010-1016.

Guo, Z., Shin, I., & Yoon, J. (2012). Recognition and sensing of various species using boronic acid derivatives. Chemical Communications, 48(48), 5956-5967.

Guo, Z. Y., Qiao, Z. S., & Feng, Y. M. (2008). The in vitro oxidative folding of the insulin superfamily. Antioxid. Redox Signal, 10, 127-140.

Guvench, O., & Brooks, C. L., 3rd. (2005). Tryptophan side chain electrostatic interactions determine edge-to-face vs parallel-displaced tryptophan side chain geometries in the designed beta-hairpin "trpzip2". J Am Chem Soc, 127(13), 4668- 4674. doi: 10.1021/ja043492e

Haataja, L., Snapp, E., Wright, J., Liu, M., Hardy, A. B., Wheeler, M. B., . . . Arvan, P. (2013). Proinsulin intermolecular interactions during secretory trafficking in pancreatic β cells. J. Biol. Chem., 288(3), 1896-1906. doi: 10.1074/jbc.M112.420018

Hagedorn, H., Jensen, B. N., Krarup, N., & Wodstrup, I. (1936). Protamine insulinate. Journal of the American Medical Association, 106(3), 177-180.

Haidar, A., Legault, L., Matteau-Pelletier, L., Messier, V., Dallaire, M., Ladouceur, M., & Rabasa-Lhoret, R. (2015). Outpatient overnight glucose control with dual- hormone artificial pancreas, single-hormone artificial pancreas, or conventional insulin pump therapy in children and adolescents with type 1 diabetes: an open-

456

label, randomised controlled trial. The lancet Diabetes & endocrinology, 3(8), 595-604.

Haidar, A., Legault, L., Messier, V., Mitre, T. M., Leroux, C., & Rabasa-Lhoret, R. (2015). Comparison of dual-hormone artificial pancreas, single-hormone artificial pancreas, and conventional insulin pump therapy for glycaemic control in patients with type 1 diabetes: an open-label randomised controlled crossover trial. The lancet Diabetes & endocrinology, 3(1), 17-26.

Halvorson, M., Carpenter, S., Kaiserman, K., & Kaufman, F. R. (2007). A pilot trial in pediatrics with the sensor-augmented pump: combining real-time continuous glucose monitoring with the insulin pump. The Journal of pediatrics, 150(1), 103- 105. e101.

Hansen, B. F., Kurtzhals, P., Jensen, A. B., Dejgaard, A., & Russell-Jones, D. (2011). Insulin X10 revisited: a super-mitogenic insulin analogue. Diabetologia, 54(9), 2226-2231. doi: 10.1007/s00125-011-2203-8

Harding, M. M., Hodgkin, D. C., Kennedy, A. F., O'Conor, A., & Weitzmann, P. D. (1966). The crystal structure of insulin. II. An investigation of rhombohedral zinc insulin crystals and a report of other crystalline forms. J. Mol. Biol., 16, 212-226.

Hartley, T., Siva, M., Lai, E., Teodoro, T., Zhang, L., & Volchuk, A. (2010). Endoplasmic reticulum stress response in an INS-1 pancreatic β-cell line with inducible expression of a folding-deficient proinsulin. BMC Cell Biol., 11, 59. doi: 10.1186/1471-2121-11-59

Hassiepen, U., Federwisch, M., Mulders, T., & Wollmer, A. (1999). The lifetime of insulin hexamers. Biophys. J., 77(3), 1638-1654. doi: 10.1016/s0006- 3495(99)77012-8

Havel, T. F., & Snow, M. E. (1991). A new method for building protein conformations from sequence alignments with homologues of known structure. J. Mol. Biol., 217, 1-7.

Havelund, S., Plum, A., Ribel, U., Jonassen, I., Vølund, A., Markussen, J., & Kurtzhals, P. (2004). The mechanism of protraction of insulin detemir, a long-acting, acylated analog of human insulin. Pharm. Res., 21, 1498-1504.

Hedeskov, C. J. (1980). Mechanism of glucose-induced insulin secretion. Physiological Reviews, 60(2), 442-509.

Heidenreich, K. A., Yip, C. C., Frank, B. H., & Olefsky, J. M. (1985). The preparation and characterization of mono-iodinated photoreactive analogs of insulin. Biochemical and biophysical research communications, 126(3), 1138-1145.

457

Heidenreich, K. A., Zahniser, N. R., Berhanu, P., Brandenburg, D., & Olefsky, J. M. (1983). Structural differences between insulin receptors in the brain and peripheral target tissues. J Biol Chem, 258(14), 8527-8530.

Heise, T., Hövelmann, U., Brøndsted, L., Adrian, C., Nosek, L., & Haahr, H. (2015). Faster‐acting insulin aspart: earlier onset of appearance and greater early pharmacokinetic and pharmacodynamic effects than insulin aspart. Diabetes, Obesity and Metabolism, 17(7), 682-688.

Heise, T., Hövelmann, U., Zijlstra, E., Stender-Petersen, K., Jacobsen, J. B., & Haahr, H. (2017). A Comparison of Pharmacokinetic and Pharmacodynamic Properties Between Faster-Acting Insulin Aspart and Insulin Aspart in Elderly Subjects with Type 1 Diabetes Mellitus. Drugs & Aging, 34(1), 29-38.

Heise, T., Nosek, L., Bottcher, S. G., Hastrup, H., & Haahr, H. (2012). Ultra-long-acting insulin degludec has a flat and stable glucose-lowering effect in type 2 diabetes. Diabetes Obes. Metab., 14(10), 944-950. doi: 10.1111/j.1463-1326.2012.01638.x

Heise, T., Pieber, T. R., Danne, T., Erichsen, L., & Haahr, H. (2017). A pooled analysis of clinical pharmacology trials investigating the pharmacokinetic and pharmacodynamic characteristics of fast-acting insulin aspart in adults with type 1 diabetes. Clinical Pharmacokinetics, 1-9.

Hellman, U., Wernstedt, C., Westermark, P., O'Brien, T. D., Rathbun, W. B., & Johnson, K. H. (1990). Amino acid sequence from degu islet amyloid-derived insulin shows unique sequence characteristics. Biochem. Biophys. Res. Commun., 169(2), 571-577.

Helms, K. L., & Kelley, K. W. (2009). Insulin glulisine: an evaluation of its pharmacodynamic properties and clinical application. Ann. Pharmacother., 43, 658-668.

Herbach, N., Rahtkolb, B., Kemter, E., Pichl, L., Klaften, M., De Angelis, M. H., . . . Wanker, R. (2007). Dominant-negative effects of a novel mutated Ins2 allele causes early-onset diabetes and severe β-cell loss in Munich Ins2C95S mutant mice. Diabetes, 56, 1268-1276.

Herbach, N., Rathkolb, B., Kemter, E., Pichl, L., Klaften, M., de Angelis, M. H., . . . Wanke, R. (2007). Dominant-negative effects of a novel mutated Ins2 allele causes early-onset diabetes and severe β-cell loss in Munich Ins2C95S mutant mice. Diabetes, 56(5), 1268-1276.

Hilgenfeld, R., Seipke, G., Berchtold, H., & D.R. Owens, D. R. (2014). The Evolution of Insulin Glargine and its Continuing Contribution to Diabetes Care. Drugs, 74 (8), 911-927.

458

Hirshberg, B. (2007). Lessons learned from the international trial of the edmonton protocol for islet transplantation. Current diabetes reports, 7(4), 301-303.

Hodish, I., Liu, M., Rajpal, G., Larkin, D., Holz, R. W., Adams, A., . . . Arvan, P. (2010). Misfolded proinsulin affects bystander proinsulin in neonatal diabetes. J. Biol. Chem., 285(1), 685-694.

Hoeg-Jensen, T., Havelund, S., & Markussen, J. (2008). United States Patent No.

Hoeg-Jensen, T., Havelund, S., Markussen, J., Ostergaard, S., Ridderberg, S., Balschmidt, P., . . . Jonassen, I. (2008). United States Patent No.

Hoeg-Jensen, T., Ridderberg, S., Havelund, S., Schäffer, L., Balschmidt, P., Jonassen, I., . . . Markussen, J. (2005). Insulins with built-in glucose sensors for glucose responsive insulin release. J. Pept. Sci., 11(6), 339-346. doi: 10.1002/psc.624

Holleman, F., & Hoekstra, J. B. (1997). Insulin lispro. N. Engl. J. Med., 337, 176-183.

Holst, B., Nygaard, R., Valentin-Hansen, L., Bach, A., Engelstoft, M. S., Petersen, P. S., . . . Schwartz, T. W. (2010). A conserved aromatic lock for the tryptophan rotameric switch in TM-VI of seven-transmembrane receptors. Journal of Biological Chemistry, 285(6), 3973-3985.

Holst, J. J., Holland, W., Gromada, J., Lee, Y., Unger, R. H., Yan, H., . . . Herrera, P. L. (2017). Insulin and Glucagon: Partners for Life. Endocrinology, 158(4), 696-701. doi: 10.1210/en.2016-1748

Horuk, R., Goodwin, P., O'Connor, K., Neville, R. W., Lazarus, N. R., & Stone, D. (1979). Evolutionary change in the insulin receptors of hystricomorph rodents. Nature, 279(5712), 439-440. doi: 10.1038/279439a0

Howell, S. L., Tyhurst, M., Duvefelt, H., Andersson, A., & Hellerstrom, C. (1978). Role of zinc and calcium in the formation and storage of insulin in the pancreatic β- cell. Cell Tissue Res., 188(1), 107-118.

Hu, S. Q., Burke, G. T., & Katsoyannis, P. G. (1993). Contribution of the B16 and B26 tyrosine residues to the biological activity of insulin. J. Protein Chem., 12(6), 741-747.

Hua, Q. (2010). Insulin: a small protein with a long journey. Protein Cell, 1(6), 537-551. doi: 10.1007/s13238-010-0069-z

Hua, Q. X., Chu, Y. C., Jia, W., Phillips, N. B., Wang, R. Y., Katsoyannis, P. G., & Weiss, M. A. (2002). Mechanism of insulin chain combination. Asymmetric roles of A-chain α-helices in disulfide pairing. J. Biol. Chem., 277, 43443-43453.

459

Hua, Q. X., Gozani, S. N., Chance, R. E., Hoffmann, J. A., Frank, B. H., & Weiss, M. A. (1995). Structure of a protein in a kinetic trap. Nat. Struct. Biol., 2, 129-138.

Hua, Q. X., Hu, S. Q., Frank, B. H., Jia, W., Chu, Y. C., Wang, S. H., . . . Weiss, M. A. (1996). Mapping the functional surface of insulin by design: structure and function of a novel A-chain analogue. J. Mol. Biol., 264, 390-403.

Hua, Q. X., Hu, S. Q., Jia, W., Chu, Y. C., Burke, G. T., Wang, S. H., . . . Weiss, M. A. (1998). Mini-proinsulin and mini-IGF-I: homologous protein sequences encoding non-homologous structures. J. Mol. Biol., 277, 103-118.

Hua, Q. X., Jia, W., Frank, B. H., Phillips, N. B., & Weiss, M. A. (2002). A protein caught in a kinetic trap: structures and stabilities of insulin disulfide isomers. Biochemistry, 41, 14700-14715.

Hua, Q. X., Jia, W., & Weiss, M. A. (2011). Conformational dynamics of insulin. Front. Endocrin., 2(48). doi: 10.3389/fendo.2011.00048

Hua, Q. X., Mayer, J., Jia, W., Zhang, J., & Weiss, M. A. (2006). The folding nucleus of the insulin superfamily: a flexible peptide model foreshadows the native state. J. Biol. Chem., 281, 28131-28142.

Hua, Q. X., Nakagawa, S. H., Hu, S. Q., Jia, W., Wang, S., & Weiss, M. A. (2006). Toward the active conformation of insulin. Stereospecific modulation of a structural switch in the B chain. J. Biol. Chem., 281, 24900-24909.

Hua, Q. X., Nakagawa, S. H., Jia, W., Hu, S. Q., Chu, Y. C., Katsoyannis, P. G., & Weiss, M. A. (2001). Hierarchical protein folding: asymmetric unfolding of an insulin analogue lacking the A7-B7 interchain disulfide bridge. Biochemistry, 40, 12299-12311.

Hua, Q. X., Nakagawa, S. H., Jia, W., Huang, K., Phillips, N. B., Hu, S. Q., & Weiss, M. A. (2008). Design of an active ultrastable single-chain insulin analog: synthesis, structure, and therapeutic implications. J. Biol. Chem., 283, 14703-14716. doi: PMC2386938

Hua, Q. X., Nakagawa, S. H., Wilken, J., Ramos, R. R., Jia, W., Bass, J., & Weiss, M. A. (2003). A divergent INS protein in Caenorhabditis elegans structurally resembles human insulin and activates the human insulin receptor. Genes Dev., 17, 826-831. doi: PMC196032

Hua, Q. X., Narhi, L., Jia, W., Arakawa, T., Rosenfeld, R., Hawkins, N., . . . Weiss, M. A. (1996). Native and non-native structure in a protein-folding intermediate: spectroscopic studies of partially reduced IGF-I and an engineered alanine model. J. Mol. Biol., 259, 297-313.

460

Hua, Q. X., Shoelson, S. E., & Weiss, M. A. (1992). Nonlocal structural perturbations in a mutant human insulin: sequential resonance assignment and 13C-isotope-aided 2D-NMR studies of [PheB24→Gly]insulin with implications for receptor recognition. Biochemistry, 31, 11940-11951.

Hua, Q. X., & Weiss, M. A. (1990). Toward the solution structure of human insulin: sequential 2D 1H NMR assignment of a des-pentapeptide analogue and comparison with crystal structure. Biochemistry, 29, 10545-10555.

Huang, K., Dong, J., Phillips, N. B., Carey, P. R., & Weiss, M. A. (2005). Proinsulin is refractory to protein fibrillation. Topological protection of a precursor protein from cross-β assembly. J. Biol. Chem., 280, 42345-42355.

Huang, K., Maiti, N. C., Phillips, N. B., Carey, P. R., & Weiss, M. A. (2006). Structure- specific effects of protein topology on cross-β assembly: studies of insulin fibrillation. Biochemistry, 45, 10278-10293.

Huang, K., Xu, B., Hu, S. Q., Chu, Y. C., Hua, Q. X., Qu, Y., . . . Weiss, M. A. (2004). How insulin binds: the B-chain a-helix contacts the L1 β-helix of the insulin receptor. J. Mol. Biol., 341, 529-550.

Ikeda, Y., SUEHIRO, T., NAKAMURA, T., KUMON, Y., & HASHIMOTO, K. (2001). Clinical significance of the insulin resistance index as assessed by homeostasis model assessment. Endocrine journal, 48(1), 81-86.

Inouye, K., Watanabe, K., Morihara, K., Tochino, Y., Kanaya, T., Emura, J., & Sakakibara, S. (1979). Enzyme-assisted semisynthesis of human insulin. J. Am. Chem. Soc., 101, 751-752.

Inouye, K., Watanabe, K., Tochino, Y., Kobayashi, M., & Shigeta, Y. (1981). Semisynthesis and properties of some insulin analogs. Biopolymers, 20(9), 1845- 1858. doi: 10.1002/bip.1981.360200909

Insel, R. A., Deecher, D. C., & Brewer, J. (2012). Juvenile Diabetes Research Foundation: mission, strategy, and priorities. Diabetes, 61(1), 30-35.

Irsigler, K., Kritz, H., Hagmüller, G., Franetzki, M., Prestele, K., Thurow, H., & Geisen, K. (1981). Long-term continuous intraperitoneal insulin infusion with an implanted remote-controlled insulin infusion device. Diabetes, 30(12), 1072- 1075.

Islam, S. A., Karplus, M., & Weaver, D. L. (2002). Application of the diffusion–collision model to the folding of three-helix bundle proteins. Journal of molecular biology, 318(1), 199-215.

461

Ito, Y., & Imanishi, Y. (1994). Protein device for glucose-sensitive release of insulin: ACS Publications.

Izumi, T., Yokota-Hashimoto, H., Zhao, S., Wang, J., Halban, P. A., & Takeuchi, T. (2003). Dominant negative pathogenesis by mutant proinsulin in the Akita diabetic mouse. Diabetes, 52, 409-416.

Jacobs, P. G., El Youssef, J., Reddy, R., Resalat, N., Branigan, D., Condon, J., . . . Edwards, C. (2016). Randomized trial of a dual‐hormone artificial pancreas with dosing adjustment during exercise compared with no adjustment and sensor‐ augmented pump therapy. Diabetes, Obesity and Metabolism, 18(11), 1110-1119.

Jacoby, E., Hua, Q. X., Stern, A. S., Frank, B. H., & Weiss, M. A. (1996). Structure and 1 dynamics of a protein assembly. H-NMR studies of the 36 kDa R6 insulin hexamer. J. Mol. Biol., 258(1), 136-157. doi: 10.1006/jmbi.1996.0239

Janssen, F. W., & Ruelius, H. W. (1968). Carbohydrate oxidase, a novel enzyme from polyporus obtusus: II. Specificity and characterization of reaction products. Biochimica et Biophysica Acta (BBA)-Enzymology, 167(3), 501-510.

Jeandidier, N., Boivin, S., Sapin, R., Rosart-Ortega, F., Uring-Lambert, B., Reville, P., & Pinget, M. (1995). Immunogenicity of intraperitoneal insulin infusion using programmable implantable devices. Diabetologia, 38, 577-584.

Jekl, V., Hauptman, K., & Knotek, Z. (2011). Diseases in pet degus: a retrospective study in 300 animals. J. Small Anim. Prac., 52(2), 107–112.

Jimenez, J. L., Nettleton, E. J., Bouchard, M., Robinson, C. V., Dobson, C. M., & Saibil, H. R. (2002). The protofilament structure of insulin amyloid fibrils. Proc. Natl. Acad. Sci. U. S. A., 99, 9196-9201. doi: PMC123117

Jimenez, J. L., Tennent, G., Pepys, M., & Saibil, H. R. (2001). Structural diversity of ex vivio amyloid fibrils studied by cryo-electron microscopy. J. Mol. Biol., 10, 241- 247.

Jin, X., Zhang, X., Wu, Z., Teng, D., Zhang, X., Wang, Y., . . . Li, C. (2009). Amphiphilic random glycopolymer based on phenylboronic acid: synthesis, characterization, and potential as glucose-sensitive matrix. Biomacromolecules, 10(6), 1337-1345.

Jonassen, I., Havelund, S., Hoeg-Jensen, T., Steensgaard, D. B., Wahlund, P. O., & Ribel, U. (2012). Design of the novel protraction mechanism of insulin degludec, an ultra-long-acting basal insulin. Pharm. Res., 29(8), 2104-2114. doi: 10.1007/s11095-012-0739-z

462

Jonassen, I. B., Havelund, S., Ribel, U., Hoeg-Jensen, T., Steensgaard, D. B., Johansen, T., . . . Kurtzhals, P. (2010). Insulin degludec is a new generation ultra-long acting basal insulin with a unique mechanism of protraction based on multi-hexamer formation. Diabetes, 59, A375.

Joosten, R. P., Womack, T., Vriend, G., & Bricogne, G. (2009). Re-refinement from deposited X-ray data can deliver improved models for most PDB entries. Acta Crystallographica Section D: Biological Crystallography, 65(2), 176-185.

Jørgensen, A. M., Kristensen, S. M., Led, J. J., & Balschmidt, P. (1992). Three- dimensional solution structure of an insulin dimer. A study of the B9(Asp) mutant of human insulin using nuclear magnetic resonance, distance geometry and restrained molecular dynamics. J. Mol. Biol., 227, 1146-1163.

Kahn, S. (2003). The relative contributions of insulin resistance and beta-cell dysfunction to the pathophysiology of type 2 diabetes. Diabetologia, 46(1), 3-19.

Kahn, S. E., Cooper, M. E., & Del Prato, S. (2014). Pathophysiology and treatment of type 2 diabetes: perspectives on the past, present, and future. The Lancet, 383(9922), 1068-1083.

Kamerlin, S. C., Haranczyk, M., & Warshel, A. (2008). Progress in ab initio QM/MM free-energy simulations of electrostatic energies in proteins: Accelerated QM/MM studies of p K a, redox reactions and solvation free energies. The journal of physical chemistry B, 113(5), 1253-1272.

Kannan, N., & Vishveshwara, S. (2000). Aromatic clusters: a determinant of thermal stability of thermophilic proteins. Protein Eng., 13(11), 753-761. doi: 10.1093/protein/13.11.753

Karamanou, M., Protogerou, A., Tsoucalas, G., Androutsos, G., & Poulakou-Rebelakou, E. (2016). Milestones in the history of diabetes mellitus: The main contributors. World journal of diabetes, 7(1), 1.

Karges, B., Rosenbauer, J., Kapellen, T., Wagner, V. M., Schober, E., Karges, W., & Holl, R. W. (2014). Hemoglobin A1c Levels and risk of severe hypoglycemia in children and young adults with type 1 diabetes from Germany and Austria: a trend analysis in a cohort of 37,539 patients between 1995 and 2012. PLoS Med, 11(10), e1001742. doi: 10.1371/journal.pmed.1001742

Kasyhap, N., Pohl, R, Bruen, K, Leone, E, Steiner S. (2010). "Smart" Basal Insulin Formulation that Releases Insulin in Response to Blood Glucose Concentrations of Diabetic Swine. Paper presented at the 2010 Annual Meeting- Controlled Release Society, Portland, OR.

463

Katsoyannis, P. G., Trakatellis, A. C., Zalut, C., Johnson, S., Tometsko, A., Schwartz, G., & Ginos, J. (1967). Studies on the synthesis of insulin from natural and synthetic A and B chains. 3. Synthetic insulins. Biochemistry, 6, 2656-2668.

Keller, D., Clausen, R., Josefsen, K., & Led, J. J. (2001). Flexibility and bioactivity of insulin: an NMR investigation of the solution structure and folding of an unusually flexible human insulin mutant with increased biological activity. Biochemistry, 40, 10732-10740.

Kerl, M. E. (2001). Diabetic ketoacidosis: pathophysiology and clinical and laboratory presentation. Compendium, 23(3), 220-228.

Khoo, C., Yang, J., Rajpal, G., Wang, Y., Liu, J., Arvan, P., & Stoffers, D. A. (2011). Endoplasmic reticulum oxidoreductin-1-like β (ERO1lβ) regulates susceptibility to endoplasmic reticulum stress and is induced by insulin flux in β-cells. Endocrinology, 152(7), 2599-2608.

Kikuchi, A., Suzuki, K., Okabayashi, O., Hoshino, H., Kataoka, K., Sakurai, Y., & Okano, T. (1996). Glucose-sensing electrode coated with polymer complex gel containing phenylboronic acid. Analytical chemistry, 68(5), 823-828.

Kilpatrick, E. S., Rigby, A. S., & Atkin, S. L. (2009). Effect of glucose variability on the long-term risk of microvascular complications in type 1 diabetes. Diabetes Care, 32(10), 1901-1903. doi: 10.2337/dc09-0109

Kim, P. S., & Baldwin, R. L. (1982). Specific intermediates in the folding reactions of small proteins and the mechanism of protein folding. Annu. Rev. Biochem., 51, 459-489.

Kim, S.-E., Harker, E. C., De Leon, A. C., Advincula, R. C., & Pokorski, J. K. (2015). Coextruded, Aligned, and Gradient-Modified Poly (ε-caprolactone) Fibers as Platforms for Neural Growth. Biomacromolecules, 16(3), 860-867.

Kim, S.-E., Wang, J., Jordan, A. M., Korley, L. T., Baer, E., & Pokorski, J. K. (2014). Surface Modification of Melt Extruded Poly (ε-caprolactone) Nanofibers: Toward a New Scalable Biomaterial Scaffold. ACS macro letters, 3(6), 585-589.

Kitano, S., Koyama, Y., Kataoka, K., Okano, T., & Sakurai, Y. (1992). A novel drug delivery system utilizing a glucose responsive polymer complex between poly (vinyl alcohol) and poly (N-vinyl-2-pyrrolidone) with a phenylboronic acid moiety. Journal of controlled release, 19(1-3), 161-170.

Knudsen, L., De Meyts, P., & Kiselyov, V. V. (2011). Insight into the molecular basis for the kinetic differences between the two insulin receptor isoforms. Biochem. J., 440(3), 397-403. doi: 10.1042/bj20110550

464

Knudsen, L., Hansen, B. F., Jensen, P., Pedersen, T. A., Vestergaard, K., Schaffer, L., . . . De Meyts, P. (2012). Agonism and antagonism at the insulin receptor. PLoS One, 7(12), e51972. doi: 10.1371/journal.pone.0051972 PONE-D-11-18081 [pii]

Kobayashi, M., Ohgaku, S., Iwasaki, M., Maegawa, H., Shigeta, Y., & Inouye, K. (1982). Supernormal insulin: [D-PheB24]-insulin with increased affinity for insulin receptors. Biochem. Biophys. Res. Commun., 107, 329-336.

Kobayashi, M., Sasaoka, T., Sugibayashi, M., Iwanishi, M., & Shigeta, Y. (1989). Receptor binding and biologic activity of biosynthetic human insulin and miniproinsulin produced by recombinant gene technology. Diabetes Res. Clin. Pract., 7, 25-28.

Koga, H., Kaushik, S., & Cuervo, A. M. (2011). Protein homeostasis and aging: The importance of exquisite quality control. Ageing Res. Rev., 10(2), 205-215. doi: 10.1016/j.arr.2010.02.001

Kolb, M., Bonniaud, P., Galt, T., Sime, P. J., Kelly, M. M., Margetts, P. J., & Gauldie, J. (2002). Differences in the fibrogenic response after transfer of active transforming growth factor-β 1 gene to lungs of “fibrosis-prone” and “fibrosis-resistant” mouse strains. American journal of respiratory cell and molecular biology, 27(2), 141- 150.

Kolb, P., Rosenbaum, D. M., Irwin, J. J., Fung, J. J., Kobilka, B. K., & Shoichet, B. K. (2009). Structure-based discovery of β2-adrenergic receptor ligands. Proc. Natl. Acad. Sci. U. S. A., 106(16), 6843-6848. doi: 10.1073/pnas.0812657106

Kramer, C., Gedeck, P., & Meuwly, M. (2012). Atomic multipoles: Electrostatic potential fit, local reference axis systems, and conformational dependence. Journal of computational chemistry, 33(20), 1673-1688.

Kristensen, C., Kjeldsen, T., Wiberg, F. C., Schaffer, L., Hach, M., Havelund, S., . . . Andersen, A. S. (1997). Alanine scanning mutagenesis of insulin. J. Biol. Chem., 272, 12978-12983.

Kurtzhals, P., Havelund, S., Jonassen, I., Kiehr, B., Larsen, U., Ribel, U., & Markussen, J. (1995). Albumin binding of insulins acylated with fatty acids: characterization of the ligand-protein interaction and correlation between binding affinity and timing of the insulin effect in vivo. Biochemical Journal, 312(3), 725-731.

Kurtzhals, P., Schaffer, L., Sorensen, A., Kristensen, C., Jonassen, I., Schmid, C., & Trub, T. (2000). Correlations of receptor binding and metabolic and mitogenic potencies of insulin analogs designed for clinical use. Diabetes, 49(6), 999-1005.

Lachin, J. M., Orchard, T. J., Nathan, D. M., & Group, D. E. R. (2014). Update on cardiovascular outcomes at 30 years of the diabetes control and complications

465

trial/epidemiology of diabetes interventions and complications study. Diabetes Care, 37(1), 39-43.

Ladenstein, R., & Antranikian, G. (1998). Proteins from hyperthermophiles: stability and enzymatic catalysis close to the boiling point of water. Adv Biochem Eng Biotechnol, 61, 37-85.

Lambers, H., Piessens, S., Bloem, A., Pronk, H., & Finkel, P. (2006). Natural skin surface pH is on average below 5, which is beneficial for its resident flora. International journal of cosmetic science, 28(5), 359-370.

Lauritzen, T., Pramming, S., Gale, E., Deckert, T., & Binder, C. (1982). Absorption of isophane (NPH) insulin and its clinical implications. BMJ, 285(6336), 159-162.

Lawrence, M. C., McKern, N. M., & Ward, C. W. (2007). Insulin receptor structure and its implications for the IGF-1 receptor. Curr. Opin. Struct. Biol., 17(6), 699-705.

Leathem, A., & Brooks, S. (1987). Enzyme binding to detect carbohydrate expression in tissue sections. The Histochemical Journal, 19(8), 405-412.

Lee, H., & Bequette, B. W. (2009). A closed-loop artificial pancreas based on model predictive control: Human-friendly identification and automatic meal disturbance rejection. Biomedical Signal Processing and Control, 4(4), 347-354.

Lee, H., Buckingham, B. A., Wilson, D. M., & Bequette, B. W. (2009). A closed-loop artificial pancreas using model predictive control and a sliding meal size estimator. Journal of diabetes science and technology, 3(5), 1082-1090.

Lee, L. C., Chou, Y. L., Chen, H. H., Lee, Y. L., & Shaw, J. F. (2009). Functional role of a non-active site residue Trp(23) on the enzyme activity of Escherichia coli thioesterase I/protease I/lysophospholipase L(1). Biochim Biophys Acta, 1794(10), 1467-1473. doi: 10.1016/j.bbapap.2009.06.008

Lee, P., Towslee, J., Maia, J., & Pokorski, J. (2015). PEGylation to Improve Protein Stability During Melt Processing. Macromolecular bioscience.

Lee, Y. M., Kim, S. H., & Cho, C. S. (1996). Synthesis and swelling characteristics of pH and thermoresponsive interpenetrating polymer network hydrogel composed of poly (vinyl alcohol) and poly (acrylic acid). Journal of applied polymer science, 62(2), 301-311.

Leibush, B., Lappova, Y., Bondareva, V., Chistyacova, O., Gutierrez, J., & Plisetskaya, E. (1998). Insulin-family peptide–receptor interaction at the early stage of vertebrate evolution. Comparative Biochemistry and Physiology Part B: Biochemistry and Molecular Biology, 121(1), 57-63.

466

Lemaire, K., Ravier, M. A., Schraenen, A., Creemers, J. W., Van de Plas, R., Granvik, M., . . . Schuit, F. C. (2009). Insulin crystallization depends on zinc transporter ZnT8 expression, but is not required for normal glucose homeostasis in mice. Proc. Natl. Acad. Sci. U. S. A., 106(35), 14872-14877. doi: 10.1073/pnas.0906587106

Li, J., Yang, L., Fan, X., Zhang, J., Wang, F., & Wang, Z. (2017). Temperature and Glucose Dual-Responsive Carriers Bearing Poly (N-Isopropylacrylamide) and Phenylboronic Acid for Insulin Controlled Release: A Review. International Journal of Polymeric Materials and Polymeric Biomaterials(just-accepted).

Li, X., Fu, M., Wu, J., Zhang, C., Deng, X., Dhinakar, A., . . . Ge, L. (2017). pH-sensitive peptide hydrogel for glucose-responsive insulin delivery. Acta biomaterialia.

Lieblich, S. A., Fang, K. Y., Cahn, J. K., Rawson, J., LeBon, J., Ku, H. T., & Tirrell, D. A. (2017). 4 S-Hydroxylation of Insulin at ProB28 Accelerates Hexamer Dissociation and Delays Fibrillation. Journal of the American Chemical Society, 139(25), 8384-8387.

Lim, F., & Sun, A. M. (1980). Microencapsulated islets as bioartificial endocrine pancreas. Science, 210(4472), 908-910.

Lipson, K. L., Fonseca, S. G., Ishigaki, S., Nguyen, L. X., Foss, E., Bortell, R., . . . Urano, F. (2006). Regulation of insulin biosynthesis in pancreatic β cells by an endoplasmic reticulum- resident protein kinase IRE1. Cell Metab., 4, 245-254.

Liu, M., Haataja, L., Wright, J., Wickramasinghe, D. N., Hua, Q. X., Phillips, N. B., . . . Arvan, P. (2010). Mutant INS-gene induced diabetes of youth: proinsulin cysteine residues impose dominant-negative inhibition on nonmutant proinsulin transport. PLos-One, 5, e13333. doi: PMC2952628

Liu, M., Hodish, I., Haataja, L., Lara-Lemus, R., Rajpal, G., Wright, J., & Arvan, P. (2010). Proinsulin misfolding and diabetes: mutant INS gene-induced diabetes of youth. Trends Endocrinol. Metab., 21(11), 652-659. doi: 10.1016/j.tem.2010.07.001

Liu, M., Hodish, I., Rhodes, C. J., & Arvan, P. (2007). Proinsulin maturation, misfolding, and proteotoxicity. Proc. Natl. Acad. Sci. U. S. A., 104, 15841-15846. doi: PMC2000385

Liu, M., Hua, Q.-X., Hu, S.-Q., Jia, W., Yang, Y., Saith, S. E., . . . Weiss, M. A. (2010). Deciphering the hidden informational content of protein sequences: foldability of proinsulin hinges on a flexible arm that is dispensable in the mature hormone. J. Biol. Chem., 285, 30989-31001.

467

Liu, M., Li, Y., Cavener, D., & Arvan, P. (2005). Proinsulin disulfide maturation and misfolding in the endoplasmic reticulum. J. Biol. Chem., 280, 13209-13212. doi: PMC2527538

Liu, M., Ramos-Castaneda, J., & Arvan, P. (2003). Role of the connecting peptide in insulin biosynthesis. J. Biol. Chem., 278, 14798-14805.

Liu, M., Sun, J., Cui, J., Chen, W., Guo, H., Barbetti, F., & Arvan, P. (2015). INS-gene mutations: From genetics and beta cell biology to clinical disease. Molecular aspects of medicine, 42, 3-18.

Liu, M., Wan, Z. L., Chu, Y. C., Aladdin, H., Klaproth, B., Choquette, M. H., . . . Weiss, M. A. (2009). Crystal structure of a "non-foldable" insulin: Impaired folding efficiency and ER stress despite native activity. J. Biol. Chem., 284, 35259-35272. doi: PMC2787385

Longo, N., Wang, Y., Smith, S. A., Langley, S. D., DiMeglio, L. A., & Giannella-Neto, D. (2002). Genotype–phenotype correlation in inherited severe insulin resistance. Human molecular genetics, 11(12), 1465-1475.

Lopes, J. R., Santos, G., Barata, P., Oliveira, R., & Lopes, C. M. (2013). Physical and chemical stimuli-responsive drug delivery systems: targeted delivery and main routes of administration. Current pharmaceutical design, 19(41), 7169-7184.

Lougheed, W. D., Zinman, B., Strack, T. R., Janis, L. J., Weymouth, A. B., Bernstein, E. A., . . . Frank, B. H. (1997). Stability of insulin lispro in insulin infusion systems. Diabetes Care, 20, 1061-1065.

Luisier, S., Avital-Shmilovici, M., Weiss, M. A., & Kent, S. B. (2010). Total chemical synthesis of human proinsulin. Chem. Commun. (Camb.), 46(43), 8177-8179. doi: 10.1039/c0cc03141k

Lum, Z.-P., Tai, I. T., Krestow, M., Norton, J., Vacek, I., & Sun, A. M. (1991). Prolonged reversal of diabetic state in NOD mice by xenografts of microencapsulated rat islets. Diabetes, 40(11), 1511-1516.

Lv, D., Kulkarni, S. D., Chan, A., Keith, S., Pettis, R., Kovatchev, B. P., . . . Breton, M. D. (2015). Pharmacokinetic Model of the Transport of Fast-Acting Insulin From the Subcutaneous and Intradermal Spaces to Blood. J Diabetes Sci Technol, 9(4), 831-840. doi: 10.1177/1932296815573864

Ly, T. T., Roy, A., Grosman, B., Shin, J., Campbell, A., Monirabbasi, S., . . . Clinton, P. (2015). Day and night closed-loop control using the integrated Medtronic hybrid closed-loop system in type 1 diabetes at diabetes camp. Diabetes Care, 38(7), 1205-1211.

468

MacKerell, A. D., Bashford, D., Bellott, M., Dunbrack, R. L., Evanseck, J. D., Field, M. J., . . . Karplus, M. (1998). All-atom empirical potential for molecular modeling and dynamics studies of proteins. J. Phys. Chem. B, 102(18), 3586-3616. doi: 10.1021/jp973084f

MacKerell, A. D., Brooks, B., Brooks, C. L., Nilsson, L., Roux, B., Won, Y., & Karplus, M. (2002). CHARMM: The Energy Function and Its Parameterization. In P. v. R. C. Schleyer, N. L. A. T. Gasteiger, J. Kollman, P. A. Schaefer, H. F., III Schreiner, P. R. S. (Ed.), Encyclopedia of Computational Chemistry (Vol. 1, pp. 271-277). Chichester: John Wiley & Sons, Ltd.

Makadia, H. K., & Siegel, S. J. (2011). Poly lactic-co-glycolic acid (PLGA) as biodegradable controlled drug delivery carrier. Polymers, 3(3), 1377-1397.

Margolis, K. L., O’Connor, P. J., Morgan, T. M., Buse, J. B., Cohen, R. M., Cushman, W. C., . . . Grimm, R. H. (2014). Outcomes of combined cardiovascular risk factor management strategies in type 2 diabetes: the ACCORD randomized trial. Diabetes Care, 37(6), 1721-1728.

Markussen, J., Diers, I., Engesgaard, A., Hansen, M. T., Hougaard, P., Langkjaer, L., . . . et al. (1987). Soluble, prolonged-acting insulin derivatives. II. Degree of protraction and crystallizability of insulins substituted in positions A17, B8, B13, B27 and B30. Protein Eng., 1(3), 215-223.

Markussen, J., Diers, I., Hougaard, P., Langkjaer, L., Norris, K., Snel, L., . . . Voight, H. O. (1988). Soluble, prolonged-acting insulin derivatives. III. Degree of protraction, crystallizability and chemical stability of insulins substituted in positions A21, B13, B23, B27 and B30. Protein Eng., 2, 157-166.

Markussen, J., Havelund, S., Kurtzhals, P., Andersen, A., Halstrøm, J., Hasselager, E., . . . Vad, K. (1996). Soluble, fatty acid acylated insulins bind to albumin and show protracted action in pigs. Diabetologia, 39(3), 281-288.

Markussen, J., Jonassen, I., Havelund, S., Brandt, J., Kurtzhals, P., Hansen, P. H., & Kaarsholm, N. C. (2003). U. S. A. Patent No. 6,620,780.

Markussen, J. M., Hougaard, P., Ribel, U., Sørensen, A. R., & Sørensen, E. (1987). Soluble, prolonged-acting insulin derivatives. I. Degree of protraction and crystallizability of insulins substituted in the termini of the B-chain. Protein Eng., 1(3), 205-213.

Marshall, G. R. (2013). Limiting assumptions in molecular modeling: electrostatics. Journal of computer-aided molecular design, 27(2), 107-114.

469

Matsumoto, A., Ishii, T., Nishida, J., Matsumoto, H., Kataoka, K., & Miyahara, Y. (2012). A Synthetic Approach Toward a Self‐Regulated Insulin Delivery System. Angewandte Chemie, 124(9), 2166-2170.

Matsumoto, A., Yoshida, R., & Kataoka, K. (2004). Glucose-responsive polymer gel bearing phenylborate derivative as a glucose-sensing moiety operating at the physiological pH. Biomacromolecules, 5(3), 1038-1045.

Mayer, J. P., Zhang, F., & DiMarchi, R. D. (2007). Insulin structure and function. Biopolymers, 88, 687-713.

Mayer, M., Kies, U., Kammermeier, R., & Buchner, J. (2000). BiP and PDI cooperate in the oxidative folding of antibodies in vitro. J. Biol. Chem., 275, 29421-29425.

Menting, J. G., Whittaker, J., Margetts, M. B., Whittaker, L. J., Kong, G. K., Smith, B. J., . . . Lawrence, M. C. (2013). How insulin engages its primary binding site on the insulin receptor. Nature, 493(7431), 241-245. doi: 10.1038/nature11781

Menting, J. G., Yang, Y., Chan, S. J., Phillips, N. B., Smith, B. J., Whittaker, J., . . . Lawrence, M. C. (2014). A structural hinge in insulin enables its receptor engagement. Proc. Natl. Acad. Sci. USA, 111(33), 1-48.

Michael, J., Carroll, R., Swift, H. H., & Steiner, D. F. (1987). Studies on the molecular organization of rat insulin secretory granules. J. Biol. Chem., 262, 16531-16535.

Michael, M. D., Kiulkarni, R. N., Postic, C., Previs, S. F., Shulman, G. I., & et. al. (2000). Loss of insulin signaling in hepatocytes leads to severe insulin resistance and progressive hepatic dysfunction. Mol. Cell, 6, 87-97.

Michael, M. D., Kulkarni, R. N., Postic, C., Previs, S. F., Shulman, G. I., Magnuson, M. A., & Kahn, C. R. (2000). Loss of insulin signaling in hepatocytes leads to severe insulin resistance and progressive hepatic dysfunction. Molecular cell, 6(1), 87- 97.

Miller, S., Nimri, R., Atlas, E., Grunberg, E. A., & Phillip, M. (2011). Automatic learning algorithm for the MD-logic artificial pancreas system. Diabetes technology & therapeutics, 13(10), 983-990.

Mirmira, R. G., Nakagawa, S. H., & Tager, H. S. (1991). Importance of the character and configuration of residues B24, B25, and B26 in insulin-receptor interactions. J. Biol. Chem., 266(3), 1428-1436.

Mirmira, R. G., & Tager, H. S. (1989). Role of the phenylalanine B24 side chain in directing insulin interaction with its receptor: Importance of main chain conformation. J. Biol. Chem., 264, 6349-6354.

470

Mirmira, R. G., & Tager, H. S. (1991). Disposition of the phenylalanine B25 side chain during insulin-receptor and insulin-insulin interactions. Biochemistry, 30, 8222- 8229.

Mitraki, A., Danner, M., King, J., & Seckler, R. (1993). Temperature-sensitive mutations and second-site suppressor substitutions affect folding of the P22 tailspike protein in vitro. J. Biol. Chem., 268, 20071-20075.

Monsellier, E., & Chiti, F. (2007). Prevention of amyloid-like aggregation as a driving force of protein evolution. EMBO Rep., 8(8), 737-742. doi: 10.1038/sj.embor.7401034

Moore, M. C., Smith, M. S., Sinha, V. P., Beals, J. M., Michael, M. D., Jacober, S. J., & Cherrington, A. D. (2014). Novel PEGylated basal insulin LY2605541 has a preferential hepatic effect on glucose metabolism. Diabetes, 63(2), 494-504.

Mudaliar, S. R., Lindberg, F. A., Joyce, M., Beerdsen, P., Strange, P., Lin, A., & Henry, R. R. (1999). Insulin Aspart (B28 Asp-Insulin): A Fast-Acting Analog of Human Insulin. Diabetes Care, 22, 1501-1506.

Mukherjee, M., Jana, J., & Chatterjee, S. (2018). A Small Molecule Impedes Insulin Fibrillation: Another New Role of Phenothiazine Derivatives. ChemistryOpen, 7(1), 68-79.

Murphy, R. M. (2002). Peptide aggregation in neurodegenerative disease. Annu. Rev. Biomed. Eng., 4, 155-174.

Myszka, D. G. (1997). Kinetic analysis of macromolecular interactions using surface plasmon resonance biosensors. Current opinion in biotechnology, 8(1), 50-57.

Naik, R. G., Brooks-Worrell, B. M., & Palmer, J. P. (2009). Latent autoimmune diabetes in adults. The Journal of Clinical Endocrinology & Metabolism, 94(12), 4635- 4644.

Nakagawa, S. H., Hua, Q. X., Hu, S. Q., Jia, W., Wang, S., Katsoyannis, P. G., & Weiss, M. A. (2006). Chiral mutagenesis of insulin. Contribution of the B20-B23 β-turn to activity and stability. J. Biol. Chem., 281, 22386-22396.

Nakagawa, S. H., & Tager, H. S. (1986). Role of the phenylalanine B25 side chain in directing insulin interaction with its receptor. Steric and conformational effects. J. Biol. Chem., 261, 7332-7341.

Nakagawa, S. H., & Tager, H. S. (1987). Role of the COOH-terminal B-chain domain in insulin-receptor interactions. Identification of perturbations involving the insulin mainchain. J. Biol. Chem., 262(25), 12054-12058.

471

Nakagawa, S. H., Tager, H. S., & Steiner, D. F. (2000). Mutational analysis of invariant ValB12 in insulin: implications for receptor binding. Biochemistry, 39, 15826- 15835.

Nakagawa, S. H., Zhao, M., Hua, Q. X., & Weiss, M. A. (1998). The importance of residue B8 in insulin activity, structure and folding. Paper presented at the Fifteenth American Peptide Symposium-Peptides: Frontiers of Peptide Science, Nashville, Tennessee.

Nanjo, K., Sanke, T., Kondo, M., Nishimura, S., Miyano, M., Linuma, J., . . . et. al. (1986). Mutant insulin syndrome: identification of two families with LeuA3- insulin and determination of its biological activity. Trans. Assoc. Am. Physicians, 99, 132-142.

Nauck, M., Stöckmann, F., Ebert, R., & Creutzfeldt, W. (1986). Reduced incretin effect in type 2 (non-insulin-dependent) diabetes. Diabetologia, 29(1), 46-52.

Neumoin, A., Mares, J., Lerch-Bader, M., Bader, R., & Zerbe, O. (2007). Probing the formation of stable tertiary structure in a model miniprotein at atomic resolution: determinants of stability of a helical hairpin. J. Am. Chem. Soc., 129(28), 8811- 8817. doi: 10.1021/ja0716960

Ngo, T. T., & Lenhoff, H. M. (1983). Amperometric assay for collagenase. Applied biochemistry and biotechnology, 8(5), 407-414.

Nicholls, M. P., & Paul, P. K. (2004). Structures of carbohydrate–boronic acid complexes determined by NMR and molecular modelling in aqueous alkaline media. Organic & biomolecular chemistry, 2(10), 1434-1441.

NIDDK. (2011). Advances and emerging opportunities in diabetes research: a Strategic Planning report of the DMICC. (NIH Publication No. 11-7572). Washington, D.C.

Nishi, M., & Steiner, D. F. (1990). Cloning of complementary DNAs encoding islet amyloid polypeptide, insulin, and glucagon precursors from a New World rodent, the degu, Octodon degus. Mol. Endocrinol., 4, 1192-1198.

Noble, J. A., & Erlich, H. A. (2012). Genetics of type 1 diabetes. Cold Spring Harbor perspectives in medicine, 2(1), a007732.

Nourse, A., & Jeffrey, P. D. (1998). A sedimentation equilibrium study of platypus insulin: the HB10D mutant does not associate beyond dimer. Biophys. Chem., 71(1), 21-34.

472

Nyambura, B. K., Kellaway, I. W., & Taylor, K. M. (2009). Insulin nanoparticles: stability and aerosolization from pressurized metered dose inhalers. International journal of pharmaceutics, 375(1-2), 114-122.

O’Halloran, T. V., Kebede, M., Philips, S. J., & Attie, A. D. (2013). Zinc, insulin, and the liver: a ménage à trois. The Journal of clinical investigation, 123(10), 4136-4139.

Okun, M. M., & Shields, D. (1992). Translocation of preproinsulin across the endoplasmic reticulum membrane. The relationship between nascent polypeptide size and extent of signal recognition particle-mediated inhibition of protein synthesis. J. Biol. Chem., 267(16), 11476-11482.

Olsen, H. B., Ludvigsen, S., & Kaarsholm, N. C. (1996). Solution structure of an engineered insulin monomer at neutral pH. Biochemistry, 35, 8836-8845.

Owens, D. R. (2011). Insulin preparations with prolonged effect. Diabetes Technol. Ther., 13 Suppl 1, S5-14. doi: 10.1089/dia.2011.0068

Owens, D. R., & Bolli, G. B. (2008). Beyond the era of NPH insulin--long-acting insulin analogs: chemistry, comparative pharmacology, and clinical application. Diabetes Technol. Ther., 10(5), 333-349. doi: 10.1089/dia.2008.0023

Pace, C. N., & Shaw, K. L. (2000). Linear extrapolation method of analyzing solvent denaturation curves. Proteins, Suppl 4, 1-7.

Pandyarajan, V., Phillips, N. B., Cox, G. P., Yang, Y., Whittaker, J., Ismail-Beigi, F., & Weiss, M. A. (2014). Biophysical Optimization of a Therapeutic Protein by Non- Standard Mutagenesis. Studies of an Iodo-Insulin Derivative. J. Biol. Chem., 289(34), 23367-23381.

Pandyarajan, V., Phillips, N. B., Rege, N. K., Lawrence, M. C., Whittaker, J., & Weiss, M. A. (2016). Contribution of TyrB26 to the Function and Stability of Insulin. Structure-activity relationships at a conserved hormone-receptor interface. J. Biol. Chem., 291(25), 12978-12990.

Pandyarajan, V., Smith, B. J., Phillips, N. B., Whittaker, L., Cox, G. P., Wickramasinghe, N., . . . Ismail-Beigi, F. (2014a). Aromatic Anchor at an Invariant Hormone- Receptor Interface Function of Insulin Residue B24 with Aplication to Protein Design. J. Biol. Chem., 289(50), 34709-34727.

Pandyarajan, V., Smith, B. J., Phillips, N. B., Whittaker, L., Cox, G. P., Wickramasinghe, N., . . . Ismail-Beigi, F. (2014b). Aromatic anchor at an invariant hormone- receptor interface function of insulin residue B24 with application to protein design. Journal of Biological Chemistry, 289(50), 34709-34727.

473

Pandyarajan, V., & Weiss, M. A. (2012). Design of non-standard insulin analogs for the treatment of diabetes mellitus. Curr. Diab. Rep., 12, 697-704. doi: 10.1007/s11892-012-0318-z

Papaioannou, A., Kuyucak, S., & Kuncic, Z. (2015). Molecular Dynamics Simulations of Insulin: Elucidating the Conformational Changes that Enable Its Binding. PloS One, 10(12), e0144058.

Patel, A., MacMahon, S., Chalmers, J., Neal, B., Billot, L., Woodward, M., . . . Travert, F. (2008). Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N. Engl. J. Med., 358(24), 2560-2572. doi: 10.1056/NEJMoa0802987

Paulová, M., Tichá, M., Entlicher, G., Koštíř, J., & Kocourek, J. (1970). Relationship between red blood cell agglutination and polysaccharide precipitation by phytohemagglutinin of Pisum sativum L. FEBS letters, 9(6), 345-347.

Pedersen, J. S., & Otzen, D. E. (2008). Amyloid a state in many guises: survival of the fittest fibril fold. Protein Sci., 17, 1-9.

Petsko, G. A. (2000). Enzyme evolution. Design by necessity. Nature, 403(6770), 606- 607. doi: 10.1038/35001176

Petznick, A. (2011). Insulin management of type 2 diabetes mellitus. Am Fam Physician, 84(2), 183-190.

Pinsker, J. E., Lee, J. B., Dassau, E., Seborg, D. E., Bradley, P. K., Gondhalekar, R., . . . Doyle, F. J. (2016). Randomized crossover comparison of personalized MPC and PID control algorithms for the artificial pancreas. Diabetes Care, 39(7), 1135- 1142.

Podual, K., Doyle, F., & Peppas, N. A. (2004). Modeling of water transport in and release from glucose-sensitive swelling-controlled release systems based on poly (diethylaminoethyl methacrylate-g-ethylene glycol). Industrial & engineering chemistry research, 43(23), 7500-7512.

Prevost, M., Wodak, S. J., Tidor, B., & Karplus, M. (1991). Contribution of the hydrophobic effect to protein stability: analysis based on simulations of the Ile- 96----Ala mutation in barnase. Proceedings of the National Academy of Sciences, 88(23), 10880-10884.

Qiao, Z. S., Guo, Z. Y., & Feng, Y. M. (2001). Putative disulfide-forming pathway of porcine insulin precursor during its refolding in vitro. Biochemistry, 40, 2662- 2668.

474

Qiao, Z. S., Guo, Z. Y., & Feng, Y. M. (2006). In vitro folding/unfolding of insulin/single-chain insulin. Protein Pept. Lett., 13, 423-429.

Qiao, Z. S., Min, C. Y., Hua, Q. X., Weiss, M. A., & Feng, Y. M. (2003). In vitro refolding of human proinsulin. Kinetic intermediates, putative disulfide-forming pathway, folding initiation site, and protential role of C-peptide in folding process. J. Biol. Chem., 278, 17800-17809.

Rabkin, R., Ryan, M., & Duckworth, W. (1984). The renal metabolism of insulin. Diabetologia, 27(3), 351-357.

Racca, J. D., Chen, Y.-S., Maloy, J. D., Wickramasinghe, N., Phillips, N. B., & Weiss, M. A. (2014). Structure-Function Relationships in Human Testis-determining Factor SRY AN AROMATIC BUTTRESS UNDERLIES THE SPECIFIC DNA- BENDING SURFACE OF A HIGH MOBILITY GROUP (HMG) BOX. Journal of Biological Chemistry, 289(47), 32410-32429.

Rahuel-Clermont, S., French, C. A., Kaarsholm, N. C., & Dunn, M. F. (1997). Mechanisms of stabilization of the insulin hexamer through allosteric ligand interactions. Biochemistry, 36(19), 5837-5845.

Raile, K., O'Connell, M., Galler, A., Werther, G., Kuhnen, P., Krude, H., & Blankenstein, O. (2011). Diabetes caused by insulin gene (INS) deletion: clinical characteristics of homozygous and heterozygous individuals. Eur. J. Endocrinol., 165(2), 255- 260. doi: 10.1530/eje-11-0208

Rajpal, G., Schuiki, I., Liu, M., Volchuk, A., & Arvan, P. (2012). Action of protein disulfide isomerase on proinsulin exit from endoplasmic reticulum of pancreaticβ- cells. J. Biol. Chem., 287(1), 43-47. doi: 10.1074/jbc.C111.279927

Ramesh, V., & Bradbury, J. H. (1986). 1H-NMR studies of insulin. Reversible transformation of 2Zn→4Zn insulin hexamer. Int. J. Pept. Protein Res., 28(2), 146-153.

Ratnaparkhi, G. S., & Varadarajan, R. (2000). Thermodynamic and structural studies of cavity formation in proteins suggest that loss of packing interactions rather than the hydrophobic effect dominates the observed energetics. Biochemistry, 39, 12365-12374.

Ravaine, V., Ancla, C., & Catargi, B. (2008). Chemically controlled closed-loop insulin delivery. J. Control Release, 132(1), 2-11. doi: 10.1016/j.jconrel.2008.08.009

Řeha, D., Kabelác, M., Ryjácek, F., Šponer, J., Šponer, J. E., Elstner, M., . . . Hobza, P. (2002). Intercalators. 1. Nature of stacking interactions between intercalators (ethidium, daunomycin, ellipticine, and 4 ‘, 6-diaminide-2-phenylindole) and DNA base pairs. Ab initio quantum chemical, density functional theory, and

475

empirical potential study. Journal of the American Chemical Society, 124(13), 3366-3376.

Renard, E. (2002). Implantable closed loop glucose sensing and insulin delivery: the future for insulin pump therapy. Curr. Opin. Pharmacol., 2, 708-716.

Renard, E., Boutleau, S., Jacques-Apostol, D., lauton, D., Gibert-Boulet, F., Costalat, G., . . . Jaffiol, C. (1996). Insulin underdelivery from implanted pumps using peritoneal route. Determinant role of insulin pump compatibility. Diabetes Care, 19, 812-817.

Renard, E., Costalat, G., & Bringer, J. (2002). [From external to implantable insulin pump, can we close the loop?]. Diabetes Metab., 28, 2S19-12S25.

Ribel, U., Hougaard, P., Drejer, K., & Sorensen, A. R. (1990). Equivalent in vivo biological activity of insulin analogues and human insulin despite different in vitro potencies. Diabetes, 39, 1033-1039.

Ribet, F., Stemme, G., & Roxhed, N. (2017). Ultra-miniaturization of a planar amperometric sensor targeting continuous intradermal glucose monitoring. Biosensors and Bioelectronics, 90, 577-583.

Riemen, M. W., Pon, L. A., & Carpenter, F. H. (1983). Preparation of semisynthetic insulin analogues from bis(tert-butyloxycarbonyl)-desoctapeptide-insulin phenylhydrazide: importance of the aromatic region B24-B26. Biochemistry, 22(6), 1507-1515.

Robertson, R. P. (2004). Islet transplantation as a treatment for diabetes—a work in progress. New England Journal of Medicine, 350(7), 694-705.

Rorsman, P., & Renström, E. (2003). Insulin granule dynamics in pancreatic beta cells. Diabetologia, 46(8), 1029-1045.

Roy, M., Brader, M. L., Lee, R. W., Kaarsholm, N. C., Hansen, J. F., & Dunn, M. F. (1989). Spectroscopic signatures of the T to R conformational transition in the insulin hexamer. J. Biol. Chem., 264(32), 19081-19085.

Salsali, A., & Nathan, M. (2006). A review of types 1 and 2 diabetes mellitus and their treatment with insulin. Am. J. Ther., 13(4), 349-361.

Saltiel, A. R., & Pessin, J. E. (2002). Insulin signaling pathways in time and space. Trends Cell Biol., 12, 65-71.

Samanta, U., Pal, D., & Chakrabarti, P. (1999). Packing of aromatic rings against tryptophan residues in proteins. Acta Crystallographica Section D: Biological Crystallography, 55(8), 1421-1427.

476

Sanger, F. (1959). Chemistry of insulin; determination of the structure of insulin opens the way to greater understanding of life processes. Science, 129(3359), 1340- 1344.

Sanger, F., & Thompson, E. O. (1952). The amino-acid sequence in the glycyl chain of insulin. Biochem. J., 52, iii.

Saudek, C. D., Duckworth, W. C., Giobbie-Hurder, A., Henderson, W. G., Henry, R. R., Kelley, D. E., . . . Morgan, N. A. (1996). Implantable insulin pump vs multiple- dose insulin for non-insulin-dependent diabetes mellitus: a randomized clinical trial. Department of Veterans Affairs Implantable Insulin Pump Study Group. JAMA, 276, 1322-1327.

Schade, D. S., & Eaton, P. (1980). The peritoneum—a potential insulin delivery route for a mechanical pancreas. Diabetes Care, 3(2), 229-234.

Schade, D. S., Eaton, R. P., Friedman, J. E., & Spencer, W. J. (1980). Normalization of plasma insulin profiles with intraperitoneal insulin in diabetic man. Diabetologia, 19, 35-39.

Scheuner, D., & Kaufman, R. J. (2008). The unfolded protein response: a pathway that links insulin demand β-cell failure and diabetes. Endocr. Rev., 29, 317-333.

Schneider, S., Feilen, P. J., Brunnenmeier, F., Minnemann, T., Zimmermann, H., Zimmermann, U., & Weber, M. M. (2005). Long-term graft function of adult rat and human islets encapsulated in novel alginate-based microcapsules after transplantation in immunocompetent diabetic mice. Diabetes, 54(3), 687-693.

Scholfield, M. R., Zanden, C. M. V., Carter, M., & Ho, P. S. (2013). Halogen bonding (X‐bonding): A biological perspective. Protein Science, 22(2), 139-152.

Schwabe, C., Gowan, L. K., & Reinig, J. W. (1982). Evolution, relaxin and insulin: a new perspective. Annals of the New York Academy of Sciences, 380(1), 6-12.

Sciacca, L., Cassarino, M. F., Genua, M., Pandini, G., LeMoli, R., Squatrito, S., & Vigneri, R. (2010). Insulin analogues differently activate insulin receptor isoforms and post-receptor signaling. Diabetologia, 53(8), 1743-1753. doi: 10.1007/s00125-010-1760-6

Seino, S., & Bell, G. I. (1989). Alternative splicing of human insulin receptor messenger RNA. Biochem. Biophys. Res. Commun., 159, 312-316.

Selkoe, D. J. (2003). Folding proteins in fatal ways. Nature, 426, 900-904.

Semple, R. K., Savage, D. B., Cochran, E. K., Gorden, P., & O'rahilly, S. (2011). Genetic syndromes of severe insulin resistance. Endocrine reviews, 32(4), 498-514.

477

Settembre, C., Fraldi, A., Jahreiss, L., Spampanato, C., Venturi, C., Medina, D., . . . Ballabio, A. (2007). A block of autophagy in lysosomal storage disorders. Human molecular genetics, 17(1), 119-129.

Shi, D., Ran, M., Zhang, L., Huang, H., Li, X., Chen, M., & Akashi, M. (2016). Fabrication of biobased polyelectrolyte capsules and their application for glucose- triggered insulin delivery. ACS applied materials & interfaces, 8(22), 13688- 13697.

Shoelson, S., Fickova, M., Haneda, M., Nahum, A., Musso, G., Kaiser, E. T., . . . Tager, H. (1983). Indentification of a mutant human insulin predicetd to contain a serine- for -phenylalanine substitition. Proc. Natl. Acad. Sci. U. S. A., 80, 7390-7394. doi: PMC389956

Shoelson, S., Haneda, M., Blix, P., Nanjo, A., Sanke, T., Inouye, K., . . . Tager, H. (1983). Three mutant insulins in man. Nature, 302, 540-543.

Shoelson, S. E., Lu, Z. X., Parlautan, L., Lynch, C. S., & Weiss, M. A. (1992). Mutations at the dimer, hexamer, and receptor-binding, surfaces of insulin independently affect insulin-insulin and insulin-receptor interactions. Biochemistry, 31, 1757- 1767.

Shoelson, S. E., Polonsky, K. S., Zeidler, A., Rubenstein, A. H., & Tager, H. S. (1984a). Human insulin B24 (Phe®Ser), secretion and metabolic clearance of the abnormal insulin in man and in a dog model. J. Clin. Invest., 73, 1351-1358. doi: PMC425157

Shoelson, S. E., Polonsky, K. S., Zeidler, A., Rubenstein, A. H., & Tager, H. S. (1984b). Human insullin B24 (Phe®Ser), secretion and metabolic clearance of the abnormal insulin in man and in a dog model. J. Clin. Invest., 73, 1351-1358. doi: PMC425157

Shojaee-Moradie, F., Powrie, J. K., Sundermann, E., Spring, M. W., Schuttler, A., Sonksen, P. H., . . . Jones, R. H. (2000). Novel hepatoselective insulin analog: studies with a covalently linked thyroxyl-insulin complex in humans. Diabetes Care, 23, 1124-1129.

Sikosek, T., & Chan, H. S. (2014). Biophysics of protein evolution and evolutionary protein biophysics. Journal of The Royal Society Interface, 11(100). doi: 10.1098/rsif.2014.0419

Simkovsky, R., & King, J. (2006). An elongated spine of buried core residues necessary for in vivo folding of the parallel β-helix of P22 tailspike adhesin. Proc. Natl. Acad. Sci. U. S. A., 103, 3575-3580. doi: PMC1383501

478

Sinnokrot, M. O., & Sherrill, C. D. (2006). High-accuracy quantum mechanical studies of π− π interactions in benzene dimers. The Journal of Physical Chemistry A, 110(37), 10656-10668.

Slieker, L. J., Brooke, G. S., DiMarchi, R. D., Flora, D. B., Green, L. K., Hoffmann, J. A., . . . Chance, R. E. (1997). Modifications in the B10 and B26-30 regions of the B chain of human insulin alter affinity for the human IGF-I receptor more than for the insulin receptor. Diabetologia, 40, S54-S61.

Snell-Bergeon, J. K., & Wadwa, R. P. (2012). Hypoglycemia, diabetes, and cardiovascular disease. Diabetes technology & therapeutics, 14(S1), S-51-S-58.

Sommerfeld, M. R., Muller, G., Tschank, G., Seipke, G., Habermann, P., Kurrle, R., & Tennagels, N. (2010). In vitro metabolic and mitogenic signaling of insulin glargine and its metabolites. PLoS One, 5, e9540. doi: PMC2832019

Sosnick, T. R., Fang, X., & Shelton, V. M. (2000). Application of circular dichroism to study RNA folding transitions. Methods Enzymol., 317, 393-409.

Springsteen, G., & Wang, B. (2002). A detailed examination of boronic acid–diol complexation. Tetrahedron, 58(26), 5291-5300. doi: http://dx.doi.org/10.1016/S0040-4020(02)00489-1

Sreerama, N., & Woody, R. W. (1993). A self-consistent method for the analysis of protein secondary structure from circular dichroism. Anal. Biochem., 209, 32-44.

Stanescu, D. E., Hughes, N., Kaplan, B., Stanley, C. A., & De León, D. D. (2012). Novel presentations of congenital hyperinsulinism due to mutations in the MODY genes: HNF1A and HNF4A. The Journal of Clinical Endocrinology & Metabolism, 97(10), E2026-E2030.

Steensgaard, D. B., Schluckebier, G., Strauss, H. M., Norrman, M., Thomsen, J. K., Friderichsen, A. V., . . . Jonassen, I. (2013). Ligand-controlled assembly of hexamers, dihexamers, and linear multihexamer structures by the engineered acylated insulin degludec. Biochemistry, 52(2), 295-309. doi: 10.1021/bi3008609

Stefani, M., & Dobson, C. M. (2003). Protein aggregation and aggregate toxicity: new insights into protein folding, misfolding diseases and biological evolution. J. Mol. Med. (Berl.), 81(11), 678-699. doi: 10.1007/s00109-003-0464-5

Steil, G., Clark, B., Kanderian, S., & Rebrin, K. (2005). Modeling insulin action for development of a closed-loop artificial pancreas. Diabetes technology & therapeutics, 7(1), 94-108.

Steiner, D. F. (1998). The proprotein convertases. Curr. Opin. Chem. Biol., 2, 31-39.

479

Steiner, D. F. (Ed.). (1990). The biosynthesis of insulin (Vol. 92). Berlin: Springer- Verlag.

Steiner, D. F., Bell, G. I., & Rubenstein, A. H. (2006). Chemistry and biosynthesis of the islet hormones: insulin, islet amyloid polypeptide (amylin), glucagon, somatostatin, and pancreatic polypeptide. In L. DeGroot & J. L. Jameson (Eds.), Endocrinology (5th ed., pp. 925-960). Philadephia, PA: W. B. Saunders.

Steiner, D. F., & Chan, S. J. (1988). An overview of insulin evolution. Horm. Metab. Res., 20, 443-444.

Steiner, D. F., Chan, S. J., Welsh, J. M., & Kwok, S. C. (1985). Structure and evolution of the insulin gene. Annu. Rev. Genet., 19, 463-484.

Steiner, D. F., Clark, J. L., Nolan, C., Rubenstein, A. H., Margoliash, E., Aten, B., & Oyer, P. E. (1969). Proinsulin and the biosynthesis of insulin. Recent Prog. Horm. Res., 25, 207-282.

Steiner, D. F., Hallund, O., Rubenstein, A., Cho, S., & Bayliss, C. (1968). Isolation and properties of proinsulin, intermediate forms, and other minor components from crystalline bovine insulin. Diabetes, 17, 725-736.

Steiner, D. F., Kemmler, W., Clark, J. L., Oyer, P. E., & Rubenstein, A. H. (Eds.). (1972). The biosynthesis of insulin. Baltimore: Williams & Wilkins.

Steiner, D. F., & Oyer, P. E. (1967). The biosynthesis of insulin and a probable precursor of insulin by a human islet cell adenoma. Proc. Natl. Acad. Sci. U. S. A., 57, 473- 480. doi: PMC335530

Steiner, D. F., Park, S. Y., Stay, J., Philipson, L. H., & Bell, G. I. (2009). A brief perspective on insulin production. Diabetes Obes. Metab., 11, 189-196. doi: PMID: 19817801

Steward, L. E., & Chamberlin, A. R. (1998). Protein engineering with non-standard amino acids. Methods Mol. Biol., 77, 325-354.

Stewart, W. J., Mcsweeney, S. M., Kellett, M. A., Faxon, D. P., & Ryan, T. J. (1984). Increased risk of severe protamine reactions in NPH insulin-dependent diabetics undergoing cardiac catheterization. Circulation, 70(5), 788-792.

Storm, M. C., & Dunn, M. F. (1985). The GluB13 carboxylates of the insulin hexamer form a cage for Cd2+ and Ca2+ ions. Biochemistry, 24(7), 1749-1756.

Stoy, J., Edghill, E. L., Flanagan, S. E., Ye, H., Paz, V. P., Pluzhnikov, A., . . . Bell, G. I. (2007). Insulin gene mutations as a cause of permanent neonatal diabetes. Proc. Natl. Acad. Sci. U. S. A., 104, 15040-15044. doi: PMC1986609

480

Stoy, J., Olsen, J., Park, S. Y., Gregersen, S., Hjorringgaard, C. U., & Bell, G. I. (2017). In vivo measurement and biological characterisation of the diabetes-associated mutant insulin p.R46Q (GlnB22-insulin). Diabetologia. doi: 10.1007/s00125-017- 4295-2

Stoy, J., Steiner, D. F., Park, S. Y., Ye, H., Philipson, L. H., & Bell, G. I. (2010). Clinical and molecular genetics of neonatal diabetes due to mutations in the insulin gene. Rev. Endocr. Metab. Disord., 11(3), 205-215. doi: 10.1007/s11154-010-9151-3

Sturtevant, J., Yu, M., Haase-Pettingell, C., & King, J. (1989). Thermostability of temperature-sensitive folding mutants of the P22 tailspike protein. Journal of Biological Chemistry, 264(18), 10693-10698.

Subbian, E., Williamson, D. M., & Shinde, U. (2015). Protein Folding Mediated by an Intramolecular Chaperone: Energy Landscape for Unimolecular Pro-Subtilisin E Maturation. Advances in Bioscience and Biotechnology, 6(02), 73.

Suissa, S., Azoulay, L., Dell'Aniello, S., Evans, M., Vora, J., & Pollak, M. (2011). Long- term effects of insulin glargine on the risk of breast cancer. Diabetologia, 54(9), 2254-2262. doi: 10.1007/s00125-011-2190-9

Sun, J., Cui, J., He, Q., Chen, Z., Arvan, P., & Liu, M. (2015). Proinsulin misfolding and endoplasmic reticulum stress during the development and progression of diabetes ☆. Molecular aspects of medicine, 42, 105-118.

Sun, L., Zhang, X., Zheng, C., Wu, Z., & Li, C. (2013). A pH gated, glucose-sensitive nanoparticle based on worm-like mesoporous silica for controlled insulin release. The journal of physical chemistry B, 117(14), 3852-3860.

Swift, B., Hawkins, P. N., Richards, C., & Gregory, R. (2002). Examination of insulin injection sites: an unexpected finding of localized amyloidosis. [Letter]. Diabet. Med., 19, 881-886.

Szyperski, T., Luginbuhl, P., Otting, G., Guntert, P., & Wüthrich, K. (1993). Protein dynamics studied by rotating frame nitrogen-15 spin relaxation times. J. Biomol. NMR, 3, 151-164.

Tamaki, M., Fujitani, Y., Hara, A., Uchida, T., Tamura, Y., Takeno, K., . . . Fukunaka, A. (2013). The diabetes-susceptible gene SLC30A8/ZnT8 regulates hepatic insulin clearance. The Journal of clinical investigation, 123(10), 4513-4524.

Tanaka, S., Kurabuchi, S., Mochida, H., Kato, T., Takahashi, S., Watanabe, T., & Nakayama, K. (1996). Immunocytochemical localization of prohormone convertases PC1/PC3 and PC2 in rat pancreatic islets. Arch. Histol. Cytol., 59(3), 261-271.

481

Thabit, H., & Hovorka, R. (2016). Coming of age: the artificial pancreas for type 1 diabetes. Diabetologia, 59(9), 1795-1805.

Thibaudeau, K., Léger, R., Huang, X., Robitaille, M., Quraishi, O., Soucy, C., . . . Castaigne, J.-P. (2005). Synthesis and Evaluation of Insulin− Human Serum Albumin Conjugates. Bioconjugate chemistry, 16(4), 1000-1008.

Thomas, G. J., Becka, R., Sargent, D., Yu, M. H., & King, J. (1990). Conformational stability of P22 tailspike proteins carrying temperature-sensitive folding mutations. Biochemistry, 29, 4181-4187.

Tirasophon, W., Welihinda, A. A., & Kaufman, R. J. (1998). A stress response pathway from the endoplasmic reticulum to the nucleus requires a novel bifunctional protein kinase/endoribonuclease (Ire1p) in mammalian cells. Genes Dev., 12, 1812-1824. doi: PMC316900

Turksoy, K., Monforti, C., Park, M., Griffith, G., Quinn, L., & Cinar, A. (2017). Use of Wearable Sensors and Biometric Variables in an Artificial Pancreas System. Sensors, 17(3), 532.

Tycko, R. (2011). Solid-state NMR studies of amyloid fibril structure. Annu. Rev. Phys. Chem., 62, 279-299. doi: 10.1146/annurev-physchem-032210-103539

Tycko, R., & Ishii, Y. (2003). Constraints on supramolecular structure in amyloid fibrils from two-dimensional solid-state NMR spectroscopy with uniform isotopic labeling. J. Am. Chem. Soc., 125, 6606-6607.

Uchio, T., Baudyš, M., Liu, F., Song, S. C., & Kim, S. W. (1999). Site-specific insulin conjugates with enhanced stability and extended action profile. Advanced drug delivery reviews, 35(2-3), 289-306.

Uchizono, Y., Alarcon, C., Wicksteed, B., Marsh, B., & Rhodes, C. J. (2007). The balance between proinsulin biosynthesis and insulin secretion: where can imbalance lead? Diabetes, Obesity and Metabolism, 9(s2), 56-66.

UKPDS, U. P. D. S. G. (1998). Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group. Lancet, 352(9131), 837-853.

Ungaro, F., di Villa Bianca, R. d. E., Giovino, C., Miro, A., Sorrentino, R., Quaglia, F., & La Rotonda, M. I. (2009). Insulin-loaded PLGA/cyclodextrin large porous particles with improved aerosolization properties: in vivo deposition and hypoglycaemic activity after delivery to rat lungs. Journal of controlled release, 135(1), 25-34.

482 van Gunsteren, W. F., Daura, X., Hansen, N., Mark, A. E., Oostenbrink, C., Riniker, S., & Smith, L. J. (2018). Validation of molecular simulation: an overview of issues. Angewandte Chemie International Edition, 57(4), 884-902.

Varghese, J., Lim, S. F., & Cohen, S. M. (2010). Drosophila miR-14 regulates insulin production and metabolism through its target, sugarbabe. Genes Dev., 24, 2748- 2753.

Vaughn, D. E., Yocum, R. C., Muchmore, D. B., Sugarman, B. J., Vick, A. M., Bilinsky, I. P., & Frost, G. I. (2009). Accelerated pharmacokinetics and glucodynamics of prandial insulins injected with recombinant human hyaluronidase. Diabetes technology & therapeutics, 11(6), 345-352.

Vegas, A. J., Veiseh, O., Gürtler, M., Millman, J. R., Pagliuca, F. W., Bader, A. R., . . . Olejnik, K. (2016). Long-term glycemic control using polymer-encapsulated human stem cell-derived beta cells in immune-competent mice. Nature medicine.

Vienberg, S. G., Bouman, S. D., Sorensen, H., Stidsen, C. E., Kjeldsen, T., Glendorf, T., . . . Nishimura, E. (2011). Receptor-isoform-selective insulin analogues give tissue- preferential effects. Biochem. J., 440(3), 301-308. doi: 10.1042/bj20110880

Vienberg, S. G., Bouman, S. D., Sørensen, H., Stidsen, C. E., Kjeldsen, T., Glendorf, T., . . . Nishimura, E. (2011). Receptor-isoform-selective insulin analogues give tissue- preferential effects. Biochemical Journal, 440(3), 301-308.

Vinik, A., & Bell, G. (1988). Mutant insulin syndromes. Horm. Metab. Res., 20, 1-10. doi: MID: 3286444

Vinther, T. N., Kjeldsen, T. B., Jensen, K. J., & Hubálek, F. (2015). The road to the first, fully active and more stable human insulin variant with an additional disulfide bond. Journal of Peptide Science, 21(11), 797-806.

Vinther, T. N., Norrman, M., Ribel, U., Huus, K., Schlein, M., Steensgaard, D. B., . . . Hubálek, F. (2013). Insulin analog with additional disulfide bond has increased stability and preserved activity. Protein Sci., 22(3), 296–305. doi: 10.1002/pro.2211

Vugmeyster, L., Kuhlman, B., & Raleigh, D. P. (1998). Amide proton exchange measurements as a probe of the stability and dynamics of the n‐terminal domain of the ribosomal protein L9: Comparison with the intact protein. Protein Science, 7(9), 1994-1997.

Wagner, A., Diez, J., Schulze-Briese, C., & Schluckebier, G. (2009). Crystal structure of ultralente - a microcrystalline insulin suspension. Proteins, 74, 1018-1027.

483

Wan, Z., Huang, K., Whittaker, J., & Weiss, M. A. (2008). The structure of a mutant insulin uncouples receptor binding from protein allostery. An electrostatic block to the TR transition. J. Biol. Chem., 283, 21198-21210. doi: PMC2475698

Wang, C., Ye, Y., Sun, W., Yu, J., Wang, J., Lawrence, D. S., . . . Gu, Z. (2017). Red Blood Cells for Glucose‐Responsive Insulin Delivery. Advanced Materials.

Wang, C. C., & Tsou, C. L. (1991). The insulin A and B chains contain sufficient structural information to form the native molecule. Trends Biochem. Sci., 16, 279- 281.

Wang, H., Yi, J., Yu, Y., & Zhou, S. (2017). NIR upconversion fluorescence glucose sensing and glucose-responsive insulin release of carbon dot-immobilized hybrid microgels at physiological pH. Nanoscale, 9(2), 509-516.

Wang, J., Onuchic, J., & Wolynes, P. (1996). Statistics of kinetic pathways on biased rough energy landscapes with applications to protein folding. Phys. Rev. Lett., 76, 4861-4864.

Wang, Y., Huang, F., Sun, Y., Gao, M., & Chai, Z. (2017). Development of shell cross- linked nanoparticles based on boronic acid-related reactions for self-regulated insulin delivery. Journal of Biomaterials Science, Polymer Edition, 28(1), 93-106.

Wang, Y., Zhang, X., Han, Y., Cheng, C., & Li, C. (2012). pH-and glucose-sensitive glycopolymer nanoparticles based on phenylboronic acid for triggered release of insulin. Carbohydrate polymers, 89(1), 124-131.

Wang, Z. X. (1995). An exact mathematical expression for describing competitive binding of two different ligands to a protein molecule FEBS Lett., 360, 111-114.

Waters, M. L. (2002). Aromatic interactions in model systems. Curr. Opin. Chem. Biol., 6(6), 736-741.

Weis, W. I., & Drickamer, K. (1996). Structural basis of lectin-carbohydrate recognition. Annual review of biochemistry, 65(1), 441-473.

Weiss, M. A. (1990). Thermal unfolding studies of a leucine zipper domain and its specific DNA complex: implications for scissor's grip recognition. Biochemistry, 29(35), 8020-8024.

Weiss, M. A. (2009a). Proinsulin and the genetics of diabetes mellitus. J. Biol. Chem., 284, 19159-19163. doi: PMC2740536

Weiss, M. A. (2009). The structure and function of insulin: decoding the TR transition. Vitamins & Hormones, 80, 33-49.

484

Weiss, M. A. (2009b). The structure and function of insulin: decoding the TR transition. In G. Litwack (Ed.), Vitamins and hormones: Advances in research and application (Vol. 80, pp. 33-49). Burlington: Academic Press.

Weiss, M. A. (2012). United States Patent No. US 2012/0316107 A1. United States Patent and Trademark Office patent full-text database: U. S. P. a. T. Office.

Weiss, M. A. (2013a). Design of ultra-stable insulin analogues for the developing world. Journal of Health Specialties 1(2), 59-70.

Weiss, M. A. (2013b). Diabetes mellitus due to the toxic misfolding of proinsulin variants. FEBS Lett., 587(13), 1942-1950. doi: 10.1016/j.febslet.2013.04.044

Weiss, M. A., Hua, Q. X., Jia, W., Chu, Y. C., Wang, R. Y., & Katsoyannis, P. G. (2000). Hierarchiacal protein "un-design": insulin's intrachain disulfide bridge tethers a recognition a-helix. Biochemistry, 39, 15429-15440.

Weiss, M. A., Hua, Q. X., Jia, W., Nakagawa, S. H., Chu, Y. C., Hu, S. Q., & Katsoyannis, P. G. (2001). Activities of monomeric insulin analogs at position A8 are uncorrelated with their thermodynamic stabilities. J. Biol. Chem., 276, 40018- 40024.

Weiss, M. A., Hua, Q. X., Jia, W., Nakagawa, S. H., Chu, Y. C., & Katsoyannis, P. G. (2002). Activities of insulin analogues at position 8 are uncorrelated with thermodynamic stability. In M. L. Dieken, M. Federwisch & P. De Meyts (Eds.), Insulin and Related Proteins - Structure to Function and Pharmacology (pp. 103- 119). Dordrecht, The Netherlands: Kluwar Academic Publishers.

Weiss, M. A., & Lawrence, M. C. (2018). A Thing of Beauty: Structure and Function of Insulin’s “Aromatic Triplet”. Obesity, Diabetes and Metab., In press.

Weiss, M. A., Nakagawa, S. H., Jia, W., Xu, B., Hua, Q. X., Chu, Y. C., . . . Katsoyannis, P. G. (2002). Protein structure and the spandrels of San Marco: insulin's receptor- binding surface is buttressed by an invariant leucine essential for protein stability. Biochemistry, 41, 809-819.

Weissman, J. S., & Kim, P. S. (1991). Reexamination of the folding of BPTI: predominance of native intermediates. Science, 253, 1386-1393.

Weissman, J. S., & Kim, P. S. (1992). Kinetic role of nonnative species in the folding of bovine pancreatic trypsin inhibitor. Proc. Natl. Acad. Sci. U. S. A., 89, 9900-9904.

Weissman, J. S., & Kim, P. S. (1992). The pro region of BPTI facilitates folding. Cell, 71(5), 841-851.

485

Westermark, P., Bellotti, V., Obici, L., Kisilevsky, R., Merlini, G., Sipe, J. D., . . . Howlett, G. J. (2005). Amyloid proteins: the beta sheet conformation and disease.

Westphal, S. A. (1996). The occurrence of diabetic ketoacidosis in non-insulin-dependent diabetes and newly diagnosed diabetic adults. The American journal of medicine, 101(1), 19-24.

White, M. F., & Kahn, C. R. (1994). The insulin signaling system. J. Biol. Chem., 269, 1- 4.

Whittaker, J., Garcia, P., Yu, G. Q., & Mynarcik, D. C. (1994). Transmembrane domain interactions are necessary for negative cooperativity of the insulin receptor. Mol. Endocrinol., 8, 1521-1527.

Willighagen, E., & Howard, M. (2007). Fast and Scriptable Molecular Graphics in Web Browsers without Java3D. Nature Precedings, from http://dx.doi.org/10.1038/npre.2007.50.1

Wilson, R., & Turner, A. (1992). Glucose oxidase: an ideal enzyme. Biosensors and Bioelectronics, 7(3), 165-185.

Wimley, W. C., & White, S. H. (1996). Experimentally determined hydrophobicity scale for proteins at membrane interfaces. Nature Structural and Molecular Biology, 3(10), 842.

Winn, M. D., Ballard, C. C., Cowtan, K. D., Dodson, E. J., Emsley, P., Evans, P. R., . . . McCoy, A. (2011). Overview of the CCP4 suite and current developments. Acta Crystallographica Section D: Biological Crystallography, 67(4), 235-242.

Winter, J., Klappa, P., Freedman, R. B., Lilie, H., & Rudolph, R. (2002). Catalytic activity and chaperone function of human protein-disulfide isomerase are required for the efficient refolding of proinsulin. J. Biol. Chem., 277, 310-317.

Winter, J., Lilie, H., & Rudolph, R. (2002). Renaturation of human proinsulin--a study on refolding and conversion to insulin. Anal. Biochem., 310, 148-155.

Wisler, J. W., Xiao, K., Thomsen, A. R., & Lefkowitz, R. J. (2014). Recent developments in biased agonism. Curr. Opin. Cell Biol., 27, 18-24. doi: 10.1016/j.ceb.2013.10.008

S0955-0674(13)00161-0 [pii]

Wollmer, A., Rannefeld, B., Johansen, B. R., Hejnaes, K. R., Balschmidt, P., & Hansen, F. B. (1987). Phenol-promoted structural transformation of insulin in solution. Biol. Chem. Hoppe Seyler, 368, 903-911.

486

Wolynes, P. G. (2005). Recent successes of the energy landscape theory of protein folding and function. Q. Rev. Biophys., 38(4), 405-410. doi: 10.1017/s0033583505004075

Wright, J., Birk, J., Haataja, L., Liu, M., Ramming, T., Weiss, M. A., . . . Arvan, P. (2013). Endoplasmic reticulum oxidoreductin-1α (Ero1α) improves folding and secretion of mutant proinsulin and limits mutant proinsulin-induced endoplasmic reticulum stress. Journal of Biological Chemistry, 288(43), 31010-31018.

Xie, S., Li, Z., & Yu, Z. (2015). Microneedles for transdermal delivery of insulin. Journal of Drug Delivery Science and Technology, 28, 11-17.

Xu, B., Hu, S. Q., Chu, Y. C., Huang, K., Nakagawa, S. H., Whittaker, J., . . . Weiss, M. A. (2004). Diabetes-associated mutations in insulin: consecutive residues in the B chain contact distinct domains of the insulin receptor. Biochemistry, 43, 8356- 8372.

Xu, B., Hu, S. Q., Chu, Y. C., Wang, S., Wang, R. Y., Nakagawa, S. H., . . . Weiss, M. A. (2004). Diabetes-associated mutations in insulin identify invariant receptor contacts. Diabetes, 53, 1599-1602.

Xu, B., Hua, Q. X., Nakagawa, S. H., Jia, W., Chu, Y. C., Katsoyannis, P. G., & Weiss, M. A. (2002). A cavity-forming mutation in insulin induces segmental unfolding of a surrounding α-helix. Protein Sci., 11, 104-116.

Xu, B., Huang, K., Chu, Y. C., Hu, S. Q., Nakagawa, S., Wang, S., . . . Weiss, M. A. (2009). Decoding the cryptic active conformation of a protein by synthetic photoscanning: Insulin inserts a detachable arm between receptor domains. J. Biol. Chem., 284, 14597-14608. doi: PMC2682907

Yamamoto, C., Miyoshi, H., Fujiwara, Y., Kameda, R., Ichiyama, M., Nomoto, H., . . . Atsumi, T. (2016). Degludec is superior to glargine in terms of daily glycemic variability in people with type 1 diabetes mellitus. Endocr J, 63(1), 53-60. doi: 10.1507/endocrj.EJ15-0438

Yan, J., Springsteen, G., Deeter, S., and Wang, B. . (2004). The relationship among pKa, pH, and binding constants in the interactions between boronic acids and diols—it is not as simple as it appears. Tetrahedron, 60, 11205–11209.

Yang, W., Gao, X., and Wang, B. (2003). Boronic acid compounds as potential pharmaceutical agents. Medicinal Research Reviews, 23, 346-268.

Yang, Y., Hua, Q. X., Liu, J., Shimizu, E. H., Choquette, M. H., Mackin, R. B., & Weiss, M. A. (2010). Solution structure of proinsulin: connecting domain flexibility and prohormone processing. J. Biol. Chem., 285, 7847-7851.

487

Yang, Y., Petkova, A., Huang, K., Xu, B., Hua, Q. X., Ye, I. J., . . . Weiss, M. A. (2010). An Achilles' Heel in an amyloidogenic protein and its repair. Insulin dynamics, misfolding, and therapeutic design. J. Biol. Chem., 285, 10806-10821. doi: PMC2856287

Ye, L., Maji, S., Sanghera, N., Gopalasingam, P., Gorbunov, E., Tarasov, S., . . . Klein- Seetharaman, J. (2017). Structure and dynamics of the insulin receptor: implications for receptor activation and drug discovery. Drug Discov Today. doi: 10.1016/j.drudis.2017.04.011

Yesilyurt, V., Webber, M. J., Appel, E. A., Godwin, C., Langer, R., & Anderson, D. G. (2016). Injectable Self‐Healing Glucose‐Responsive Hydrogels with pH‐ Regulated Mechanical Properties. Advanced Materials, 28(1), 86-91.

Yilmaz, M. D., Xue, M., Ambrogio, M. W., Buyukcakir, O., Wu, Y., Frasconi, M., . . . Zink, J. I. (2015). Sugar and pH dual-responsive mesoporous silica nanocontainers based on competitive binding mechanisms. Nanoscale, 7(3), 1067-1072.

Yu, J., Qian, C., Zhang, Y., Cui, Z., Zhu, Y., Shen, Q., . . . Gu, Z. (2017). Hypoxia and H2O2 Dual-Sensitive Vesicles for Enhanced Glucose-Responsive Insulin Delivery. Nano Letters.

Yu, M., & King, J. (1988). Surface amino acids as sites of temperature-sensitive folding mutations in the P22 tailspike protein. Journal of Biological Chemistry, 263(3), 1424-1431.

Žáková, L., Kletvíková, E., Veverka, V., Lepsík, M., Watson, C. J., Turkenburg, J. P., . . . Brzozowski, A. M. (2013). Structural integrity of the B24 site in human insulin is important for hormone functionality. J. Biol. Chem., 288(15), 10230-10240. doi: 10.1074/jbc.M112.448050

Zaykov, A. N., Mayer, J. P., & DiMarchi, R. D. (2016). Pursuit of a perfect insulin. Nature Reviews Drug Discovery.

Zhang, J. X., & Goldenberg, D. P. (1993). Amino acid replacement that eliminates kinetic traps in the folding pathway of pancreatic trypsin inhibitor. Biochemistry, 32, 14075-14081.

Zhuravlev, P. I., Materese, C. K., & Papoian, G. A. (2009). Deconstructing the native state: energy landscapes, function, and dynamics of globular proteins. The Journal of Physical Chemistry B, 113(26), 8800-8812.

Zimmerman, A. E., Moule, M. L., & Yip, C. C. (1974). Guinea pig insulin II. Biological activity. Journal of Biological Chemistry, 249(13), 4026-4029.

488

Zimmet, P., Alberti, K. G., & Shaw, J. (2001). Global and societal implications of the diabetes epidemic. Nature, 414(6865), 782-787. doi: 10.1038/414782a

Zion, T. C., Tsang, H. H., & Ying, J. Y. (2003). Glucose-sensitive nanoparticles for controlled insulin delivery.

489