Computational Evaluation and Structure-based Design for Potentiation of

Nicotine Vaccines

Kyle Saylor

Dissertation submitted to the faculty of Virginia Polytechnic Institute and State University

in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

In

Biological Systems Engineering

Chenming Zhang, Chair

Xin M. Luo

Xiang-Jin Meng

Ryan S. Senger

September 3rd, 2020

Blacksburg, VA

Keywords: nicotine vaccine, virus-like particle, physiologically-based pharmacokinetic

model, epitope prediction, structural vaccinology Computational Evaluation and Structure-based Design for Potentiation of

Nicotine Vaccines

Kyle Saylor

Abstract (Academic)

Existing therapeutic options for the alleviation of nicotine addiction have been largely ineffective at stemming the tide of tobacco use. Immunopharmacotherapy, or vaccination, is a promising, alternate therapy that is currently being explored. Results from previous studies indicate that nicotine vaccines (NVs) are effective in subjects that achieve high drug-specific antibody titers, though overall efficacy has not been observed. Consequently, improvement of these vaccines is necessary before they can achieve approval for human use. In this report, three separate approaches towards NV potentiation are explored.

The first approach applied physiologically-based pharmacokinetic (PBPK) modeling to better assess NV potential. Rat and human physiological and pharmacological parameters were obtained from literature and used to construct compartmentalized models for nicotine and cotinine distribution. These models were then calibrated and validated using data obtained from literature. The final models verified the therapeutic potential of the NV concept, identified four key parameters associated with vaccine success, and established correlates for success that could be used to evaluate future NVs prior to clinical trials.

In the second approach, conjugate NV scaffoldings were engineered by using wild-type (WT) and chimeric human papilloma (HPV) 16 L1 protein virus-like particles

(VLPs). The chimeric protein was created by removing the last 34 C-terminal residues from the WT protein and then incorporating a multi-epitope insert that could universally target major histocompatibility complex (MHC) class II molecules. The proteins were subsequently expressed in E. coli and purified using a multi-step process. Comparisons between the separation outcomes revealed that the insert was able to modulate individual process outcomes and improve overall yield without inhibiting VLP assembly.

In the third approach, commonly used carrier proteins were computationally mined for their MHC class II epitope content using human leukocyte antigen (HLA) population frequency data and MHC epitope prediction software. The most immunogenic epitopes were concatenated with interspacing cathepsin cleavage sequences and the resulting protein was re-evaluated using the earlier methods. This work represents the first ever in silico design of chimeric antigens that could potentially target all of the major HLA DQ and HLA DR allotypes found in humans.

Computational Evaluation and Structure-based Design for Potentiation of

Nicotine Vaccines

Kyle Saylor

Abstract (General Audience)

Existing treatment options for addressing nicotine addiction have been largely ineffective at preventing tobacco use. Vaccination, on the other hand, is a promising, alternate treatment option that is currently being explored. Previous studies have shown that nicotine vaccines (NVs) are effective in the subjects that respond well to the vaccine. Effectiveness in the majority of vaccine recipients, however, has not been observed. Consequently, improvement of these vaccines is necessary before they can be used in humans. In this report, three separate approaches for improving NV effectiveness are explored.

The first approach applied physiologically-based pharmacokinetic (PBPK) modeling to better assess NV potential. Parameters were obtained from literature and used to construct models that could predict NV effectiveness in rats and humans. These models were then calibrated and validated using data obtained from literature. The final models verified that NVs could work if certain conditions were met, identified four key parameters associated with vaccine success, and allowed for estimation of NV efficacy prior to their evaluation in humans.

In the second approach, protein carriers for conjugate NVs were constructed using the human papilloma (HPV) 16 L1 protein. This protein is known for its ability to form virus-like particles (VLPs). Both a modified and an unmodified (wild-type) protein were constructed. The modified HPV 16 L1 protein was created by replacing the last 34

C-terminal amino acids with a polypeptide insert that could enhance the immunogenicity of the vaccine. The modified and unmodified proteins were then expressed in E. coli and purified. Results indicated that the insert was able to modulate individual process outcomes and improve overall process yield without preventing VLP assembly.

In the third approach, commonly used carrier proteins were computationally mined for their MHC class II epitope content using human gene frequency data and

MHC epitope prediction software. The epitopes that were predicted to be the most immunogenic were linked together with interspacing protease recognition sequences and the immunogenicity of the resulting protein was re-evaluated using the prediction software. This work represents the first computational design of antigens that could potentially allow a vaccine to be effective in a large portion of human population regardless of the genetic variability.

v

Dedication

I would like to acknowledge all of the individuals, both known and unknown to me, that in some capacity may have helped bring me to this point. This acknowledgement is not only directed at family, friends, teachers, coaches, mentors, etc., all of whom I greatly appreciate, but also at the individuals and groups, past and present and willing and unwilling, that suffered any tribulations that I may have benefited from. I am free and I am privileged and I hope that over the course of my life I can at least make a dent in the debt I owe to these outstanding people.

I would not be here, on many different accounts, if it weren’t for my parents. I can’t thank you enough for all of the support you’ve given me over the years. I love you both very much.

I would like to thank my sister. I’ll never forget how you saved me from internment, and sorry for beating you up and framing you when we were little. BFFs for life.

I would like to thank my partner. Your support has kept me going when I considered giving up, and you’ve somehow managed to put up with my constant indignation about menial (and sometimes less menial) offenses I see in the world. You’re a keeper.

vi

Acknowledgements

I would like to thank Dr. Mike Zhang, for having faith in my abilities and supporting my ideas. I will forever be grateful for the amount of creativity and independence you allowed me to have while working on this PhD.

I would like to thank my committee members, Drs. Xin Luo, Xiang-Jin Meng, and Ryan

Senger. Your support and insight have helped immensely throughout this process.

I would like to thank my lab mates, both past and present; Yuanzhi Bian, Dr. Frank

Gillam, Dr. Andy Hu, Dr. Yun Hu, Taylor Lohneis, Dr. Song Lou, Dr. Yi Lu, Dr. Debra

Walter, and Dr. Zongmin Zhao. I am particularly indebted to Frank, who showed me the ropes when I first started. I owe you, dude.

I would like to thank all of my instructors and colleagues that taught me anything over the years. I came here to learn and you helped facilitate that learning.

I would like to thank the Biological Systems Engineering Department, the National

Institute of Health, New Horizon Graduate Scholars, the American Association of

Immunologists, and the Virginia Tech Open Access Subvention fund for their academic and/or financial support.

vii

Table of Contents

Abstract (Academic) ...... ii

Abstract (General Audience) ...... iv

Dedication ...... vi

Acknowledgements ...... vii

Table of Contents ...... viii

List of Figures ...... xv

List of Tables ...... xviii

List of Equations ...... xx

Attribution ...... xxii

Chapter 1: Introduction ...... 1

References ...... 9

Chapter 2: Literature Review: Tobacco Use, Nicotine, and Nicotine Vaccines ..... 13

2.1. Synopsis of the Recurring Tobacco use Problem ...... 13

2.2. Nicotine and its Role in Addiction ...... 15

2.3. Pharmaceutical Treatment Options for Nicotine Addiction ...... 18

2.4. Immunopharmacotherapeutic Treatment Options for Nicotine Addiction ...... 22

2.5. An Alternate Strategy for Nicotine Vaccine Assessment ...... 28

2.6. Alternate Strategies for Nicotine Vaccine Potentiation – Chimeric Virus-like

Particle Carriers ...... 30

viii

2.7. Alternate Strategies for Nicotine Vaccine Potentiation – Universal Carriers ...... 34

2.8. Conclusions ...... 35

References ...... 37

Chapter 3: Supplementary Literature Review: Designs of Antigen Structure and

Composition for Improved Protein-based Vaccine Efficacy ...... 57

Abstract ...... 58

Keywords ...... 58

3.1. Introduction ...... 59

3.2. Targeting Pattern Recognition Receptors ...... 63

3.2.1. PBVs, PRRs, and the Innate Immune System ...... 63

3.2.2. Rational Incorporation of PTMs as PAMPs ...... 64

3.2.3. Facilitating PAMP Proximity to PBV Via Fusion ...... 65

3.2.4. Facilitating PAMP Proximity to PBV Via Encapsulation ...... 68

3.3. Targeting B Cell Receptors ...... 69

3.3.1. PBVs and the B Cell ...... 69

3.3.2. Mechanisms Behind BCR Recognition of PBV ...... 70

3.3.3. Chemical Conjugation of BCR Epitope to PBV ...... 71

3.3.4. Genetic Fusion of BCR Epitope to PBV ...... 72

3.3.5. Carrier Induced Epitopic Suppression ...... 74

3.4. Targeting T Cell Receptors ...... 76

ix

3.4.1. PBVs and the T Cell ...... 76

3.4.2. The Impact of APC Uptake of PBV on T Cell Activation ...... 76

3.4.3. The Impact of Cellular Localization of PBV Within APCs on T Cell Activation

...... 77

3.4.4. The Impact of PBV Stability on APC Processing and T Cell Activation ...... 78

3.4.5. The Importance of the T Cell Epitopes in PBV Efficacy and Design ...... 80

3.4.6. Targeting TCR Via Recombinant and Conjugate Approaches ...... 82

3.4.7. Limitations of Experimentally and Computationally Determined UTEs ...... 84

3.5. Additional Considerations ...... 84

3.5.1. PBV Safety Considerations...... 84

3.5.2. Animal Model Considerations ...... 85

3.5.3. Additional Stability Considerations ...... 87

3.5.4. PBV Targeting Strategies for Antigenically Variable Pathogens ...... 88

3.5.5. The Chimeric-Conjugate Approach ...... 90

3.5.6. Additional PBV Modification Techniques ...... 91

3.6. Conclusions ...... 92

Funding ...... 94

Conflicts of Interest ...... 95

Tables and Figures ...... 96

References ...... 106

x

Chapter 4: A Simple Physiologically Based Pharmacokinetic Model Evaluating the

Effect of Anti-Nicotine Antibodies on Nicotine Disposition in the Brains of Rats and Humans ...... 137

Abstract ...... 138

Keywords ...... 138

Abbreviations ...... 138

4.1. Background ...... 140

4.2. Methods ...... 143

4.2.1. Approach to Model Construction ...... 143

4.2.2 Major Model Assumptions...... 144

4.2.3. Compartmental Design ...... 145

4.2.4. Model Parameters ...... 148

4.2.5. Vaccine Efficacy Factors...... 149

4.2.6. Calibration ...... 151

4.2.7. Human Vaccine Efficacy Predictions ...... 152

4.3. Results ...... 154

4.3.1. Rat Model Calibration ...... 154

4.3.2. Human Model Calibration and Testing ...... 155

4.3.3. Sensitivity Analysis Results ...... 156

4.3.4. Maximal Antibody Effect Prediction Results...... 157

xi

4.4. Discussion ...... 158

4.5. Conclusions ...... 163

Conflict of Interest Statement ...... 164

Acknowledgments ...... 164

Tables and Figures ...... 165

References ...... 185

Chapter 5: Multi-epitope Insert Modulates Solubility-based and Chromatographic

Purification of Human Papilloma Virus 16 L1-based Vaccine Without Inhibiting

Virus-like Particle Assembly ...... 193

Abstract ...... 194

Highlights ...... 195

Keywords ...... 195

5.1. Introduction ...... 196

5.2. Materials and Methods ...... 198

5.2.1. Genetic Engineering ...... 198

5.2.2. Protein Expression ...... 199

5.2.3. Cell Processing ...... 200

5.2.4. Protein Purification ...... 201

5.2.5. Protein Refolding and VLP Assembly ...... 202

5.2.6. Quantitative Analyses ...... 203

xii

5.2.7. Statistical Analysis ...... 206

5.2.8. Purification System Scoring ...... 206

5.3. Results ...... 206

5.3.1. Genetic Engineering and Expression ...... 206

5.3.2. Purification ...... 207

5.3.3. Final Protein Quantification ...... 208

5.4. Discussion ...... 209

5.5. Conclusions ...... 215

Conflicts of Interest ...... 216

Acknowledgements ...... 216

Tables and Figures ...... 217

References ...... 224

Chapter 6: Computational Mining of MHC class II Epitopes for the Development of a Universal Carrier Protein ...... 229

Abstract ...... 230

Highlights ...... 230

Keywords ...... 231

6.1. Introduction ...... 232

6.2. Methods ...... 235

6.2.1. HLA-DQB1 Allele Frequency Analysis ...... 235

xiii

6.2.2. Source Immunogen Selection ...... 235

6.2.3. MHC Class II Epitope Predictions ...... 236

6.2.4. Epitope Scoring and Anchor Residue Identification ...... 236

6.2.5. Unweighted Epitope Score Analyses, Ranking, and Excision ...... 237

6.2.6. Conception and Analysis of Isotype-specific, Universal Immunogens ...... 238

6.2.7. Comparing Prediction Methods and Outputs ...... 239

6.3. Results ...... 239

6.3.1. Input Data Collection and Setup ...... 239

6.3.2. MHC Class II Epitope Predictions, Scoring, and Analysis ...... 240

6.3.3. Isotype-specific UCA Design and Analysis Results ...... 242

6.3.4. Inter-method Comparisons ...... 242

6.4. Discussion ...... 243

Conflicts of Interest ...... 248

Acknowledgements ...... 248

Tables and Figures ...... 249

References ...... 257

Chapter 7: General Conclusions ...... 262

Appendix A: Supplementary Material for Chapter 4 ...... 267

Appendix B: Supplementary Material for Chapter 5 ...... 276

Appendix C: Supplementary Material for Chapter 6 ...... 280

xiv

List of Figures

Figure 3.1. Recombinant toxins...... 96

Figure 3.2. Immunological mechanisms of recombinant, protein-based vaccination...... 98

Figure 3.3. PBV modification principles...... 101

Figure 3.4. The impact of PBV stability on immune response...... 104

Figure 4.1. Overall schematic for the rat and human PBPK models...... 173

Figure 4.2. Rat whole blood nicotine concentration – PBPK model calibration. . 174

Figure 4.3. Rat whole blood nicotine concentration – PBPK model validation. .. 175

Figure 4.4. Rat brain nicotine concentration – PBPK model calibration...... 176

Figure 4.5. Human venous blood nicotine and cotinine concentrations – PBPK model calibration...... 177

Figure 4.6. Human venous blood nicotine concentrations over varying dosages –

PBPK model validation...... 178

Figure 4.7. Human venous blood nicotine concentrations – PBPK model validation...... 179

Figure 4.8. Human venous blood cotinine concentrations – PBPK model validation...... 180

Figure 4.9. Evaluating the impact of antibody concentration, antibody affinity, and antibody cross-reactivity with cotinine on nicotine-specific antibody binding capacity – smokers with high resting cotinine levels...... 181

xv

Figure 4.10. Evaluating the impact of antibody concentration, antibody affinity, and antibody cross-reactivity with cotinine on nicotine-specific antibody binding capacity – non-smokers with negligible resting cotinine levels...... 182

Figure 4.11. Vaccine efficacy predictions in smokers and non-smokers – average antibody affinity for nicotine...... 183

Figure 4.12. Vaccine efficacy predictions in smokers and non-smokers – maximal antibody affinity for nicotine...... 184

Figure 5.1. Primary structure for the WT and WTΔC34-2TEp proteins...... 217

Figure 5.2. IB wash step yield, enrichment ratio, and overall score...... 218

Figure 5.3. Representative DEAE chromatograms...... 219

Figure 5.4. DEAE chromatography step yield, enrichment ratio, and overall score.

...... 220

Figure 5.5. Overall separation performance for WT and WTΔC34-2TEp proteins.

...... 221

Figure 5.6. Overall and regional hydrophobic and ionic characterization of the WT and WTΔC34-2TEp proteins...... 222

Figure 5.8. TEM images of WT and WTΔC34-2TEp VLPs...... 223

Figure 6.1. HLA population frequencies...... 252

Figure 6.2. HLA-DQ DT epitope analysis results...... 253

Figure 6.3. Design and assessment of DQ- and DR-specific UCAs and UCnAs. . 254

Figure 6.4. Comparing between prediction methods...... 255

Figure B.1. DNA Sequences for the WT and WTΔC34-2TEp proteins...... 276

Figure B.2. Amino acid sequences for the WT and WTΔC34-2TEp proteins...... 277

xvi

Figure B.3. Solubilized IB purity (SDS-PAGE), solubilized IB content (Western

Blot), and final purity (SDS-PAGE)...... 278

Figure B.4. Recoveries over all separation steps...... 279

Figure C.1. MHC epitope analysis results for immunogens / benchmarks 1-10. . 281

Figure C.2. MHC epitope analysis results for immunogens / benchmarks 11-17.283

Figure C.3. Combined MHC epitope analysis results for all immunogens / benchmarks...... 285

Figure C.4. Design and assessment of IAd- and IEd-specific UCAs and UCnAs. 287

xvii

List of Tables

Table 4.1. Nicotine-cotinine pharmacokinetic model parameters for human and rat model...... 165

Table 4.2. Nicotine–cotinine pharmacokinetic model parameters for human and rat model, cont...... 166

Table 4.3. Nicotine–cotinine pharmacokinetic model parameters for human and rat model, cont...... 167

Table 4.4. Keyler et al. 1999 published dose and antibody parameters...... 167

Table 4.5. Hieda et al. 1999 published dose and antibody parameters...... 168

Table 4.6. Benowitz et al. 1991 published dosing parameters...... 168

Table 4.7. Benowitz et al. 1982 published dosing parameters...... 168

Table 4.8. Benowitz et al. 1983 estimated dosing parameters...... 169

Table 4.9. Human serum nicotine-specific antibody concentrations from clinical trials...... 169

Table 4.10. Human vaccine efficacy dosing parameters #1...... 169

Table 4.11. Human vaccine efficacy prediction parameters...... 170

Table 4.12. Human vaccine efficacy dosing parameters #2...... 170

Table 4.13. Human vaccine concentration effect parameters...... 171

Table 4.14. Human vaccine concentration effect parameters...... 172

Table 6.1. Common immunogen information...... 249

Table 6.2. Overview of prediction, scoring, and analysis results...... 250

Table 6.3. HLA-DQ epitope ranking and excision results...... 251

xviii

Table C.1. Epitope ranking and excision results for HLR-DR, IAd (NetMHC), IAd

(SMM), and IEd (SMM) predictions...... 280

xix

List of Equations

Equation A.1. Tissue compartment(s) material balance (non-lungs, non-liver, non- brain, non-skin)...... 267

Equation A.2. Lungs compartment material balance...... 267

Equation A.3. Brain compartment material balance...... 267

Equation A.4. Liver compartment material balance...... 268

Equation A.5. Skin compartment material balance...... 268

Equation A.6. Arterial blood compartment material balance...... 268

Equation A.7. Venous blood compartment material balance...... 268

Equation A.8. Tissue compartment(s) material balance (non-lungs, non-liver, non- brain, non-skin)...... 270

Equation A.9. Lungs compartment material balance...... 270

Equation A.10. Brain compartment material balance...... 270

Equation A.11. Liver compartment material balance...... 271

Equation A.12. Skin compartment material balance...... 271

Equation A.13. Arterial blood compartment material balance...... 271

Equation A.14. Venous blood compartment material balance...... 271

Equation A.15. Nicotine-antibody association constant...... 273

Equation A.16. Nicotine-antibody dissociation constant...... 273

Equation A.17. Nicotine-antibody association kinetics...... 273

Equation A.18. Nicotine-antibody dissociation kinetics...... 273

Equation A.19. Cotinine-antibody association constant...... 273

Equation A.20. Cotinine-antibody dissociation constant...... 273

xx

Equation A.21. Cotinine-antibody association kinetics...... 273

Equation A.22. Cotinine-antibody dissociation kinetics...... 274

Equation A.23. Brain nicotine dispersion kinetics...... 274

Equation A.24. Brain nicotine agglomeration kinetics...... 274

Equation A.25. Nicotine hepatic clearance (other)...... 274

Equation A.26. Nicotine hepatic clearance to cotinine...... 274

Equation A.27. Nicotine renal clearance...... 275

Equation A.28. Cotinine hepatic clearance...... 275

Equation A.29. Cotinine renal clearance...... 275

xxi

Attribution

Chapter 3: Kyle Saylor and Frank Gillam initiated the review. All authors contributed to the writing and revision of the manuscript.

Chapter 4: Kyle Saylor carried out the data acquisition, model design and construction, and model output analysis, in addition to drafting the manuscript. Chenming Zhang contributed to project conception, manuscript revision, and final approval of the manuscript contents for submission.

Chapter 5: Kyle Saylor conducted experiments, analyzed results, and drafted the manuscript. Frank Gillam and Alison Waldman conducted experiments and provided assistance with manuscript drafting. Frank Gillam and Chenming Zhang contributed to project conception. Chenming Zhang contributed to manuscript revision, and approved the final manuscript for submission. Kathy Lowe helped with the TEM images.

Chapter 6: Kyle Saylor collected the data, ran predictions, wrote the code for processing the data, processed the data, and drafted the manuscript. Ben Donnan ran predictions, processed the data, and provided assistance with coding. Chenming Zhang contributed to manuscript drafting, revision, and approved the final manuscript for submission. The IEDB Workshop was exceptionally helpful when becoming familiar with the website and prediction methods.

xxii

Chapter 1: Introduction

The use of tobacco products is one of the largest causes of morbidity and mortality in the world today (1). Additionally, it is possible that the direct and indirect economic damages suffered in the United States over the past four decades as result of tobacco smoking is in excess of $7 trillion USD (2). Existing therapies aiming to alleviate nicotine addiction have been largely ineffective, and in some cases, even harmful (3).

As such, the search for safer, more effective therapies has become critical as more and more people are negatively impacted by tobacco use. One of these prospective therapies, vaccinating against nicotine, has yielded promising results in preclinical and clinical trials.

The premise behind nicotine vaccines is simple. Nicotine, the main addictive compound in tobacco products, has been shown to 1) induce rewarding behavior through mesolimbic dopaminergic neuron stimulation and 2) instigate the withdrawal symptoms observed in those trying to quit (4). If the drug’s ability to reach the brain could be impeded, however, the neuropharmacological effects of the drug would be attenuated. This, in turn, would assist those trying to quit. Nicotine vaccines are able to do this by eliciting the production of highly specific, high affinity anti-nicotine antibodies.

These antibodies distribute throughout the blood and extracellular fluid, and when nicotine is introduced to the system, bind the drug to form drug-antibody complexes.

Ultimately, these antibodies soak up nicotine like a sponge, and since the drug-antibody complexes are too large to enter the brain, they attenuate neuropharmacological effects

(5). It is important to note that these antibodies do not eliminate the drug from the body.

1

They simply facilitate a temporary (but long-lived) barrier between nicotine and the brain and a controlled, time-release of the drug after exposure (6).

Many preclinical studies have successfully demonstrated the nicotine vaccine concept in a variety of animal models (7). Both passive administration (when infusions of anti-nicotine antibodies are given to the subject) and active administration (when an immunogen is used to elicit the native production of anti-nicotine antibodies) has been successfully employed (8). The most common application of the approach thus far, however, and by far the most economical, has been active immunization (9). Active immunization can be achieved via many different mechanisms / approaches, though these approaches are limited for therapeutic, nicotine vaccines when compared to prophylactic, traditional vaccines (those that target disease vectors). This is because nicotine vaccines require the incorporation of nicotine haptens (a nicotine-like molecule that can be chemically tethered to an immunogenic scaffolding molecule) with the vaccine.

Out of the many preclinical studies that have evaluated nicotine vaccines, a few have gone on to initiate clinical trials. Though many of these trials showed promise, and at least one was able to make it Phase III clinical trials, they all ultimately resulted in failure and/or discontinuation (9). Additionally, the results from these failed studies were not inconsequential, as they were able correlate high anti-nicotine antibody titers with smoking cessation in humans (10-12). The majority of vaccine recipients in these trials, however, did not produce enough antibodies to show significant benefits from the therapy. Ultimately, these early studies serve as a proof of concept for nicotine vaccine

2 potential in humans. They also show that future nicotine vaccines will need greater population coverage if they hope to surpass their progenitors’ efficacies.

Vaccines, much like biologics and pharmaceuticals, require extensive in vivo assessment in order to fully quantify their prophylactic and/or therapeutic potential.

Efficacy and safety must be well established before these drugs can make it to market.

This process, though very necessary, makes it incredibly expensive and time consuming to fully evaluate vaccine candidates. Additionally, difficulties arise when attempting to translate results from animal studies into potential efficacy in humans due to physiological differences between species (13). As such, the development of new methods for predicting nicotine vaccine efficacy in humans during the earliest stages of in vivo evaluation would be exceptionally valuable. In recent years, in silico modeling has shown promise as a method that could accomplish this task. These models commonly take the form of physiologically-based pharmacokinetic models (PBPK), an approach characterized by the application of species-specific tissue and blood compartments, tissue-to-blood partition coefficients, tissue volumes, blood flow rates, and metabolic activities towards the prediction of drug pharmacokinetics (14). PBPK models can be simple or complex, depending upon the available data and the needs of the designer, and they have already been developed for systems absent of anti-nicotine antibodies (2, 15).

In Chapter 4, two PBPK models, one for rats and one for humans, were constructed and calibrated such that nicotine disposition in the presence of anti-nicotine antibodies could be predicted. These models, which were validated by the good fit between predicted and published values, were able to identify key parameters that

3 adversely impact nicotine efficacy. The models were also able to identify inadequacies in past nicotine vaccines that may have prevented efficacy in humans and confirm that nicotine vaccines could work in humans if certain antibody characteristics were achieved.

Nicotine by itself is too small to be immunogenic. It must be attached to a larger immunogen before it can be recognized by the immune system (7). Additionally, nicotine’s native structure lacks the necessary components to facilitate chemical attachment to other biomolecules (16). Consequently, the first and one of the most important steps when aiming to develop a successful nicotine vaccine is the engineering of an effective nicotine hapten. An “effective” hapten provides easy and efficient options for scaffolding attachment while also presenting optimal post-conjugation structure when considering the immune system’s ability to 1) recognize the hapten and 2) elicit the production of antibodies with high specificity and affinity for the native nicotine molecule (17). While it is always possible that better nicotine haptens could be engineered, extensive time and effort has already been devoted towards the design of the current compendium of nicotine haptens. The current hapten designs already elicit the production of anti-nicotine antibodies with high specificity and affinity (18). As such, most would agree that the onus of developing more effective nicotine vaccines does not fall on the redesigning of the hapten, but rather the reimagining of the scaffolding (the other major piece to the efficacy puzzle).

Proteins, polymeric nanoparticles, liposome nanoparticles, DNA, and various combinations of these materials have all been investigated as potential scaffolds in nicotine vaccine formulations. The cheapest, most common, and most straight-forward

4 of these scaffolds is the protein. The nicotine vaccines that showed promise in past clinical trials were all protein-based, and non-protein nicotine vaccines commonly employ proteins or peptides in some capacity in order to potentiate their effectiveness

(9). Immunologically, effective nicotine vaccines require three events to occur: activation of the innate immune system, helper T cell activation, and B cell activation.

Antibodies with high affinity and specificity cannot be effectively produced if any one of these components are missing from the immunological response to vaccine. The activation of helper T cells is particularly important because their effector functions 1) directly facilitate the activation of other lymphocytes that are critical to adaptive immunity and 2) indirectly assist in the overall immune response via the secretion of various pro-inflammatory cytokines. Antigen presenting cells (APCs), such as dendritic cells and B cells, are the only cells capable of activating helper T cells. Additionally, they can only facilitate this activation once they themselves have been activated by a protein/polypeptide antigen/epitope (13). For these reasons, it is reasonable to assume that proteins will play an important role in the formulation of any future nicotine vaccines that prove to be successful.

In the past, successful conjugate vaccines have typically utilized either 1) toxoid protein or 2) virus-like particle (VLP) protein as the hapten carrier (7). In general, both types of protein, when sufficiently long, provide the benefit of major histocompatibility complex (MHC) class II epitopes, while VLPs have the added benefit of extensive quaternary structure that promotes recognition by APCs (13). The success of these protein types in past conjugate vaccines led researchers to incorporate them into the early nicotine vaccine formulations that showed promise in clinical trials (9). The issue

5 of population coverage needs to be addressed, however, if protein-based nicotine vaccines ever hope to achieve clinical efficacy. To do this, the benefits provided by both toxoid vaccines (epitope content) and VLP vaccines (quaternary structure) should be combined. Additional factors such as hapten coverage and the strategic usance of immunological memory should also be addressed. All of this can be achieved by implementing structural vaccinology approaches, though the economic viability of the process should also be considered when changes are made to product structure.

In Chapter 5, WT and chimeric HPV 16 L1 VLP-based nicotine vaccines are engineered. HPV 16 L1 was chosen as the carrier protein due to its 1) modifiability, 2) sequence and structure, 3) conjugation site density, 4) ubiquitous nature, and 5) proven efficacy and safety in past clinical trials (15, 19-25). The chimeric version of the protein was generated by replacing the last 34 C-terminal residues from the WT protein with a multi-epitope tag. This multi-epitope tag was created by splicing together two experimentally validated universal MHC class II epitopes that were derived from tetanus toxin (aa 947-967) and diphtheria toxin (aa 271-290) with interspacing flexible cleavage sequences (GGVVRGG) (26, 27). The proteins were expressed in E. coli and then optimal purification conditions for quaternary structure retention and economical production were quantified.

The human leucocyte antigen (HLA) gene complex contains many genes that are essential to immunity. The genes that code for all of the MHC molecule classes reside here. One of these classes, MHC class II, plays an especially important role in the activation of humoral immune responses, the type of response that is most critical to the efficacy of nicotine vaccines. Each class is made up of many isotypes, and each isotype

6 is characterized by a vast number of allotypes that appear with varying frequencies throughout the human population (28). These frequencies can be genealogically and demographically linked, but even within related groups, the level of variation between individuals is immense (29). It is highly possible that this genetic variation has played a role in the failure of past nicotine vaccines, as each individual’s response to vaccine is largely dictated by the immune receptors they express (30). The importance of the MHC molecules in cellular and humoral immune responses, coupled with the vast amounts of data associating epitopes with individual MHC phenotypes, has led to a situation where accurate prediction of MHC epitopes using computational methods is now possible. In fact, multiple software packages have been developed that utilize different algorithms and datasets to predict the most likely immunogenic regions within a protein (31). The application of these tools to the nicotine vaccine concept remains to be explored.

In Chapter 6, an in silico approach to designing epitope-based nicotine vaccines that could universally target the human population was investigated. MHC molecules associated with the HLA-DQ and HLA-DR isotypes were selected based on beta allele population frequency data. The goal was to achieve at least 99% population coverage with the selected alleles. IAd and IEd isotypes were also investigated when mining for epitopes that could potentiate vaccine immunogenicity in mice. Common carrier proteins that have been used in past conjugate vaccine formulations were then ‘mined’ for their

MHC class II epitope content using multiple different epitope prediction software packages. The output from these predictions were then processed and analyzed using in-house methods, and universal chimeric antigens were constructed using the

7 processed output. The study was specifically designed in order to accommodate inter- species and inter-method prediction output comparisons.

8

References

1. World Health Organization. WHO report on the global tobacco epidemic, 2013 : enforcing bans on tobacco advertising, promotion and sponsorship.(in IRIS). Geneva:

World Health Organization; 2013. 202 p. p.

2. Saylor K, Zhang C. A simple physiologically based pharmacokinetic model evaluating the effect of anti-nicotine antibodies on nicotine disposition in the brains of rats and humans. Toxicol Appl Pharmacol. 2016;307:150-64.

3. Messer K, Trinidad DR, Al-Delaimy WK, Pierce JP. Smoking cessation rates in the United States: a comparison of young adult and older smokers. Am J Public Health.

2008;98(2):317-22.

4. Corrigall WA, Franklin KB, Coen KM, Clarke PB. The mesolimbic dopaminergic system is implicated in the reinforcing effects of nicotine. Psychopharmacology (Berl).

1992;107(2-3):285-9.

5. Raupach T, Hoogsteder PH, Onno van Schayck CP. Nicotine vaccines to assist with smoking cessation: current status of research. Drugs. 2012;72(4):e1-16.

6. Pentel P, Malin D. A vaccine for nicotine dependence: targeting the drug rather than the brain. Respiration. 2002;69(3):193-7.

7. Pentel PR, LeSage MG. New directions in nicotine vaccine design and use. Adv

Pharmacol. 2014;69:553-80.

8. Pentel PR, Malin DH, Ennifar S, Hieda Y, Keyler DE, Lake JR, et al. A nicotine conjugate vaccine reduces nicotine distribution to brain and attenuates its behavioral and cardiovascular effects in rats. Pharmacol Biochem Behav. 2000;65(1):191-8.

9

9. Hu Y, Zhao Z, Saylor K, Zhang C. Nicotine Vaccines: The Past, the Present, and the Future. Neuroscience of nicotine : mechanisms and treatment. 1st edition. ed.

Cambridge, MA: Elsevier; 2019.

10. Cornuz J, Zwahlen S, Jungi WF, Osterwalder J, Klingler K, van Melle G, et al. A vaccine against nicotine for smoking cessation: a randomized controlled trial. PLoS

One. 2008;3(6):e2547.

11. Esterlis I, Hannestad JO, Perkins E, Bois F, D'Souza DC, Tyndale RF, et al.

Effect of a nicotine vaccine on nicotine binding to beta2*-nicotinic acetylcholine receptors in vivo in human tobacco smokers. Am J Psychiatry. 2013;170(4):399-407.

12. Hatsukami DK, Rennard S, Jorenby D, Fiore M, Koopmeiners J, de Vos A, et al.

Safety and immunogenicity of a nicotine conjugate vaccine in current smokers. Clin

Pharmacol Ther. 2005;78(5):456-67.

13. Saylor K, Gillam F, Lohneis T, Zhang C. Designs of Antigen Structure and

Composition for Improved Protein-Based Vaccine Efficacy. Front Immunol.

2020;11:283.

14. Yang RSH, Andersen ME, Dennison JE, Ou YC, Liao KH, Reisfeld. B.

Physiologically based pharmacokinetic and pharmacodynamic modeling. Mouse models of human cancer. Hoboken, N.J.: Wiley-Liss; 2004. p. p. 391-405.

15. Schwartz SL, Gastonguay MR, Robinson DE, Balter NJ. Physiologically based pharmacokinetic modelling of nicotine. In: Gorrod JW, Wahren J, editors. Nicotine and

Related Alkaloids: Springer Netherlands; 1993. p. 255-74.

10

16. de Villiers SH, Lindblom N, Kalayanov G, Gordon S, Baraznenok I, Malmerfelt A, et al. Nicotine hapten structure, antibody selectivity and effect relationships: results from a nicotine vaccine screening procedure. Vaccine. 2010;28(10):2161-8.

17. Bremer PT, Janda KD. Conjugate Vaccine Immunotherapy for Substance Use

Disorder. Pharmacol Rev. 2017;69(3):298-315.

18. Collins KC, Janda KD. Hapten Design for Anti-addiction Vaccine Development.

2016. In: Biologics to treat substance use disorders : vaccines, monoclonal antibodies, and enzymes [Internet]. Cham: Springer,; [1 online resource ()]. Available from: http://dx.doi.org/10.1007/978-3-319-23150-1 MIT Access Only.

19. Gillam F, Zhang J, Zhang C. Hepatitis B core antigen based novel vaccine against porcine epidemic diarrhea virus. J Virol Methods. 2018;253:61-9.

20. Mohsen MO, Zha L, Cabral-Miranda G, Bachmann MF. Major findings and recent advances in virus-like particle (VLP)-based vaccines. Semin Immunol. 2017;34:123-32.

21. Fraser CC, Altreuter DH, Ilyinskii P, Pittet L, LaMothe RA, Keegan M, et al.

Generation of a universal CD4 memory T cell recall peptide effective in humans, mice and non-human primates. Vaccine. 2014;32(24):2896-903.

22. Jennings GT, Bachmann MF. The coming of age of virus-like particle vaccines.

Biol Chem. 2008;389(5):521-36.

23. Sadeyen JR, Tourne S, Shkreli M, Sizaret PY, Coursaget P. Insertion of a foreign sequence on capsid surface loops of human papillomavirus type 16 virus-like particles reduces their capacity to induce neutralizing antibodies and delineates a conformational neutralizing epitope. Virology. 2003;309(1):32-40.

11

24. Pinto LA, Edwards J, Castle PE, Harro CD, Lowy DR, Schiller JT, et al. Cellular immune responses to human papillomavirus (HPV)-16 L1 in healthy volunteers immunized with recombinant HPV-16 L1 virus-like particles. J Infect Dis.

2003;188(2):327-38.

25. Muller M, Zhou J, Reed TD, Rittmuller C, Burger A, Gabelsberger J, et al.

Chimeric papillomavirus-like particles. Virology. 1997;234(1):93-111.

26. Diethelm-Okita BM, Okita DK, Banaszak L, Conti-Fine BM. Universal epitopes for human CD4+ cells on tetanus and diphtheria toxins. J Infect Dis. 2000;181(3):1001-9.

27. Bromme D, Bonneau PR, Lachance P, Storer AC. Engineering the S2 subsite specificity of human cathepsin S to a cathepsin L- and cathepsin B-like specificity. J Biol

Chem. 1994;269(48):30238-42.

28. Dendrou CA, Petersen J, Rossjohn J, Fugger L. HLA variation and disease. Nat

Rev Immunol. 2018;18(5):325-39.

29. Maiers M, Gragert L, Klitz W. High-resolution HLA alleles and haplotypes in the

United States population. Hum Immunol. 2007;68(9):779-88.

30. Jain A, Pasare C. Innate Control of Adaptive Immunity: Beyond the Three-Signal

Paradigm. J Immunol. 2017;198(10):3791-800.

31. Moore TV, Nishimura MI. Improved MHC II epitope prediction - a step towards personalized medicine. Nat Rev Clin Oncol. 2020;17(2):71-2.

12

Chapter 2: Literature Review: Tobacco Use, Nicotine, and Nicotine Vaccines

2.1. Synopsis of the Recurring Tobacco use Problem

Tobacco use has manifested in nearly every country on Earth. Past studies have estimated that 1.337 billion people worldwide partake and more recent studies report increasing prevalence in some developing nations, though there has been an overall worldwide decline in tobacco use for the past twenty years. Smoking is the most common form of tobacco use, accounting for approximately 82% of all users (1).

Particularly in the US, the Centers for Disease Control and Prevention (CDC) reports that there were approximately 49.1 million tobacco users, of which 83.8% reported using combustibles. Additional CDC data shows that tobacco use is most prevalent in males (25.8% of males and 14.1% of females), a large proportion of those in the

“working” age groups (17.1% of 18-24 YOs, 23.8% of 25-44 YOs, 21.3% of 45-64 YOs,

11.9% of 65+ YOs), in individuals of impoverished background, lesbian/gay/bisexual disposition, and/or those that were disabled or had experienced serious psychological distress at some point in their life (2). The trends apparent in these demographics, which are supported by complimentary studies, show that smoking prevalence correlates with past, present, and perceived future stressors that are largely beyond the control of the user (3, 4).

Studies have indirectly and directly linked tobacco use to a myriad of diseases. It has been associated with infertility in both men and women and birth defects in babies born to women who used while pregnant. Smoking tobacco has been shown to be particularly dangerous, as the act of combusting the chemicals naturally found in tobacco leaves generates a host of additional free radicals and carcinogenic

13 compounds that are not normally present in the plant. After being inhaled, these toxic molecules are absorbed into the body and can cause devastating effects with continued use of the substance (5).

Diseases that are directly associated with smoking include cardiovascular disease, pulmonary diseases, and many cancers. Smoking is linked to approximately 25% of all cardiovascular disease cases (largely comprised of, but not limited to, atheroscleroses, coronary heart diseases, strokes, peripheral arterial diseases, and abdominal aortic aneurysms) and approximately 33% of all cancer cases (largely comprised of, but not limited to, oral, respiratory, gastrointestinal, colorectal, bladder, pancreatic, cervical, and liver cancers). Chronic obstructive pulmonary disease (COPD) is the most common pulmonary disease associated with smoking, causing nearly 80% of all cases. Other pulmonary issues associated with smoking, most likely caused by the poor health of lung tissue and mucosa, include stunted lung growth, increased risk of infections such as tuberculosis, and performance debilitating conditions such as asthma. Finally, tobacco use has been linked to poor overall health and the increased incidence of various infectious and autoimmune diseases due to the cumulative physiological stresses instigated by the practice (6).

Evidence associating tobacco use with health risks has existed for over 60 years. The propaganda machine of Big Tobacco, however, was able to convince the public to continue trying tobacco products by using engineered science to combat negative findings concerning health implications (7). The effects of these propaganda are still visible today; Even with scientific opinion about the harms of tobacco use nearing complete consensus, dissenting perceptions and opinions in young users still exist (6,

14

8). Unfortunately, these dissenting perceptions and opinions have become self- perpetuating with the rise of internet pseudoscience, conspiracy theories, and denialism

(9). Additionally, the effects of this past, junk science on the safety of tobacco use are amplified by the nature of the product, since simply trying the products strongly impels continued use due to tobacco’s addictive nature and stress-relieving properties (10).

Summarily, tobacco use is the leading cause of preventable morbidity and mortality in the world today (11). In particular, smoking tobacco can be directly attributed to the death of approximately 6 million people annually and could lead to 1 billion people dying from smoking-related diseases in the 21st century (12). This astonishing health bill translates into an equally astonishing financial burden, where in the US it is estimated that the direct and indirect costs associated with tobacco smoking could exceed $7 trillion USD over the past four decades (13). Perceptions of the practice have been discombobulated by the rampant mis-information born of the fight between ethical science and the tobacco industry, and this confusion is likely to conditionally continue while the industry remains profitable (14). Discrimination against tobacco users by insurance agencies and/or employers is also worthy of mention when considering motives for the development of therapies that can help people quit (15, 16).

2.2. Nicotine and its Role in Addiction

The global prevalence of tobacco use can largely be attributed to nicotine, a small alkaloid (molecular weight =162.23 Da) that’s produced in plants of the Solanaceae, or

Nightshade, family. The genus Nicotiana, a member of the Solanaceae family, produces the most nicotine within the family by a factor of more than 100,000. Humans have cultivated many of the species within this genus because of their high nicotine content,

15 and collectively these plants have become known as tobacco (17). The cultivation and use of tobacco has occurred for more than 5000 years, starting in the Andes of South

America and slowly spreading throughout the world (18). Today, commercial and subsistent production of tobacco, particularly the species N. tobacum and N. rustica, is widespread. The expansion of tobacco production outside of the Americas has made global access to tobacco products much easier. As such, tobacco use is less limited by logistics when compared with other agricultural commodities (19). For most, inceptive use is merely a choice and continued use is effectuated by addiction (20).

Nicotine is a potent neurotoxin (human LD50 of ~0.8 mg/kg) that works by activating the nicotinic acetylcholine receptors (nAChRs) found on neuronal and non-neuronal cells throughout the body (21, 22). Like many other neurotoxins, nicotine has exploitable pharmacologic properties when applied in appropriate doses. Specifically, nicotine can initiate the domimenergic reward system via the activation of neuronal nAChRs in the mid-brain (23). Depending on how the drug is dosed, this process can happen rapidly or gradually. Dosing via smoking, vaping, and nasal spray occurs quickly because nicotine is absorbed directly into arterial blood via the lungs. On the other hand, lozenges, gum, and patches are subjected to the first-pass effect and as such are characterized by slower pharmacokinetics and lower bioavailability. The rate of nAChR activation is important because more expedient transport to the brain culminates in 1) higher brain nicotine concentrations and 2) minimal feedback lag between dosing and rewarding effects (21). The high brain nicotine concentration achieved during smoking effectively saturates all available nAChRs, resulting in desensitization of existing receptors and the subsequent upregulation of receptor expression. Ultimately, this process establishes the

16 physiologically basis for diminishing returns and withdrawal (24). This effect is also seen when using slower dosing methods, though it is less prevalent due to the brain’s ability to build up tolerance to nicotine during intake. Minimal feedback lag when dosing allows smokers to self-titrate; that is, they are able to dose nicotine such that they immediately experience the pharmacological effect they desire. The immediate satisfaction achieved when using the drug further establishes the reward system and reinforces addiction

(21).

Neurological and physiological effects of tobacco use are generally positive and can be described as pleasurable, relaxing, stimulating, anxiety-relieving, and promoting focus

(25). The effects also alleviate the symptoms of many functional neurological / psychiatric disorders and as such tobacco is commonly thought to be self-medicated

(26). In fact, it would not be wrong to state that tobacco use is self-medicating in-so- much as it is used to alleviate symptoms of nicotine withdrawal. These symptoms can manifest as anxiety, irritability, depression, problems focusing, insomnia, and restlessness (25). If tobacco is being used to self-medicate for pre-existing disorder(s), it is common for those trying to quit to feel more extreme symptoms from these disorders upon ceasing tobacco use (27).

Physiologically, withdrawal symptoms occur when the brain chemistry of a smoker is interrupted by a lack of nicotine. They begin 4-24 hours after quitting, peak at 2-3 days, and can last for over a month (28). Over time, recurring tobacco use conditions the brain to oppose the effects of nicotine by implementing tolerizing, neuroadaptive mechanisms. Eventually, the brain is able to achieve a new equilibrium between the stimulatory (nicotine) and depressing (tolerizing) effects that are present. When this

17 equilibrium is interrupted by the removal of nicotine, however, the new depressing effects are able to dominate brain chemistry and cause withdrawal (29). It is proposed that re-sensitization of upregulated nAChRs, and possibly also dopaminergic receptors, in the absence of nicotine leads to withdrawal. Speculatively, the saturation of nAChRs that occurs when using tobacco insures that 1) nicotine-receptor binding occurs at levels that support the current neurotransmitter equilibrium (i.e. stimulation vs. depression) and 2) complete saturation of nAChRs with nicotine insures continued desensitization of the receptors (30). Psychologically, withdrawal symptoms occur when the brain associates the good feelings achieved through tobacco use or the bad feeling achieved through withdrawal with environmental and/or social cues. This phenomenon is most commonly referred to as conditioned behavior and, unfortunately it 1) cannot be solved pharmacologically and 2) can potentially continue into perpetude after quitting

(31).

In summary of this information, nicotine addiction is characterized by both positive and negative reinforcement and has both physiological and psychological components. This may be why many tobacco users find it so hard to quit, even when they have a strong desire to do so; The CDC reports that in 2015 only 7.4% of smokers were able to quit

(6-12 months cessation) when 68.0% expressed a desire to do so (32). Due to the exceptional addictiveness of nicotine, medications are needed to assist those that want quit but find it hard to do so on their own.

2.3. Pharmaceutical Treatment Options for Nicotine Addiction

The physiological and psychological withdrawal symptoms experienced by those trying to quit tobacco are the most common cause for relapse. The severity of these

18 symptoms can be lessened by the application of pharmaceutical and counseling therapies, respectively (33). However, nearly 68.8% of those trying to quit smoking in

2015 did so without the help of medication or counseling. Of the 32.2% that sought help,

6.8% attended counseling, 29.0% used medication, and 4.7% used medication and attended counseling. Among the 29.0% of smokers that used medication to assist their quit attempts, 16.6% used nicotine patches, 7.9% used varenicline, 12.5% used nicotine gum or lozenges, and 2.4% used nicotine spray or inhaler (32). The success rate for smokers that do not seek help (in the form of counseling or medications) is small, with

7-8% sustaining abstinence (6). Medical and psychological therapies have all been shown to improve the success rate for smokers motivated to quit, though greatly varying levels of success are reported throughout literature. Combination therapies (NRT-other and counseling-other pairs) have proven to be effective, as these approaches have synergizing effects (34-36). Overall, the hierarchy of long-term (12 months) efficacy for individual treatments seems to be nicotine replacement therapies (NRTs; patches, sprays, and lozenges; ~20.8-52.8%) ≅ varenicline (~19.1-32.5%) > bupropion (~23-

29.4%) > counseling (~9.8-14.4%) > no assistance (~7-8%), though multiple studies with (and some without) conflicts of interest conclude that varenicline performs better than NRTs (36-42).

NRTs were the first pharmacotherapeutic approved by the FDA for use as a smoking cessation aid. The concept originally took the form of a gum and was marketed by

GlaxoSmithKline under the name of “Nicorette” in 1984 (43). Today, they come in many forms, including gums, patches, lozenges, sprays, tablets, and vapor inhalers. NRTs work by providing the brain with enough nicotine to reduce withdrawal symptoms. Some

19 withdrawal will always be present when appropriately using NRTs, however, due to the nature of the therapy (the goal being to wean the individual off of nicotine so dosages will be lower) and/or the nature of the dosage (NRTs that aren’t directly absorbed into the arterial blood won’t produce the same rapid and high brain nicotine concentrations that smoking does) (44). Overall, NRTs have proven to be a successful and safe option as a smoking cessation aid, though they still carry some pitfalls. First, they are associated with many minor side effects, such as heart palpitations and chest pain, gastrointestinal complaints, nausea and vomiting, and insomnia. Many route-specific adverse events also exist, such as skin irritations when using the patch, throat soreness, mouth ulcers, and hiccups when using orally administered NRT, and lung inflammation when using vaporizers (44-46). Second, they require frequent administration, a characteristic that drives up costs and causes many users that are less motivated to give up on the therapy (47). Third, NRTs contain nicotine. As such, their misuse can facilitate either addiction (when using too much) or relapse (when using too little, or when the therapy course ends) (48, 49).

Varenicline is the most recent pharmacotherapeutic to have achieved FDA approval. It was developed by Pfizer and reached the market under the name of Chantix in 2006.

Efficacy of the drug as a cessation aid can be attributed to its ability to partially agonize the α4β2 subclass of neuronal nAChRs (50, 51). This subclass is particularly important among the nAChRs implicated in nicotine addiction due to its high affinity for nicotine and high level of expression in the mid-brain (52). By partially agonizing this receptor, varenicline is able to function much like NRTs by stimulating enough dopaminergic neurons to reduce withdrawal symptoms. In the event of a lapse in tobacco use,

20 varenicline also prevents the binding of nicotine to α4β2 nAChRs, further assisting those trying to quit. Varenicline has been associated with many minor side effects, including nausea, insomnia, drowsiness, and constipation (53). Additionally, anecdotal evidence gathered in 2008 was able to convince the FDA that varenicline was associated with severe psychiatric side effects, including depression, hostility, and thinking about or attempting suicide. In their report, the FDA stated that as “review of the data has progressed it has become increasingly likely that the severe changes in mood and behavior may be related to Chantix” (54). The FDA requested a year later that Chantix be labeled with a black box warning notifying users of these serious side effects. Since the application of this label, however, studies evaluating the safety of varenicline

(multiple observation studies, at least one randomized study, and many studies with conflicts of interest) have been unable to establish a connection between the drug and these side effects (55). As a result, the FDA decided that the black warning label was not warranted and had it removed in 2016 (56). In review of all the available information, it appears that varenicline is safer than the original follow-ups suggested. It would be unwise, however, to completely ignore the FDAs early opinions on varenicline in favor of the industry-funded science that exonerated the drug.

The last tobacco cessation aid that will be discussed, bupropion, was originally marketed as an antidepressant in 1985 (57). Like varenicline, bupropion had safety issues early on. It was found that the drug significantly increased the incidence of epileptic seizures when taken in high dosages. Lower doses, however, were found to be pharmacologically effective without triggering the seizures. As a result, the FDA allowed reintroduction of bupropion to the market it 1989 (58). The drug was later discovered to

21 have potential as a tobacco use cessation aid and was marketed as Zyban for that purpose (59). Bupropion works by antagonizing nAChRs and selectively inhibiting norepinephrine-dopamine uptake. This prevents nicotine from binding with certain subtypes of nAChRs (reducing the level of reward achieved upon lapsing) and inhibits norepinephrine-dopamine reuptake after neuronal release (increasing extracellular norepinephrine-dopamine levels, thus mimicking the dopamine-rich, extracellular environment found in the brains of those using nicotine and subsequently reducing withdrawal symptoms), respectively (60). Again, like varenicline (and NRTs, for that matter), bupropion is associated with many side effects. Minor ones include nausea, insomnia, dizziness, headache, malaise, and rash, while major ones include seizures, urticaria, angioedema, and adverse effects leading to death (59). Due to high percentage of major side effects, the FDA required drugs containing bupropion to include a black box warning on its label. This requirement was rescinded in 2016, however, along with the same requirement for varenicline, due to new evidence suggesting that bupropion was safer than earlier studies suggested (61). Just like with varenicline, however, the FDA’s early opinions on bupropion should be considered alongside more recent findings.

2.4. Immunopharmacotherapeutic Treatment Options for Nicotine Addiction

The devastating impact that tobacco use has on the world, the incredible willpower it takes quit, and the failure of existing pharmacotherapies to safely and completely tackle the crisis all justify continued research into new tobacco cessation aids. Intuitively, the addictiveness of tobacco could be mitigated if nicotine were blocked from reaching the brain. An alternative therapeutic option for tobacco cessation, immunopharmacotherapy

22

(or vaccination), uses this concept to alter the pharmacokinetics of nicotine such that its easier for users to quit (62). The mechanism behind this alteration is simple; haptens (or nicotine-like molecules) are attached to the surface of an immunogenic carrier and presented to the immune system (63). In response, the immune system produces antibodies specific for the carrier and the hapten. The hapten-specific antibodies share a specificity for nicotine (having been specifically designed for this purpose) and as such they are able to bind nicotine when the drug is introduced systemically (64). Upon binding, the nicotine-antibody complexes that form are too large to enter the brain.

Consequently, the antibodies effectively function as a barrier between the blood and brain, preventing the majority of nicotine from traversing the blood-brain barrier in the event of a relapse into tobacco use. In addition to dampening nicotine’s ability to enter the brain, the antibodies also act as a sponge by soaking up nicotine in the blood and extracellular fluid. The bound nicotine is released slowly over time, providing enough free nicotine to the system (both in the blood and in the brain) to prevent withdrawal symptoms (13). Unlike existing therapeutics for tobacco cessation, nicotine vaccines are exceptionally safe, highly specific, and require a less assiduous dosing regimen (65).

Additionally, their mechanism of action complements other pharmacotherapies well

(66).

A plethora of studies have successfully evaluated nicotine vaccines in animal models

(67-86). In most of these studies, proteins were used as the immunogenic carrier and hapten structure consisted of a nicotine molecule modified with a flexible linker on its pyrrolidine or pyrodine ring (74). Success of these studies was quantified using both behavioral (self-administration) and immunological (antibody titer) assays and the

23 results were promising enough to spawn multiple clinical studies (87-92). Though all of the early clinical trials ultimately failed, they were able to prove that the concept works in those that respond well to the vaccines (87, 89, 90). In fact, they made it quite clear that the concept would be successful if the binding capacity of nicotine-specific antibodies could be improved (88). The inability of these early nicotine vaccines to achieve overall clinical efficacy, however, illuminated a need for new approaches towards their development. Consequently, research into the three components comprising nicotine vaccines that have the most impact on their efficacy, the adjuvant, the hapten, and the carrier, has continued in recent years (93).

Vaccines require the use of an adjuvant in order to effectively activate the innate arm of the immune system. This activation helps 1) recruit key immune cells, such as antigen presenting cells (APCs), to the site of vaccine administration and 2) create a proinflammatory environment that’s conducive to proper activation of the adaptive immune system (94). Adjuvants are functionally essential to all vaccines, though some vaccine have self-adjuvanting properties. There are three major mechanisms by which adjuvants can achieve activation of the innate immune system; targeting pathogen- associated molecular pattern (PAMP) receptors, targeting damage-associated molecular pattern (DAMP) receptors, or facilitating a depot effect (63). In the past, most nicotine vaccines have typically only employed the use of aluminum salts (62). These salts were the first adjuvant to be discovered and their use today comprises one of the most common adjuvanting approaches for all vaccines (both prophylactic and therapeutic). In terms of mechanisms of action, aluminum salts work through the activation of DAMPs. More specifically, they facilitate lysosomal rupture upon cellular

24 uptake, leading to the release of reactive oxygen species that go on to activate NLR- subset inflammasomes (94). NIC7, the newest protein-based nicotine vaccine to undergo clinical evaluation, has taken a two-pronged, adjuvanting approach by co- administering CpG oligodeoxynucleotide (ODN) with aluminum salts. CpG ODN is

PAMP that works through toll-like receptors (TLRs, specifically TLR9). By using both of these adjuvants, NIC7 can facilitate the activation of the innate immune system via both

PAMP and DAMP receptors (95). There are many different types of pattern recognition receptors (PRRs, the proinflammatory receptors that bind PAMPs and DAMPs) that can be targeted with adjuvanting strategies, including but not limited to the previously mentioned TLRs (10 types) and NLRs (2 types), in addition to C-type lectin receptors

(CLRs, 3 types) and RIG-I-like receptors (RLRs, 3 types) (96). Following in NIC7’s footsteps, recent efforts to produce more effective nicotine vaccines have led to co- administration of a plethora of adjuvants with vaccine, typically a variety of TLR agonists

(97, 98). Additionally, other nicotine vaccines have been incorporating novel adjuvants in an attempt to improve immunogenicity (78, 99). When considering the large number of known adjuvants, the inevitability of more being discovered, the somewhat continuous nature of adjuvant and vaccine dosing (there are an infinite number of levels), and many, many other factors, it seems likely that research into the topic, especially for combination approaches, will continue perpetually.

Haptens are typically synthesized by chemically attaching a functional linker to a drug molecule in such a way that the drug's structure is not significantly altered. The location where the linker is attached to the drug, the length and structure of the linker, and the difficulty of the synthesis (i.e. the number of steps and their time and material costs) are

25 all major considerations when trying to design a hapten that will effectively present a drug molecule to B cell receptors (BCRs) (100). Extensive work has already been done in the search for effective drug haptens (101). Two particularly successful nicotine haptens, CMUNic and 3’-AmNic, have consistently elicited the production of antibodies with high affinities (KD<30 nm) and specificities (<1% cross-reactivity with major nicotine metabolites) for nicotine when used in conjugate vaccine formulations (82). High antibody affinity is typically defined as KD values in the nM range, though it is possible for these values to drop into the pM range (102). Interestingly, it may not be beneficial for anti-nicotine antibodies to achieve this level of affinity. If the strength of association between nicotine and antibody becomes too high, binding could become practically irreversible. In this situation, all of the anti-nicotine antibodies would quickly become permanently saturated with nicotine and the vaccine concept would be rendered useless. For this reason, it's unclear how much more work needs to be done in regards to improving the anti-nicotine antibody affinity. There are still questions that can be answered in regards to hapten design, however, that could improve binding capacity of anti-nicotine antibodies. Some of these include the utilization of deuterated, enantiopure, and/or clustered hapten in order to extend coverage of drug specific antibodies to key neuroactive metabolites, to specifically target the most neuroactive isomers when a drug comes as a racemic mixture, and to improve BCR activation and/or facilitate the generation of antibodies specific for more than two nicotine molecules, respectively (77, 103, 104).

Proof-of-concept work for most nicotine vaccines has utilized blood proteins as carriers.

Specifically, keyhole limpet hemocyanin (KLH), bovine serum albumin (BSA), and

26 ovalbumin (OVA) have been very popular due to their safety and capacity for various conjugation chemistries. After advancing beyond the proof-of-concept stage, studies have typically adopted either 1) toxoid proteins or 2) virus-like particles (VLPs) as the carrier in their vaccine formulations in order to improve vaccine immunogenicity (64). In fact, the carriers used in the nicotine vaccines that made it to clinical trials were all protein-based. NicVAX, the only nicotine vaccine to have made it to phase III clinical trials, was fabricated by attaching the 3’-AmNic nicotine hapten to recombinant

Pseudomonas aeruginosa exoprotein A (rEPA) via a succinyl linker. Though the vaccine ultimately failed, the top 30% of responders showed significantly increased abstinence rates when compared to placebo (89, 90). NicQb, a nicotine vaccine that used the capsid protein of the Qb bacteriophage (a protein that spontaneously forms VLPs) as a carrier and O-succinyl-3’-hydroxy-methyl-(±)-nicotine as a hapten, also showed promising results in phase II clinical trials. Similar to NicVAX, the top 33% of responders were shown to have significantly higher abstinence rates, though this success was not enough to propel the vaccine to the next stage of the clinical assessment process (87,

91). Vaccines that showed less promise in clinical trials include Niccine, a tetanus toxin and nicotine hapten conjugate vaccine, and TA-Nic, a recombinant cholera toxin B subunit (rCTB) and nicotine hapten conjugate. The evaluation of both of these vaccines was halted in phase II clinical trials due to an early lack of efficacy (93). NIC7 is the last protein-based conjugate nicotine vaccine to have undergone clinical trials. The formulation originally consisted of diphtheria toxoid (DT) carrier and 3’-AmNic hapten, though a lack of efficacy in early evaluation led to reformulation to cross-reactive material 197 (CRM197) carrier and 5’-amino-ethoxy-nicotine hapten (84, 105). Pfizer,

27 the sponsor for the NIC7 vaccine, evaluated two formulations (NIC7-001 and NIC7-003) in a phase I clinical trial that ended in 2015 (106). The results from this study could not be found during this literature review and it is unclear whether clinical research on NIC7 is still ongoing.

Many interpretations on why early nicotine vaccines failed have been proposed and the most plausible explanations have spawned a resurgence of research regarding the concept. Research into hapten design has continued, as has been discussed. The most-voiced rationale when concerning vaccine carrier has been that previous protein- hapten conjugate vaccines did not offer enough control over key vaccine characteristics, such as particulate nature, hapten density, and vaccine size. For this reason, lipid, polymeric, and DNA-scaffold nanoparticle-based nicotine vaccines, as well as hybrid methods using more than one of these approaches, have recently been hot topics (73,

76, 107, 108). In terms of progress, the polymeric nanoparticle-based nicotine vaccine

SEL-068 is the furthest along, having completed phase I clinical trials (109). This vaccine applies several next-generation approaches to the nicotine vaccine concept, including toll-like receptor (TLR) targeting, T cell receptor (TCR) targeting, and the structural use of biodegradable polymer scaffolding (110). The trial was completed in

March 2013, though results from the trial have not been published (109).

2.5. An Alternate Strategy for Nicotine Vaccine Assessment

It is likely that over $100 million USD has been invested into the nicotine vaccine concept. The fruits of this investment, other than general knowledge, amount to the two partially successful clinical trials, two failed clinical trials, two clinical trials in limbo, and an abundance of animal studies (111). The partially successful studies, when

28 considering the enormous health and financial impact tobacco has on society, justify continued support of the research. Evidently, however, there is a need for better ways to assess vaccine potential prior to costly clinical trials. As more and more data on the subjects of nicotine vaccination, nicotine metabolism, and mathematical modeling techniques become available in literature, in silico methods are becoming an increasingly relevant approach to nicotine vaccine evaluation outside of living systems.

Simple pharmacokinetic (PK) models have been used in the past to achieve this purpose, with the most common approach being the calculation of theoretical maximum binding capacity of nicotine-specific antibodies. This is achieved by making the assumptions that 1) an antibody can only bind two nicotine molecules, 2) all nicotine- specific antibodies have a molecular weight of 150 kDa (i.e. they are all of the IgG subclass), 3) the antibodies are 100% specific for nicotine, and 4) maximum drug- antibody interaction is possible (i.e. instantaneous mixing within one compartment)

(112, 113). While all PK models require the use of assumptions, this simple model is particularly limiting because it fails to consider the dynamic nature of drug pharmacokinetics, antibody distributions, and drug-antibody interactions (114, 115).

More value would be derived from a model that takes these factors into account, such as a physiologically-based pharmacokinetic (PBPK) model.

PBPK models are typically multi-compartmental constructs that use transport equations built around venous blood, arterial blood, and various tissues compartments, tissue-to- blood partition coefficients, compartment volumes and/or weights (w/ densities), and tissue blood flow rates to predict xenobiotic pharmacokinetics and disposition. Their complexity greatly varies and usually depends on the intended purpose of the model

29

(116). When applied to nicotine vaccine development efforts, PBPK modelling would allow for the evaluation of human-specific factors that are unaccounted for in animal models, such as the impact of cotinine (the major metabolite of nicotine in humans), nicotine elimination rates (hepatic and renal), and physiological differences in tissue volumes, tissue permeabilities, and blood flow rates (117). PBPK modeling has been used in many instances to predict nicotine distribution and disposition in animal and human systems (118-124). As such, similar application of the concept to the development of nicotine vaccines would merely be a step up on the scientific ladder of progress. In direct comparison with the extended animal and human studies that are currently implemented to determine the efficacy of potential nicotine vaccines, an in silico approach to vaccine assessment could effectively save time, money, and resources while also preserving subject life and wellbeing by accurately estimating nicotine kinetics and disposition. Additionally, a nicotine vaccine PBPK model could be used to more accurately establish the antibody characteristic threshold values (i.e. affinity for nicotine, specificity for nicotine, and concentration) that would need to be achieved in vaccinated subjects before clinical effectiveness of potential nicotine vaccines could be realized. Ultimately, this approach to early nicotine vaccine assessment could be used to direct the development of future vaccines.

2.6. Alternate Strategies for Nicotine Vaccine Potentiation – Chimeric Virus-like

Particle Carriers

The rationale behind the development of non-protein vaccines may be unsound. VLPs, a previously explored protein-based vaccine platform that showed much promise in clinical trials, offer many of the same advantages that lipid, polymeric, DNA-scaffold,

30 and hybrid nanoparticle vaccines do, such as a particulate nature and methods for controlling hapten density, adjuvant payload, and MHC class II (helper T cell) epitope content. They are usually easier and cheaper to produce (after the upstream genetic engineering has been completed and purification methods have been established) and they contain all the components necessary to initiate a strong immune response (63).

Summarily, it would be unwise to write off the technology when only one VLP-based nicotine vaccine has been evaluated in clinical trials, it having turned out to be one of the most successful to date (62).

Many alternate explanations can be posited to account for the partial success of the

NicQb vaccine. Issues with the Qb capsid protein, and not the VLP approach, are evident. First, the Qb capsid protein is small at 133 residues in length. Considering its small size, it is unlikely that the protein contains a diverse array of MHC class II epitopes (63, 125). Additionally, there are only 7 lysine residues (5.2% of total), of which an average of only 2.5 can be used for conjugation chemistry. Since the quaternary structure of the assembled VLP consists of 180 subunits (creating particles ~30 nm diameter), only ~450 lysine-linked haptens can be displayed on each VLP (126).

Second, the Qb capsid protein does not seem to be as readily modifiable as other commonly used VLPs (127). Third, the Qb virus is not a human pathogen. As such, it is likely that there has been little to no evolutionary pressure for the human immune system to recognize the Qb capsid protein. This may diminish the overall immunogenicity of the protein in humans when compared to capsid proteins derived from human pathogens (128). Fourth, the Qb virus infects E. coli, a microbe that is a member of the human microbiome. It is plausible to believe that this close proximity to

31 humans, in the absence of a proinflammatory environment, could lead to tolerance

(129). Fifth, if carrier-induced epitopic suppression (CIES) can be avoided, the use of material derived from ubiquitous human pathogens and/or materials already found in widespread vaccine formulations (i.e. VLPs derived from viruses such as hepatitis B virus and human papillomaviruses) could allow for the utilization of existing memory follicular helper CD4+ T cells and subsequently accelerate the humoral immune response to hapten / vaccine (130).

One VLP that could accommodate the application and/or improvement of all these factors is the HPV 16 L1 protein (131). First, this protein is much larger than the Qb capsid protein at 505 residues. Consequently, it should have a much more diverse array of MHC class II epitopes (132, 133). The wild-type (WT) HPV 16 L1 protein also contains 44 lysine residues (8.3% of total, ~60% more than the Qb capsid protein) and the post-assembly quaternary structure of HPV 16 L1 consists of 72 subunits, typically creating a particle of ~55 nm diameter. This diameter can range from 15-60 nm, however, depending upon the expression system (132, 134-139). If a comparable percentage of lysines on HPV 16 L1 VLPs are accessible to conjugation chemistries when compared to the Qb capsid protein, ~1131 haptens could be chemically attached to each HPV 16 L1 VLP. Second, HPV 16 L1 proteins can be readily modified within immunodominant loop regions and at both the N- and C-termini without inhibiting the formation of VLPs (139, 140). Third, HPV 16 is a ubiquitous and occasionally fatal human pathogen. Evolutionarily speaking, this should further insure the immunogenicity of the L1 protein in human subjects (128). Additionally, since HPV 16 is a ubiquitous human pathogen and the HPV 16 L1 protein is used in the two multi-valent HPV

32 vaccines available on the market, the use of HPV 16 L1 as a carrier protein in conjugate nicotine vaccine formulations could accelerate humoral immune responses (130).

Again, CIES would need to be avoided for this to be possible. Since the HPV 16 L1 protein is readily modifiable, changes could be made to it immunodominant loop regions in order to simultaneously 1) hide epitopes that may have cognate antibodies remnant from previous immune responses and 2) insert new amino acid sequences that can accommodate conjugation chemistry.

When developing pharmaceuticals, two of the manufacturer’s primary concerns are always efficacy and marketability. These concerns are naturally interconnected in the case of proteins (whose manufacturing costs are generally dominated by downstream processes), as they are all directly and profoundly impacted by purity (141). As such, companies are always on the lookout for ways to improve purification processes for their products. Various innovations and inventions have been made over the years to improve purification methods. Of these advancements, affinity chromatography and protein tags have arguably been the most impactful. With their application, protein purification processes can be improved to the extent that it only takes a single step to achieve target purity (142). The fine line between clinical trial success and failure, along with the limitations in marketability that accompany high purification costs, creates incentive for scientists to tackle both the efficacy and the purity concerns at once. That is, it is worthwhile to explore protein modification that can simultaneously improve purity and efficacy. This approach to improving protein production outcomes would not have been applicable years ago, as the recombinant expression of proteins was a laborious and expensive process. Today, however, advancements in genetic engineering and

33 expression techniques make this dual approach to improving protein production outcomes a real opportunity.

2.7. Alternate Strategies for Nicotine Vaccine Potentiation – Universal Carriers

Another plausible reason for the failure of past protein-based nicotine vaccines is that they lacked the necessary MHC class II epitope content to facilitate follicular helper T cell activation. One of the most diverse gene systems found in humans, the human leucocyte antigen (HLA) system, largely dictates how our adaptive immune systems respond to pathogens and vaccines (143-146). These molecules activate helper T cells when abnormal extracellular proteins are encountered, processed, and presented on the surface of antigen presenting cells (APCs). Since helper T cells largely facilitate the adaptive immune response, the presence of sufficient MHC class II epitopes within vaccine formulations is critical to their success (147). There are three classical HLA isotypes that express MHC class II molecules; HLA-DR, HLA-DQ, and HLA-DP. Within these HLA isotypes, there are 1092, 1700, and 1824 αβ allele combinations (each receptor is made up of an αβ chain heterodimer), respectively, making the theoretical maximum number of possible haplotypes in excess of 3 billion (148). This immense level of diversity is beneficial from an evolutionary standpoint, as it effectively insures that pathogens cannot go undetected throughout the entire population. Unfortunately, it also insures that adaptive immune responses to protein antigens will be variable in magnitude. Nicotine vaccines, whose modus operandi requires the potentiation of antibody expression, are particularly vulnerable to this variability (149).

Recent years have seen the tools and data needed to overcome this variability become available. The importance of MHCs in transplant rejection has become well established

34

(as well as their roles in many other diseases), and as a consequence there is ample genotype data available for establishing population frequency dynamics (148, 150, 151).

Additionally, more and more data is becoming available on the interactions between specific MHCs and their polypeptide ligands (152). This data has been used to create and train a plethora of bioinformatics tools that can predict how well sections within a protein can bind to MHC molecules. As such, it is now possible to computationally generate concatenated and/or chimeric antigens based only on the portions of their sequence that are predicted to perform the best as MHC epitopes (153, 154). When combined with the population frequency data, these bioinformatics tools make it possible to design universal antigens that can target the overall population and/or portions based on demographic factors (155). It is likely that such computational methods will become a staple in future vaccine designs and formulations, especially when considering the lack of similarity between human and animal HLA genotypes (13).

2.8. Conclusions

Tobacco use is both directly and indirectly devastating public health and the world economy. Existing pharmacotherapeutics have been shown to help some quit, but they are associated with moderate to severe side effects and fail to work in the majority of those who use them. As such, there is a need for more efficacious tobacco cessation aid options. Immunopharmacotherapy, an alternate and complimentary approach to existing therapeutics, has shown promise in past studies. All of the clinical trials that attempted to bring a nicotine vaccine to the market, however, have been unable to do so. The rationale behind the failings of past nicotine vaccines is broad and many have ultimately decided to abandon a protein-based conjugate vaccine approach. A multitude

35 of evidence suggests that more thorough assessment of the nicotine vaccine concept using PBPK modeling and the application of new approaches towards the design of protein-based nicotine vaccines, such as the chimeric HPV 16 L1 VLPs and computationally-conceived universal antigens previously discussed, could finally lead to a clinically successful vaccine.

36

References

1. Commar AP, V.; Tursan d’Espaignet, E.; Ong, L. WHO global report on trends in prevalence of tobacco use 2000-2025. 3rd ed ed. Geneva: World Health Organization;

2019 2019.

2. Creamer MR, Wang TW, Babb S, Cullen KA, Day H, Willis G, et al. Tobacco

Product Use and Cessation Indicators Among Adults - United States, 2018. MMWR

Morb Mortal Wkly Rep. 2019;68(45):1013-9.

3. Torres OV, O'Dell LE. Stress is a principal factor that promotes tobacco use in females. Prog Neuropsychopharmacol Biol Psychiatry. 2016;65:260-8.

4. Slopen N, Kontos EZ, Ryff CD, Ayanian JZ, Albert MA, Williams DR.

Psychosocial stress and cigarette smoking persistence, cessation, and relapse over 9-

10 years: a prospective study of middle-aged adults in the United States. Cancer

Causes Control. 2013;24(10):1849-63.

5. How Tobacco Smoke Causes Disease: The Biology and Behavioral Basis for

Smoking-Attributable Disease: A Report of the Surgeon General. Publications and

Reports of the Surgeon General. Atlanta (GA)2010.

6. Warren GW, Alberg AJ, Kraft AS, Cummings KM. The 2014 Surgeon General's report: "The health consequences of smoking--50 years of progress": a paradigm shift in cancer care. Cancer. 2014;120(13):1914-6.

7. Brandt AM. Inventing conflicts of interest: a history of tobacco industry tactics.

Am J Public Health. 2012;102(1):63-71.

8. United States. Public Health Service. Office of the Surgeon General., National

Center for Chronic Disease Prevention and Health Promotion (U.S.). Office on Smoking

37 and Health. Preventing tobacco use among youth and young adults : we can make the next generation tobacco-free : a report of the Surgeon General. Atlanta, Georgia:

Centers for Disease Control and Prevention, Office on Smoking and Health; 2012. 16 pages p.

9. Diethelm P, McKee M. Denialism: what is it and how should scientists respond?

Eur J Public Health. 2009;19(1):2-4.

10. Zabor EC, Li Y, Thornton LM, Shuman MR, Bulik CM, Lichtenstein P, et al. Initial reactions to tobacco use and risk of future regular use. Nicotine Tob Res.

2013;15(2):509-17.

11. Samet JM. Tobacco smoking: the leading cause of preventable disease worldwide. Thorac Surg Clin. 2013;23(2):103-12.

12. World Health Organization. WHO report on the global tobacco epidemic, 2013 : enforcing bans on tobacco advertising, promotion and sponsorship.(in IRIS). Geneva:

World Health Organization; 2013. 202 p. p.

13. Saylor K, Zhang C. A simple physiologically based pharmacokinetic model evaluating the effect of anti-nicotine antibodies on nicotine disposition in the brains of rats and humans. Toxicol Appl Pharmacol. 2016;307:150-64.

14. Proctor R. Golden holocaust : origins of the cigarette catastrophe and the case for abolition. Berkeley: University of California Press; 2011. x, 737 p., 24 p. of plates p.

15. Mello MM, Rosenthal MB. Wellness programs and lifestyle discrimination--the legal limits. N Engl J Med. 2008;359(2):192-9.

16. Stuber J, Galea S, Link BG. Smoking and the emergence of a stigmatized social status. Soc Sci Med. 2008;67(3):420-30.

38

17. Siegmund B, Leitner E, Pfannhauser W. Determination of the nicotine content of various edible nightshades (Solanaceae) and their products and estimation of the associated dietary nicotine intake. J Agric Food Chem. 1999;47(8):3113-20.

18. Musk AW, de Klerk NH. History of tobacco and health. Respirology.

2003;8(3):286-90.

19. Swinnen JFM. Global supply chains, standards and the poor : how the globalization of food systems and standards affects rural development and poverty.

Wallingford, UK ; Cambridge, MA: CABI; 2007. xv, 322 p. p.

20. Bock G, Goode J. Understanding nicotine and tobacco addiction. Chichester, UK

; Hoboken, NJ: John Wiley & Sons; 2006. viii, 275 p. p.

21. Hukkanen J, Jacob P, 3rd, Benowitz NL. Metabolism and disposition kinetics of nicotine. Pharmacol Rev. 2005;57(1):79-115.

22. Mayer B. How much nicotine kills a human? Tracing back the generally accepted lethal dose to dubious self-experiments in the nineteenth century. Arch Toxicol.

2014;88(1):5-7.

23. De Biasi M, Dani JA. Reward, addiction, withdrawal to nicotine. Annu Rev

Neurosci. 2011;34:105-30.

24. Govind AP, Vezina P, Green WN. Nicotine-induced upregulation of nicotinic receptors: underlying mechanisms and relevance to nicotine addiction. Biochem

Pharmacol. 2009;78(7):756-65.

25. Benowitz NL. Neurobiology of nicotine addiction: implications for smoking cessation treatment. Am J Med. 2008;121(4 Suppl 1):S3-10.

39

26. Nisell M, Nomikos GG, Svensson TH. Nicotine dependence, midbrain dopamine systems and psychiatric disorders. Pharmacol Toxicol. 1995;76(3):157-62.

27. Taylor G, McNeill A, Girling A, Farley A, Lindson-Hawley N, Aveyard P. Change in mental health after smoking cessation: systematic review and meta-analysis. BMJ.

2014;348:g1151.

28. McLaughlin I, Dani JA, De Biasi M. Nicotine withdrawal. Curr Top Behav

Neurosci. 2015;24:99-123.

29. Hughes JR. Effects of abstinence from tobacco: valid symptoms and time course.

Nicotine Tob Res. 2007;9(3):315-27.

30. Benowitz NL. Pharmacology of nicotine: addiction, smoking-induced disease, and therapeutics. Annu Rev Pharmacol Toxicol. 2009;49:57-71.

31. Prochaska JJ, Benowitz NL. Current advances in research in treatment and recovery: Nicotine addiction. Sci Adv. 2019;5(10):eaay9763.

32. Babb S, Malarcher A, Schauer G, Asman K, Jamal A. Quitting Smoking Among

Adults - United States, 2000-2015. MMWR Morb Mortal Wkly Rep. 2017;65(52):1457-

64.

33. Gollust SE, Schroeder SA, Warner KE. Helping smokers quit: understanding the barriers to utilization of smoking cessation services. Milbank Q. 2008;86(4):601-27.

34. Patnode CD, Henderson JT, Thompson JH, Senger CA, Fortmann SP, Whitlock

EP. Behavioral Counseling and Pharmacotherapy Interventions for Tobacco Cessation in Adults, Including Pregnant Women: A Review of Reviews for the U.S. Preventive

Services Task Force. Ann Intern Med. 2015;163(8):608-21.

40

35. Ebbert JO, Hays JT, Hurt RD. Combination pharmacotherapy for stopping smoking: what advantages does it offer? Drugs. 2010;70(6):643-50.

36. Jorenby DE, Leischow SJ, Nides MA, Rennard SI, Johnston JA, Hughes AR, et al. A controlled trial of sustained-release bupropion, a nicotine patch, or both for smoking cessation. N Engl J Med. 1999;340(9):685-91.

37. Swan GE, McAfee T, Curry SJ, Jack LM, Javitz H, Dacey S, et al. Effectiveness of bupropion sustained release for smoking cessation in a health care setting: a randomized trial. Arch Intern Med. 2003;163(19):2337-44.

38. Salisbury-Afshar E. Individual Behavioral Counseling for Smoking Cessation. Am

Fam Physician. 2018;98(1):21-2.

39. Stapleton J, West R, Hajek P, Wheeler J, Vangeli E, Abdi Z, et al. Randomized trial of nicotine replacement therapy (NRT), bupropion and NRT plus bupropion for smoking cessation: effectiveness in clinical practice. Addiction. 2013;108(12):2193-201.

40. Taylor GMJ, Taylor AE, Thomas KH, Jones T, Martin RM, Munafo MR, et al. The effectiveness of varenicline versus nicotine replacement therapy on long-term smoking cessation in primary care: a prospective cohort study of electronic medical records. Int J

Epidemiol. 2017;46(6):1948-57.

41. Baker TB, Piper ME, Stein JH, Smith SS, Bolt DM, Fraser DL, et al. Effects of

Nicotine Patch vs Varenicline vs Combination Nicotine Replacement Therapy on

Smoking Cessation at 26 Weeks: A Randomized Clinical Trial. JAMA. 2016;315(4):371-

9.

41

42. Yilmazel Ucar E, Araz O, Yilmaz N, Akgun M, Meral M, Kaynar H, et al.

Effectiveness of pharmacologic therapies on smoking cessation success: three years results of a smoking cessation clinic. Multidiscip Respir Med. 2014;9(1):9.

43. Oster G, Huse DM, Delea TE, Colditz GA. Cost-effectiveness of nicotine gum as an adjunct to physician's advice against cigarette smoking. JAMA. 1986;256(10):1315-

8.

44. Wadgave U, Nagesh L. Nicotine Replacement Therapy: An Overview. Int J

Health Sci (Qassim). 2016;10(3):425-35.

45. Chaumont M, van de Borne P, Bernard A, Van Muylem A, Deprez G, Ullmo J, et al. Fourth generation e-cigarette vaping induces transient lung inflammation and gas exchange disturbances: results from two randomized clinical trials. Am J Physiol Lung

Cell Mol Physiol. 2019;316(5):L705-L19.

46. Mills EJ, Wu P, Lockhart I, Wilson K, Ebbert JO. Adverse events associated with nicotine replacement therapy (NRT) for smoking cessation. A systematic review and meta-analysis of one hundred and twenty studies involving 177,390 individuals. Tob

Induc Dis. 2010;8:8.

47. Etter JF, Stapleton JA. Nicotine replacement therapy for long-term smoking cessation: a meta-analysis. Tob Control. 2006;15(4):280-5.

48. Medioni J, Berlin I, Mallet A. Increased risk of relapse after stopping nicotine replacement therapies: a mathematical modelling approach. Addiction.

2005;100(2):247-54.

42

49. Hughes JR, Adams EH, Franzon MA, Maguire MK, Guary J. A prospective study of off-label use of, abuse of, and dependence on nicotine inhaler. Tob Control.

2005;14(1):49-54.

50. Jorenby DE, Hays JT, Rigotti NA, Azoulay S, Watsky EJ, Williams KE, et al.

Efficacy of varenicline, an alpha4beta2 nicotinic acetylcholine receptor partial agonist, vs placebo or sustained-release bupropion for smoking cessation: a randomized controlled trial. JAMA. 2006;296(1):56-63.

51. Gonzales D, Rennard SI, Nides M, Oncken C, Azoulay S, Billing CB, et al.

Varenicline, an alpha4beta2 nicotinic acetylcholine receptor partial agonist, vs sustained-release bupropion and placebo for smoking cessation: a randomized controlled trial. JAMA. 2006;296(1):47-55.

52. McGranahan TM, Patzlaff NE, Grady SR, Heinemann SF, Booker TK. alpha4beta2 nicotinic acetylcholine receptors on dopaminergic neurons mediate nicotine reward and anxiety relief. J Neurosci. 2011;31(30):10891-902.

53. Fagerstrom K, Hughes J. Varenicline in the treatment of tobacco dependence.

Neuropsychiatr Dis Treat. 2008;4(2):353-63.

54. (FDA) FaDA. Public Health Advisory: Important Information on Chantix

(varenicline) 2008 [updated 2 November 2017. Available from: https://wayback.archive- it.org/7993/20171115034327/https://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSa fetyInformationforPatientsandProviders/ucm051136.htm.

55. Davies NM, Thomas KH. The Food and Drug Administration and varenicline: should risk communication be improved? Addiction. 2017;112(4):555-8.

43

56. Desai RJ, Good MM, San-Juan-Rodriguez A, Henriksen A, Cunningham F,

Hernandez I, et al. Varenicline and Nicotine Replacement Use Associated With US

Food and Drug Administration Drug Safety Communications. JAMA Netw Open.

2019;2(9):e1910626.

57. Huecker MR, Smiley A, Saadabadi A. Bupropion. StatPearls. Treasure Island

(FL)2020.

58. Davidson J. Seizures and bupropion: a review. J Clin Psychiatry. 1989;50(7):256-

61.

59. Boshier A, Wilton LV, Shakir SA. Evaluation of the safety of bupropion (Zyban) for smoking cessation from experience gained in general practice use in England in

2000. Eur J Clin Pharmacol. 2003;59(10):767-73.

60. Paterson NE, Balfour DJ, Markou A. Chronic bupropion attenuated the anhedonic component of nicotine withdrawal in rats via inhibition of dopamine reuptake in the nucleus accumbens shell. Eur J Neurosci. 2007;25(10):3099-108.

61. Shah D, Shah A, Tan X, Sambamoorthi U. Trends in utilization of smoking cessation agents before and after the passage of FDA boxed warning in the United

States. Drug Alcohol Depend. 2017;177:187-93.

62. Pentel PR, LeSage MG. New directions in nicotine vaccine design and use. Adv

Pharmacol. 2014;69:553-80.

63. Saylor K, Gillam F, Lohneis T, Zhang C. Designs of Antigen Structure and

Composition for Improved Protein-Based Vaccine Efficacy. Front Immunol.

2020;11:283.

44

64. Bremer PT, Janda KD. Conjugate Vaccine Immunotherapy for Substance Use

Disorder. Pharmacol Rev. 2017;69(3):298-315.

65. Orson FM, Kinsey BM, Singh RA, Wu Y, Gardner T, Kosten TR. Substance abuse vaccines. Ann N Y Acad Sci. 2008;1141:257-69.

66. Hoogsteder PH, Kotz D, van Spiegel PI, Viechtbauer W, Brauer R, Kessler PD, et al. The efficacy and safety of a nicotine conjugate vaccine (NicVAX(R)) or placebo co-administered with varenicline (Champix(R)) for smoking cessation: study protocol of a phase IIb, double blind, randomized, placebo controlled trial. BMC Public Health.

2012;12:1052.

67. Cornish KE, Harris AC, LeSage MG, Keyler DE, Burroughs D, Earley C, et al.

Combined active and passive immunization against nicotine: minimizing monoclonal antibody requirements using a target antibody concentration strategy. Int

Immunopharmacol. 2011;11(11):1809-15.

68. de Villiers SH, Cornish KE, Troska AJ, Pravetoni M, Pentel PR. Increased efficacy of a trivalent nicotine vaccine compared to a dose-matched monovalent vaccine when formulated with alum. Vaccine. 2013;31(52):6185-93.

69. de Villiers SH, Lindblom N, Kalayanov G, Gordon S, Baraznenok I, Malmerfelt A, et al. Nicotine hapten structure, antibody selectivity and effect relationships: results from a nicotine vaccine screening procedure. Vaccine. 2010;28(10):2161-8.

70. Hieda Y, Keyler DE, Ennifar S, Fattom A, Pentel PR. Vaccination against nicotine during continued nicotine administration in rats: immunogenicity of the vaccine and effects on nicotine distribution to brain. Int J Immunopharmacol. 2000;22(10):809-19.

45

71. Hieda Y, Keyler DE, Vandevoort JT, Kane JK, Ross CA, Raphael DE, et al.

Active immunization alters the plasma nicotine concentration in rats. J Pharmacol Exp

Ther. 1997;283(3):1076-81.

72. Hieda Y, Keyler DE, VanDeVoort JT, Niedbala RS, Raphael DE, Ross CA, et al.

Immunization of rats reduces nicotine distribution to brain. Psychopharmacology (Berl).

1999;143(2):150-7.

73. Hu Y, Zheng H, Huang W, Zhang C. A novel and efficient nicotine vaccine using nano-lipoplex as a delivery vehicle. Hum Vaccin Immunother. 2014;10(1):64-72.

74. Keyler DE, Hieda Y, St Peter J, Pentel PR. Altered disposition of repeated nicotine doses in rats immunized against nicotine. Nicotine Tob Res. 1999;1(3):241-9.

75. Keyler DE, Roiko SA, Earley CA, Murtaugh MP, Pentel PR. Enhanced immunogenicity of a bivalent nicotine vaccine. Int Immunopharmacol. 2008;8(11):1589-

94.

76. Lockner JW, Ho SO, McCague KC, Chiang SM, Do TQ, Fujii G, et al. Enhancing nicotine vaccine immunogenicity with liposomes. Bioorg Med Chem Lett.

2013;23(4):975-8.

77. Lockner JW, Lively JM, Collins KC, Vendruscolo JC, Azar MR, Janda KD. A conjugate vaccine using enantiopure hapten imparts superior nicotine-binding capacity.

J Med Chem. 2015;58(2):1005-11.

78. Miller KD, Roque R, Clegg CH. Novel Anti-Nicotine Vaccine Using a Trimeric

Coiled-Coil Hapten Carrier. PLoS One. 2014;9(12):e114366.

46

79. Moreno AY, Azar MR, Warren NA, Dickerson TJ, Koob GF, Janda KD. A critical evaluation of a nicotine vaccine within a self-administration behavioral model. Mol

Pharm. 2010;7(2):431-41.

80. Pentel P, Malin D. A vaccine for nicotine dependence: targeting the drug rather than the brain. Respiration. 2002;69(3):193-7.

81. Pentel PR, Malin DH, Ennifar S, Hieda Y, Keyler DE, Lake JR, et al. A nicotine conjugate vaccine reduces nicotine distribution to brain and attenuates its behavioral and cardiovascular effects in rats. Pharmacol Biochem Behav. 2000;65(1):191-8.

82. Pravetoni M, Keyler DE, Pidaparthi RR, Carroll FI, Runyon SP, Murtaugh MP, et al. Structurally distinct nicotine immunogens elicit antibodies with non-overlapping specificities. Biochem Pharmacol. 2012;83(4):543-50.

83. Pravetoni M, Keyler DE, Raleigh MD, Harris AC, Lesage MG, Mattson CK, et al.

Vaccination against nicotine alters the distribution of nicotine delivered via cigarette smoke inhalation to rats. Biochem Pharmacol. 2011;81(9):1164-70.

84. Pryde DC, Jones LH, Gervais DP, Stead DR, Blakemore DC, Selby MD, et al.

Selection of a novel anti-nicotine vaccine: influence of antigen design on antibody function in mice. PLoS One. 2013;8(10):e76557.

85. Satoskar SD, Keyler DE, LeSage MG, Raphael DE, Ross CA, Pentel PR. Tissue- dependent effects of immunization with a nicotine conjugate vaccine on the distribution of nicotine in rats. International Immunopharmacology. 2003;3(7):957-70.

86. Zheng H, Hu Y, Huang W, de Villiers S, Pentel P, Zhang J, et al. Negatively

Charged Carbon Nanohorn Supported Cationic Liposome Nanoparticles: A Novel

47

Delivery Vehicle for Anti-Nicotine Vaccine. J Biomed Nanotechnol. 2015;11(12):2197-

210.

87. Cornuz J, Zwahlen S, Jungi WF, Osterwalder J, Klingler K, van Melle G, et al. A vaccine against nicotine for smoking cessation: a randomized controlled trial. PLoS

One. 2008;3(6):e2547.

88. Esterlis I, Hannestad JO, Perkins E, Bois F, D'Souza DC, Tyndale RF, et al.

Effect of a nicotine vaccine on nicotine binding to beta2*-nicotinic acetylcholine receptors in vivo in human tobacco smokers. Am J Psychiatry. 2013;170(4):399-407.

89. Hatsukami DK, Jorenby DE, Gonzales D, Rigotti NA, Glover ED, Oncken CA, et al. Immunogenicity and smoking-cessation outcomes for a novel nicotine immunotherapeutic. Clin Pharmacol Ther. 2011;89(3):392-9.

90. Hatsukami DK, Rennard S, Jorenby D, Fiore M, Koopmeiners J, de Vos A, et al.

Safety and immunogenicity of a nicotine conjugate vaccine in current smokers. Clin

Pharmacol Ther. 2005;78(5):456-67.

91. Maurer P, Jennings GT, Willers J, Rohner F, Lindman Y, Roubicek K, et al. A therapeutic vaccine for nicotine dependence: preclinical efficacy, and Phase I safety and immunogenicity. Eur J Immunol. 2005;35(7):2031-40.

92. Wagena EJ, de Vos A, Horwith G, van Schayck CP. The immunogenicity and safety of a nicotine vaccine in smokers and nonsmokers: results of a randomized, placebo-controlled phase 1/2 trial. Nicotine Tob Res. 2008;10(1):213-8.

93. Hu Y, Zhao Z, Saylor K, Zhang C. Nicotine Vaccines: The Past, the Present, and the Future. Neuroscience of nicotine : mechanisms and treatment. 1st edition. ed.

Cambridge, MA: Elsevier; 2019.

48

94. Awate S, Babiuk LA, Mutwiri G. Mechanisms of action of adjuvants. Front

Immunol. 2013;4:114.

95. McCluskie MJ, Pryde DC, Gervais DP, Stead DR, Zhang N, Benoit M, et al.

Enhancing immunogenicity of a 3'aminomethylnicotine-DT-conjugate anti-nicotine vaccine with CpG adjuvant in mice and non-human primates. Int Immunopharmacol.

2013;16(1):50-6.

96. Takeuchi O, Akira S. Pattern recognition receptors and inflammation. Cell.

2010;140(6):805-20.

97. Zhao Z, Harris B, Hu Y, Harmon T, Pentel PR, Ehrich M, et al. Rational incorporation of molecular adjuvants into a hybrid nanoparticle-based nicotine vaccine for immunotherapy against nicotine addiction. Biomaterials. 2018;155:165-75.

98. Zhao Z, Hu Y, Harmon T, Pentel P, Ehrich M, Zhang C. Effect of Adjuvant

Release Rate on the Immunogenicity of Nanoparticle-Based Vaccines: A Case Study with a Nanoparticle-Based Nicotine Vaccine. Mol Pharm. 2019;16(6):2766-75.

99. Fraleigh NL, Oliva R, Lewicky JD, Martel AL, Acevedo R, Dagmar GR, et al.

Assessing the immunogenicity and toxicity of the AFPL1-conjugate nicotine vaccine using heterologous and homologous vaccination routes. PLoS One.

2019;14(8):e0221708.

100. Isomura S, Wirsching P, Janda KD. An immunotherapeutic program for the treatment of nicotine addiction: hapten design and synthesis. J Org Chem.

2001;66(12):4115-21.

101. Collins KC, Janda KD. Hapten Design for Anti-addiction Vaccine Development.

2016. In: Biologics to treat substance use disorders : vaccines, monoclonal antibodies,

49 and enzymes [Internet]. Cham: Springer,; [1 online resource ()]. Available from: http://dx.doi.org/10.1007/978-3-319-23150-1 MIT Access Only.

102. Landry JP, Ke Y, Yu GL, Zhu XD. Measuring affinity constants of 1450 monoclonal antibodies to peptide targets with a microarray-based label-free assay platform. J Immunol Methods. 2015;417:86-96.

103. Belz TF, Bremer PT, Zhou B, Ellis B, Eubanks LM, Janda KD. Enhancement of a

Heroin Vaccine Through Hapten Deuteration. J Am Chem Soc. 2020.

104. Collins KC, Janda KD. Investigating hapten clustering as a strategy to enhance vaccines against drugs of abuse. Bioconjug Chem. 2014;25(3):593-600.

105. McCluskie MJ, Thorn J, Gervais DP, Stead DR, Zhang N, Benoit M, et al. Anti- nicotine vaccines: Comparison of adjuvanted CRM197 and Qb-VLP conjugate formulations for immunogenicity and function in non-human primates. Int

Immunopharmacol. 2015;29(2):663-71.

106. A Study to Assess the Safety and Tolerability of Different Doses of PF-05402536 and PF-06413367 in Healthy Adult Smokers. https://ClinicalTrials.gov/show/NCT01672645.

107. Zhao Z, Powers K, Hu Y, Raleigh M, Pentel P, Zhang C. Engineering of a hybrid nanoparticle-based nicotine nanovaccine as a next-generation immunotherapeutic strategy against nicotine addiction: A focus on hapten density. Biomaterials.

2017;123:107-17.

108. Liu X, Xu Y, Yu T, Clifford C, Liu Y, Yan H, et al. A DNA nanostructure platform for directed assembly of synthetic vaccines. Nano Lett. 2012;12(8):4254-9.

50

109. Safety and Pharmacodynamics of SEL-068 Vaccine in Smokers and Non-

Smokers. https://ClinicalTrials.gov/show/NCT01478893.

110. Desai RI, Bergman J. Effects of the Nanoparticle-Based Vaccine, SEL-068, on

Nicotine Discrimination in Squirrel Monkeys. Neuropsychopharmacology.

2015;40(9):2207-16.

111. Pravetoni M. Biologics to treat substance use disorders: Current status and new directions. Hum Vaccin Immunother. 2016;12(12):3005-19.

112. Tonstad S, Heggen E, Giljam H, Lagerback PA, Tonnesen P, Wikingsson LD, et al. Niccine(R), a nicotine vaccine, for relapse prevention: a phase II, randomized, placebo-controlled, multicenter clinical trial. Nicotine Tob Res. 2013;15(9):1492-501.

113. Raupach T, Hoogsteder PH, Onno van Schayck CP. Nicotine vaccines to assist with smoking cessation: current status of research. Drugs. 2012;72(4):e1-16.

114. Basu S, Lien YTK, Vozmediano V, Schlender JF, Eissing T, Schmidt S, et al.

Physiologically Based Pharmacokinetic Modeling of Monoclonal Antibodies in Pediatric

Populations Using PK-Sim. Front Pharmacol. 2020;11:868.

115. Jones H, Rowland-Yeo K. Basic concepts in physiologically based pharmacokinetic modeling in drug discovery and development. CPT Pharmacometrics

Syst Pharmacol. 2013;2:e63.

116. Zhuang X, Lu C. PBPK modeling and simulation in drug research and development. Acta Pharm Sin B. 2016;6(5):430-40.

117. Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Physiological parameter values for physiologically based pharmacokinetic models. Toxicol Ind Health.

1997;13(4):407-84.

51

118. Gabrielsson J, Bondesson U. Constant-rate infusion of nicotine and cotinine. I. A physiological pharmacokinetic analysis of the cotinine disposition, and effects on clearance and distribution in the rat. J Pharmacokinet Biopharm. 1987;15(6):583-99.

119. Gajewska M, Worth A, Urani C, Briesen H, Schramm KW. The acute effects of daily nicotine intake on heart rate--a toxicokinetic and toxicodynamic modelling study.

Regul Toxicol Pharmacol. 2014;70(1):312-24.

120. Plowchalk DR, Andersen ME, deBethizy JD. A physiologically based pharmacokinetic model for nicotine disposition in the Sprague-Dawley rat. Toxicol Appl

Pharmacol. 1992;116(2):177-88.

121. Robinson DE, Balter NJ, Schwartz SL. A physiologically based pharmacokinetic model for nicotine and cotinine in man. J Pharmacokinet Biopharm. 1992;20(6):591-

609.

122. Schwartz SL, Gastonguay MR, Robinson DE, Balter NJ. Physiologically based pharmacokinetic modelling of nicotine. In: Gorrod JW, Wahren J, editors. Nicotine and

Related Alkaloids: Springer Netherlands; 1993. p. 255-74.

123. Teeguarden JG, Housand CJ, Smith JN, Hinderliter PM, Gunawan R, Timchalk

CA. A multi-route model of nicotine-cotinine pharmacokinetics, pharmacodynamics and brain nicotinic acetylcholine receptor binding in humans. Regul Toxicol Pharmacol.

2013;65(1):12-28.

124. Yamazaki H, Horiuchi K, Takano R, Nagano T, Shimizu M, Kitajima M, et al.

Human blood concentrations of cotinine, a biomonitoring marker for tobacco smoke, extrapolated from nicotine metabolism in rats and humans and physiologically based pharmacokinetic modeling. Int J Environ Res Public Health. 2010;7(9):3406-21.

52

125. Kozlovska TM, Cielens I, Dreilinna D, Dislers A, Baumanis V, Ose V, et al.

Recombinant RNA phage Q beta capsid particles synthesized and self-assembled in

Escherichia coli. Gene. 1993;137(1):133-7.

126. Akache B, Weeratna RD, Deora A, Thorn JM, Champion B, Merson JR, et al.

Anti-IgE Qb-VLP Conjugate Vaccine Self-Adjuvants through Activation of TLR7.

Vaccines (Basel). 2016;4(1).

127. Brown SD, Fiedler JD, Finn MG. Assembly of hybrid bacteriophage Qbeta virus- like particles. Biochemistry. 2009;48(47):11155-7.

128. Kaufman J. Evolution and immunity. Immunology. 2010;130(4):459-62.

129. Belkaid Y, Hand TW. Role of the microbiota in immunity and inflammation. Cell.

2014;157(1):121-41.

130. Fraser CC, Altreuter DH, Ilyinskii P, Pittet L, LaMothe RA, Keegan M, et al.

Generation of a universal CD4 memory T cell recall peptide effective in humans, mice and non-human primates. Vaccine. 2014;32(24):2896-903.

131. Saylor K, Waldman A, Gillam F, Zhang C. Multi-epitope insert modulates solubility-based and chromatographic purification of human papilloma virus 16 L1-based vaccine without inhibiting virus-like particle assembly. Journal of Chromatography A.

2020:461567.

132. Zahin M, Joh J, Khanal S, Husk A, Mason H, Warzecha H, et al. Scalable

Production of HPV16 L1 Protein and VLPs from Tobacco Leaves. PLoS One.

2016;11(8):e0160995.

133. Del Rio-Ospina L, Camargo M, Soto-De Leon SC, Sanchez R, Moreno-Perez

DA, Patarroyo ME, et al. Identifying the HLA DRB1-DQB1 molecules and predicting

53 epitopes associated with high-risk HPV infection clearance and redetection. Sci Rep.

2020;10(1):7306.

134. Pineo CB, Hitzeroth, II, Rybicki EP. Immunogenic assessment of plant-produced human papillomavirus type 16 L1/L2 chimaeras. Plant Biotechnol J. 2013;11(8):964-75.

135. Chabeda A, van Zyl AR, Rybicki EP, Hitzeroth, II. Substitution of Human

Papillomavirus Type 16 L2 Neutralizing Epitopes Into L1 Surface Loops: The Effect on

Virus-Like Particle Assembly and Immunogenicity. Front Plant Sci. 2019;10:779.

136. Bazan SB, de Alencar Muniz Chaves A, Aires KA, Cianciarullo AM, Garcea RL,

Ho PL. Expression and characterization of HPV-16 L1 capsid protein in Pichia pastoris.

Arch Virol. 2009;154(10):1609-17.

137. Deschuyteneer M, Elouahabi A, Plainchamp D, Plisnier M, Soete D, Corazza Y, et al. Molecular and structural characterization of the L1 virus-like particles that are used as vaccine antigens in Cervarix, the AS04-adjuvanted HPV-16 and -18 cervical cancer vaccine. Hum Vaccin. 2010;6(5):407-19.

138. Kirnbauer R, Taub J, Greenstone H, Roden R, Durst M, Gissmann L, et al.

Efficient self-assembly of human papillomavirus type 16 L1 and L1-L2 into virus-like particles. J Virol. 1993;67(12):6929-36.

139. Wei M, Wang D, Li Z, Song S, Kong X, Mo X, et al. N-terminal truncations on L1 proteins of human papillomaviruses promote their soluble expression in Escherichia coli and self-assembly in vitro. Emerg Microbes Infect. 2018;7(1):160.

140. Sadeyen JR, Tourne S, Shkreli M, Sizaret PY, Coursaget P. Insertion of a foreign sequence on capsid surface loops of human papillomavirus type 16 virus-like particles

54 reduces their capacity to induce neutralizing antibodies and delineates a conformational neutralizing epitope. Virology. 2003;309(1):32-40.

141. Rudge SR, Ladisch MR. Industrial Challenges of Recombinant Proteins. Adv

Biochem Eng Biotechnol. 2020;171:1-22.

142. Kimple ME, Brill AL, Pasker RL. Overview of affinity tags for protein purification.

Curr Protoc Protein Sci. 2013;73:9 1-9 23.

143. Yao Y, Yang H, Shi L, Liu S, Li C, Chen J, et al. HLA Class II Genes HLA-DRB1,

HLA-DPB1, and HLA-DQB1 Are Associated With the Antibody Response to Inactivated

Japanese Encephalitis Vaccine. Front Immunol. 2019;10:428.

144. Moss AJ, Gaughran FP, Karasu A, Gilbert AS, Mann AJ, Gelder CM, et al.

Correlation between human leukocyte antigen class II alleles and HAI titers detected post-influenza vaccination. PLoS One. 2013;8(8):e71376.

145. Haralambieva IH, Ovsyannikova IG, Pankratz VS, Kennedy RB, Jacobson RM,

Poland GA. The genetic basis for interindividual immune response variation to measles vaccine: new understanding and new vaccine approaches. Expert Rev Vaccines.

2013;12(1):57-70.

146. Gartland AJ, Li S, McNevin J, Tomaras GD, Gottardo R, Janes H, et al. Analysis of HLA A*02 association with vaccine efficacy in the RV144 HIV-1 vaccine trial. J Virol.

2014;88(15):8242-55.

147. Dendrou CA, Petersen J, Rossjohn J, Fugger L. HLA variation and disease. Nat

Rev Immunol. 2018;18(5):325-39.

148. Shiina T, Hosomichi K, Inoko H, Kulski JK. The HLA genomic loci map: expression, interaction, diversity and disease. J Hum Genet. 2009;54(1):15-39.

55

149. Chowell D, Krishna C, Pierini F, Makarov V, Rizvi NA, Kuo F, et al. Evolutionary divergence of HLA class I genotype impacts efficacy of cancer immunotherapy. Nat

Med. 2019;25(11):1715-20.

150. Jones EY, Fugger L, Strominger JL, Siebold C. MHC class II proteins and disease: a structural perspective. Nat Rev Immunol. 2006;6(4):271-82.

151. Gragert L, Madbouly A, Freeman J, Maiers M. Six-locus high resolution HLA haplotype frequencies derived from mixed-resolution DNA typing for the entire US donor registry. Hum Immunol. 2013;74(10):1313-20.

152. Vita R, Mahajan S, Overton JA, Dhanda SK, Martini S, Cantrell JR, et al. The

Immune Epitope Database (IEDB): 2018 update. Nucleic Acids Res. 2019;47(D1):D339-

D43.

153. Fleri W, Paul S, Dhanda SK, Mahajan S, Xu X, Peters B, et al. The Immune

Epitope Database and Analysis Resource in Epitope Discovery and Synthetic Vaccine

Design. Front Immunol. 2017;8:278.

154. Martini S, Nielsen M, Peters B, Sette A. The Immune Epitope Database and

Analysis Resource Program 2003-2018: reflections and outlook. Immunogenetics.

2020;72(1-2):57-76.

155. Saylor KL, Donnan B, Zhang C. In silico design of a universal carrier protein: A potential solution to the variable immunogenicity observed in past vaccines against drugs of abuse. The Journal of Immunology. 2020;204:169.2-.2.

56

Chapter 3: Supplementary Literature Review: Designs of Antigen Structure and

Composition for Improved Protein-based Vaccine Efficacy

Kyle Saylor1, Frank Gillam1,2, Taylor Lohneis1,3, Chenming Zhang1,*

1Department of Biological Systems Engineering, Virginia Tech, Blacksburg, VA

2Process Development Department, Locus Biosciences, Morrisville, NC

3Biopharmaceutical Technology Department, GlaxoSmithKline, Rockville, MD

*Corresponding Author: Chenming (Mike) Zhang

302D, HABB1, 1230 Washington St., S.W.

Blacksburg, VA 24061

Voice: (540)-231-7601

Fax: (540)-231-3199

Email: [email protected]

This manuscript has been published in Frontiers in Immunology: Vaccines and

Molecular Therapeutics, 2020 11:283.

57

Abstract

Today, vaccinologists have come to understand that the hallmark of any protective immune response is the antigen. However, it is not the whole antigen that dictates the immune response, but rather the various parts comprising the whole that are capable of influencing immunogenicity. Protein-based antigens hold particular importance within this structural approach to understanding immunity because, though different molecules can serve as antigens, only proteins are capable of inducing both cellular and humoral immunity. This fact, coupled with the versatility and customizability of proteins when considering vaccine design applications, makes protein-based vaccines (PBVs) one of today’s most promising technologies for artificially inducing immunity. In this review, we follow the development of PBV technologies through time and discuss the antigen- specific receptors that are most critical to any immune response: pattern recognition receptors, B cell receptors, and T cell receptors. Knowledge of these receptors and their ligands has become exceptionally valuable in the field of vaccinology, where today it is possible to make drastic modifications to PBV structure, from primary to quaternary, in order to promote recognition of target epitopes, potentiate vaccine immunogenicity, and prevent antigen-associated complications. Additionally, these modifications have made it possible to control immune responses by modulating stability and targeting PBV to key immune cells. Consequently, careful consideration should be given to protein structure when designing PBVs in the future in order to potentiate PBV efficacy.

Keywords vaccine, immunity, antigen, epitope, modification, vaccine composition, vaccine structure

58

3.1. Introduction

Physicians have been aware of the effects of vaccination for over a millennium, though they had little understanding of the mechanisms through which immunity was achieved until the early 19th century. From the simple and dangerous live-pathogen variolation employed in ancient and medieval times to the recombinant protein and DNA vaccines we use today, vaccine development has followed a path of improving efficacy and safety due to an ever-increasing understanding of these mechanisms. The hallmarks of this understanding have been that an adaptive immune response cannot take place without an antigen and that the most effective antigens tend to be proteins. Ultimately, these two facts have led the field of vaccinology to focus on the research and development of protein-based vaccine (PBVs). As such, we now have a dense literature pool that can be used to recapitulate past and present PBV design concepts and illuminate fundamental structural vaccinology principles in PBV design.

Bacterial toxin vaccines were the first PBVs to be developed. They originally consisted of antitoxin isolated from animals inoculated with small quantities of unmodified toxin, but later it was discovered that active immunization could be safely achieved if toxin was either 1) co-administered with a sufficient amount of antitoxin (partial neutralization) or 2) treated chemically or thermally prior to vaccine administration (denaturation).

Chemically inactivated toxin, dubbed toxoid, would go on to garner widespread celebrity due to its success in World War II and eventually become the primary means for immunization against Diphtheria and Tetanus (1). The development of these toxoid vaccines, along with the gradual acceptance of the side chain theory, the establishment

59 of immunological proteomics, and the application of genetic engineering technologies, paved the way for the emergence of the PBVs we know today (2).

Recombinant vaccines, as their name suggests, are PBVs that are produced using recombinant DNA technology. The first recombinant PBV was produced in yeast in

1986 and targeted the surface antigen of the Hepatitis B virus (HBV) (3). Though it was one of the first recombinant proteins to be approved for use in humans, its discovery was largely overlooked due to perceived limitation in likely impact coupled with the fact that a serum-derived vaccine for HBV already existed (3). However, the success of this virus-like particle-based (VLP) PBV spurred an increased interest in whether the same recombinant strategy could be implemented with other viruses. This idea flourished over time, and today recombinant vaccine approaches have been attempted for nearly every viral structural protein identified. Results from this research have been considerable, with five VLP-based PBVs having been commercialized and dozens having reached clinical trials (4). It is important to note, however, that recombinant PBVs are not limited to VLP formulations, as many bacterial pathogen-derived recombinant PBVs have also been developed (5).

Conjugate vaccines, which consist of carrier protein – subunit complex, were conceived in the late 1980s in order to address a growing need for more effective vaccines against encapsulated bacterial pathogens from the Neisseria, Streptococcus, Staphylococcus,

Haemophilus, and Pseudomonas generas. Before their discovery, vaccine formulations targeting these pathogens singularly consisted of polysaccharide (typically the exposed glycan from encapsulated bacterial surfaces). Although these polysaccharide vaccines were shown to elicit the production of protective antibodies, they proved to be

60 tremendously ineffective at conferring protection in young and immunocompromised individuals and largely failed to elicit immunological memory (6). The limited success of the first subunit polysaccharide vaccines was eventually concomitant to the discovery that polysaccharide vaccines are unable to recruit the assistance of T helper cells and thus rely on T cell-independent activation alone (7). Protein-based, subunit vaccines, in contrast, were found to have all the components necessary to initiate T cell-dependent activation of B cells, a process characterized by a more robust immune response, affinity maturation, and immunological memory (8).

Toxoids have traditionally been used as carrier proteins in conjugate PBV formulations because of their excellent immunogenicity, availability, and simplicity (9). Many of the conjugate PBVs being developed today, however, use recombinantly produced carrier proteins that have been specifically designed to maximize efficacy while simultaneously maintaining a good safety profile (10). The first carrier protein of this type, cross-reactive material 197 (CRM197), was discovered upon the random, mutagenic conversion of glutamic acid to glycine at position 52 of diphtheria toxin (DT, Figure 3.1, A). Though distal to the ADP-ribosyltransferase active site found on the A subunit of DT, this single point mutation on the B subunit was able to completely eliminate DT’s toxicity without negatively impacting its ability to stimulate the immune system (11-13). The discovery of

CRM197 ultimately led to the realization that the inherent toxicity of the antigens typically employed in conjugate PBV formulations could be modulated using precise structural modifications as opposed to broad-based chemical and thermal denaturation.

Thus, the idea of structure-based vaccinology was born and a growing trend in research involving ‘designer vaccines’ began. Since its conception, this concept has been applied

61 to a plethora of pathogenic determinants, specifically toxins. It was observed that the use of cholera toxin B subunit (CTB) in PBV formulations, as opposed to complete toxin, could lead to improved safety profiles with little-to-no decline in overall immunogenicity

(Figure 3.1, B). The improved safety was attributed to the missing A1 domain, the portion of cholera toxin responsible for intracellular activity that leads to disease symptoms (14). A similar discovery was made for tetanus toxin when it was revealed that the heavy chain C fragment (TTc), when used as an immunogen, could confer protection upon toxin challenge in a mouse model without eliciting the neurotoxic effects of its parent protein (Figure 3.1, C) (15). Unfortunately, the use of TTc in modern vaccines may be discouraged by its capacity to bind neurons, though researchers have undertaken structural and conformational approaches to the modulation of this interaction (15, 16). Similar methods to those outlined here have also be employed with other toxins, such as heat-liable enterotoxin (a close relative of cholera toxin) and botulinum toxin (a close relative of tetanus toxin) (17, 18).

Today, interest in ‘designer vaccines’ has been increasingly fueled by advancements in our understanding of the mechanisms behind innate and adaptive immunity, specifically the role of antigen composition in PBV immunogenicity (Figure 3.2, A-C). X-ray crystallography, genetic sequencing, epitope prediction algorithms, and in vivo studies of adjuvant properties have all lead to a better understanding of why some proteins are simply more immunogenic than others. Ultimately, differences in protein structure can result in different capacities for antigen to interact with cells and receptors that are key to triggering an immune response (19). Knowledge of this phenomenon has led to a situation where vaccines are no longer conceptualized based on whole antigen, but

62 rather on immune receptor epitopes/ligands and propensity of an antigen (based on structural motifs) for uptake by antigen presenting cells (APCs). As such, many of today’s PBVs are engineered using structural vaccinology principles and rationally target APCs and the three receptor groups key to any adaptive immune response; the pattern recognition receptors (PRRs), the B cell receptors (BCRs), and the T cell receptors (Figure 3.3, A-C).

3.2. Targeting Pattern Recognition Receptors

3.2.1. PBVs, PRRs, and the Innate Immune System

PRRs are specific for highly conserved molecular signals indicating the presence of cell damage and/or pathogens. They activate the innate arm of the immune system, resulting in the production of pro-inflammatory cytokines and chemokines. This activation assists in the maturation of the adaptive immune response and can be critical to the success or failure of a PBV. There are a myriad of PRRs that have been discovered that recognize either pathogen associated molecular patterns (PAMPs) or damage associated molecular patterns (DAMPs) (20). For the purpose of this review, however, only PRRs that recognize PAMPs will be considered, as these are routinely incorporated in the vaccine design process. Generally, it is the responsibility of the adjuvant in most vaccine formulations to activate the innate immune system. To this effect, researchers usually co-administer free PRR ligand, such as toll-like receptor

(TLR) agonists, lipopolysaccharide (LPS) derivatives, and liposomes with the target antigen when they want to potentiate immunogenicity of a vaccine (21). It is also possible to rationally incorporate PRR ligands with PBVs using various techniques, an

63 approach that imparts a distinct advantage over simple co-formulation in that the immunogen maintains proximity to the antigen.

3.2.2. Rational Incorporation of PTMs as PAMPs

One strategy for the incorporation of PRR agonist is the manipulation of post translational modifications (PTMs) through the selection of the expression host (22).

Since PTM types and sites vary widely between species, they are believed to direct the innate immune response against antigen (23). The PTM we know most about when concerning vaccine immunogenicity is glycosylation. With the exception of E. coli, all expression hosts that are commonly used for recombinant PBV production incorporate glycans on protein surfaces (24). These glycans can serve as ligands for many PRRs, specifically those of the C-type lectin family (CLRs) (25). In fact, glycans have long been known to elicit immune responses to recombinant therapeutic proteins (26-28), and many of the potent adjuvants co-formulated with today’s PBVs are polysaccharides (29-

33). In one instance, glycosylation patterns were even shown to potentiate immune response to subunit vaccine (34). A thorough literature search, however, indicated that the modulation of PBV immunogenicity via the rational selection of the expression host has never been directly assessed. Considering that PTMs are incorporated in vivo on specific, surface-exposed amino acids found on most proteins, it should be possible to exploit expression host glycosylation patterns when designing PBVs. Likewise, the potentiation of post-translational modification through recombinant addition of these residues within protein primary structure where modification is likely to occur might prove to be an effective means of improving PBV immunogenicity. Detrimentally, however, PRR ligands also have the capacity to act as BCR and TCR epitopes (also

64 known as antigenic determinants) and stereometrically crowd antigen surface. This can result in masked immunogenicity of and/or immune responses being redirected away from important epitopes contained within PBVs. A perfect example of this phenomenon was reported by Ansari et al. when they observed improved in vitro viral neutralization and in vivo epitope-specific antibody response to glycoprotein 5 (GP5) of the porcine reproductive and respiratory syndrome virus (PRRSV) when key N-linked glycosylation sites were eliminated from the immunodominant, N-terminal ectodomain (35).

3.2.3. Facilitating PAMP Proximity to PBV Via Fusion

Another way to engineer PBVs that can directly engage a wide variety of PRRs is to covalently fuse PRR agonists to the PBV. This can be accomplished during protein synthesis via the inclusion of the gene for a protein adjuvant within the open reading frame of the target antigen (chimeric approach) or post synthesis by using a number of biochemical techniques (conjugate approach). Mechanistically, fusion of protein-based

PRR ligands to PBVs ensures proximity of adjuvant to antigen, thus increasing the likelihood of antigen-adjuvant co-delivery to key immune cells. Ultimately, this approach to adjuvanting is intended to improve immune response profile and intensity while simultaneously minimizing off-target effects (36). It is important to note, however, that chimeric and conjugate modification of antigen with PRR ligand can have detrimental impacts similar to those mentioned previously for natural PRR ligand incorporation. For example, one study was able to show that the chemical conjugation of imidazoquinoline compound 3M-012, a TLR7/8 agonist, to HIV envelope glycoprotein gp120 results in improved in vitro expression of IFNα by peripheral dendritic cells (DCs) while simultaneously abrogating the binding of critical, broadly neutralizing antibodies (37).

65

3.2.3.1. Chemical conjugation of PAMP to PBV

Chemical conjugation of PRR to PBV surface has been explored by many research groups. Chang et al. attached flagellin to the surface of ovalbumin (OVA) nanoparticles

(NPs) and observed improved TLR5 activation in HeLa cells when compared with OVA

NPs without adjuvant. However, there was no statistical difference when comparing

OVA NPs that were conjugated to flagellin with those that were co-administered. These results indicated that conjugating adjuvant to antigen is a viable means of activating the innate immune system, but failed to implicate an effect of adjuvant proximity to antigen on PRR signaling (38). Alternatively, Kastenmüller et al. found that conjugation of

TLR7/8 agonist to OVA improved DC uptake of antigen and subsequent innate immune activation when compared to co-administered antigen and adjuvant (39). A similar study targeted TLR7 by chemically conjugating the adenine-based adjuvant SA-26E to recombinantly expressed group 2 allergen from the house dust mite. Results indicated that conjugation not only improved innate immune response when compared to co- administered antigen and adjuvant, but also that conjugation was capable of redirecting immune response away from the Th2 cell subtype associated with hypersensitivity (40).

Tighe et al. observed comparable results when conjugation, but not coadministration, of

CpG oligonucleotide (CpG ODN, a TLR9 agonist) with the major short ragweed allergen

Amb a 1 led to polarization of Th1 response in mice and higher IgG antibody titers in rabbits and monkeys (41). Schulke et al. observed enhanced secretion of all evaluated pro-inflammatory cytokines without bias towards the activation of any one T cell subtype when monophosphoryl lipid A (MPLA) was chemically conjugated to OVA and used to stimulate DC / T cell co-cultures (42). Finally, results from two additional studies indicate

66 that adjuvant-mediated activation of the innate immune system is variable between species and that LPS and MPLA, two PRR ligands that signal through the same TLR4 receptor, has differing capacities to activate the innate immune system in vitro (43, 44).

Together, these results indicate that PRR conjugation is a viable means of potentiating innate immune system activation, though it appears that there is still more to understand when considering the exact nature of the resulting immune response.

3.2.3.2. Genetic Fusion of PAMP to PBV

One example of recombinant fusion of PRR with PBVs are flaggelin-fused antigens.

These vaccines have intrinsic self-adjuvanting properties through their activation of

TLR5 (45). In one study assessing this approach, an 11.9x increase in hemagglutination inhibition assay titers was observed when a flagellin-fused hemagglutinin vaccine was compared with commercial influenza PBVs (46). Another promising study reported improved antibody titers (~60% increase), improved viral neutralization titers (~3x increase), and improved Th1 cytokine profile (~2x increase in IFNγ and TNFα expression) when comparing the effectiveness of flagellin-modified and wild-type porcine circovirus type 2 Cap protein (47). Additionally, many other studies evaluating the effectiveness of recombinantly fusing flagellin to PBV have shown promising results, with the majority reporting improved overall immunogenicity and/or survival when compared to antigen administered alone (48-54). Unsurprisingly, results like these have propelled multiple flaggelin-fused PBVs targeting influenza to clinical trials (55-58).

None of these candidates achieved commercialization, however, possibly due to the cytokine storm (a phenomenon that occurs when the immune system uncontrollably releases proinflammatory signals) observed upon vaccine administration in clinical trials

67

(57). Nonetheless, structurally modified flaggelin is still being investigated as a potential

PRR fusion partner for antigen within recombinant PBV formulations (59, 60).

Another example of the PBV-PRR fusion approach to vaccination are recombinant and synthetic lipoprotein-based vaccines (LPBVs). These have become a popular research topic due to their ability to self-adjuvant via TLR2/1-6 heterodimers (61). LPBVs consist of a polypeptide sequence with a PRR-activating, N-terminal diacyl or triacyl lipid attachment (62). Incorporation of target epitopes using chimeric and conjugate approaches have been demonstrated in LPBVs targeting tuberculosis (TB), human papilloma virus, hepatitis C virus, influenza A virus, human immunodeficiency virus

(HIV), and cancer (63-68). Generally, augmentation of immunogenicity has been reported in these studies, though it is important to note that prophylactic efficacy of the

TB vaccine and therapeutic efficacy of the HIV vaccine was not observed (64, 67).

Expression host may be an important factor in LPBV production as fatty acid incorporation in vivo not only varies among species but also results in a mixture of different lipoprotein structures that may have varying immunomodulating properties

(69). This principle, however, has only been demonstrated when comparing the immunogenicity of naturally and synthetically produced LPBVs (70).

3.2.4. Facilitating PAMP Proximity to PBV Via Encapsulation

PRR agonists can also be encapsulated by antigen when working with proteins that self-assemble into organized matrixes, as has been demonstrated by the ‘packaging’ of nucleic acids and proteins within various VLPs (71-76). Post-expression encapsulation can be achieved via simple diffusion through vaccine matrix pores when working with nucleic acids, though assembly and disassembly cycling may be necessary to

68 encapsulate larger particles, such as proteins. The efficiency of this process, at least when considering nucleic acids, is attributed to electrostatic interactions between PBV interior and PRR agonists (77). The natural encapsulation of ssRNA upon expression of recombinant, RNA-containing bacteriophage VLPs in E. coli has been observed, illuminating another mode in which choice of expression host can influence antigen composition (78). Ultimately, encapsulation has been shown to improve the half-life of nucleic acids (by over 9x in some instances), most likely due to the reduced access of endonucleases mediated by vaccine matrix (79, 80). Additionally, as vaccine matrix payload won’t be accessible to PRRs until after cellular uptake of vaccine, this approach ensures that encapsulated PRR agonists are more effectively delivered to endosomal and/or cytosolic PRRs (4). This is especially important for nucleic acid PRR agonists, as all PRRs recognizing nucleic acids are intracellular (81).

3.3. Targeting B Cell Receptors

3.3.1. PBVs and the B Cell

BCRs (membrane bound immunoglobulin-CD79 protein complexes) and the B cell maturation process are used by the adaptive immune system to identify and neutralize linear and conformational epitopes exposed on the surface of antigens. Their activation eventually results in the proliferation of plasma cells and the subsequent secretion of antibodies that are highly specific for their target epitope. This, in turn, confers protection through various mechanisms that eventually result in antigen clearance and further stimulation of both the adaptive and innate immune system (82). Targeting the

BCR in a vaccination strategy can therefore greatly impact the efficacy of a vaccine, especially if antibody-mediated neutralization or sequestration is required.

69

3.3.2. Mechanisms Behind BCR Recognition of PBV

Many structures can serve as BCR epitopes in PBV designs, though there are some limitations associated with BCR complimentary determining region (CDR, the portion of an immunoglobulin or TCR that interacts with antigen) binding (83). To start, initial BCR

CDR binding is largely beyond the control of PBV design. This is because BCR CDR structure, and therefore binding, is randomly determined in pro-B cells via variable, diversity, and joining (VDJ) gene recombination (84). PBV design can influence BCR

CDR structure during subsequent, somatic hypermutation rearrangements through many different mechanisms, though efficient control over these rearrangements is difficult to achieve (85). These mechanisms are mostly epitope-specific and include epitope localization within antigen, competition between epitopes, epitope shape, and epitope size. BCR epitope localization is an important consideration in PBV design because exposed regions on antigen surface will always have the highest probability of becoming antibody immunodominant regions (AIRs) due to increased BCR access (86).

Epitope competition refers to situations in which a PBV has multiple AIRs that compete with each other for BCR recognition. This results in the generation of multiple, polyclonal memory B cell pools that compete with each other for PBV upon boosting, an effect that can be detrimental when attempting to target or avoid specific AIRs (87-89).

Epitope shape refers to the inherent ability of a conformational epitope to fit within the

CDR of a BCR. The sequence of a linear epitope similarly influences BCR recognition and can be included in this category. Prediction software has been developed for the purpose of recognizing both conformational and linear BCR epitopes, indicating that receptors do preferentially interact with epitopes displaying certain quantifiable

70 properties (i.e. patterns in sequence) (90). However, the utility of these predictions when considering PBV design is limited due to small and inconsistent datasets, difficulties predicting antigen 3D structure, and the structural heterogeneity exhibited by some

PBVs (90-93). The last consideration, epitope size, consists of constraints that are somewhat vague. Conceivably, any molecule or portion of a larger antigen that is soluble and large enough to initiate a BCR signaling cascade can be considered a BCR epitope. As such, there is no upper limit on antigen size outside of reasonable physiological constraints and any lower size limitation most likely coincides with major histocompatibility complex (MHC) binding limitations (where 10-20mer polypeptides are necessary). BCR epitope size, on the other hand, is only limited by the diameter of recognition on a BCR paratope (approximately 40 Å). A lower threshold for epitope size must also exist, but where this falls is subjectively based on the extent of non-covalent interaction between paratope and epitope.

3.3.3. Chemical Conjugation of BCR Epitope to PBV

Inherently non-immunogenic molecules that fall below the antigen size threshold but have the capacity to sufficiently interact with antibody paratope can be made immunogenic via covalent attachment to a larger immunogen. These molecules, dubbed haptens, are the premise behind conjugate PBVs. It is conceivable that any molecule could be forced into the role of a BCR epitope using this approach. This rationale is exemplified by conjugate PBVs that have successfully elicited humoral immune responses against glycans, self-antigens, and drugs of abuse (94-96). When designing a conjugate PBV, the most important considerations are choice of protein carrier and choice of hapten. Choice of protein carrier is crucial because it determines

71 the number of potential conjugation sites and where they are located. These properties affect 1) conjugation efficiency, 2) conjugation number, and 3) masking of carrier protein

AIRs. This, in turn, has a profound influence on the ability of conjugate PBVs to redirect the humoral immune response away from carrier protein and towards hapten (97).

Additionally, the number of MHC class II epitopes found within carrier protein primary sequence and their affinity for receptors can potentiate humoral immune response through 1) the targeting of specific human leukocyte antigen (HLA) allotypes within a population and 2) the activation of helper T (TH) cells (19, 98). This applies more to the targeting of TCR, however, and as such will be discussed in future sections. Choice of hapten has been shown to have a sizeable impact on the number and binding characteristics of antibodies elicited by conjugate PBVs (99, 100). This effect is exemplified in nicotine vaccines by the difference in efficacy observed when only the attachment position to nicotine is changed (101). Consequently, hapten design has been extensively investigated for PBV formulations targeting drugs of abuse, as these vaccines require large quantities of high affinity antibodies in order to effectively sequester drug in the blood and extracellular fluid (102, 103).

3.3.4. Genetic Fusion of BCR Epitope to PBV

A similar embodiment to the conjugate PBV is the chimeric, fusion PBV. By definition, the flaggelin and lipoprotein fusion proteins described previously can be considered chimeric PBVs. However, within this context, the term most often refers to immunogens that have been rationally modified with recombinant epitopes. With this approach, instead of chemically attaching epitopes to immunogen surface, genes are recombined such that conformational or linear polypeptide epitopes are inserted within PBV

72 immunodominant regions. These regions should be surface exposed if the intent is to activate a humoral immune response (i.e. AIRs). If epitope insertions can be made within AIRs without negatively influencing protein folding, it ultimately results in B cell responses being redirected away from the previous AIR and towards the introduced epitope. This concept was demonstrated by Gillam et al. when they reported increased antibody titers to recombinantly inserted YSNIGVCK epitope and decreased titers to

HBcAg VLP when evaluating a porcine epidemic diarrhea PBV in mice (104). A maximal insert size exists for all AIRs in which proper protein folding can still be accommodated.

Successful inserts within HBcAg have been reported at greater than 200 residues, whereas inserts within HPV 16 L1 protein, even when at the C-terminus, rarely exceed

60 residues without negatively influencing PBV structure (105, 106). Additionally,

Varsani et al. observed that epitopes of identical length had variable effects on the ability of chimeric HPV 16 L1 PBVs to form VLPs when they were inserted within different AIRs (107). Together, these results indicate that maximum insert size is variable and largely influenced by the properties of the protein, the insert, and the insert location. It also appears that larger epitopes may influence protein structure less than smaller ones in some cases, but this is not the norm (108).

The fusion approach to targeting epitopes is most commonly employed using VLPs as scaffolding, a technology which comprises the assemblage of multiple protein copies all containing the same immunodominant regions. In this way, high epitope densities per antigen can be achieved without sacrificing the intrinsic, immunogenic advantages provided by VLP shape, size, and structure (4). Chimeric VLP PBVs have been successfully developed to target a variety of infections that plague humans and

73 livestock species. Examples include a HPV 16 L1-based vaccine targeting influenza A virus (108), a MS2-based vaccine targeting HIV (109), and a HBcAg-based vaccine targeting porcine reproductive and respiratory syndrome virus (110), though many other studies implementing this technology exist (111). Chimeric, fusion PBVs have also been specially designed to convey cross-protective immunity via the insertion of broadly neutralizing and/or multiple BCR epitopes (112). It is important to note, however, that

VLP-forming proteins are not the only proteins that can support this approach.

Alternatively, toxoid proteins have also been used in chimeric vaccines targeting insert- specific antibody production, though much less frequently than their VLP counterparts as they do not offer many of the same advantages mentioned earlier (113-116).

3.3.5. Carrier Induced Epitopic Suppression

Recent strides in vaccine development have been accompanied by increased interest in conjugate and chimeric BCR epitope PBVs. When combined, they present an outstanding opportunity for vaccinologists to generate vaccines that can target nearly all conceivable chemicals, biologicals, and polypeptides. In addition, they provide a means through which epitope densities can be increased on antigen surface such that BCR cross-linking and subsequent T cell-independent activation of B cells becomes more likely. However, the promise of a ‘vaccine for anything’ provided by conjugate and chimeric PBVs has elicited various complications. Of these complications, carrier- induced epitopic suppression (CIES) is the most noteworthy issue. Both conjugate and chimeric PBVs require the use of existing immunogens as carrier proteins in order to elicit an immune response. This, along with the fact that there are a limited number of suitable immunogens available for use in vaccine formulations due to factors such as

74 immunogenicity, toxicity, stability, and availability, leads to a situation where a select group of ‘preferred’ immunogens are most often used. CIES refers to a phenomenon that occurs when the same immunogen is used in sequential, independent vaccine administrations targeting different epitopes, resulting in the sequestration, elimination, and/or inhibition of vaccine response to target epitope by pre-existing, immunogen specific antibodies and lymphocytes (117). As expected, this can lead to inhibition of hapten-specific lymphocyte recognition of vaccine and an ultimate reduction in vaccine efficacy (118).

Though CIES is likely to always occur to some extent when simply boosting a conjugate or chimeric PBV, prevention of antibody specificity for new vaccines is paramount to the creation of more successful immunizations (119). It has been shown that CIES can be overcome by increasing vaccine dosage and/or including more booster injections within a vaccine regimen. More interesting, however, is the positive effect that increasing hapten density has had on the occurrence of CIES, presumably through the crowding of carrier-specific BCR epitopes (120). Interruption, removal, or blocking of AIRs in chimeric VLP formulations via recombinant and/or chemical means has been shown to improve immunogenicity towards target epitopes and minimize immunogenicity towards carrier, further corroborating this presumption (104, 121, 122). However, chemical modification, such as PEGylation, is largely non-specific, thus resulting in unpredictable outcomes for vaccines. For this reason, it is mainly employed to reduce therapeutic protein antigenicity (123-126). Ideally, immunogen surface should be considered for each individual vaccine in order to simultaneously direct antibody-mediated immune

75 response towards important epitopes and prevent pre-existing antibody recognition of carrier-specific immunodominant regions.

3.4. Targeting T Cell Receptors

3.4.1. PBVs and the T Cell

The most critical component in any adaptive immune response is arguably the T cell, as it serves as a key facilitator of both cell-mediated and humoral immunity. T cells assist with B cell maturation (TH cells), destroy infected and malfunctioning cells (cytotoxic T

(TC) cells), prevent T cell autoreactivity and terminate T cell activity at the end of an immune response (regulatory T (TReg) cells), provide tissue, effector, central, and virtual antigen memory (memory T (TM) cells), in addition to assuming many other roles

(natural killer T (NKT) cells, mucosal associated invariant T (MAIT) cells, and gamma delta T (Tγσ) cells) (127). At the heart of this broad functionality lies the TCR, the associated MHCs (class I and class II), and TCR epitopes. When a PBV is administered, cellular processing of antigen leads to MHC display of small, linear, antigen-derived peptides, also known as TCR epitopes, on cell surface. TCR CDR recognition of these MHC-peptide complexes, in turn, leads to T cell activation and proliferation. More specifically, activation of CD8+ TC cells is facilitated by MHC class I molecules whereas activation of CD4+ TH and TReg cells is facilitated by MHC class II molecules. CD4+ and CD8+ TM cells can be activated via either MHC class I or MHC class II pathway (128, 129). Many antigen-associated factors influence the type and magnitude of T cell response. The most important of these factors include antigen uptake, localization, processing, and T cell epitope content.

3.4.2. The Impact of APC Uptake of PBV on T Cell Activation

76

The first step in the T cell activation process is uptake of antigen by cells that express

MHC molecules. For the purposes of PBVs, this process is generally orchestrated by specialized APCs, though many other cell types can participate in the event of cellular damage or infection. In fact, the MHC class I pathway can be initiated by any nucleated cell. Alternately, activation of the MHC class II pathway can only be achieved after phagocytosis of antigen by APCs (130). For this reason, fusion of PBVs to antibodies and antibody fragments specific for APC surface markers has been extensively investigated as means of potentiating both helper and cytotoxic T cell activation. This approach has routinely demonstrated success, with antibody-mediated, APC targeting generally resulting in a subsequent increase in cellular and humoral mediated immune response (131-139). PBVs have also aimed to rationally target T cell activation through the selective incorporation of ligands specific for specialized receptors found on APCs.

These receptors include PRRs such as integrins and CLRs, MHC class II molecules, and Fcγ receptors, among others. However, most research has focused on CLRs such as DC-SIGN, langerin, and the DECTIN-1 subfamily (140, 141). This approach presents a major advantage over the old method of tethered antibody mediated targeting in that incorporation of glycans specific for CLRs do not require any additional steps when appropriate expression hosts are used.

3.4.3. The Impact of Cellular Localization of PBV Within APCs on T Cell Activation

Antigen localization is one of the major controlling factors behind the type of T cell response initiated by exogenously administered PBVs. MHC class I pathway requires internalization and processing of antigen into short peptide sequences (8-10 amino acids) within the cytoplasm. In contrast, MHC class II pathway requires phagocytosis of

77 antigen by APCs and subsequent lysosomal processing into somewhat longer peptide sequences (15-24 amino acids) (130). As such, attempts have been made in the past to direct PBV to cytoplasm (to target MHC class I pathway) or lysosome (to target MHC class II pathway) using co-administration and fusion protein approaches. Results have been mixed, with some studies observing increases in pathway-specific immune response and others showing no improvement. Targeting lysosomal degradation by tethering antigen to lysosome-associated membrane protein (LAMP) generally has shown no significant increase in antigen-specific antibody population and only marginal increases in CD4+ T cell population (142, 143). Invariant chain (Ii) has also been fused to antigen in an attempt to target lysosome and subsequent MHC class II pathway.

While fusion of full-length Ii to endogenous antigen results in inconsistent activation of

CD4+ T cells, most likely due to the presence of class II associated Ii peptide (CLIP) on the C-terminus, fusion of Ii derivatives accounting for various lengths of N-terminal sequence have been shown to effectively direct antigen to lysosome (144, 145).

Exogenously administered fusion proteins that have been tagged with an Ii portion thought to assist with MHC class II loading (LRMK, Ii-Key) have also been explored, resulting in potentiation of both CD4+ and CD8+ T cell immune response (146-151).

Finally, since APCs are the only cell type capable of processing antigen via the MHC class II pathway, the targeting of APC uptake is also a viable means of targeting CD4+

T cell activation. This approach to controlling antigen localization has generally proven to be non-specific to humoral or cytotoxic pathway, however, as it dually potentiates activation of both CD4+ and CD8+ T cells (131-139).

3.4.4. The Impact of PBV Stability on APC Processing and T Cell Activation

78

APCs, specifically dendritic cells (DCs), are unique in their ability to activate TC cells after exposure to extracellular antigen. This cross-presentation (CP) of antigen is key to the initiation of cellular immunity to many cancers, viruses, and exogenously administered PBVs (152). The majority of CP in APCs is due to antigen stability and slowed lysosomal digestion, though other factors, such as the vacuolar pathway, cell maturation stage, and immunostimulatory environment, do play a role (153-160). PBVs can be engineered to improve stability, as has been demonstrated by the introduction of inter-capsomeric di-sulfide bonds within hepatitis B core antigen (HBcAg) VLPs and

MS2 bacteriophage VLPs (104, 161-163). The effect that this engineered stability has on MHC pathway, however, has not been thoroughly investigated, though it appears that stabilization of antigen promotes CP and the activation of TC cells as destabilization has been shown to reduce cross-presentation efficiency (164). One study conducted by Schliehe et al. was able to observe that recombinant fusion of ubiquitin (Ub) or Ub-like modifier Fat10 to vaccinia virus nucleoprotein (NP) resulted in abrogated protein stability and TC and TM cell response to select immunodominant NP epitopes (157). Another study conducted by Delamarre et al. showed that the slight structural difference between RNase-A and RNase-S (where RNase-S has a peptide bond cleaved between A20 and S21) caused RNase-S to be more susceptible to lysosomal proteolysis in vitro and that this susceptibility ultimately resulted in reduced ability for mice to mount a humoral immune response upon vaccination. The T cell activation upon incubation of RNase-A or RNase-S with splenocytes harvested from

RNase-S vaccinated mice was near zero, however, which seems to indicate that

RNase-S failed to reach APCs upon in vivo administration (165). When considering

79

MHC class II pathway, increased antigen complexity has been shown to slow in vitro

DC processing and decrease in vivo antibody response (166). For example, So et al. observed that the stability of two recombinantly modified hen-egg lysozyme proteins

(one with a deleted di-sulfide bond and the other with a selectively added intramolecular ester bond) was inversely correlated with CD4+ T cell activation and both cytokine and antibody production in a mouse model (167). When paired with evidence suggesting that these observations are not due to differences in T cell epitope content, these results indicate that PBV stability plays a crucial role in managing the adaptive immune system during an immune response (156, 168). Whether or not the extent of PBV processing is orchestrated by the cell with the intent of modulating humoral and cellular immune responses, however, remains to be seen.

3.4.5. The Importance of the T Cell Epitopes in PBV Efficacy and Design

After cellular processing of antigen, presentation of MHC-peptide complex to TCR can’t take place if there are no antigen-derived, T cell epitopes specific for the MHC molecules expressed by the cell, as only MHC-peptide complexes can bind TCRs and activate T cells (128). For this reason, toxoid proteins and toxoid protein derivatives have been used for many decades in the form of adjuvants and fusion proteins to supplement antigens that are lacking in T cell response. Eventually, the improved immunogenicity of these combinations was associated with the density and promiscuity of T cell epitopes contained within the toxoid proteins (169). Since this realization, scientists have searched for “universal” T cell epitopes (UTEs), or rather epitopes that can bind with a large portion of the MHC phenotypes found within the human population

80

(170). This search has been rewarding, with UTEs having been discovered in a plethora of pathogen-associated proteins.

UTEs come in two forms; those that activate TH cells via MHC class II pathway (UhTEs) and those that active TC cells via MHC class I pathway (UcTEs). Synthetic UhTEs, such as the Pan DR Epitope (PADRE) peptides, have also been engineered. Their development has typically been based on the fact that properties of residues at certain positions within an epitope are more important for MHC binding than others (171, 172).

This principle has also led to the construction of data-trained algorithms that can help predict MHC binding preferences (173, 174). As such, it is now possible to dissect an antigen in silico, identifying key sequences that are most likely to serve as epitopes for

MHC molecules. Taking this approach to PBV design now makes it possible to direct the immune system toward either a cell-mediated or humoral response and increase in vivo T cell response to PBV by targeting more MHC phenotypes. Ultimately, the use of these algorithms is paving the way towards vaccine designs that target predominant

HLA allotypes, further broadening the range of subjects that respond to PBVs (175-

178).

Though TCRs aren’t subjected to the same post-activation, somatic hypermutation that

BCRs are, they do undergo VDJ recombination and as such their affinity for MHC- peptide complex should theoretically be random (128). This is not the case for BCRs, however, and considering the similarity between BCR and TCR CDRs, it is reasonable to assume that TCRs, like BCRs, may also preferentially bind polypeptide motifs displaying certain quantifiable properties. This hypothesis was recently supported by

Chowell et al. when they observed that TCR contact residues exhibit a strong bias for

81 hydrophobic amino acids contained within MHC class I epitopes (179). They postulated that this bias was due to the favorable thermodynamics associated with TCR covering hydrophobic residues on MHC-epitope complexes. Parrish et al. also presented evidence that TCRs are germline encoded to have intrinsic specificity for unloaded MHC molecules (largely independent of allele, class, and polypeptide sequence) (180).

Furthermore, the existence of superantigens (SAgs), atypical immunogens that allosterically interact with MHC class II molecules, crosslink the variable region on TCR

β-chain, and have the capacity to activate up to 20000x more T cells than are activated by typical antigens, provides additional evidence for this principle. The binding of SAgs by TCR is largely nonspecific when compared with interactions between TCR and peptide-loaded MHC molecules, indicating that, at least under certain circumstances, structurally distinct TCRs have a propensity to bind specific antigenic motifs (181).

Together, these results strongly suggest that interactions between TCR and MHC- epitope complexes are not random and that there may be opportunities in the future to develop algorithms, like those modeling MHC-epitope interactions, that can predict TCR binding of MHC-epitope complexes.

3.4.6. Targeting TCR Via Recombinant and Conjugate Approaches

Many groups have targeted T cell activation via chimeric or covalent attachment of naturally, synthetically, or computationally derived UTEs to protein and peptide-based vaccines. A recombinantly modified rabbit haemorrhagic disease virus-like particle- based vaccine developed by Jemon et al. that incorporated the universal T cell epitope

PADRE and an MHC I-restricted epitope derived from the HPV 16 E6 protein (aa 48-57) showed promise as an anaphylactic HPV 16 vaccine when it reduced the tumor burden

82 and improved the survival time of HPV tumor-bearing mice (182). In another study,

Percival-Alwyn et al. observed that CD1 mice were able to mount an autoimmune response to self-protein ST2 only once it had been tethered to the A fragment of DT

(DTA) or dual TH epitopes derived from TT. Interestingly, the dual epitope performed better than DTA at eliciting autoimmunity when comparing antibody titer differentials between ST2 and control fusion proteins. When considering the size of the two inserts, this result supports the earlier consideration that inserts have the capacity to redirect immune response away from target epitopes (in this case, antibody responses seem to have been redirected away from ST2 and towards DTA) (183). The epitome of the epitopic approach to vaccination was explored by Wu et al. when their group fused a single BCR epitope derived from epidermal growth factor receptor (EGFR, aa 237-267) to a UTE derived from measles virus fusion protein (MVF, aa 288-302) and saw a sizeable antibody response in a mouse model (184). A computational approach to vaccine design was evaluated by Hurtgen et al. when their group used software to predict the UTEs found within three Coccidioides posadasii antigens, Pep1, Amn1, and

PIb. Using in vitro assays to screen for immunogenicity and MHC II affinity, five epitopes were selected for incorporation within a recombinant epitope-based vaccine (along with murine Ii-Key and spacer sequences) and subsequently evaluated in an HLA-DR4 transgenic mouse model. Recall epitope assay indicated that 4 of the 5 epitopes successfully activated T cells and challenge assay results showed early activation of TH cells, elevated interferon and interleukin expression, and prolonged survival rates in vaccinated mice (185). Other studies have observed similar, positive results when employing this approach to targeting T cell activation (186-189).

83

3.4.7. Limitations of Experimentally and Computationally Determined UTEs

UTEs provide a rational means of potentiating humoral and/or cytotoxic immunogenicity of both peptide and protein-based vaccines. Their incorporation into PBV designs could also provide a means of overcoming the effects T cell competition when common antigens are used as vaccine (190). Experimental discovery of UTEs, however, is made difficult by the fact that the HLA system is one of the most variable gene complexes found in humans. For example, at least 3.82e8 combinations exist for MHC class I- associated alleles, and that number is far greater for MHC class II-associated alleles

(191, 192). Additionally, the HLA complex shows considerable difference between humans and common animal models such as mice (193). As such, computational approaches to predicting UhTEs for specific MHC phenotypes are oftentimes plagued by insufficient amounts of raw data on MHC-epitope interactions, data that are needed in order to establish and validate prediction algorithms. As a result, most methods’ predictions can establish what peptide stretches qualify as binders and non-binders, but for the most part these predictions fall short of the quality needed to become the basis for sizeable PBV decisions (168). Ultimately, however, predictions will improve as more data become available for algorithm training and calibration, making it likely that epitope prediction software will become a common tool in the vaccinology toolbox at some point in future.

3.5. Additional Considerations

3.5.1. PBV Safety Considerations

One of the primary reasons PBVs were explored as an alternative to live attenuated and inactivated vaccines in the past was their improved safety profile. This, however, was

84 before genetic modification of antigen had become feasible. Today, modifications made to antigen structure can influence PBV safety profile both positively and negatively. For example, attempts to improve PBV immunogenicity via the many fusion protein approaches previously described could backfire in the form of adverse events, such as cytokine storm and molecular mimicry (autoimmunity instigated by epitopic similarities between foreign and self-immunogens) (57, 194). On the opposite end of the spectrum, the need for safer carrier proteins in subunit vaccine formulations has led to modifications being made to many bacterial toxins with hopes of finding safer alternatives (11, 14). Ultimately, the extreme variations observed in PBV efficacy and safety when even the smallest changes are made in antigen structure leads to a situation where safety evaluation becomes all-the-more critical throughout the vaccine development process. This is compounded by the fact that vaccines are generally administered to healthy subjects, making adverse events even less acceptable when compared with other medicines being developed to maintain non-maleficence. As such, additional steps should be made beginning with the very first animal studies when assessing the safety and efficacy profiles of modified PBVs outside of the standard establishment of correlates of protection and the full characterization of vaccine formulation.

3.5.2. Animal Model Considerations

PBVs must first prove themselves in numerous animal studies before they can be used in humans. Unfortunately, however, immunological and physiological interspecies differences make it unwise to extrapolate pre-clinical results to human studies even when using the most optimal animal model for the PBV being developed (195-198).

85

Outside of obvious issues with dose scaling, a perfect example of this is illustrated when one considers that the animal models typically used in vaccine research are inbred to the point of isogenecity. This can be considered determinantal when performing vaccine research because some of the most diverse genes in humans, such as those located on the HLA gene complex, code for important immune molecules that have a major impact on the nature and magnitude of immune responses to antigen and vaccine (199). Even when using humanized HLA transgenic (Tg) mice as an animal model in vaccine studies, cytotoxic T cell epitope recognition concordance rates with humans have only been reported at 47% for vaccinia virus (following immunization with full virus) and 68% for HIV (following immunization with peptide) (200, 201). Though the sample sizes for these studies were small, these results demonstrate that both epitope recognition and antigen processing are quite different between animal models and humans. The effectiveness of some adjuvants can also vary widely between species, making it difficult to ascertain the true value of vaccine efficacy studies done in animals. For example, the antibody response to HBcAg adjuvanted with oil-in-water MF59 adjuvant system is nearly 10x more potent in humans than in mice and approximately 4x more potent in humans than in baboons (202). Summarily, the correlative utility of potential animal models when assessing PBV efficacy, especially when investigating the effects of TCR epitopes and adjuvants, should be carefully considered before in vivo testing in order to insure the best translation from animal to human success. Only in this way can we hope to consistently elucidate the impact of PBV modification and formulation on vaccine safety and performance in humans. It is safe to assume, however, that there will always been some level of uncertainty prior to starting in vivo studies.

86

3.5.3. Additional Stability Considerations

As has already been mentioned, conformation and stability play an important role in the immunogenicity of PBVs. This importance, however, goes beyond the ability of APCs to cross-present antigen (Figure 3.4). Recombinant proteins, especially when expressed in lower-order systems and/or as inclusion bodies, are notorious for their inclination to misfold (203). This often results in unstable, heterogenous mixtures of protein-derived particles that may 1) inefficiently present BCR epitopes, 2) denature prior to encountering immune cells, or 3) fail to degrade within endosomal compartments (156).

Many components commonly found in PBV formulations can also negatively affect conformation and stability. Specifically, alum, oil-in-water, and TLR agonist adjuvants have the potential to cause PBV structural changes due to electrostatic, hydrophobic, and/or coordination interactions with amino acid side chains (204). Since these immunostimulatory molecules are regularly necessary to improve PBV efficacy and help prevent peripheral tolerance, an event that occurs when the adaptive immune system encounters antigen in low doses or outside of an inflammatory setting, simple removal is not an option. Protein-protein interactions can also become a problem when PBVs are rich in reactive side chains. Stabilizing agents such as polysorbate 80 and sucrose can be employed to help prevent protein-protein and protein-adjuvant interactions in PBV formulations, but these molecules also have the capacity to detrimentally interact with certain amino acid side chains via glycation and oxidation (205). Protein aggregation, however, does not always have a negative impact on immunogenicity. Aggregation of therapeutic proteins has been shown to augment the formation of anti-drug antibodies in a phenomenon that is likely caused when immune epitopes / ligands become easier to

87 access / process (206). Additionally, it seems that conjugate PBVs may benefit from some level of aggregation. This is because, while the conjugation approach to PBV design insures proper BCR epitope presentation, aggregation has also been shown to improve DC uptake by more than 3x and facilitate DC drainage to lymph nodes (39). Of course, improving PBV immunogenicity via engineered aggregation would not be beneficial when employing a chimeric approach to PBV design as the aggregated vaccine structure would most likely present immunological determinants that are much different from those that were intended. Evidently, many important structural and formulative considerations must be made when designing modified PBVs.

3.5.4. PBV Targeting Strategies for Antigenically Variable Pathogens

The development of PBVs targeting antigenically variable pathogens (AVPs) such as

HIV and influenza is complicated by the effects of epitope masking, antigenic drift (the natural accumulation of mutations within existing immunological determinants), and antigenic shift (the natural recombination of immunological determinants between independent influenza A strains). These events are characterized by mutations within antigenic determinants that make it difficult for an adaptive immune response to hone in on antigen (207). Multiple strategies to overcoming these complications have been investigated. For PBVs targeting influenza, the most common approach is biannual reformulation of multivalent, recombinant hemagglutinin PBV such that the flu strains predicted to have the most impact on public health are targeted (208). Unfortunately, this approach to combating AVPs cannot be applied to HIV due to the severity of complications associated with and the chronic nature of HIV infection. One exceptionally promising vaccination approach to protecting against HIV, however, appears to be the

88 reverse engineering of PBVs based on broadly neutralizing antibody (bNAb) populations either found in seropositive individuals or discovered by using massively parallel mapping techniques (209, 210). The premise behind this approach is simple. During the initial stages of HIV infection, the war between mutating antigen and adaptive immune response results in the generation of bNAbs that have been shown to confer powerful protection upon passive immunization in various animal models (211). Intuitively, it should be possible to use structural vaccinology principles and reverse engineer PBVs that can elicit protective antibodies with similar specificities to previously identified bNAbs using an active immunization approach. Scientists hoping to use this approach to develop an effective, prophylactic HIV PBV have identified five key sites associated with bNAbs and work is underway to construct PBVs that can target these vulnerable sites (212-214).

It is worth noting that any active vaccination attempt against HIV will most likely have to consider more than just BCR epitope presentation in order to be effective. This is because, in addition to masking and mutating BCR epitopes, HIV is notorious for its ability to escape immune recognition via alterations in TCR epitopes, specifically those that are recognized by MHC class I molecules (215). A post-hoc analysis of the most successful HIV vaccine study conducted to date, which used an RV144 pox virus prime, recombinant HIV-1 gp120 (rgp120) boosts, and had an estimated 31% overall protection rate, supports this conclusion. Within the vaccinated population, the HLA

A*02 genotype was a marker for success, with A*02+ individuals showing significantly greater protection than A*02- individuals (54% vs. 3% effective). Other studies have made similar observations (216, 217). These results not only illuminate the importance

89 of making HLA-targeting considerations in the PBV design process, but also indicate how impactful MHC escape mutations can be when using PBVs to protect against AVPs

(even when only one set of HLA-specific epitopes is affected). In this same study there was also a significant increase in vaccine efficacy within the A*02+ population when the presence of a lysine at position 169 in the V2 region of rgp120 was factored in (74% vs.

15% effective) (218). Since this region of the HIV-1 proteome is known as one of the five key bNAb eliciting sites, this result further highlights the challenge associated with targeting AVPs via antibody-mediated approaches (219). Ultimately, it is very possible that the success of future HIV vaccines will be dictated by how well antigenic determinants can be targeted and controlled using PBV formulations.

3.5.5. The Chimeric-Conjugate Approach

Residues commonly used for chemical crosslinking can be recombinantly incorporated onto protein surface in order to increase conjugation capacity of PBV and ultimately improve vaccine performance (chimeric-conjugate approach). In one study evaluating this approach, researchers were able to recombinantly substitute threonine-15 of the

MS2 coat protein with a cysteine residue without influencing folding efficiency (220).

Another study observed that replacement of proline-79 and alanine-80 with the peptide

GGKGG within the immunodominant region of modified HBcAg resulted in increased titers against conjugated M2 and improved survival rates upon viral challenge when compared to WT HBcAg that had been recombinantly fused to the same protein (221).

Similar research using lysine-modified tobacco mosaic virus (TMV) coat protein virus- like particles (VLPs) reported an increase in immunogenicity when compared to co- administered antigen and VLP (222) and another study that used lysine-modified and

90 cysteine-modified TMV VLPs observed that coat protein assembly around RNA scaffold could be modulated by altering the ratio of the two mutant proteins during in vitro assembly (223). Together, these results indicate that a chimeric-conjugate approach could easily be employed to incorporate epitope and/or PAMP ligands with PBV, though more research will need to be done if we wish to fully understand the effect of this method on loading capacity and immunogenicity.

3.5.6. Additional PBV Modification Techniques

Additional PBV modification techniques exist that could be employed to modulate immune responses and target the key immune receptors discussed here. The incorporation of unnatural amino acids (UAAs), such as p-nitrophenylalanine, within

PBV structure has been used to overcome autoimmune tolerance in vaccines against

RANKL and TNF-α (224, 225). Incorporation of the UAA azidohomoalanine have also been used to enable click chemistry on PBV surface (226). As opposed to traditional carbodiimide and maleimide conjugation techniques, click chemistry allows for rapid reaction kinetics, selective ligand attachment, and high yields of successfully utilized attachment sites (227). Sortase-mediated conjugation allows for the highly specific attachment of LPETG(G) tagged molecules to PBV bearing recombinantly inserted multi-glycine stretches, though coupling efficiencies (~30 %) tend to be low when using this technique (228, 229). Polyhistidine (pHis) tags are well known for ability to interact with immobilized metal ions in affinity chromatography applications. At least one research group has attempted to utilize this interaction when conjugating nickel-loaded, tris-nitrilotriacetic acid (tNTA) ligand to norovirus (NoV) VLPs that had been C-terminally modified with pHis tags of various lengths. Results confirmed the utility of the approach

91 as a means of attaching ligand to PBV surface and suggested that the degree of tNTA loading correlated with the number of NoV VLP subunits that contained a pHis tag

(230). It may be unwise to employ this conjugation technique when designing future vaccines, however, as nickel has been established as both an allergen and a carcinogen (231). Protein affinity-tag interactions have also been explored as an alternative means of tethering ligand to PBV (232). In one example of this approach,

Thrane et al. screened multiple chimeric HPV 16 L1 vaccines that recombinantly displayed a biotin acceptor sequence on one of the protein’s surface loops. Insertion did not prevent formation of VLPs in all but one construct, and biotinylation and subsequent attachment of cVLP (HI loop insert) with monovalent streptavidin (mSA)-fused

VAR2CSA ligand (a portion of Plasmodium falciparum erythrocyte membrane protein 1 that can be implicated in malaria pathogenicity) showed significant improvement in antigenicity over free ligand early on in vaccination timeline and similar efficacy in final blood draws when the two PBVs were evaluated in a mouse model (233). Clearly, the vaccinologist toolbox is deep when it comes to ways of tethering ligand to PBV.

3.6. Conclusions

Modified PBVs present an effective means of overcoming many of the limitations encountered by today’s vaccines. First, the addition of PAMP ligands to PBV structure insures that sufficient immune activation signals are co-delivered with vaccine upon administration. PBV modifications of this type have been shown to increase overall vaccine immunogenicity and should reduce the likelihood of initiating immune tolerance.

Second, recombinant and/or chemical modification of PBVs with BCR epitopes can modulate humoral immune response to vaccine. Specifically, this type of modification

92 makes it possible to 1) direct antibody specificity towards select epitopes, 2) design cross-reactive vaccines that can neutralize multiple epitopes, and 3) prevent CIES when the subject has previously been exposed to antigen. Third, the recombinant incorporation of TCR epitopes within PBV structure is useful in that it can 1) allow targeting of specific HLA genotypes / MHC phenotypes, 2) influence the type of cellular immune response that’s initiated, and 3) overcome the effects of T cell antigenic competition. Finally, modifications can be made to PBV primary structure that influence stability and cellular uptake, both important factors when it comes to the magnitude and type of immune response initiated.

Limitations associated with PBVs also exist. These are generally structural in nature and can result in the negative modulation of both safety and efficacy. First, it is possible for modification and/or misfolding of PBV to result in the incorporation of unintended

PAMPs and antigenemic determinants, an effect that could lead to issues with immunotoxicity and/or autoimmunity. Second, the extensive structural variability cognate with PBV production processes and the general weakness of animal models as success correlates also makes it difficult and expensive to quantify just how well PBVs will work prior to direct evaluation in their target species. Along these same lines, issues with protein stability and interactions between vaccine species when making modifications to PBV formulations can add an additional confounding element to the design process, as the margins between protective immunity and immune tolerance are often very fine. Lastly, AVPs present an additional challenge when concerning the utility of PBVs due to the difficulty of targeting ever-changing, antigenic determinants.

93

From prophylactic use aimed at preventing hard-to-treat diseases such as HIV to anaphylactic use aimed at alleviating chronic conditions such as addiction, allergies, and autoimmunity, the utility of PBVs has advanced well beyond what was previously imagined possible. Here, we have described the mechanisms behind PBV immunogenicity and listed many structural modifications that have been explored in the past as a means of modulating and/or potentiating PBV in vivo effects. The results from these studies have shown promise on a case-by-case basis, but thus far we have yet to realize the dream of a silver bullet, modified, PBV vaccination approach that would allow the rational targeting of any epitope in any species with a high response rate and without complications. It is possible that we never will, due to the complexity of the immune system and the sheer number of interacting variables that influence the outcome of each immune response. Ultimately, however, the state of PBV research and the sizeable impact that even case-by-case breakthroughs have on the veterinary and medical worlds more than justify continued research into structural vaccinology and the design of modified PBVs.

Funding

This work was partially supported by the National Institute on Drug Abuse

(U01DA036850) and the American Association of Immunologists Careers in

Immunology Fellowship Program, both of which provided support for KS. The authors declare that this study received funding from Smithfield Foods and Murphy-Brown LLC, which provided support for FG. The funder was not involved in the study design, collection, analysis, interpretation of data, the writing of this article or the decision to submit it for publication. The department of Biological Systems Engineering, which

94 provided support for TL. Funds were also received from Virginia Tech Open Access

Subvention Fund to cover the publication fee.

Conflicts of Interest

FG is employed by Locus Biosciences. TL is employed by GlaxoSmithKline. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

95

Tables and Figures

Figure 3.1. Recombinant toxins.

(A) Diphtheria toxin (DT), when replacing glycine with glutamic acid at position 52, loses its toxicity without affecting its antigenicity. The highlighted residues (red) indicate the exact residue (sphere) and area (licorice) where this substitution would occur on monomeric DT. (B) Cholera toxin (CT) is composed of six subunits; one A subunit and five B subunits. B subunit (monomer in red, remaining subunits in pink), which lacks the toxicity of its partner A subunit, has proven to be extremely immunogenic and is used as a carrier protein and adjuvant. B subunit of heat-labile enterotoxin, which shares much of the same homology as B subunit of cholera toxin, has been similarly investigated

(17). (C) Tetanus toxin (TT) is comprised of two chains, a light chain and a heavy chain, of which the light chain is responsible for the protein's toxicity. In the past, proteolytic digestion of TT with papain yielded two fragments, a light chain-containing, toxic B fragment and a non-toxic C fragment (red). Vaccination with the non-toxic C fragment was found to be protective against lethal toxin dose in a mouse model, and today PBVs comprised of recombinant C fragment are being investigated as a potential replacement for TT vaccines. Botulinum toxin, which shares much of the same homology as tetanus toxin, has been similarly investigated (234). The 3D protein structures for DT, CT, and

TT used in this image were rendered in PyMOL 2.3.0 and accessed via the Protein Data

Bank (235-239).

96

97

Figure 3.2. Immunological mechanisms of recombinant, protein-based vaccination.

(A) PBV structure, as illustrated here for the model protein hepatitis B core antigen

(HBcAg, 183aa long, non-truncated form, Accession number P03146), is ultimately determined by primary sequence. Vaccine can comprise monomeric antigen (i.e. toxoid protein) or multimeric antigen (i.e. virus-like particles), though multimeric antigen is used for demonstration purposes here. T cell and B cell antigenic determinants can be identified in primary sequence using various in vitro and in silico methods. The linear

MHC epitopes illustrated here were predicted using Epitope Analysis Resources on the

Immune Epitope Database (IEDB) website. More specifically, MHC epitopes were predicted for HLA-A*02:01 (class I molecules) and all heterodimer combinations of

DQB1*02:01 (class II molecules) using IEDB recommended methods. Linear B cell epitopes, on the other hand, were assigned using frequency analysis results from the

IEDB website (240). PBV structures are color coded to represent epitope content. (B)

Cell processing and activation in response to PBV is generally orchestrated by antigen presenting cells, of which the most important are dendritic cells. APCs sample their environment via endocytosis, specifically via receptor-mediated endocytosis when antigen presents glycan and/or protein pathogen-associated molecular patterns such as high mannose glycans or bacterial flagellin. Depending on the structural and compositional characteristics of the antigen, APCs will either process antigen via MHC class II pathway or MHC class I pathway using a mechanism known as cross- presentation, respectively resulting in activation of either CD4+ (helper) or CD8+

(cytotoxic) T cell response. CD4+ T cell activation requires co-activating signals and

98 results in the proliferation of effector and memory CD4+ T cell pools. Effector CD4+ T cells go on to assist with the activation of B cells (T cell dependent activation) and provide survival signals to activated CD8+ T cells, whereas CD8+ T cells have immediate effector functionality. B cells can also undergo T cell independent activation when antigen cross-links multiple BCRs on B cells surface (TI-2 activation) or co-signals via PRR (TI-1 activation) (82, 128). (C) Activation results in the proliferation of memory and effector cytotoxic T cell and B cell pools. Memory CD8+ cells remain dormant until they encounter cells presenting MHC class I molecules loaded with cognate epitope, upon which they begin mounting an effector response. Effector CD8+ T cells go on to instruct apoptosis in cells presenting MHC class I molecules loaded with cognate epitope. B cells activated via T cell independent pathway generally proliferate into short- lived plasmablasts that express low affinity IgM antibodies (not shown). B cells activated via T cell dependent pathway, on the other hand, result in the proliferation of memory B cells and long-lived plasma cells expressing high-affinity IgA, IgE, or IgG antibodies.

Antibodies secreted by plasms cells (and plasmablasts) go on to bind vaccine and pathogen and initiate antibody effector functions. Memory B cells remain dormant until they encounter antigen presenting cognate epitope, upon which they rapidly proliferate and clones either class switch to become antibody secreting plasma cells or re-enter germinal centers and restart affinity maturation processes (82, 128). The 3D protein structure for HBcAg used in this image was rendered in PyMOL 2.3.0 and accessed via the Protein Data Bank (235, 239, 241).

99

100

Figure 3.3. PBV modification principles.

(A) Potential fusion modifications sites for the model protein hepatitis B core antigen

(HBcAg, 149aa long, truncated form of AN P03146 used for vaccine purposes) as represented by primary structure. Residues are color-coded in gray scale, with darker residues indicating more exposed insertion locations. Polypeptide termini and random coil loop regions are primary targets for PBV fusion modification, as they generally have the least effect on protein structure. Within these considerations, surface regions that do not participate in intra- and intermolecular interactions are preferred. Specifically, most

HBcAg fusion PBVs presenting foreign B cell epitopes have been modified within the

α3α4 loop (AIR between amino acids 75 and 85), though modifications are also routinely made at the C and N termini. A different approach to insertion site selection should be taken when creating fusion PBVs targeting T cell immune responses, as antibody response to epitope becomes detrimental. Towards this goal, inserting epitopes within loop regions that are less exposed and less likely to negatively influence protein stability is optimal (242, 243). (B) Potential fusion modification and conjugation sites for truncated HBcAg model protein as represented by higher order folded and assembled structure. Residues are color-coded in gray scale, with darker residues indicating more exposed insertion locations. Natural conjugation sites (lysine and cysteine) have also been highlighted in red. (C) PBV modifications are generally orchestrated via fusion, conjugation, or encapsulation. Each type of modification occurs at a different level of protein structure, with fusion inserts occurring within primary structure, conjugated inserts occurring within secondary/tertiary structure, and encapsulated inserts occurring within the quaternary structure of proteins that form

101 enclosed, organized matrixes. As such, both monomeric and multimeric protein can accommodate fusion and conjugation modifications, whereas only multimeric protein can accommodate encapsulation. The 3D protein structure for HBcAg used in this image was rendered in PyMOL 2.3.0 and accessed via the Protein Data Bank (235,

239, 241).

102

103

Figure 3.4. The impact of PBV stability on immune response.

PBV stability has a profound impact on conformation, immunogenicity, and vaccination outcome. Outside of inherent fold stability that’s dictated by protein primary structure, many factors contribute to final PBV conformation and stability. Upstream and downstream processes, such as expression and purification, have a sizeable impact on the capacity of PBV to form higher-order structures and can ultimately lead to unintended PBV surface modification, aggregation, and/or decomposition. These issues can also present during formulation as a result of protein-protein, protein-adjuvant, and protein-container interactions. Incompatibilities between physiological conditions and

PBV formulation can result in poor extracellular stability, a phenomenon that often presents as excessive local inflammation and poor transport of antigen to secondary lymphoid organs. Finally, cellular stability, which is a function of all the factors mentioned previously, largely dictates MHC processing and the nature of T cell activation orchestrated by APCs (cellular vs. humoral, Th1 vs. Th2). The 3D protein structure for HBcAg used in this image was rendered in PyMOL 2.3.0 and accessed via the Protein Data Bank (235, 239, 241).

104

105

References

1. Centers for Disease Control and Prevention (U.S.), National Immunization

Program (Centers for Disease Control and Prevention). Epidemiology and prevention of vaccine-preventable diseases. Atlanta, Ga.?: Dept. of Health & Human Services, Public

Health Service. p. v.

2. Talmage DW. The acceptance and rejection of immunological concepts. Annu

Rev Immunol. 1986;4:1-11.

3. Huzair F, Sturdy S. Biotechnology and the transformation of vaccine innovation:

The case of the hepatitis B vaccines 1968-2000. Stud Hist Philos Biol Biomed Sci.

2017;64:11-21.

4. Mohsen MO, Zha L, Cabral-Miranda G, Bachmann MF. Major findings and recent advances in virus-like particle (VLP)-based vaccines. Semin Immunol. 2017;34:123-32.

5. Lavelle EC, Leavy O, Mills KHG. Modified Bacterial Toxins. In: Hackett CJ, Harn

DA, editors. Vaccine Adjuvants: Immunological and Clinical Principles. Totowa, NJ:

Humana Press; 2006. p. 111-53.

6. Ada G, Isaacs D. Carbohydrate-protein conjugate vaccines. Clin Microbiol Infect.

2003;9(2):79-85.

7. Robbins JB. Vaccines for the prevention of encapsulated bacterial diseases: current status, problems and prospects for the future. Immunochemistry. 1978;15(10-

11):839-54.

8. Katz DH, Benacerraf B. The regulatory influence of activated T cells on B cell responses to antigen. Adv Immunol. 1972;15:1-94.

106

9. Pichichero ME. Protein carriers of conjugate vaccines: characteristics, development, and clinical trials. Hum Vaccin Immunother. 2013;9(12):2505-23.

10. Vartak A, Sucheck SJ. Recent Advances in Subunit Vaccine Carriers. Vaccines

(Basel). 2016;4(2).

11. Giannini G, Rappuoli R, Ratti G. The amino-acid sequence of two non-toxic mutants of diphtheria toxin: CRM45 and CRM197. Nucleic Acids Res.

1984;12(10):4063-9.

12. Uchida T, Pappenheimer AM, Jr., Harper AA. Diphtheria toxin and related proteins. II. Kinetic studies on intoxication of HeLa cells by diphtheria toxin and related proteins. J Biol Chem. 1973;248(11):3845-50.

13. Malito E, Bursulaya B, Chen C, Lo Surdo P, Picchianti M, Balducci E, et al.

Structural basis for lack of toxicity of the diphtheria toxin mutant CRM197. Proc Natl

Acad Sci U S A. 2012;109(14):5229-34.

14. Baldauf KJ, Royal JM, Hamorsky KT, Matoba N. Cholera toxin B: one subunit with many pharmaceutical applications. Toxins (Basel). 2015;7(3):974-96.

15. Figueiredo D, Turcotte C, Frankel G, Li Y, Dolly O, Wilkin G, et al.

Characterization of recombinant tetanus toxin derivatives suitable for vaccine development. Infect Immun. 1995;63(8):3218-21.

16. Yu R, Yi S, Yu C, Fang T, Liu S, Yu T, et al. A conformational change of C fragment of tetanus neurotoxin reduces its ganglioside-binding activity but does not destroy its immunogenicity. Clin Vaccine Immunol. 2011;18(10):1668-72.

107

17. Millar DG, Hirst TR, Snider DP. Escherichia coli heat-labile enterotoxin B subunit is a more potent mucosal adjuvant than its vlosely related homologue, the B subunit of cholera toxin. Infect Immun. 2001;69(5):3476-82.

18. Smith LA. Botulism and vaccines for its prevention. Vaccine. 2009;27 Suppl

4:D33-9.

19. Tontini M, Romano MR, Proietti D, Balducci E, Micoli F, Balocchi C, et al.

Preclinical studies on new proteins as carrier for glycoconjugate vaccines. Vaccine.

2016;34(35):4235-42.

20. Takeuchi O, Akira S. Pattern recognition receptors and inflammation. Cell.

2010;140(6):805-20.

21. Shah RR, Hassett KJ, Brito LA. Overview of Vaccine Adjuvants: Introduction,

History, and Current Status. Methods Mol Biol. 2017;1494:1-13.

22. Karch CP, Burkhard P. Vaccine technologies: From whole organisms to rationally designed protein assemblies. Biochem Pharmacol. 2016;120:1-14.

23. Liu J, Qian C, Cao X. Post-Translational Modification Control of Innate Immunity.

Immunity. 2016;45(1):15-30.

24. Clausen H, Wandall HH, Steentoft C, Stanley P, Schnaar RL. Glycosylation

Engineering. In: rd, Varki A, Cummings RD, Esko JD, Stanley P, Hart GW, et al., editors. Essentials of Glycobiology. Cold Spring Harbor (NY)2015. p. 713-28.

25. Dambuza IM, Brown GD. C-type lectins in immunity: recent developments. Curr

Opin Immunol. 2015;32:21-7.

108

26. Kuriakose A, Chirmule N, Nair P. Immunogenicity of Biotherapeutics: Causes and Association with Posttranslational Modifications. J Immunol Res.

2016;2016:1298473.

27. Ghaderi D, Taylor RE, Padler-Karavani V, Diaz S, Varki A. Implications of the presence of N-glycolylneuraminic acid in recombinant therapeutic glycoproteins. Nat

Biotechnol. 2010;28(8):863-7.

28. Chung CH, Mirakhur B, Chan E, Le QT, Berlin J, Morse M, et al. Cetuximab- induced anaphylaxis and IgE specific for galactose-alpha-1,3-galactose. N Engl J Med.

2008;358(11):1109-17.

29. Li X, Min M, Du N, Gu Y, Hode T, Naylor M, et al. Chitin, chitosan, and glycated chitosan regulate immune responses: the novel adjuvants for cancer vaccine. Clin Dev

Immunol. 2013;2013:387023.

30. Lin CC, Pan IH, Li YR, Pan YG, Lin MK, Lu YH, et al. The adjuvant effects of high-molecule-weight polysaccharides purified from Antrodia cinnamomea on dendritic cell function and DNA vaccines. PLoS One. 2015;10(2):e0116191.

31. Li P, Wang F. Polysaccharides: Candidates of promising vaccine adjuvants. Drug

Discov Ther. 2015;9(2):88-93.

32. Petrovsky N, Cooper PD. Carbohydrate-based immune adjuvants. Expert Rev

Vaccines. 2011;10(4):523-37.

33. Miyake Y, Toyonaga K, Mori D, Kakuta S, Hoshino Y, Oyamada A, et al. C-type lectin MCL is an FcRgamma-coupled receptor that mediates the adjuvanticity of mycobacterial cord factor. Immunity. 2013;38(5):1050-62.

109

34. Li D, von Schaewen M, Wang X, Tao W, Zhang Y, Li L, et al. Altered

Glycosylation Patterns Increase Immunogenicity of a Subunit Hepatitis C Virus Vaccine,

Inducing Neutralizing Antibodies Which Confer Protection in Mice. J Virol.

2016;90(23):10486-98.

35. Ansari IH, Kwon B, Osorio FA, Pattnaik AK. Influence of N-linked glycosylation of porcine reproductive and respiratory syndrome virus GP5 on virus infectivity, antigenicity, and ability to induce neutralizing antibodies. J Virol. 2006;80(8):3994-4004.

36. Batista-Duharte A, Martinez DT, Carlos IZ. Efficacy and safety of immunological adjuvants. Where is the cut-off? Biomed Pharmacother. 2018;105:616-24.

37. Feng Y, Forsell MN, Flynn B, Adams W, Lore K, Seder R, et al. Chemical cross- linking of HIV-1 Env for direct TLR7/8 ligand conjugation compromises recognition of conserved antigenic determinants. Virology. 2013;446(1-2):56-65.

38. Chang TZ, Diambou I, Kim JR, Wang B, Champion JA. Host- and pathogen- derived adjuvant coatings on protein nanoparticle vaccines. Bioeng Transl Med.

2017;2(1):120-30.

39. Kastenmuller K, Wille-Reece U, Lindsay RW, Trager LR, Darrah PA, Flynn BJ, et al. Protective T cell immunity in mice following protein-TLR7/8 agonist-conjugate immunization requires aggregation, type I IFN, and multiple DC subsets. J Clin Invest.

2011;121(5):1782-96.

40. Fili L, Vultaggio A, Cardilicchia E, Manuelli C, Casini A, Nencini F, et al. A novel allergen-adjuvant conjugate suitable for specific immunotherapy of respiratory allergy. J

Allergy Clin Immunol. 2013;132(1):84-92.

110

41. Tighe H, Takabayashi K, Schwartz D, Van Nest G, Tuck S, Eiden JJ, et al.

Conjugation of immunostimulatory DNA to the short ragweed allergen amb a 1 enhances its immunogenicity and reduces its allergenicity. J Allergy Clin Immunol.

2000;106(1 Pt 1):124-34.

42. Schulke S, Vogel L, Junker AC, Hanschmann KM, Flaczyk A, Vieths S, et al. A

Fusion Protein Consisting of the Vaccine Adjuvant Monophosphoryl Lipid A and the

Allergen Ovalbumin Boosts Allergen-Specific Th1, Th2, and Th17 Responses In Vitro. J

Immunol Res. 2016;2016:4156456.

43. Soudi S, Zavaran-Hosseini A, Muhammad Hassan Z, Soleimani M, Jamshidi

Adegani F, Hashemi SM. Comparative study of the effect of LPS on the function of

BALB/c and C57BL/6 peritoneal macrophages. Cell J. 2013;15(1):45-54.

44. Schulke S, Flaczyk A, Vogel L, Gaudenzio N, Angers I, Loschner B, et al. MPLA shows attenuated pro-inflammatory properties and diminished capacity to activate mast cells in comparison with LPS. Allergy. 2015;70(10):1259-68.

45. Kawasaki T, Kawai T. Toll-like receptor signaling pathways. Front Immunol.

2014;5:461.

46. Liu G, Tarbet B, Song L, Reiserova L, Weaver B, Chen Y, et al. Immunogenicity and efficacy of flagellin-fused vaccine candidates targeting 2009 pandemic H1N1 influenza in mice. PLoS One. 2011;6(6):e20928.

47. Zhang C, Zhu S, Wei L, Yan X, Wang J, Quan R, et al. Recombinant Flagellin-

Porcine Circovirus Type 2 Cap Fusion Protein Promotes Protective Immune Responses in Mice. PLoS One. 2015;10(6):e0129617.

111

48. Liu G, Song L, Beasley DW, Putnak R, Parent J, Misczak J, et al.

Immunogenicity and efficacy of flagellin-envelope fusion dengue vaccines in mice and monkeys. Clin Vaccine Immunol. 2015;22(5):516-25.

49. Mizel SB, Graff AH, Sriranganathan N, Ervin S, Lees CJ, Lively MO, et al.

Flagellin-F1-V fusion protein is an effective plague vaccine in mice and two species of nonhuman primates. Clin Vaccine Immunol. 2009;16(1):21-8.

50. Kitzmuller C, Kalser J, Mutschlechner S, Hauser M, Zlabinger GJ, Ferreira F, et al. Fusion proteins of flagellin and the major birch pollen allergen Bet v 1 show enhanced immunogenicity, reduced allergenicity, and intrinsic adjuvanticity. J Allergy

Clin Immunol. 2018;141(1):293-9 e6.

51. Cuadros C, Lopez-Hernandez FJ, Dominguez AL, McClelland M, Lustgarten J.

Flagellin fusion proteins as adjuvants or vaccines induce specific immune responses.

Infect Immun. 2004;72(5):2810-6.

52. McDonald WF, Huleatt JW, Foellmer HG, Hewitt D, Tang J, Desai P, et al. A

West Nile virus recombinant protein vaccine that coactivates innate and adaptive immunity. J Infect Dis. 2007;195(11):1607-17.

53. Farajnia S, Peerayeh SN, Tanomand A, Majidi J, Goudarzi G, Naghili B, et al.

Protective efficacy of recombinant exotoxin A--flagellin fusion protein against

Pseudomonas aeruginosa infection. Can J Microbiol. 2015;61(1):60-4.

54. Stepanova LA, Mardanova ES, Shuklina MA, Blokhina EA, Kotlyarov RY,

Potapchuk MV, et al. Flagellin-fused protein targeting M2e and HA2 induces potent humoral and T-cell responses and protects mice against various influenza viruses a subtypes. J Biomed Sci. 2018;25(1):33.

112

55. Treanor JT, Albano FR, Sawlwin DC, Graves Jones A, Airey J, Formica N, et al.

Immunogenicity and safety of a quadrivalent inactivated influenza vaccine compared with two trivalent inactivated influenza vaccines containing alternate B strains in adults:

A phase 3, randomized noninferiority study. Vaccine. 2017;35(15):1856-64.

56. Treanor JJ, Taylor DN, Tussey L, Hay C, Nolan C, Fitzgerald T, et al. Safety and immunogenicity of a recombinant hemagglutinin influenza-flagellin fusion vaccine

(VAX125) in healthy young adults. Vaccine. 2010;28(52):8268-74.

57. Turley CB, Rupp RE, Johnson C, Taylor DN, Wolfson J, Tussey L, et al. Safety and immunogenicity of a recombinant M2e-flagellin influenza vaccine (STF2.4xM2e) in healthy adults. Vaccine. 2011;29(32):5145-52.

58. Taylor DN, Treanor JJ, Sheldon EA, Johnson C, Umlauf S, Song L, et al.

Development of VAX128, a recombinant hemagglutinin (HA) influenza-flagellin fusion vaccine with improved safety and immune response. Vaccine. 2012;30(39):5761-9.

59. Yoon SI, Kurnasov O, Natarajan V, Hong M, Gudkov AV, Osterman AL, et al.

Structural basis of TLR5-flagellin recognition and signaling. Science.

2012;335(6070):859-64.

60. Song WS, Jeon YJ, Namgung B, Hong M, Yoon SI. A conserved TLR5 binding and activation hot spot on flagellin. Sci Rep. 2017;7:40878.

61. O'Neill LA, Golenbock D, Bowie AG. The history of Toll-like receptors - redefining innate immunity. Nat Rev Immunol. 2013;13(6):453-60.

62. Moyle PM. Biotechnology approaches to produce potent, self-adjuvanting antigen-adjuvant fusion protein subunit vaccines. Biotechnol Adv. 2017;35(3):375-89.

113

63. Ingale S, Wolfert MA, Gaekwad J, Buskas T, Boons GJ. Robust immune responses elicited by a fully synthetic three-component vaccine. Nat Chem Biol.

2007;3(10):663-7.

64. Gartner T, Baeten M, Otieno S, Revets H, De Baetselier P, Huygen K. Mucosal prime-boost vaccination for tuberculosis based on TLR triggering OprI lipoprotein from

Pseudomonas aeruginosa fused to mycolyl-transferase Ag85A. Immunol Lett.

2007;111(1):26-35.

65. Chua BY, Eriksson EM, Brown LE, Zeng W, Gowans EJ, Torresi J, et al. A self- adjuvanting lipopeptide-based vaccine candidate for the treatment of hepatitis C virus infection. Vaccine. 2008;26(37):4866-75.

66. Zeng W, Tan AC, Horrocks K, Jackson DC. A lipidated form of the extracellular domain of influenza M2 protein as a self-adjuvanting vaccine candidate. Vaccine.

2015;33(30):3526-32.

67. Seth A, Yasutomi Y, Jacoby H, Callery JC, Kaminsky SM, Koff WC, et al.

Evaluation of a lipopeptide immunogen as a therapeutic in HIV type 1-seropositive individuals. AIDS Res Hum Retroviruses. 2000;16(4):337-43.

68. Huang CY, Chen JJ, Shen KY, Chang LS, Yeh YC, Chen IH, et al. Recombinant lipidated HPV E7 induces a Th-1-biased immune response and protective immunity against cervical cancer in a mouse model. PLoS One. 2012;7(7):e40970.

69. Basto AP, Piedade J, Ramalho R, Alves S, Soares H, Cornelis P, et al. A new cloning system based on the OprI lipoprotein for the production of recombinant bacterial cell wall-derived immunogenic formulations. J Biotechnol. 2012;157(1):50-63.

114

70. Song YC, Liu HH, Chen IH, Chen HW, Chong P, Leng CH, et al. A purified recombinant lipopeptide as adjuvant for cancer immunotherapy. Biomed Res Int.

2014;2014:349783.

71. Dhason MS, Wang JC, Hagan MF, Zlotnick A. Differential assembly of Hepatitis

B Virus core protein on single- and double-stranded nucleic acid suggest the dsDNA- filled core is spring-loaded. Virology. 2012;430(1):20-9.

72. Porterfield JZ, Dhason MS, Loeb DD, Nassal M, Stray SJ, Zlotnick A. Full-length hepatitis B virus core protein packages viral and heterologous RNA with similarly high levels of cooperativity. J Virol. 2010;84(14):7174-84.

73. Strods A, Ose V, Bogans J, Cielens I, Kalnins G, Radovica I, et al. Preparation by alkaline treatment and detailed characterisation of empty hepatitis B virus core particles for vaccine and gene therapy applications. Sci Rep. 2015;5:11639.

74. Cadena-Nava RD, Comas-Garcia M, Garmann RF, Rao AL, Knobler CM, Gelbart

WM. Self-assembly of viral capsid protein and RNA molecules of different sizes: requirement for a specific high protein/RNA mass ratio. J Virol. 2012;86(6):3318-26.

75. Storni T, Ruedl C, Schwarz K, Schwendener RA, Renner WA, Bachmann MF.

Nonmethylated CG motifs packaged into virus-like particles induce protective cytotoxic

T cell responses in the absence of systemic side effects. J Immunol. 2004;172(3):1777-

85.

76. Comellas-Aragones M, Engelkamp H, Claessen VI, Sommerdijk NA, Rowan AE,

Christianen PC, et al. A virus-based single-enzyme nanoreactor. Nat Nanotechnol.

2007;2(10):635-9.

115

77. Rohovie MJ, Nagasawa M, Swartz JR. Virus-like particles: Next-generation nanoparticles for targeted therapeutic delivery. Bioeng Transl Med. 2017;2(1):43-57.

78. Pickett GG, Peabody DS. Encapsidation of heterologous RNAs by bacteriophage

MS2 coat protein. Nucleic Acids Res. 1993;21(19):4621-6.

79. Pasloske BL, Walkerpeach CR, Obermoeller RD, Winkler M, DuBois DB.

Armored RNA technology for production of ribonuclease-resistant viral RNA controls and standards. J Clin Microbiol. 1998;36(12):3590-4.

80. Perrault SD, Shih WM. Virus-inspired membrane encapsulation of DNA nanostructures to achieve in vivo stability. ACS Nano. 2014;8(5):5132-40.

81. Sparrer KM, Gack MU. Intracellular detection of viral nucleic acids. Curr Opin

Microbiol. 2015;26:1-9.

82. LeBien TW, Tedder TF. B lymphocytes: how they develop and function. Blood.

2008;112(5):1570-80.

83. Sharon J, Rynkiewicz MJ, Lu Z, Yang CY. Discovery of protective B-cell epitopes for development of antimicrobial vaccines and antibody therapeutics. Immunology.

2014;142(1):1-23.

84. Li Z, Woo CJ, Iglesias-Ussel MD, Ronai D, Scharff MD. The generation of antibody diversity through somatic hypermutation and class switch recombination.

Genes Dev. 2004;18(1):1-11.

85. Doria-Rose NA, Joyce MG. Strategies to guide the antibody affinity maturation process. Curr Opin Virol. 2015;11:137-47.

86. Frank SA. Immunodominance within Hosts. Immunology and evolution of infectious disease. Princeton,: Princeton University Press; 2002. p. x, 348 p.

116

87. Slifka MK, Amanna I. How advances in immunology provide insight into improving vaccine efficacy. Vaccine. 2014;32(25):2948-57.

88. Schutze MP, Deriaud E, Przewlocki G, LeClerc C. Carrier-induced epitopic suppression is initiated through clonal dominance. J Immunol. 1989;142(8):2635-40.

89. Brauner S, Folkersen L, Kvarnstrom M, Meisgen S, Petersen S, Franzen-

Malmros M, et al. H1N1 vaccination in Sjogren's syndrome triggers polyclonal B cell activation and promotes autoantibody production. Ann Rheum Dis. 2017;76(10):1755-

63.

90. Sun P, Ju H, Liu Z, Ning Q, Zhang J, Zhao X, et al. Bioinformatics resources and tools for conformational B-cell epitope prediction. Comput Math Methods Med.

2013;2013:943636.

91. Zhao Q, Li S, Yu H, Xia N, Modis Y. Virus-like particle-based human vaccines: quality assessment based on structural and functional properties. Trends Biotechnol.

2013;31(11):654-63.

92. Gallagher JR, Torian U, McCraw DM, Harris AK. Characterization of the disassembly and reassembly of the HBV glycoprotein surface antigen, a pliable nanoparticle vaccine platform. Virology. 2017;502:176-87.

93. McCraw DM, Gallagher JR, Harris AK. Characterization of Influenza Vaccine

Hemagglutinin Complexes by Cryo-Electron Microscopy and Image Analyses Reveals

Structural Polymorphisms. Clin Vaccine Immunol. 2016;23(6):483-95.

94. Krumm SA, Doores KJ. Targeting Glycans on Human Pathogens for Vaccine

Design. Curr Top Microbiol Immunol. 2018.

117

95. Bremer PT, Janda KD. Conjugate Vaccine Immunotherapy for Substance Use

Disorder. Pharmacol Rev. 2017;69(3):298-315.

96. Bot A, Berinstein EM, Berinstein NL. Cancer Vaccines. Plotkin's Vaccines.

Seventh edition. ed. Philadelphia, PA: Elsevier; 2018. p. xxi, 1691 pages.

97. Garrity RR, Rimmelzwaan G, Minassian A, Tsai WP, Lin G, de Jong JJ, et al.

Refocusing neutralizing antibody response by targeted dampening of an immunodominant epitope. J Immunol. 1997;159(1):279-89.

98. Jorgensen KW, Buus S, Nielsen M. Structural properties of MHC class II ligands, implications for the prediction of MHC class II epitopes. PLoS One. 2010;5(12):e15877.

99. Collins KC, Janda KD. Hapten Design for Anti-addiction Vaccine Development.

2016. In: Biologics to treat substance use disorders : vaccines, monoclonal antibodies, and enzymes [Internet]. Cham: Springer,; [1 online resource ()]. Available from: http://dx.doi.org/10.1007/978-3-319-23150-1 MIT Access Only.

100. Astronomo RD, Burton DR. Carbohydrate vaccines: developing sweet solutions to sticky situations? Nat Rev Drug Discov. 2010;9(4):308-24.

101. Pryde DC, Jones LH, Gervais DP, Stead DR, Blakemore DC, Selby MD, et al.

Selection of a novel anti-nicotine vaccine: influence of antigen design on antibody function in mice. PLoS One. 2013;8(10):e76557.

102. Saylor K, Zhang C. A simple physiologically based pharmacokinetic model evaluating the effect of anti-nicotine antibodies on nicotine disposition in the brains of rats and humans. Toxicol Appl Pharmacol. 2016;307:150-64.

103. Pentel PR, LeSage MG. New directions in nicotine vaccine design and use. Adv

Pharmacol. 2014;69:553-80.

118

104. Gillam F, Zhang C. Epitope selection and their placement for increased virus neutralization in a novel vaccination strategy for porcine epidemic diarrhea virus utilizing the Hepatitis B virus core antigen. Vaccine. 2018;36(30):4507-16.

105. Kratz PA, Bottcher B, Nassal M. Native display of complete foreign protein domains on the surface of hepatitis B virus capsids. Proc Natl Acad Sci U S A.

1999;96(5):1915-20.

106. Muller M, Zhou J, Reed TD, Rittmuller C, Burger A, Gabelsberger J, et al.

Chimeric papillomavirus-like particles. Virology. 1997;234(1):93-111.

107. Varsani A, Williamson AL, de Villiers D, Becker I, Christensen ND, Rybicki EP.

Chimeric human papillomavirus type 16 (HPV-16) L1 particles presenting the common neutralizing epitope for the L2 minor capsid protein of HPV-6 and HPV-16. J Virol.

2003;77(15):8386-93.

108. Matic S, Rinaldi R, Masenga V, Noris E. Efficient production of chimeric human papillomavirus 16 L1 protein bearing the M2e influenza epitope in Nicotiana benthamiana plants. BMC Biotechnol. 2011;11:106.

109. Peabody DS, Manifold-Wheeler B, Medford A, Jordan SK, do Carmo Caldeira J,

Chackerian B. Immunogenic display of diverse peptides on virus-like particles of RNA phage MS2. J Mol Biol. 2008;380(1):252-63.

110. Murthy AM, Ni Y, Meng X, Zhang C. Production and evaluation of virus-like particles displaying immunogenic epitopes of porcine reproductive and respiratory syndrome virus (PRRSV). Int J Mol Sci. 2015;16(4):8382-96.

111. Yan D, Wei YQ, Guo HC, Sun SQ. The application of virus-like particles as vaccines and biological vehicles. Appl Microbiol Biotechnol. 2015;99(24):10415-32.

119

112. Gillam F, Zhang J, Zhang C. Hepatitis B core antigen based novel vaccine against porcine epidemic diarrhea virus. J Virol Methods. 2018;253:61-9.

113. Zhang C, Knudsen DE, Liu M, Robertson DC, Zhang W, Group STTVC. Toxicity and immunogenicity of Enterotoxigenic Escherichia coli heat-labile and heat-stable toxoid fusion 3xSTa(A14Q)-LT(S63K/R192G/L211A) in a murine model. PLoS One.

2013;8(10):e77386.

114. Zeinalzadeh N, Salmanian AH, Goujani G, Amani J, Ahangari G, Akhavian A, et al. A Chimeric protein of CFA/I, CS6 subunits and LTB/STa toxoid protects immunized mice against enterotoxigenic Escherichia coli. Microbiol Immunol. 2017;61(7):272-9.

115. Gil LA, da Cunha CE, Moreira GM, Salvarani FM, Assis RA, Lobato FC, et al.

Production and evaluation of a recombinant chimeric vaccine against clostridium botulinum neurotoxin types C and D. PLoS One. 2013;8(7):e69692.

116. Hollingshead S, Jongerius I, Exley RM, Johnson S, Lea SM, Tang CM. Structure- based design of chimeric antigens for multivalent protein vaccines. Nat Commun.

2018;9(1):1051.

117. McCluskie MJ, Evans DM, Zhang N, Benoit M, McElhiney SP, Unnithan M, et al.

The effect of preexisting anti-carrier immunity on subsequent responses to CRM197 or

Qb-VLP conjugate vaccines. Immunopharmacol Immunotoxicol. 2016;38(3):184-96.

118. Dagan R, Poolman J, Siegrist CA. Glycoconjugate vaccines and immune interference: A review. Vaccine. 2010;28(34):5513-23.

119. Pecetta S, Tontini M, Faenzi E, Cioncada R, Proietti D, Seubert A, et al. Carrier priming effect of CRM197 is related to an enhanced B and T cell activation in

120 meningococcal serogroup A conjugate vaccination. Immunological comparison between

CRM197 and diphtheria toxoid. Vaccine. 2016;34(20):2334-41.

120. Jegerlehner A, Wiesel M, Dietmeier K, Zabel F, Gatto D, Saudan P, et al. Carrier induced epitopic suppression of antibody responses induced by virus-like particles is a dynamic phenomenon caused by carrier-specific antibodies. Vaccine.

2010;28(33):5503-12.

121. Sadeyen JR, Tourne S, Shkreli M, Sizaret PY, Coursaget P. Insertion of a foreign sequence on capsid surface loops of human papillomavirus type 16 virus-like particles reduces their capacity to induce neutralizing antibodies and delineates a conformational neutralizing epitope. Virology. 2003;309(1):32-40.

122. Sominskaya I, Skrastina D, Dislers A, Vasiljev D, Mihailova M, Ose V, et al.

Construction and immunological evaluation of multivalent hepatitis B virus (HBV) core virus-like particles carrying HBV and HCV epitopes. Clin Vaccine Immunol.

2010;17(6):1027-33.

123. Xu Y, Shi Y, Zhou J, Yang W, Bai L, Wang S, et al. Structure-based antigenic epitope and PEGylation improve the efficacy of staphylokinase. Microb Cell Fact.

2017;16(1):197.

124. Zhang T, Yu W, Wang Y, Hu T. Moderate PEGylation of the carrier protein improves the polysaccharide-specific immunogenicity of meningococcal group A polysaccharide conjugate vaccine. Vaccine. 2015;33(28):3208-14.

125. Gefen T, Vaya J, Khatib S, Harkevich N, Artoul F, Heller ED, et al. The impact of

PEGylation on protein immunogenicity. Int Immunopharmacol. 2013;15(2):254-9.

121

126. Dozier JK, Distefano MD. Site-Specific PEGylation of Therapeutic Proteins. Int J

Mol Sci. 2015;16(10):25831-64.

127. Kumar BV, Connors TJ, Farber DL. Human T Cell Development, Localization, and Function throughout Life. Immunity. 2018;48(2):202-13.

128. Smith-Garvin JE, Koretzky GA, Jordan MS. T cell activation. Annu Rev Immunol.

2009;27:591-619.

129. Livingstone AM, Fathman CG. The structure of T-cell epitopes. Annu Rev

Immunol. 1987;5:477-501.

130. Blum JS, Wearsch PA, Cresswell P. Pathways of antigen processing. Annu Rev

Immunol. 2013;31:443-73.

131. Grodeland G, Mjaaland S, Tunheim G, Fredriksen AB, Bogen B. The specificity of targeted vaccines for APC surface molecules influences the immune response phenotype. PLoS One. 2013;8(11):e80008.

132. Kratzer R, Mauvais FX, Burgevin A, Barilleau E, van Endert P. Fusion proteins for versatile antigen targeting to cell surface receptors reveal differential capacity to prime immune responses. J Immunol. 2010;184(12):6855-64.

133. Lu L, Palaniyandi S, Zeng R, Bai Y, Liu X, Wang Y, et al. A neonatal Fc receptor- targeted mucosal vaccine strategy effectively induces HIV-1 antigen-specific immunity to genital infection. J Virol. 2011;85(20):10542-53.

134. Loureiro S, Ren J, Phapugrangkul P, Colaco CA, Bailey CR, Shelton H, et al.

Adjuvant-free immunization with hemagglutinin-Fc fusion proteins as an approach to influenza vaccines. J Virol. 2011;85(6):3010-4.

122

135. Soleimanpour S, Farsiani H, Mosavat A, Ghazvini K, Eydgahi MR, Sankian M, et al. APC targeting enhances immunogenicity of a novel multistage Fc-fusion tuberculosis vaccine in mice. Appl Microbiol Biotechnol. 2015;99(24):10467-80.

136. Ooms K, Van Gorp H, Van Gaever T, Nauwynck HJ, Delputte PL. Development of a recombinant antibody to target peptides and proteins to sialoadhesin-expressing macrophages. BMC Biotechnol. 2013;13:33.

137. Tagliani E, Guermonprez P, Sepulveda J, Lopez-Bravo M, Ardavin C, Amigorena

S, et al. Selection of an antibody library identifies a pathway to induce immunity by targeting CD36 on steady-state CD8 alpha+ dendritic cells. J Immunol.

2008;180(5):3201-9.

138. Caminschi I, Maraskovsky E, Heath WR. Targeting Dendritic Cells in vivo for

Cancer Therapy. Front Immunol. 2012;3:13.

139. Mi W, Wanjie S, Lo ST, Gan Z, Pickl-Herk B, Ober RJ, et al. Targeting the neonatal fc receptor for antigen delivery using engineered fc fragments. J Immunol.

2008;181(11):7550-61.

140. Blander JM. Regulation of the Cell Biology of Antigen Cross-Presentation. Annu

Rev Immunol. 2018;36:717-53.

141. Plato A, Willment JA, Brown GD. C-type lectin-like receptors of the dectin-1 cluster: ligands and signaling pathways. Int Rev Immunol. 2013;32(2):134-56.

142. Simon GG, Hu Y, Khan AM, Zhou J, Salmon J, Chikhlikar PR, et al. Dendritic cell mediated delivery of plasmid DNA encoding LAMP/HIV-1 Gag fusion immunogen enhances T cell epitope responses in HLA DR4 transgenic mice. PLoS One.

2010;5(1):e8574.

123

143. Dobano C, Rogers WO, Gowda K, Doolan DL. Targeting antigen to MHC Class I and Class II antigen presentation pathways for malaria DNA vaccines. Immunol Lett.

2007;111(2):92-102.

144. Fujii S, Senju S, Chen YZ, Ando M, Matsushita S, Nishimura Y. The CLIP- substituted invariant chain efficiently targets an antigenic peptide to HLA class II pathway in L cells. Hum Immunol. 1998;59(10):607-14.

145. Carstens C, Newman DK, Bohlen H, Konig A, Koch N. Invariant chains with the class II binding site replaced by a sequence from influenza virus matrix protein constrain low-affinity sequences to MHC II presentation. Int Immunol. 2000;12(11):1561-8.

146. Wu C, Zhang DG, Chen FF, Liu XL, Liu SJ, Yu WY. N-terminal functional region of the invariant chain efficiently targets the binding of a CTL epitope to MHC class I molecules during cross-presentation. Genet Mol Res. 2014;13(2):2438-50.

147. Capone S, Naddeo M, D'Alise AM, Abbate A, Grazioli F, Del Gaudio A, et al.

Fusion of HCV nonstructural antigen to MHC class II-associated invariant chain enhances T-cell responses induced by vectored vaccines in nonhuman primates. Mol

Ther. 2014;22(5):1039-47.

148. Perez SA, Peoples GE, Papamichail M, Baxevanis CN. Invariant chain-peptide fusion vaccine using HER-2/neu. Methods Mol Biol. 2014;1139:321-36.

149. Kallinteris NL, Lu X, Blackwell CE, von Hofe E, Humphreys RE, Xu M. Ii-

Key/MHC class II epitope hybrids: a strategy that enhances MHC class II epitope loading to create more potent peptide vaccines. Expert Opin Biol Ther.

2006;6(12):1311-21.

124

150. Liu XL, Shan WJ, Xu SS, Zhang JJ, Xu FZ, Xia SL, et al. The efficacy of chimeric vaccines constructed with PEP-1 and Ii-Key linking to a hybrid epitope from heterologous viruses. Biologicals. 2015;43(5):377-82.

151. Spencer AJ, Cottingham MG, Jenks JA, Longley RJ, Capone S, Colloca S, et al.

Enhanced vaccine-induced CD8+ T cell responses to malaria antigen ME-TRAP by fusion to MHC class ii invariant chain. PLoS One. 2014;9(6):e100538.

152. Moron G, Dadaglio G, Leclerc C. New tools for antigen delivery to the MHC class

I pathway. Trends Immunol. 2004;25(2):92-7.

153. Belizaire R, Unanue ER. Targeting proteins to distinct subcellular compartments reveals unique requirements for MHC class I and II presentation. Proc Natl Acad Sci U

S A. 2009;106(41):17463-8.

154. Gutierrez-Martinez E, Planes R, Anselmi G, Reynolds M, Menezes S, Adiko AC, et al. Cross-Presentation of Cell-Associated Antigens by MHC Class I in Dendritic Cell

Subsets. Front Immunol. 2015;6:363.

155. Colbert JD, Farfan-Arribas DJ, Rock KL. Substrate-induced protein stabilization reveals a predominant contribution from mature proteins to peptides presented on MHC class I. J Immunol. 2013;191(11):5410-9.

156. Scheiblhofer S, Laimer J, Machado Y, Weiss R, Thalhamer J. Influence of protein fold stability on immunogenicity and its implications for vaccine design. Expert Rev

Vaccines. 2017;16(5):479-89.

157. Schliehe C, Bitzer A, van den Broek M, Groettrup M. Stable antigen is most effective for eliciting CD8+ T-cell responses after DNA vaccination and infection with recombinant vaccinia virus in vivo. J Virol. 2012;86(18):9782-93.

125

158. Thai R, Moine G, Desmadril M, Servent D, Tarride JL, Menez A, et al. Antigen stability controls antigen presentation. J Biol Chem. 2004;279(48):50257-66.

159. Trombetta ES, Ebersold M, Garrett W, Pypaert M, Mellman I. Activation of lysosomal function during dendritic cell maturation. Science. 2003;299(5611):1400-3.

160. Lennon-Dumenil AM, Bakker AH, Maehr R, Fiebiger E, Overkleeft HS,

Rosemblatt M, et al. Analysis of protease activity in live antigen-presenting cells shows regulation of the phagosomal proteolytic contents during dendritic cell activation. J Exp

Med. 2002;196(4):529-40.

161. Schumacher J, Bacic T, Staritzbichler R, Daneschdar M, Klamp T, Arnold P, et al. Enhanced stability of a chimeric hepatitis B core antigen virus-like-particle (HBcAg-

VLP) by a C-terminal linker-hexahistidine-peptide. J Nanobiotechnology. 2018;16(1):39.

162. Lu Y, Chan W, Ko BY, VanLang CC, Swartz JR. Assessing sequence plasticity of a virus-like nanoparticle by evolution toward a versatile scaffold for vaccines and drug delivery. Proc Natl Acad Sci U S A. 2015;112(40):12360-5.

163. Ashcroft AE, Lago H, Macedo JM, Horn WT, Stonehouse NJ, Stockley PG.

Engineering thermal stability in RNA phage capsids via disulphide bonds. J Nanosci

Nanotechnol. 2005;5(12):2034-41.

164. Delamarre L, Pack M, Chang H, Mellman I, Trombetta ES. Differential lysosomal proteolysis in antigen-presenting cells determines antigen fate. Science.

2005;307(5715):1630-4.

165. Delamarre L, Couture R, Mellman I, Trombetta ES. Enhancing immunogenicity by limiting susceptibility to lysosomal proteolysis. J Exp Med. 2006;203(9):2049-55.

126

166. Zhao Z, Hu Y, Harmon T, Pentel PR, Ehrich M, Zhang C. Hybrid nanoparticle- based nicotine nanovaccines: Boosting the immunological efficacy by conjugation of potent carrier proteins. Nanomedicine. 2018;14(5):1655-65.

167. So T, Ito H, Hirata M, Ueda T, Imoto T. Contribution of conformational stability of hen lysozyme to induction of type 2 T-helper immune responses. Immunology.

2001;104(3):259-68.

168. Nielsen M, Lund O, Buus S, Lundegaard C. MHC class II epitope predictive algorithms. Immunology. 2010;130(3):319-28.

169. da Silva Antunes R, Paul S, Sidney J, Weiskopf D, Dan JM, Phillips E, et al.

Correction: Definition of Human Epitopes Recognized in Tetanus Toxoid and

Development of an Assay Strategy to Detect Ex Vivo Tetanus CD4+ T Cell Responses.

PLoS One. 2018;13(2):e0193382.

170. Panina-Bordignon P, Tan A, Termijtelen A, Demotz S, Corradin G, Lanzavecchia

A. Universally immunogenic T cell epitopes: promiscuous binding to human MHC class

II and promiscuous recognition by T cells. Eur J Immunol. 1989;19(12):2237-42.

171. Alexander J, Sidney J, Southwood S, Ruppert J, Oseroff C, Maewal A, et al.

Development of high potency universal DR-restricted helper epitopes by modification of high affinity DR-blocking peptides. Immunity. 1994;1(9):751-61.

172. Jensen KK, Andreatta M, Marcatili P, Buus S, Greenbaum JA, Yan Z, et al.

Improved methods for predicting peptide binding affinity to MHC class II molecules.

Immunology. 2018;154(3):394-406.

127

173. Wang P, Sidney J, Dow C, Mothe B, Sette A, Peters B. A systematic assessment of MHC class II peptide binding predictions and evaluation of a consensus approach.

PLoS Comput Biol. 2008;4(4):e1000048.

174. Wang P, Sidney J, Kim Y, Sette A, Lund O, Nielsen M, et al. Peptide binding predictions for HLA DR, DP and DQ molecules. BMC Bioinformatics. 2010;11:568.

175. Li Pira G, Ivaldi F, Moretti P, Manca F. High throughput T epitope mapping and vaccine development. J Biomed Biotechnol. 2010;2010:325720.

176. Patronov A, Doytchinova I. T-cell epitope vaccine design by immunoinformatics.

Open Biol. 2013;3(1):120139.

177. Bui HH, Sidney J, Dinh K, Southwood S, Newman MJ, Sette A. Predicting population coverage of T-cell epitope-based diagnostics and vaccines. BMC

Bioinformatics. 2006;7:153.

178. Sanchez-Trincado JL, Gomez-Perosanz M, Reche PA. Fundamentals and

Methods for T- and B-Cell Epitope Prediction. J Immunol Res. 2017;2017:2680160.

179. Chowell D, Krishna S, Becker PD, Cocita C, Shu J, Tan X, et al. TCR contact residue hydrophobicity is a hallmark of immunogenic CD8+ T cell epitopes. Proc Natl

Acad Sci U S A. 2015;112(14):E1754-62.

180. Parrish HL, Deshpande NR, Vasic J, Kuhns MS. Functional evidence for TCR- intrinsic specificity for MHCII. Proc Natl Acad Sci U S A. 2016;113(11):3000-5.

181. Petersson K, Forsberg G, Walse B. Interplay between superantigens and immunoreceptors. Scand J Immunol. 2004;59(4):345-55.

128

182. Jemon K, Young V, Wilson M, McKee S, Ward V, Baird M, et al. An enhanced heterologous virus-like particle for human papillomavirus type 16 tumour immunotherapy. PLoS One. 2013;8(6):e66866.

183. Percival-Alwyn JL, England E, Kemp B, Rapley L, Davis NH, McCarthy GR, et al.

Generation of potent mouse monoclonal antibodies to self-proteins using T-cell epitope

"tags". MAbs. 2015;7(1):129-37.

184. Wu M, Zhao L, Zhu L, Chen Z, Li H. Expression and purification of chimeric peptide comprising EGFR B-cell epitope and measles virus fusion protein T-cell epitope in Escherichia coli. Protein Expr Purif. 2013;88(1):7-12.

185. Hurtgen BJ, Hung CY, Ostroff GR, Levitz SM, Cole GT. Construction and evaluation of a novel recombinant T cell epitope-based vaccine against

Coccidioidomycosis. Infect Immun. 2012;80(11):3960-74.

186. Wang H, Su X, Zhang P, Liang J, Wei H, Wan M, et al. Recombinant heat shock protein 65 carrying PADRE and HBV epitopes activates dendritic cells and elicits HBV- specific CTL responses. Vaccine. 2011;29(12):2328-35.

187. Feng GD, Xue XC, Gao ML, Wang XF, Shu Z, Mu N, et al. Therapeutic effects of

PADRE-BAFF autovaccine on rat adjuvant arthritis. Biomed Res Int. 2014;2014:854954.

188. Ghaffari-Nazari H, Tavakkol-Afshari J, Jaafari MR, Tahaghoghi-Hajghorbani S,

Masoumi E, Jalali SA. Improving Multi-Epitope Long Peptide Vaccine Potency by Using a Strategy that Enhances CD4+ T Help in BALB/c Mice. PLoS One.

2015;10(11):e0142563.

189. Bergmann-Leitner ES, Chaudhury S, Steers NJ, Sabato M, Delvecchio V,

Wallqvist AS, et al. Computational and experimental validation of B and T-cell epitopes

129 of the in vivo immune response to a novel malarial antigen. PLoS One.

2013;8(8):e71610.

190. Kedl RM, Rees WA, Hildeman DA, Schaefer B, Mitchell T, Kappler J, et al. T cells compete for access to antigen-bearing antigen-presenting cells. J Exp Med.

2000;192(8):1105-13.

191. Robinson J, Halliwell JA, Hayhurst JD, Flicek P, Parham P, Marsh SG. The IPD and IMGT/HLA database: allele variant databases. Nucleic Acids Res.

2015;43(Database issue):D423-31.

192. Maiers M, Gragert L, Klitz W. High-resolution HLA alleles and haplotypes in the

United States population. Hum Immunol. 2007;68(9):779-88.

193. Shiina T, Blancher A, Inoko H, Kulski JK. Comparative genomics of the human, macaque and mouse major histocompatibility complex. Immunology. 2017;150(2):127-

38.

194. Segal Y, Shoenfeld Y. Vaccine-induced autoimmunity: the role of molecular mimicry and immune crossreaction. Cell Mol Immunol. 2018;15(6):586-94.

195. Mestas J, Hughes CC. Of mice and not men: differences between mouse and human immunology. J Immunol. 2004;172(5):2731-8.

196. Huang M, Zhang W, Guo J, Wei X, Phiwpan K, Zhang J, et al. Improved

Transgenic Mouse Model for Studying HLA Class I Antigen Presentation. Sci Rep.

2016;6:33612.

197. Sesma L, Alvarez I, Marcilla M, Paradela A, Lopez de Castro JA. Species- specific differences in proteasomal processing and tapasin-mediated loading influence peptide presentation by HLA-B27 in murine cells. J Biol Chem. 2003;278(47):46461-72.

130

198. Burgevin A, Saveanu L, Kim Y, Barilleau E, Kotturi M, Sette A, et al. A detailed analysis of the murine TAP transporter substrate specificity. PLoS One.

2008;3(6):e2402.

199. Shiina T, Hosomichi K, Inoko H, Kulski JK. The HLA genomic loci map: expression, interaction, diversity and disease. J Hum Genet. 2009;54(1):15-39.

200. Kotturi MF, Assarsson E, Peters B, Grey H, Oseroff C, Pasquetto V, et al. Of mice and humans: how good are HLA transgenic mice as a model of human immune responses? Immunome Res. 2009;5:3.

201. Corbet S, Nielsen HV, Vinner L, Lauemoller S, Therrien D, Tang S, et al.

Optimization and immune recognition of multiple novel conserved HLA-A2, human immunodeficiency virus type 1-specific CTL epitopes. J Gen Virol. 2003;84(Pt 9):2409-

21.

202. Davis HL. Novel vaccines and adjuvant systems: the utility of animal models for predicting immunogenicity in humans. Hum Vaccin. 2008;4(3):246-50.

203. Baneyx F, Mujacic M. Recombinant protein folding and misfolding in Escherichia coli. Nat Biotechnol. 2004;22(11):1399-408.

204. Fox CB, Kramer RM, Barnes VL, Dowling QM, Vedvick TS. Working together: interactions between vaccine antigens and adjuvants. Ther Adv Vaccines. 2013;1(1):7-

20.

205. Manning MC, Chou DK, Murphy BM, Payne RW, Katayama DS. Stability of protein pharmaceuticals: an update. Pharm Res. 2010;27(4):544-75.

206. Ratanji KD, Derrick JP, Dearman RJ, Kimber I. Immunogenicity of therapeutic proteins: influence of aggregation. J Immunotoxicol. 2014;11(2):99-109.

131

207. Servin-Blanco R, Zamora-Alvarado R, Gevorkian G, Manoutcharian K. Antigenic variability: Obstacles on the road to vaccines against traditionally difficult targets. Hum

Vaccin Immunother. 2016;12(10):2640-8.

208. Grohskopf LA, Sokolow LZ, Broder KR, Walter EB, Fry AM, Jernigan DB.

Prevention and Control of Seasonal Influenza with Vaccines: Recommendations of the

Advisory Committee on Immunization Practices-United States, 2018-19 Influenza

Season. MMWR Recomm Rep. 2018;67(3):1-20.

209. Gao Y, McKay PF, Mann JFS. Advances in HIV-1 Vaccine Development.

Viruses. 2018;10(4).

210. Dingens AS, Haddox HK, Overbaugh J, Bloom JD. Comprehensive Mapping of

HIV-1 Escape from a Broadly Neutralizing Antibody. Cell Host Microbe. 2017;21(6):777-

87 e4.

211. Pegu A, Hessell AJ, Mascola JR, Haigwood NL. Use of broadly neutralizing antibodies for HIV-1 prevention. Immunol Rev. 2017;275(1):296-312.

212. Haynes BF, Mascola JR. The quest for an antibody-based HIV vaccine. Immunol

Rev. 2017;275(1):5-10.

213. Jardine J, Julien JP, Menis S, Ota T, Kalyuzhniy O, McGuire A, et al. Rational

HIV immunogen design to target specific germline B cell receptors. Science.

2013;340(6133):711-6.

214. Kwong PD, Mascola JR. HIV-1 Vaccines Based on Antibody Identification, B Cell

Ontogeny, and Epitope Structure. Immunity. 2018;48(5):855-71.

215. Kloverpris HN, Leslie A, Goulder P. Role of HLA Adaptation in HIV Evolution.

Front Immunol. 2015;6:665.

132

216. Crawford H, Lumm W, Leslie A, Schaefer M, Boeras D, Prado JG, et al.

Evolution of HLA-B*5703 HIV-1 escape mutations in HLA-B*5703-positive individuals and their transmission recipients. J Exp Med. 2009;206(4):909-21.

217. Sun X, Shi Y, Akahoshi T, Fujiwara M, Gatanaga H, Schonbach C, et al. Effects of a Single Escape Mutation on T Cell and HIV-1 Co-adaptation. Cell Rep.

2016;15(10):2279-91.

218. Gartland AJ, Li S, McNevin J, Tomaras GD, Gottardo R, Janes H, et al. Analysis of HLA A*02 association with vaccine efficacy in the RV144 HIV-1 vaccine trial. J Virol.

2014;88(15):8242-55.

219. Doria-Rose NA, Schramm CA, Gorman J, Moore PL, Bhiman JN, DeKosky BJ, et al. Developmental pathway for potent V1V2-directed HIV-neutralizing antibodies.

Nature. 2014;509(7498):55-62.

220. Peabody DS. A Viral Platform for Chemical Modification and Multivalent Display.

J Nanobiotechnology. 2003;1(1):5.

221. Jegerlehner A, Tissot A, Lechner F, Sebbel P, Erdmann I, Kundig T, et al. A molecular assembly system that renders antigens of choice highly repetitive for induction of protective B cell responses. Vaccine. 2002;20(25-26):3104-12.

222. Smith ML, Lindbo JA, Dillard-Telm S, Brosio PM, Lasnik AB, McCormick AA, et al. Modified tobacco mosaic virus particles as scaffolds for display of protein antigens for vaccine applications. Virology. 2006;348(2):475-88.

223. Geiger FC, Eber FJ, Eiben S, Mueller A, Jeske H, Spatz JP, et al. TMV nanorods with programmed longitudinal domains of differently addressable coat proteins.

Nanoscale. 2013;5(9):3808-16.

133

224. Wu T, Li F, Sha X, Li F, Zhang B, Ma W, et al. A novel recombinant RANKL vaccine prepared by incorporation of an unnatural amino acid into RANKL and its preventive effect in a murine model of collagen-induced arthritis. Int Immunopharmacol.

2018;64:326-32.

225. Grunewald J, Hunt GS, Dong L, Niessen F, Wen BG, Tsao ML, et al. Mechanistic studies of the immunochemical termination of self-tolerance with unnatural amino acids.

Proc Natl Acad Sci U S A. 2009;106(11):4337-42.

226. Patel KG, Swartz JR. Surface functionalization of virus-like particles by direct conjugation using azide-alkyne click chemistry. Bioconjug Chem. 2011;22(3):376-87.

227. Brune KD, Howarth M. New Routes and Opportunities for Modular Construction of Particulate Vaccines: Stick, Click, and Glue. Front Immunol. 2018;9:1432.

228. Patterson D, Schwarz B, Avera J, Western B, Hicks M, Krugler P, et al. Sortase-

Mediated Ligation as a Modular Approach for the Covalent Attachment of Proteins to the Exterior of the Bacteriophage P22 Virus-like Particle. Bioconjug Chem.

2017;28(8):2114-24.

229. Therien A, Bedard M, Carignan D, Rioux G, Gauthier-Landry L, Laliberte-Gagne

ME, et al. A versatile papaya mosaic virus (PapMV) vaccine platform based on sortase- mediated antigen coupling. J Nanobiotechnology. 2017;15(1):54.

230. Koho T, Ihalainen TO, Stark M, Uusi-Kerttula H, Wieneke R, Rahikainen R, et al.

His-tagged norovirus-like particles: A versatile platform for cellular delivery and surface display. Eur J Pharm Biopharm. 2015;96:22-31.

231. Clancy HA, Costa M. Nickel: a pervasive carcinogen. Future Oncol.

2012;8(12):1507-9.

134

232. Frietze KM, Peabody DS, Chackerian B. Engineering virus-like particles as vaccine platforms. Curr Opin Virol. 2016;18:44-9.

233. Thrane S, Janitzek CM, Agerbaek MO, Ditlev SB, Resende M, Nielsen MA, et al.

A Novel Virus-Like Particle Based Vaccine Platform Displaying the Placental Malaria

Antigen VAR2CSA. PLoS One. 2015;10(11):e0143071.

234. Pellizzari R, Rossetto O, Schiavo G, Montecucco C. Tetanus and botulinum neurotoxins: mechanism of action and therapeutic uses. Philos Trans R Soc Lond B Biol

Sci. 1999;354(1381):259-68.

235. Schrodinger, LLC. The PyMOL Molecular Graphics System, Version 2.3.0. 2019.

236. Bell CE, Eisenberg D. Crystal structure of nucleotide-free diphtheria toxin.

Biochemistry. 1997;36(3):481-8.

237. Zhang RG, Scott DL, Westbrook ML, Nance S, Spangler BD, Shipley GG, et al.

The three-dimensional crystal structure of cholera toxin. J Mol Biol. 1995;251(4):563-73.

238. Masuyer G, Conrad J, Stenmark P. The structure of the tetanus toxin reveals pH- mediated domain dynamics. EMBO Rep. 2017;18(8):1306-17.

239. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, et al. The

Protein Data Bank. Nucleic Acids Res. 2000;28(1):235-42.

240. Vita R, Mahajan S, Overton JA, Dhanda SK, Martini S, Cantrell JR, et al. The

Immune Epitope Database (IEDB): 2018 update. Nucleic Acids Res. 2019;47(D1):D339-

D43.

241. Bottcher B, Nassal M. Structure of Mutant Hepatitis B Core Protein Capsids with

Premature Secretion Phenotype. J Mol Biol. 2018;430(24):4941-54.

135

242. Wynne SA, Crowther RA, Leslie AG. The crystal structure of the human hepatitis

B virus capsid. Mol Cell. 1999;3(6):771-80.

243. Pumpens P, Ulrich RG, Sasnauskas K, Kazaks A, Ose V, Grens E. Construction of Novel Vaccines on the Basis of Virus-Like Particles: Hepatitis B Virus Proteins as

Vaccine Carriers. In: Khudyakov YE, editor. Medicinal protein engineering. Boca Raton:

CRC Press; 2009. p. xvii, 563 p., 4 p. of plates.

136

Chapter 4: A Simple Physiologically Based Pharmacokinetic Model Evaluating the

Effect of Anti-Nicotine Antibodies on Nicotine Disposition in the Brains of Rats

and Humans

Kyle Saylor1, Chenming Zhang1,*

1Department of Biological Systems Engineering, Virginia Tech, Blacksburg, VA

*Corresponding Author: Chenming (Mike) Zhang

302D, HABB1, 1230 Washington St., S.W.

Blacksburg, VA 24061

Voice: (540)-231-7601

Fax: (540)-231-3199

Email: [email protected]

This manuscript has been published in Toxicology and Applied Pharmacology, 2016

307:150-64.

137

Abstract

Physiologically based pharmacokinetic (PBPK) modeling was applied to investigate the effects of anti-nicotine antibodies on nicotine disposition in the brains of rats and humans. Successful construction of both rat and human models was achieved by fitting model outputs to published nicotine concentration time course data in the blood and in the brain. Key parameters presumed to have the most effect on the ability of these antibodies to prevent nicotine from entering the brain were selected for investigation using the human model. These parameters, which included antibody affinity for nicotine, antibody cross-reactivity with cotinine, and antibody concentration, were broken down into different, clinically-derived in silico treatment levels and fed into the human PBPK model. Model predictions suggested that all three parameters, in addition to smoking status, have a sizable impact on anti-nicotine antibodies’ ability to prevent nicotine from entering the brain and that the antibodies elicited by current human vaccines do not have sufficient binding characteristics to reduce brain nicotine concentrations. If the antibody binding characteristics achieved in animal studies can similarly be achieved in human studies, however, nicotine vaccine efficacy in terms of brain nicotine concentration reduction is predicted to meet threshold values for alleviating nicotine dependence.

Keywords

Physiologically based pharmacokinetic model; PBPK; nicotine vaccine; anti-nicotine antibodies; nicotine disposition; nicotine dependence

Abbreviations

138

PBPK, physiologically based pharmacokinetic model; USD, United States dollar; CDC,

Centers for Disease Control and Prevention; IC50, half maximal inhibitory concentration; CMUNic, 6-(carboxymethylureido)-(+/-)-nicotine

139

4.1. Background

The use of tobacco products is the leading preventable cause of death in the world today, resulting in nearly six million deaths annually and assisting in the manifestation of many other diseases (1). In addition, the economic damages caused by tobacco smoking in the US alone over the past four decades could exceed $7 trillion USD

(extrapolated from CDC reported data, unpublished). Existing therapeutic treatments have proven to be more and more ineffective at assisting those with the desire to quit

(2) and, if current trends continue, more than 1 billion people may die from smoking related diseases in the 21st century (1).

Nicotine is the main addictive compound found in tobacco products (3). It has been shown to induce rewarding behavior through stimulation of brain mesolimbic dopamine neurons (4), prompting addiction and directly instigating the withdrawal symptoms experienced by those trying to quit (5). Intuitively, the addictive properties of tobacco could be avoided if nicotine molecules were prevented from entering the brain. The resulting effect could be a significant reduction in the number of potential quitters that eventually relapse. One possible therapeutic option that applies this concept as a mode of stemming nicotine addiction is the nicotine vaccine.

There are currently a number of studies claiming the successful application of a conjugate nicotine vaccine in animals (5-24). In the majority of these studies, immunogenic carrier proteins are conjugated to nicotine molecules via a linker modification added to nicotine’s pyrrolidine ring or pyrodine ring (12). These newly formed protein-hapten conjugates are then injected into patients with an accompanying adjuvant, eliciting an immune response and prompting the generation of antibodies with

140 a high affinity for nicotine (25). After successful vaccination, exposure to tobacco products or byproducts results in nicotine being sequestered in the blood and extracellular fluid by the anti-nicotine antibodies (18). It is important to note that these antibodies do not “eliminate” nicotine from the body, but rather only temporarily impede the mobility and pharmacokinetic activity of nicotine molecules by forming a nicotine- antibody complex (9). Ultimately, the formation of these nicotine-antibody complexes induces a dampening effect that reduces spikes of nicotine concentration in the brain and provides a controlled, incremental release of nicotine in the blood (12). Dampening nicotine kinetics in this way has been proven to aid the dissolution of nicotine addiction in humans who achieve high immunogenic responses to conjugate nicotine vaccines

(26-28).

Appreciable anti-nicotine antibodies have been produced using protein-hapten conjugates, with some studies even proving successful enough to proceed to human clinical trials (26-31). However, none of these studies have proceeded past Phase III clinical trials due to vaccinated subjects’ failure to exhibit increased abstinence rates when compared to subjects given a placebo (25). The overall lack of positive results from human studies illuminates a need for new approaches to the nicotine vaccine design process. Therefore, the majority of current nicotine vaccine research focuses on the investigation of the key variables thought to have the most impact on this dampening effect, such as vaccine immunogenicity and vaccine-elicited antibodies’ affinity for nicotine and its major metabolites (25).

A review of factors that are suspected to have a significant impact on nicotine vaccine efficacy prior to additional animal and human experimentation could aid in

141 understanding the success gap between these trials. Additionally, it might prove particularly useful when evaluating human-specific factors that cannot be evaluated in animal studies, such as nicotine elimination rates, cross reactivity of the major nicotine metabolite cotinine with anti-nicotine antibodies, and those factors that have the most impact on the ability of nicotine to enter the brain, such as cerebral volume and blood flow rate. Now that sufficient public data exists on the subject of nicotine vaccination and nicotine metabolism, it may be prudent to employ an in silico method outside of in vivo experimentation to evaluate all of these variables.

As computational capacities continue to improve, in silico methods are becoming an increasingly relevant approach to drug evaluation outside of living systems. When paired with a pharmacokinetic model, they can effectively save time, money, and resources by predicting a drug’s disposition and pharmacokinetics prior to animal and human trials. Currently, the most widely used form of pharmacokinetic model for in silico research is the physiologically based pharmacokinetic model (PBPK) (32). These models are usually multi-compartmental constructs that utilize species-specific, tissue- to-blood partition coefficients coupled with tissue volumes or weights, average tissue blood flow rates, and mathematical relationships to make predictions about xenobiotic kinetics and disposition. Complexity of PBPK models varies greatly between individual studies, as their design is based upon the complexity of output variables and the desired scale and accuracy of model predictions. Many pharmacokinetic and toxicokinetic models of nicotine and cotinine disposition kinetics have been proposed in literature (33-39). However, to date there have been no models constructed that

142 evaluate the effect of anti-nicotine antibodies on nicotine disposition in the blood and in the brain.

In this study, data relevant to nicotine and cotinine metabolism, nicotine and cotinine affinity to anti-nicotine antibodies, maximal concentrations of antibody achieved in humans using clinically approved materials and procedures, and various other parameters necessary for the generation of a compartmental model were gathered for both rats and humans. All of this information was applied to two PBPK models (one rat and one human) in order to investigate key issues that are thought to negatively affect nicotine vaccine efficacy in terms of percent decrease in brain nicotine concentration.

The ultimate goal of this project was to provide a novel approach to evaluating the efficacy of potential nicotine vaccines prior to human trials and extended animal trials.

4.2. Methods

4.2.1. Approach to Model Construction

Human and rat pharmacokinetic data were obtained from literature and applied to two mechanistic, multi-compartmental nicotine and cotinine PBPK models constructed in

Matlab’s SimBiology application. Similar to existing studies, it was assumed that simple mass action kinetics and flow-limited tissue distributions could accurately predict nicotine and cotinine systemic distribution (34, 35, 38). Model design was inspired by previous works used to predict nicotine and cotinine pharmacokinetics and pharmacodynamics in both rats and humans, with the most notable works being those of Plowchalk (35) and Teeguarden (38). It was initially determined that a simple model consisting of only a blood, brain, and liver compartment would be sufficient to simulate nicotine metabolism and evaluate the effect of anti-nicotine antibodies on nicotine

143 systemic distribution, similar to the model proposed by Yamazaki (39). This approach was abandoned, however, due to evidence presented by Satoskar (23) that suggests reduced nicotine concentrations in the brain not only result from anti-nicotine antibodies sequestering nicotine in the blood, but also from the distribution of nicotine to the tissues being disproportionally redirected away from the brain. This effect could not be accounted for in the three compartment model.

Parameters acquired from literature were calibrated using existing nicotine and cotinine blood and tissue concentration time course data in rats (10, 12) and in humans (40-42).

Model accuracy simulations were subsequently performed. Calibration was achieved via

SimBiology’s data fit task followed by manual adjustment of parameters to provide an optimal visual fit. Less emphasis was placed on the accuracy of tissue concentrations while more emphasis was placed on the accuracy of both the venous and arterial blood concentrations due to the large variability in both calculated (43-45) and experimentally determined (33, 35, 46, 47) partition coefficients. However, a single exception to this approach was made for the nicotine brain-to-blood partition coefficient, since the evaluation of nicotine concentration in the brain was one of the main foci of this study.

4.2.2 Major Model Assumptions

Many key assumptions were made that allowed for the construction of the PBPK models described in this study. In order to include the effect of anti-nicotine antibodies on nicotine distribution, it was assumed that peak antibody concentrations had already been achieved via vaccination. These antibody concentration values were based off of previously published data and did not represent immunogenic capacity of nicotine vaccines as a whole. Making this assumption led to model predictions overestimating

144 vaccine efficacy at all time points other than when peak anti-nicotine antibodies had been achieved. This result was intentional, as these predictions could be viewed as good indicators of vaccine potential. Alternatively, different antibody concentrations could be applied to these models in order to predict vaccine efficacy at different time points during the course of the vaccination process. It was assumed that the anti- nicotine antibodies’ affinity for cotinine would be an important factor in the efficacy of any nicotine vaccine in humans, an assumption that was based on the large proportion of serum nicotine that is metabolized into cotinine, the long residence time of serum cotinine, and the structural similarities between the two molecules (5). Blood composition was assumed to be 55 parts serum to 45 parts red blood cells (48), values that allowed for the conversion of the serum antibody concentrations reported in literature to the whole blood antibody concentrations that were defined in the model.

Antibody binding capacities were calculated using whole blood antibody concentrations and the assumptions that each antibody can only bind a maximum of two nicotine molecules at any given time and that anti-nicotine antibodies have an average molecular weight of 150kDa (10, 12). Additionally, many more important assumptions were included in the model (following the conventions set by previous PBPK models), such as assuming instantaneous mixing within compartments (35, 38), partition coefficients independent of time and concentration parameters (35), minimal cellular uptake and protein binding of chemical species (42), complete access of entire blood volume to circulatory flow (35, 38), and even distribution of nicotine and cotinine between red blood cells and serum (39).

4.2.3. Compartmental Design

145

The majority of tissues included in the model were chosen based on their perceived relevance to overall nicotine and cotinine distribution and the availability of necessary data. Partitioning between compartments was achieved by using published tissue-to- blood partition coefficients (defined in this study as steady-state tissue to venous blood concentration ratios), tissue volumes, and tissue blood flow rates. The eleven compartments selected for use in this model are shown in Figure 4.1. Blood flow was modeled in a cyclic manner, with the lung compartment accounting for all blood flow between the venous and arterial compartments. Returning blood flow was distributed between the remaining interior compartments and was based on published and calibrated tissue blood flow rates. Partition coefficients were used to account for the flow-limited (42) distribution of nicotine and cotinine between blood and tissue.

The antibody sub-compartments illustrated in Figure 4.1 represent compartment specific anti-nicotine antibody species and the arrows represent antibody association and dissociation with nicotine (rat and human models) and cotinine (human model). Tissue retention kinetics (aggregation and dispersion) were conceptualized for the brain compartment in order to account for tissue concentration changes that could not be accounted for by using a partition coefficient alone. Antibody mobility between compartments was not accounted for in either the rat or the human model due to lack of necessary parameters, such as the tissue-to-blood partition coefficients associated with antibody systemic distribution. Blood volume was divided between arterial and venous blood, with arterial blood accounting for approximately three quarters of the total blood volume (49). Dosing was applied directly to the lung compartment when published, experimental data was obtained from cigarette smoking or nicotine inhalation. Following

146 similar logic, when published, experimental data were obtained from intravenous infusion, the simulation dose was applied to the venous blood compartment.

4.2.3.1. Clearance Organ Compartments

The final model design applied hepatic and renal elimination directly to the arterial blood compartment, which is similar to the approach by Teeguarden (38). This method can effectively account for these processes because the hepatic and renal elimination rates of both nicotine and cotinine are measured as a function of changes in blood concentration, as opposed to changes in liver and kidney concentrations (42). The metabolism of nicotine to cotinine in the human model was directed to the liver compartment to avoid immediate metabolism of cotinine. Being stripped of their primary physiological function, both the liver and the kidney compartments only served as partitions between the venous and arterial blood compartments. It was assumed that this function would still provide a significant contribution to model predictive accuracy, however, due to the large portion of cardiac output supplied to these tissues.

4.2.3.2. Calibration Compartment

Parameters associated with the ‘other’ compartment served as a means of model calibration for venous nicotine concentrations in both the human and rat model and for venous cotinine concentration in the human model when antibodies were absent. This seemed to be the most logical approach to calibration when considering that the ‘other’ compartment represents organs and tissues that are absent from both models due to lack of published parameters. Calibration of the blood flow rate in the rat model resulted in blood flow fractions that added to a value greater than one, indicating a

147 physiologically impossible situation in which blood flow to the tissues was greater than cardiac output. This issue could be corrected by providing adjusted blood flow fractions, but no adjustment was made due to priority being placed on providing references for all physiological parameters. Blood flow rate through the ‘other’ compartment in the human model was determined by calculating the cardiac output and then balancing that value versus all of the blood flow rates published for the remaining tissue compartments.

Partition coefficients for the ‘other’ compartment were also calibrated in both the rat and the human model.

4.2.3.3. Brain Compartment

Brain partition coefficient, blood flow rate, and retention parameters in the rat PBPK model were calibrated using data from Hieda (10). This calibration step resulted in brain partition coefficients that do not physiologically represent a true steady state tissue to blood concentration ratio but rather represent overall nicotine transport to the brain when paired with the tissue retention parameters. Published brain nicotine concentrations were converted from units of mass nicotine per mass tissue to units of moles nicotine per volume tissue using a mean brain tissue density value of 1.04 g/L

(50) so that they could be compared with brain nicotine concentrations predicted by the models. Identical calibration steps to those performed in the rat model involving the brain could not be completed in the human model due to the lack of available data. As such, the brain partition coefficient and retention parameters calibrated in the rat model were also applied to the human model.

4.2.4. Model Parameters

148

Parameters necessary for the construction of both the rat and human PBPK models were collected from various sources, with the majority consisting of averaged values

(non-specific for sex, ethnicity, breed, etc.). Mean tissue volume fractions were published as a function of subject mass and were obtained from literature for both the rat and human compartmental models (50). Nicotine partition coefficients associated with each tissue were obtained from Teeguarden (38) for both the rat and human compartmental models, with the exception of the calibrated values in the brain and

‘other’ compartments. Cotinine partition coefficients for all tissue compartments were calibrated visually using published starting values (47) and were only used in the human model where cotinine is a major metabolite. Cardiac output for the rat and human models were calculated based on subject mass (34, 50). Tissue blood flow rates were obtained from literature and based on a percentage of overall cardiac output (50), with the only exception being the calibrated ‘other’ compartment blood flow rate. Rat brain nicotine retention kinetics was accounted for by using dispersion and aggregation parameters. These values were calibrated by fitting model predictions for rat brain concentration with Hieda (10) brain nicotine concentration time course data via the data fit task in SimBiology. Human metabolism rates were obtained from literature (42) and rat metabolism rates were calibrated using Keyler (12) blood nicotine concentration time course data. Parameters used by the human and rat PBPK models can be found in

Tables 4.1-4.3.

4.2.5. Vaccine Efficacy Factors

The key variables that were used to evaluate vaccine efficacy in the human model were antibody concentration, antibody affinity for nicotine, and antibody cross-reactivity with

149 cotinine. These variables have been identified in previous studies as being the most important factors influencing nicotine access to the brain (25). The effect of resting cotinine levels (which are correlated with smoking status) on vaccine efficacy was also investigated. Increasing systemic antibody concentration allows more nicotine to be bound in the blood, decreasing the likelihood that free nicotine will be able to traverse the blood-brain barrier (13). Antibody affinity for nicotine characterizes nicotine-antibody interactions, which determine how quickly nicotine associates with and dissociates from anti-nicotine antibodies. Antibody cross-reactivity with cotinine may have a profound effect on the number of free antibodies able to bind nicotine when cotinine levels are elevated in the blood (5, 42). Finally, high resting cotinine levels could have a detrimental effect on vaccine efficacy, especially when anti-nicotine antibodies have a high affinity for cotinine. Different treatment levels of these key variables were selected using published values in human clinical trials as references and then applied to the human PBPK model in order to determine their effect on vaccine efficacy. Due to programing limitations in SimBiology, antibody affinity had to be divided into associative

(on) and dissociative (off) terms. This was achieved by estimating a reasonable antibody-nicotine dissociation rate using limitations described by Poulsen (51) and then using this value along with a published antibody affinity value to calculate the association rate. The cotinine dissociation rate was estimated to be double that of the nicotine dissociation rate. Anti-nicotine antibody affinity for cotinine had to be estimated using cotinine and nicotine IC50 (half maximal inhibitory concentration) ratios paired with nicotine antibody affinity values reported in literature. Since the ratio between off and on rates is what determines the extent of antibody binding, setting an antibody

150 dissociation rate that slightly deviates from the true pharmacokinetic value should have little influence on model predictions.

4.2.6. Calibration

Nicotine concentration time course data were obtained from published sources along with dosing, antibody concentration range, and mean antibody affinity data. The calibration of the rat PBPK model parameters was then performed in three subsequent steps. The first step involved calibrating the nicotine metabolism rate, the ‘other’ blood flow rate, and the ‘other’ partition coefficient using data published by Keyler (12). The second step involved taking these newly calibrated values and applying them to a second calibration of brain blood flow, partition coefficient, and retention parameters using data published by Hieda (10). The values obtained in the first two steps were then applied to a third and final calibration step in which antibody concentration was adjusted within the published range for each study. Calibrated values were set as primary parameters in the model and then simulations were run using dosing and mean antibody affinity data published alongside the nicotine time course concentration data in each respective study.

The human PBPK model was constructed by substituting physiological parameters in the rat model with human values obtained from literature. There was only one calibration step and it involved adjusting nicotine and cotinine partition coefficients for the ‘other’ compartment so that simulation predictions would fit nicotine and cotinine venous concentration time course data published by Benowitz (40). As was done with the rat model, calibrated values were set as primary parameters in the human model and then simulations were run using dosing data published alongside nicotine and

151 cotinine time course data in order to test model accuracy in unvaccinated subjects.

Accuracy of venous nicotine concentration predictions were tested using time course nicotine concentrations reported by Benowitz (41, 52) and accuracy of venous cotinine concentrations were tested using time course cotinine concentrations also reported by

Benowitz (52). Nicotine concentration time course data that included anti-nicotine antibody effect was unavailable, so model accuracy could not be verified for vaccinated subjects.

4.2.7. Human Vaccine Efficacy Predictions

One of the main goals of this project was to predict human vaccine efficacy via in silico methods without needing to commit to costly clinical trials. In order to do this, pre- existing data from human studies that quantified achievable antibody binding characteristics were paired with the human PBPK model described earlier and multiple simulations were run. A crude sensitivity analysis was designed in an attempt to identity antibody binding characteristics that would have the most impact on preventing nicotine from reaching the brain, and this analysis was followed by several simulations aimed at predicting the maximum effect antibodies could have on the distribution of nicotine to the brain in humans under plausible physiologic conditions.

4.2.7.1. Sensitivity Analysis

Simulations using the human model were conducted using all possible combinations of the following treatments and treatment levels: high and low antibody binding capacity, high and low antibody affinity for nicotine, and high and low cross reactivity with cotinine. High and low antibody affinity for nicotine was defined as the maximum and

152 minimum values of the range reported by Maurer (30), respectively. In order to quantify antibody affinity for cotinine, two cross reactivity factors were defined using the ratio of cotinine and nicotine IC50 values for the worst performing hapten (GK56KLH) and the best performing hapten (IP31KLH) described by de Villiers (5). These cross reactivity factors were then multiplied by the antibody affinity for nicotine treatment levels, resulting in four test cotinine affinities for anti-nicotine antibodies. High and low antibody binding capacities were concluded using the mean and standard deviation for antibody concentrations reported in multiple human trials (see Table 4.9) (27, 28, 31). The low antibody binding capacity was defined as the mean value from these studies while the high antibody binding capacity was defined as this same mean value plus three standard deviations. Eight simulations were run using all possible variable combinations for smokers (who maintain baseline cotinine concentrations) and non-smokers (who do not maintain baseline cotinine concentrations). These simulations were also paired with control simulations in which anti-nicotine antibodies were not present. Reduction in brain nicotine concentration with respect to control concentrations were calculated and plotted versus cigarette number.

4.2.7.2. Predicting Maximal Antibody Effect in Humans

The initial vaccine efficacy predictions made by the human model seemed less promising than expected when compared to results published from rat studies in which brain nicotine concentration reductions of up to 90% were reported (25). As a result, additional tests were designed in order to predict the maximum reduction in human brain nicotine concentration that could theoretically be achieved using first generation nicotine vaccination technologies and approaches. Geometric mean antibody

153 concentration time course data paired with vaccine and booster shot application times published by Hastukami (27) were used to predict maximal antibody concentrations when utilizing the most beneficial time intervals between booster applications

(measured as the average time between each booster application and the resulting maximum serum antibody concentration). The total number of booster applications was varied using this prediction method in order to generate plausible serum antibody concentration test values. These data can be found in Table 4.6 (6B16W refers to administration of vaccine and six subsequent booster shots over the course of sixteen weeks, with 10B24W and 14B32W following identical nomenclature). The total dose per cigarette was not changed from the assigned value in earlier tests, but the dose time, dose rate, and recorded total dose were modified to make predictions better represent smokers who are trying to quit (the most likely candidate for clinical use of a nicotine vaccine). Other alterations to the model included changing the affinity for nicotine value to that of the average value in one set of simulations and the maximal value for a second set of simulations (values were derived from the range reported by Maurer (30)) and using only the smallest IC50 ratio reported by de Villiers (5) as a cotinine cross- reactivity factor. Simulations were designed around three variables (antibody concentration, antibody affinity for nicotine, and smoking status) and were run in addition to a control simulation for each consecutive dose. Reductions in brain nicotine concentration with respect to control concentration were calculated and plotted versus cigarette number.

4.3. Results

4.3.1. Rat Model Calibration

154

The results from the rat model calibration simulations can be seen in Figures 4.2-4.4.

Figure 4.2 shows the successful calibration of the rat model using venous nicotine concentration time course data and antibody binding characteristics (see Table 4.4) reported by Keyler (12) in response to both the CMUNic vaccine as well as the negative control. The model simulation output was able to remain well within the standard deviations defined for each data point, indicating proximity of predictions to reality over the range of time described in the data set. The success of this calibration step was confirmed in Figure 4.3 by the close proximity of model predictions to the venous nicotine time course data reported by Hieda (10) when using the newly calibrated values and the antibody binding characteristics defined for the new data set (see Table

4.5). Figure 4.4 shows the successful calibration of the rat model in terms of brain nicotine concentration using data again described by Hieda (10). The close proximity of the model output to the above-mentioned time course data (values are within the standard deviation defined for each data point) again indicates proximity of predictions to reality.

4.3.2. Human Model Calibration and Testing

The results from human model calibration and accuracy testing simulations can be seen in Figures 4.5-4.8. Figure 4.5 shows the successful calibration of human model venous nicotine and cotinine concentration predictions using time-course data reported by

Benowitz (40) (dosing information can be found in Table 4.6). Predictions remain within one standard deviation of each data point with the exception of two cotinine concentration values at 2 and 3 hours that appear to be outliers within the published data set. Accuracy of these newly calibrated values was confirmed using controlled-

155 dose smoking data (see Table 4.7) reported by Benowitz (41) and unrestricted-dose smoking data (see Table 4.8) reported by Benowitz (52). Figure 4.6 shows that the calibrated human model can approximate nicotine venous concentrations in smokers when applying lower and average nicotine doses (0.4 mg and 1.2 mg of nicotine per cigarette), but there appears to be sizeable discrepancy between model predictions and published time-course data when applying higher nicotine doses (2.5 mg of nicotine per cigarette). This is most likely due to self-regulation of nicotine uptake during the administering of higher doses (41). Figure 4.7 and Figure 4.8 show that the calibrated human model can accurately approximate nicotine and cotinine venous concentrations, respectively, at dosing rates comparable to those that smokers would typically experience. Additionally, the simulation results in these figures have not been adjusted to averaged peak-trough values, so they may serve as better representatives of nicotine and cotinine concentrations in venous blood upon nicotine inhalation.

4.3.3. Sensitivity Analysis Results

The sensitivity analysis results can be seen in Figure 4.9 (active smokers with high resting cotinine systemic concentrations) and Figure 4.10 (non-smokers with negligible resting cotinine systemic concentrations). Together, Figures 4.9 and 4.10 show that antibody binding capacity, affinity for nicotine, and cross reactivity with cotinine have a sizeable impact on the ability of potential nicotine vaccines to prevent nicotine from entering the brain. Additionally, the ability of vaccine-elicited antibodies to impede nicotine’s passage into the brain was reduced upon sequential, in silico dosing with nicotine (for dosing information, see Table 4.10). The percent decrease in brain nicotine concentration never reaches zero, however, but rather approaches a constant value

156 that appears to depend on the initial antibody binding characteristic treatment combination and the nicotine dosing rate. When comparing Figure 4.9 and Figure 4.10, the sensitivity analysis results showed that resting cotinine levels and smoking status have a detrimental effect on nicotine vaccine performance, especially when the elicited antibodies have a high affinity for cotinine. The results obtained from simulations using low cotinine cross-reactivity factors were similar in models describing both smokers and non-smokers. Also, the largest antibody effect predicted in both smokers and non- smokers was obtained when a high antibody concentration, a high antibody affinity for nicotine, and a low antibody affinity for cotinine were used in the simulation. In contrast, the smallest antibody effect was predicted in simulations that included a low antibody concentration, a low antibody affinity for nicotine, and a high antibody affinity for cotinine. Legend nomenclature found in each figure refers to the combination of various levels for each test parameter, with NhChAbh representing the simulation run using high antibody affinity for nicotine (Nh), high cotinine cross-reactivity factor value (Ch), and high blood antibody concentration (Abh). Similarly, NlClAbl represents the simulation run using low antibody affinity for nicotine (Nl), low cotinine cross-reactivity factor value

(Cl), and low blood antibody concentration (Abl).

4.3.4. Maximal Antibody Effect Prediction Results

When maximal antibody effect predictions were made using hypothetical antibody concentration thresholds within the human model, the largest reductions in brain nicotine concentration were ~22 percent in both smokers and non-smokers when using an averaged antibody-nicotine affinity value (Figure 4.11) and ~26 percent and ~27 percent in smokers and non-smokers, respectively, when using a maximized antibody

157 affinity for nicotine value (Figure 4.12). The ability of vaccine elicited antibodies to impede nicotine’s passage into the brain was reduced after each successive nicotine dose, a similar effect to that seen in the sensitivity analysis results. When comparing

Figures 4.11 and 4.12, it appears that there are fewer discrepancies between model predictions in smokers and non-smokers when a low cotinine cross-reactivity factor is supplied to the model. This result was also observed when performing the sensitivity analysis. Additionally, increasing the affinity of anti-nicotine antibodies for nicotine from

44.05 µM-1 to 71.43 µM-1 resulted in an additional decrease in brain nicotine concentration of ~3, ~2.5, and ~2 percent for the first, second, and third nicotine doses, respectively.

4.4. Discussion

Predictions provided by the rat PBPK model fit well with published, experimental data, confirming that this model could be a good predictor of whole blood and brain nicotine concentrations in rats with and without anti-nicotine antibodies present. It is worth noting that the model predicted the steady-state brain nicotine half-life in rats without antibody action to be approximately 40 min, which is comparable to the 52 min described by

Ghosheh (53). The data presented herein indicates a successful conceptualization and construction of a rat PBPK model that can evaluate nicotine vaccine efficacy. However, additional confirmation of model prediction accuracy should be performed in the future when more data becomes available.

Agreement between simulations and experimental data confirmed the validity of the human PBPK model in terms of predicting venous nicotine and cotinine concentrations without the presence of anti-nicotine antibodies. Given the accuracy of the predictions

158 made by the rat model and the similarity between rat and human nicotine binding kinetics in the brain reported by Court (54), it was assumed that a similar level of accuracy would be achieved in the human model. However, assuming that the brain partition coefficient and retention parameters found in rats are comparable to those found in humans may lead to some inaccuracies in model predictions due to variable interspecies brain pharmacokinetics (55). While it is unfortunate that the physiological accuracy of key human model predictions cannot be confirmed, applying the brain pharmacokinetic parameters calibrated in the rat model to the human model was necessary in order to draw conclusions about nicotine vaccine efficacy in humans.

Various pieces of published data support the relative accuracy of the brain nicotine concentration predictions described in this study. First, the nicotine concentration half- life predicted in human brains when antibody effect is not applied (~43 min) is comparable to that illustrated in Nyback (56) for the active stereochemical form of nicotine in nonsmokers (~39 min). Second, the range of brain nicotine concentration reductions predicted during the maximal antibody effect study (17-26 percent) encompasses the value of 23.6 percent reported by Esterlis (29). Therefore, percent reduction in nicotine brain concentrations using this model could remain accurate even in the event that predictions produced by the human model vary from their true physiological values. While these results certainly do not confirm the accuracy of the human PBPK model predictions, it can be argued that they do present evidence of model proximity to reality.

The investigation of antibody specific factors that contribute to nicotine vaccine efficacy in humans provided results indicating that smoking status appears to have a sizable

159 impact on nicotine access to the brain. This is especially true when a high cotinine cross-reactivity factor is defined for the simulation, indicating that increased antibody affinity for cotinine is detrimental to the effectiveness of any nicotine vaccine that is intended for use in humans. Since it is assumed that active smokers will occasionally

“cheat” during quitting attempts, controlling this factor could prove to be vital to vaccine success in humans. Consequently, results from these simulations strongly support the hypothesis that choice of nicotine hapten could play a significant role in vaccine efficacy due to its profound effect on the affinity of anti-nicotine antibodies for both nicotine and cotinine. The human PBPK model predicts that increasing antibody affinity for nicotine and antibody concentration leads to decreased nicotine access to the brain, a result that is intuitive in nature. The greatest reduction in brain nicotine concentration that could be achieved using the most optimal antibody pharmacokinetic parameters published from human trials was ~15 percent in smokers and ~16 percent in non-smokers after the first in silico dose, which is much lower than many of the reductions that have been published in animal studies. A steady drop off in the amount of nicotine prevented from entering the brain occurred over the next doses, indicating antibody saturation and competition with newly formed cotinine for antibody binding sites. This drop off did not approach zero when additional doses were applied, however, but rather converged on a constant value, indicating that antibody saturation does not completely nullify the impact anti-nicotine antibodies have on brain nicotine concentrations. The results from these simulations could explain the issues that have been encountered in past clinical trials.

Additionally, these results seem to indicate that if future nicotine vaccines cannot elicit

160 the production of antibodies with more beneficial pharmacokinetic properties, achieving desired clinical effectiveness might be difficult.

The brain nicotine concentration reductions achieved after the initial in silico dose in the maximal antibody effect prediction simulations are above the 23.6 percent described by

Esterlis (29) that coincides with a 50 percent reduction in cigarette use. The model predicted brain nicotine levels to decline by ~20 percent in non-smokers and ~16 percent in smokers when compared with controls after applying the third and final nicotine dose, values that were below the 50 percent effectiveness threshold described earlier. This result indicates that the effectiveness of nicotine vaccines may be reduced upon the application of additional nicotine doses. It is worth noting that the antibody binding capacities applied within these simulations were much higher than those that have been achieved in humans to date. Serum binding capacities of greater than 1 µM

(~75 µg/mL) have been achieved in many rat studies (12, 14, 19), so it is not a stretch to assume that the same can be accomplished in humans after extensive optimization of the nicotine vaccine concept. Until this binding capacity is achieved, however, the reduction in brain nicotine concentration elicited by nicotine vaccines in humans is predicted to remain below the threshold described by Esterlis (29).

Most variables used in this study were obtained from literature and in many cases they were not checked for physiological accuracy. Additionally, parameters that were not obtained from literature were either assumed or calculated/estimated using available data. Great effort was made to create two PBPK models that could predict the concentration of nicotine in the blood and brain of rats and humans, but the accuracy of concentration predictions in other tissues included in these models was largely ignored.

161

Consequently, the accuracies of all nicotine concentration predictions in the rat model compartments other than those of the brain and mixed venous/arterial blood are questionable and should not be used to make conclusions about physiological processes involving nicotine. This is equally true for the cotinine and nicotine concentration predictions outside of the venous blood compartment in the human model.

To the best of our knowledge, this has been the first attempt to quantify anti-nicotine antibody effect on nicotine systemic distribution in both rats and in humans by using

PBPK modeling. The work outlined in this study is preliminary in nature and will require refinement in the future. The lack of published, comparative nicotine concentration data in humans vaccinated against nicotine makes it impossible to determine prediction accuracy, and it is possible that there will never be sufficient data available to completely validate the human model in terms of the effect anti-nicotine antibodies have on nicotine distribution to the brain. Also, there may be additional biological mechanisms that affect the distribution of nicotine in the presence of anti-nicotine antibodies that have not been considered in either of the models described in this study.

Until the accuracy of the human PBPK model outputs can be validated, predictions should only be viewed as guidelines that can direct the development of future human nicotine vaccines and not as true representations of physiological concentrations.

However, this model could potentially be used to predict brain nicotine concentration reductions in rats and humans without requiring brain removal based on measurements in antibody titer and affinity. A possible benefit of the in silico modeling in combination with animal trials would be the reduction in study costs and the number of experiments

162 necessary to quantify antibody effect. The model may also be used to assess brain nicotine levels in humans upon nicotine challenge using indirectly measured parameters. Applying the model to different species for further vaccine analysis is also possible through slight modification, as has already been demonstrated with the construction of the human PBPK model within the scope of this study.

4.5. Conclusions

In this study, two physiologically based pharmacokinetic models were successfully constructed in order to investigate the effects of anti-nicotine antibodies on nicotine disposition in the brains of rats and humans. The good fit between predicted and published blood and brain nicotine concentrations in the rat model suggests that this approach can effectively quantify the complex, biological processes associated with nicotine pharmacokinetics and nicotine-antibody interactions. Investigation of the key parameters presumed to have the most effect on the ability of anti-nicotine antibodies to prevent nicotine from entering the brain in the human model, such as antibody affinity for nicotine, antibody cross-reactivity with cotinine, and overall antibody concentration, confirmed the conclusions made in previous studies about antibody binding characteristics necessary for a successful vaccine. Predictions made by the human model also suggest that the antibodies elicited by current vaccines do not have sufficient binding characteristics to reduce brain nicotine concentrations to clinically significant levels. However, the human model also predicts that if antibody binding characteristics similar to those achieved in animal studies can be achieved in human studies, human vaccine efficacy in terms of brain nicotine concentration reduction should meet the threshold values for smoking cessation that have been recently

163 described in literature. Ultimately, both of the models presented in this study can be used to predict nicotine vaccine efficacy in terms of brain nicotine concentration reduction prior to human trials and extended animal studies, a feature that can save both time and money in future endeavors for nicotine vaccine development.

Conflict of Interest Statement

The authors declare that they have no conflicts of interest.

Acknowledgments

The authors are indebted to Frank Gillam for the insight he has provided for this project, as well as to Dr. Yun Hu and Zongmin Zhao for supplementing our understanding of the nicotine vaccine concept. This work was financially supported by the National Institutes of Health, more specifically, the National Institute on Drug Abuse (U01DA036850).

164

Tables and Figures

Table 4.1. Nicotine-cotinine pharmacokinetic model parameters for human and rat

model. Table 1 Nicotine–cotinine pharmacokinetic model parameters for human and rat model. Model parameter Human Value Source Rat Value Source Tissue volume (fraction of body weight) Lung 0.0076 Brown et al. 1997 0.0048 Brown et al. 1997 Heart 0.0050 Brown et al. 1997 0.0029 Brown et al. 1997 Brain 0.0193 Brown et al. 1997 0.0058 Brown et al. 1997 Adipose 0.2326 Brown et al. 1997 0.0761 Brown et al. 1997 Liver 0.0248 Brown et al. 1997 0.0324 Brown et al. 1997 Kidney 0.0038 Brown et al. 1997 0.0067 Brown et al. 1997 Skin 0.0339 Brown et al. 1997 0.1743 Brown et al. 1997 Muscle 0.3883 Brown et al. 1997 0.3922 Brown et al. 1997 Other 0.1940 Brown et al. 1997 0.1965 Brown et al. 1997 Arterial Blood 0.0190 Brown et al. 1997 0.0178 Brown et al. 1997 Venous Blood 0.0570 Brown et al. 1997 0.0534 Brown et al. 1997 Nicotine partition coefficients Lung 0.870 Teeguarden et al. 2013 0.870 Teeguarden et al. 2013 Heart 0.490 Teeguarden et al. 2013 0.490 Teeguarden et al. 2013 Brain 2.650 Rat Data 2.650 Hieda Calibration Adipose 0.190 Teeguarden et al. 2013 0.190 Teeguarden et al. 2013 Liver 7.600 Teeguarden et al. 2013 7.600 Teeguarden et al. 2013 Kidney 7.600 Teeguarden et al. 2013 7.600 Teeguarden et al. 2013 Skin 0.970 Teeguarden et al. 2013 0.970 Teeguarden et al. 2013 Muscle 1.000 Teeguarden et al. 2013 1.000 Teeguarden et al. 2013 Other 13.721 Benowitz Infusion Calibration 7.100 Keyler Calibration Cotinine partition coefficients Lung 0.901 Optimized, ʎapriori starting values from Thompson et al. 2012

Heart 0.832 Optimized, ʎapriori starting values from Thompson et al. 2012

Brain 0.892 Optimized, ʎapriori starting values from Thompson et al. 2012

Adipose 0.828 Optimized, ʎapriori starting values from Thompson et al. 2012

Liver 0.770 Optimized, ʎapriori starting values from Thompson et al. 2012

Kidney 0.822 Optimized, ʎapriori starting values from Thompson et al. 2012

Skin 1.019 Optimized, ʎapriori starting values from Thompson et al. 2012

Muscle 1.178 Optimized, ʎapriori starting values from Thompson et al. 2012 Other 0.658 Benowitz Infusion Calibration

165

Table 4.2. Nicotine–cotinine pharmacokinetic model parameters for human and

rat model, cont.

Model parameter Human Value Source Rat Value Source Mean Body Weight, BW (kg) 70 - 0.35 - 0.74 0.75 Cardiac Output (L/hr) 15*BW(kg) Gajewska et al. 2014 14.2*BW(kg) Brown et al. 1997 Blood flow (fraction of cardiac output) Lung 1.000 Brown et al. 1997 1.000 Brown et al. 1997 Heart 0.040 Brown et al. 1997 0.051 Brown et al. 1997 Brain 0.114 Brown et al. 1997 0.046 Hieda Calibration Adipose 0.052 Brown et al. 1997 0.070 Brown et al. 1997 Liver 0.227 Brown et al. 1997 0.183 Brown et al. 1997 Kidney 0.175 Brown et al. 1997 0.141 Brown et al. 1997 Skin 0.058 Brown et al. 1997 0.058 Brown et al. 1997 Muscle 0.191 Brown et al. 1997 0.278 Brown et al. 1997 Other 0.143 Flow Balance 0.178 Keyler Calibration Tissue Ab Conc. (fraction of serum Ab conc.) Lung 0.029 Satoskar et al. 2003 0.029 Satoskar et al. 2003 Heart 0.042 Satoskar et al. 2003 0.042 Satoskar et al. 2003 Brain 0.002 Satoskar et al. 2003 0.002 Satoskar et al. 2003 Adipose 0.125 Satoskar et al. 2003 0.125 Satoskar et al. 2003 Liver 0.029 Satoskar et al. 2003 0.029 Satoskar et al. 2003 Kidney 0.03 Satoskar et al. 2003 0.03 Satoskar et al. 2003 Skin 0.01 Satoskar et al. 2003 0.01 Satoskar et al. 2003 Muscle 0.05 Satoskar et al. 2003 0.05 Satoskar et al. 2003 Other 0.047 Satoskar et al. 2003 0.047 Satoskar et al. 2003

Nicotine tissue retention (1/sec) Brain nicotine dispersion 0.00343 Rat Value 0.00343 Hieda Calibration Brain nicotine aggregation 0.00201 Rat Value 0.00201 Hieda Calibration

166

Table 4.3. Nicotine–cotinine pharmacokinetic model parameters for human and

rat model, cont.

Model parameter Value Source Human Clearances (L/h) Nicotine total clearance 79.0720 Hukkanen et al. 2005 Nicotine renal clearance 3.7520 Hukkanen et al. 2005 Nicotine hepatic clearance 75.3200 Hukkanen et al. 2005 Fraction nicotine hepatic clearance to cotinine 0.8000 Hukkanen et al. 2005 Cotinine total clearance 3.2720 Hukkanen et al. 2005 Cotinine hepatic clearance 2.9120 Hukkanen et al. 2005 Cotinine renal clearance 0.3600 Hukkanen et al. 2005 Rat Clearances (L/h) Nicotine total clearance 0.756 Keyler Calibration Molecular Weights (g/mol) MW of nicotine 162.23 MW of cotinine 172.22 MW of antibody 150000 Other parameters Fraction plasma in whole blood 0.55 Dill et al. 1974 Fraction nicotine protein binding 0.05 Hukkanen et al. 2005 Fraction cotinine protein binding 0.05 Hukkanen et al. 2005 Nicotine concentration ratio of blood to plasma 1 Yamazaki et al. 2010 Cotinine concentration ratio of blood to plasma 1 Yamazaki et al. 2010 Antibody concentration ratio of blood to plasma 0.55 Percent total blood volume in arterial compartment 0.25 Percent total blood volume in venous compartment 0.75 Baseline systemic cotinine in smokers (µM) 1.2000 Simulation Prediction Antibody decomposition rate (day-1) 0.01155 Maurer et al.

Table 4.4. Keyler et al. 1999 published dose and antibody parameters.

Keyler et al. 1999 Published Dose and Antibody Parameters: Dose (mg nicotine/kg rat mass) 0.1 Dose time (sec) 60 Mean antibody serum concentration (mg/ml) 0.13±0.10 Calibrated antibody serum concentration (mg/ml) 0.149 Mean antibody binding affinity (M-1) 1.83x108 Dose compartment Venous Blood

167

Table 4.5. Hieda et al. 1999 published dose and antibody parameters.

Hieda et al. 1999 Published Dose and Antibody Parameters: Dose (mg nicotine/kg rat mass) 0.03 Dose time (sec) 10 Mean antibody serum concentration (mg/ml) 0.38±0.32 Calibrated antibody serum concentration (mg/ml) 0.423 Mean antibody binding affinity (M-1) 1.4x107 Dose compartment Venous Blood

Table 4.6. Benowitz et al. 1991 published dosing parameters.

Benowitz et al. 1991 Published Dosing Parameters Dose (mg nicotine/kg human mass) 0.288 Dose time (hours) 24 Dose compartment Venous Blood

Table 4.7. Benowitz et al. 1982 published dosing parameters.

Benowitz et al. 1982 Published Dosing Parameters High Dose (mg/cigarette) 2.5 Usual Brand Dose (mg/cigarette) 1.2 Low Dose (mg/cigarette) 0.6 Dose time (min) 4 Dose rate (cigarette/hour) 2 Total Dose (cigarette) 30 Dose compartment Lungs

168

Table 4.8. Benowitz et al. 1983 estimated dosing parameters.

Benowitz et al. 1983 Estimated Dosing Parameters Dose (mg/cigarette) 1.2 Dose time (min) 4 Dose rate (cigarette/hour) 1.5 Total Dose (cigarette) 24 Dose compartment Lungs

Table 4.9. Human serum nicotine-specific antibody concentrations from clinical

trials.

Human Serum Nicotine-Specific Antibody Concentrations from Clinical Trials Source Max Serum Concentration (µg/mL) Hatsukami et al. 2005 27 Hatsukami et al. 2011 44 Wagena et al. 2008 10.8 Parameter Value Mean (µg/mL) 27.27 Standard Deviation (µg/mL) 16.60

Table 4.10. Human vaccine efficacy dosing parameters #1.

Human Vaccine Efficacy Dosing Parameters #1 Dose (mg/cigarette) 1.2 Dose time (min) 5 Dose rate (cigarette/hour) 1.5 Total Dose (cigarette) 20 Dose compartment Lungs

169

Table 4.11. Human vaccine efficacy prediction parameters. Human Vaccine Efficacy Prediction Parameters Model parameter Value Source Serum Concentration (µg/mL) High 77.07 Mean + 3*SD Low 27.27 Mean Binding Capacity (µM) High 0.57 Calculation Low 0.2 Calculation -1 Affinity for Nicotine, KA,Nic (µM ) High 71.43 Maurer et al. 2005 Low 16.67 Maurer et al. 2005 Cotinine Cross-reactivity Factor, CRFCot High 1.183E-01 de Villiers et al. 2010 Low 7.321E-03 de Villiers et al. 2010 -1 Affinity for Cotinine, KA,Cot (µM ) High KA,Nic*CRFCot Estimate

Low KA,Nic*CRFCot Estimate -1 Estimated Antibody Off Rates (sec ) Nicotine, koff,Nic 5.50E-04 Estimate

Cotinine, koff,Cot 1.10E-03 Estimate -1 -1 Antibody On Rates (µM sec ) Nicotine, kon,Nic koff,Nic*KA,Nic Calculation

Cotinine, kon,Cot koff,Cot*KA,Cot Calculation

Table 4.12. Human vaccine efficacy dosing parameters #2.

Human Vaccine Efficacy Dosing Parameters #2 Dose (mg/cigarette) 1.2 Dose time (min) 5 Dose rate (cigarette/hour) 1 Total Dose (cigarette) 3 Dose comparment Lungs

170

Table 4.13. Human vaccine concentration effect parameters.

Human Vaccine Concentration Effect Parameters Model Parameter Value Source Serum Antibody Concentration (µg/mL) 6B16W 94.3 Estimate 10B24W 128.6 Estimate 14B32W 162.9 Estimate Blood Antibody Binding Capacity (µM) 6B16W 0.6915 Calculation 10B24W 0.9431 Calculation 14B32W 1.195 Calculation -1 Affinity for Nicotine (µM ) KA,Nic 44.05 Maurer et al. 2005 Cotinine Cross-reactivity Factor CRFCot 7.321E-03 de Villiers et al. 2010 -1 Affinity for Cotinine (µM ) KA,Cot 0.3225 Estimate -1 Estimated Antibody Off Rates (sec ) Nicotine, koff,Nic 5.50E-04 Estimate

Cotinine, koff,Cot 1.10E-03 Estimate -1 -1 Antibody On Rates (µM sec ) Nicotine, kon,Nic 2.42E-02 Calculation

Cotinine, kon,Cot 3.55E-04 Calculation

171

Table 4.14. Human vaccine concentration effect parameters.

Model Parameter Value Source Serum Antibody Concentration (µg/mL) 6B16W 94.3 Estimate 10B24W 128.6 Estimate 14B32W 162.9 Estimate Blood Antibody Binding Capacity (µM) 6B16W 0.6915 Calculation 10B24W 0.9431 Calculation 14B32W 1.195 Calculation -1 Affinity for Nicotine (µM ) KA,Nic 71.43 Maurer et al. 2005 Cotinine Cross-reactivity Factor CRFCot 7.321E-03 de Villiers et al. 2010 -1 Affinity for Cotinine (µM ) KA,Cot 0.5229 Estimate -1 Estimated Antibody Off Rates (sec ) Nicotine, koff,Nic 5.50E-04 Estimate

Cotinine, koff,Cot 1.10E-03 Estimate -1 -1 Antibody On Rates (µM sec ) Nicotine, kon,Nic 3.93E-02 Calculation

Cotinine, kon,Cot 5.75E-04 Calculation

172

Figure 4.1. Overall schematic for the rat and human PBPK models.

Grey boxes represent antibody binding nicotine (and cotinine in the human model), blue boxes represent tissue retention of nicotine, and arrows represent nicotine and cotinine mass action kinetics or metabolism.

173

Figure 4.2. Rat whole blood nicotine concentration – PBPK model calibration.

Rat whole blood nicotine concentration prediction using parameters listed in Tables 4.1-

4.3 along with serum nicotine concentration time course data reported by Keyler (12).

Whole blood refers to the pooling of both the venous and arterial blood compartments and then re-evaluation of nicotine blood concentration. Serum was acquired from trunk blood (mixed venous and arterial blood). Simulation predictions were compared to experimental data collected with (CMUNic) and without (control) the presence of anti- nicotine antibodies. Additional simulation-specific parameters can be viewed in Table

4.4.

174

Figure 4.3. Rat whole blood nicotine concentration – PBPK model validation.

Rat whole blood nicotine concentration prediction using parameters listed in Tables 4.1-

4.3 along with serum nicotine concentration time course data reported by Hieda (10).

Whole blood refers to the pooling of both the venous and arterial blood compartments and then re-evaluation of nicotine blood concentration. Serum was acquired from trunk blood (mixed venous and arterial blood). Simulation predictions were compared to experimental data collected with (CMUNic) and without (control) the presence of anti- nicotine antibodies. Additional simulation-specific parameters can be viewed in Table

4.5.

175

Figure 4.4. Rat brain nicotine concentration – PBPK model calibration.

Rat brain nicotine concentration prediction using parameters listed in Tables 4.1-4.3 along with brain nicotine concentration time course data reported by Hieda (10).

Simulation predictions were compared to experimental data collected with (CMUNic) and without (control) the presence of anti-nicotine antibodies. Additional simulation- specific parameters can be viewed in Table 4.5.

176

Figure 4.5. Human venous blood nicotine and cotinine concentrations – PBPK model calibration.

Human model simulation of venous nicotine and cotinine concentrations was calibrated using published time course data reported by Benowitz (40) for a 24 hour intravenous nicotine infusion. Hour zero represents the start of infusion. Additional simulation- specific parameters can be viewed in Table 4.6.

177

Figure 4.6. Human venous blood nicotine concentrations over varying dosages –

PBPK model validation.

Venous nicotine time course data were reported by Benowitz (41). Simulation data points are presented as the average of trough (prior to dose) and peak (post dose) nicotine concentrations. Hour zero represents the start of smoking. Additional simulation-specific parameters can be viewed in Table 4.7.

178

Figure 4.7. Human venous blood nicotine concentrations – PBPK model validation.

Venous nicotine time course data that were originally reported by Benowitz (52) were obtained from Hukkanen (42). Nicotine concentrations were achieved during

‘unrestricted smoking’ and no dosing regimen was described. As such, a reasonable dosing regimen was estimated. Simulation data are presented with trough (prior to dose) and peak (post dose) nicotine concentrations. The simulation was conducted for a week in silico and results were recorded during the final day. Hour zero represents the start of smoking. Additional simulation-specific parameters can be viewed in Table 4.8.

179

Figure 4.8. Human venous blood cotinine concentrations – PBPK model validation.

Venous cotinine time course data that were originally reported by Benowitz (52) were obtained from Hukkanen (42). Cotinine concentrations were achieved during

‘unrestricted smoking’ and no dosing regimen was described. As such, a reasonable dose regimen was estimated. Simulation data is presented with trough (prior to dose) and peak (post dose) cotinine concentrations. The simulation was conducted for a week in silico and results were recorded during the final day. Hour zero represents the start of smoking. Additional simulation-specific parameters can be viewed in Table 4.8.

180

Figure 4.9. Evaluating the impact of antibody concentration, antibody affinity, and antibody cross-reactivity with cotinine on nicotine-specific antibody binding capacity – smokers with high resting cotinine levels.

Vaccine efficacy predictions for active smokers presented as the percent decrease in peak brain nicotine concentration when compared with control values. Cigarette number represents sequential dosing of nicotine to the lung compartment over time (refer to Table 4.10). Additional simulation-specific parameters can be viewed in Table 4.11.

181

Figure 4.10. Evaluating the impact of antibody concentration, antibody affinity, and antibody cross-reactivity with cotinine on nicotine-specific antibody binding capacity – non-smokers with negligible resting cotinine levels.

Vaccine efficacy predictions for non-smokers presented as the percent decrease in peak brain nicotine concentration when compared with control values. Cigarette number represents sequential dosing of nicotine to the lung compartment over time (refer to Table 4.10). Additional simulation-specific parameters can be viewed in Table 4.11.

182

Figure 4.11. Vaccine efficacy predictions in smokers and non-smokers – average antibody affinity for nicotine.

Vaccine efficacy predictions for smokers (S) and non-smokers (NS) using various antibody concentration values. Level one (L1), level two (L2), and level three (L3) antibody concentrations coincide with the 6B16W, 10B24W, and 14B32W values presented in Table 4.6, respectively. Data are presented as the percent decrease in peak brain nicotine concentration when compared with control values. The average antibody affinity for nicotine value was used in each simulation. Cigarette number represents sequential dosing of nicotine to the lung compartment over time (refer to

Table 4.12). Additional simulation-specific parameters can be viewed in Table 4.13.

183

Figure 4.12. Vaccine efficacy predictions in smokers and non-smokers – maximal antibody affinity for nicotine.

Vaccine efficacy predictions for smokers (S) and non-smokers (NS) using various antibody concentration values. Level one (L1), level two (L2), and level three (L3) antibody concentrations coincide with the 6B16W, 10B24W, and 14B32W values presented in Table 4.6, respectively. Data are presented as the percent decrease in peak brain nicotine concentration when compared with control values. The maximal antibody affinity for nicotine value (Ka) was used in each simulation. Cigarette number represents sequential dosing of nicotine to the lung compartment over time (refer to Table 4.12). Additional simulation-specific parameters can be viewed in Table 4.14.

184

References

1. World Health Organization. WHO report on the global tobacco epidemic, 2013 : enforcing bans on tobacco advertising, promotion and sponsorship.(in IRIS). Geneva:

World Health Organization; 2013. 202 p. p.

2. Messer K, Trinidad DR, Al-Delaimy WK, Pierce JP. Smoking cessation rates in the United States: a comparison of young adult and older smokers. Am J Public Health.

2008;98(2):317-22.

3. Benowitz NL. Pharmacology of nicotine: addiction, smoking-induced disease, and therapeutics. Annu Rev Pharmacol Toxicol. 2009;49:57-71.

4. Corrigall WA, Franklin KB, Coen KM, Clarke PB. The mesolimbic dopaminergic system is implicated in the reinforcing effects of nicotine. Psychopharmacology (Berl).

1992;107(2-3):285-9.

5. de Villiers SH, Lindblom N, Kalayanov G, Gordon S, Baraznenok I, Malmerfelt A, et al. Nicotine hapten structure, antibody selectivity and effect relationships: results from a nicotine vaccine screening procedure. Vaccine. 2010;28(10):2161-8.

6. Cornish KE, Harris AC, LeSage MG, Keyler DE, Burroughs D, Earley C, et al.

Combined active and passive immunization against nicotine: minimizing monoclonal antibody requirements using a target antibody concentration strategy. Int

Immunopharmacol. 2011;11(11):1809-15.

7. de Villiers SH, Cornish KE, Troska AJ, Pravetoni M, Pentel PR. Increased efficacy of a trivalent nicotine vaccine compared to a dose-matched monovalent vaccine when formulated with alum. Vaccine. 2013;31(52):6185-93.

185

8. Hieda Y, Keyler DE, Ennifar S, Fattom A, Pentel PR. Vaccination against nicotine during continued nicotine administration in rats: immunogenicity of the vaccine and effects on nicotine distribution to brain. Int J Immunopharmacol. 2000;22(10):809-19.

9. Hieda Y, Keyler DE, Vandevoort JT, Kane JK, Ross CA, Raphael DE, et al.

Active immunization alters the plasma nicotine concentration in rats. J Pharmacol Exp

Ther. 1997;283(3):1076-81.

10. Hieda Y, Keyler DE, VanDeVoort JT, Niedbala RS, Raphael DE, Ross CA, et al.

Immunization of rats reduces nicotine distribution to brain. Psychopharmacology (Berl).

1999;143(2):150-7.

11. Hu Y, Zheng H, Huang W, Zhang C. A novel and efficient nicotine vaccine using nano-lipoplex as a delivery vehicle. Hum Vaccin Immunother. 2014;10(1):64-72.

12. Keyler DE, Hieda Y, St Peter J, Pentel PR. Altered disposition of repeated nicotine doses in rats immunized against nicotine. Nicotine Tob Res. 1999;1(3):241-9.

13. Keyler DE, Roiko SA, Earley CA, Murtaugh MP, Pentel PR. Enhanced immunogenicity of a bivalent nicotine vaccine. Int Immunopharmacol. 2008;8(11):1589-

94.

14. Lockner JW, Ho SO, McCague KC, Chiang SM, Do TQ, Fujii G, et al. Enhancing nicotine vaccine immunogenicity with liposomes. Bioorg Med Chem Lett.

2013;23(4):975-8.

15. Lockner JW, Lively JM, Collins KC, Vendruscolo JC, Azar MR, Janda KD. A conjugate vaccine using enantiopure hapten imparts superior nicotine-binding capacity.

J Med Chem. 2015;58(2):1005-11.

186

16. Miller KD, Roque R, Clegg CH. Novel Anti-Nicotine Vaccine Using a Trimeric

Coiled-Coil Hapten Carrier. PLoS One. 2014;9(12):e114366.

17. Moreno AY, Azar MR, Warren NA, Dickerson TJ, Koob GF, Janda KD. A critical evaluation of a nicotine vaccine within a self-administration behavioral model. Mol

Pharm. 2010;7(2):431-41.

18. Pentel P, Malin D. A vaccine for nicotine dependence: targeting the drug rather than the brain. Respiration. 2002;69(3):193-7.

19. Pentel PR, Malin DH, Ennifar S, Hieda Y, Keyler DE, Lake JR, et al. A nicotine conjugate vaccine reduces nicotine distribution to brain and attenuates its behavioral and cardiovascular effects in rats. Pharmacol Biochem Behav. 2000;65(1):191-8.

20. Pravetoni M, Keyler DE, Pidaparthi RR, Carroll FI, Runyon SP, Murtaugh MP, et al. Structurally distinct nicotine immunogens elicit antibodies with non-overlapping specificities. Biochem Pharmacol. 2012;83(4):543-50.

21. Pravetoni M, Keyler DE, Raleigh MD, Harris AC, Lesage MG, Mattson CK, et al.

Vaccination against nicotine alters the distribution of nicotine delivered via cigarette smoke inhalation to rats. Biochem Pharmacol. 2011;81(9):1164-70.

22. Pryde DC, Jones LH, Gervais DP, Stead DR, Blakemore DC, Selby MD, et al.

Selection of a novel anti-nicotine vaccine: influence of antigen design on antibody function in mice. PLoS One. 2013;8(10):e76557.

23. Satoskar SD, Keyler DE, LeSage MG, Raphael DE, Ross CA, Pentel PR. Tissue- dependent effects of immunization with a nicotine conjugate vaccine on the distribution of nicotine in rats. International Immunopharmacology. 2003;3(7):957-70.

187

24. Zheng H, Hu Y, Huang W, de Villiers S, Pentel P, Zhang J, et al. Negatively

Charged Carbon Nanohorn Supported Cationic Liposome Nanoparticles: A Novel

Delivery Vehicle for Anti-Nicotine Vaccine. J Biomed Nanotechnol. 2015;11(12):2197-

210.

25. Raupach T, Hoogsteder PH, Onno van Schayck CP. Nicotine vaccines to assist with smoking cessation: current status of research. Drugs. 2012;72(4):e1-16.

26. Cornuz J, Zwahlen S, Jungi WF, Osterwalder J, Klingler K, van Melle G, et al. A vaccine against nicotine for smoking cessation: a randomized controlled trial. PLoS

One. 2008;3(6):e2547.

27. Hatsukami DK, Jorenby DE, Gonzales D, Rigotti NA, Glover ED, Oncken CA, et al. Immunogenicity and smoking-cessation outcomes for a novel nicotine immunotherapeutic. Clin Pharmacol Ther. 2011;89(3):392-9.

28. Hatsukami DK, Rennard S, Jorenby D, Fiore M, Koopmeiners J, de Vos A, et al.

Safety and immunogenicity of a nicotine conjugate vaccine in current smokers. Clin

Pharmacol Ther. 2005;78(5):456-67.

29. Esterlis I, Hannestad JO, Perkins E, Bois F, D'Souza DC, Tyndale RF, et al.

Effect of a nicotine vaccine on nicotine binding to beta2*-nicotinic acetylcholine receptors in vivo in human tobacco smokers. Am J Psychiatry. 2013;170(4):399-407.

30. Maurer P, Jennings GT, Willers J, Rohner F, Lindman Y, Roubicek K, et al. A therapeutic vaccine for nicotine dependence: preclinical efficacy, and Phase I safety and immunogenicity. Eur J Immunol. 2005;35(7):2031-40.

188

31. Wagena EJ, de Vos A, Horwith G, van Schayck CP. The immunogenicity and safety of a nicotine vaccine in smokers and nonsmokers: results of a randomized, placebo-controlled phase 1/2 trial. Nicotine Tob Res. 2008;10(1):213-8.

32. Yang RSH, Andersen ME, Dennison JE, Ou YC, Liao KH, Reisfeld. B.

Physiologically based pharmacokinetic and pharmacodynamic modeling. Mouse models of human cancer. Hoboken, N.J.: Wiley-Liss; 2004. p. p. 391-405.

33. Gabrielsson J, Bondesson U. Constant-rate infusion of nicotine and cotinine. I. A physiological pharmacokinetic analysis of the cotinine disposition, and effects on clearance and distribution in the rat. J Pharmacokinet Biopharm. 1987;15(6):583-99.

34. Gajewska M, Worth A, Urani C, Briesen H, Schramm KW. The acute effects of daily nicotine intake on heart rate--a toxicokinetic and toxicodynamic modelling study.

Regul Toxicol Pharmacol. 2014;70(1):312-24.

35. Plowchalk DR, Andersen ME, deBethizy JD. A physiologically based pharmacokinetic model for nicotine disposition in the Sprague-Dawley rat. Toxicol Appl

Pharmacol. 1992;116(2):177-88.

36. Robinson DE, Balter NJ, Schwartz SL. A physiologically based pharmacokinetic model for nicotine and cotinine in man. J Pharmacokinet Biopharm. 1992;20(6):591-

609.

37. Schwartz SL, Gastonguay MR, Robinson DE, Balter NJ. Physiologically based pharmacokinetic modelling of nicotine. In: Gorrod JW, Wahren J, editors. Nicotine and

Related Alkaloids: Springer Netherlands; 1993. p. 255-74.

38. Teeguarden JG, Housand CJ, Smith JN, Hinderliter PM, Gunawan R, Timchalk

CA. A multi-route model of nicotine-cotinine pharmacokinetics, pharmacodynamics and

189 brain nicotinic acetylcholine receptor binding in humans. Regul Toxicol Pharmacol.

2013;65(1):12-28.

39. Yamazaki H, Horiuchi K, Takano R, Nagano T, Shimizu M, Kitajima M, et al.

Human blood concentrations of cotinine, a biomonitoring marker for tobacco smoke, extrapolated from nicotine metabolism in rats and humans and physiologically based pharmacokinetic modeling. Int J Environ Res Public Health. 2010;7(9):3406-21.

40. Benowitz NL, Chan K, Denaro CP, Jacob P, 3rd. Stable isotope method for studying transdermal drug absorption: the nicotine patch. Clin Pharmacol Ther.

1991;50(3):286-93.

41. Benowitz NL, Kuyt F, Jacob P, 3rd. Circadian blood nicotine concentrations during cigarette smoking. Clin Pharmacol Ther. 1982;32(6):758-64.

42. Hukkanen J, Jacob P, 3rd, Benowitz NL. Metabolism and disposition kinetics of nicotine. Pharmacol Rev. 2005;57(1):79-115.

43. Peyret T, Poulin P, Krishnan K. A unified algorithm for predicting partition coefficients for PBPK modeling of drugs and environmental chemicals. Toxicol Appl

Pharmacol. 2010;249(3):197-207.

44. Poulin P, Theil FP. A priori prediction of tissue:plasma partition coefficients of drugs to facilitate the use of physiologically-based pharmacokinetic models in drug discovery. J Pharm Sci. 2000;89(1):16-35.

45. Schmitt W. General approach for the calculation of tissue to plasma partition coefficients. Toxicol In Vitro. 2008;22(2):457-67.

46. Benowitz NL, Porchet H, Jacob P, 3rd. Pharmacokinetics, metabolism, and pharmacodynamics of nicotine. Nicotine psychopharmacology : molecular, cellular, and

190 behavioural aspects. Oxford England ; New York: Oxford University Press; 1990. p. pp.

112-57.

47. Thompson MD, Beard DA, Wu F. Use of partition coefficients in flow-limited physiologically-based pharmacokinetic modeling. J Pharmacokinet Pharmacodyn.

2012;39(4):313-27.

48. Ballard HS. The hematological complications of alcoholism. Alcohol Health Res

World. 1997;21(1):42-52.

49. Dill DB, Costill DL. Calculation of percentage changes in volumes of blood, plasma, and red cells in dehydration. J Appl Physiol. 1974;37(2):247-8.

50. Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Physiological parameter values for physiologically based pharmacokinetic models. Toxicol Ind Health.

1997;13(4):407-84.

51. Poulsen TR, Jensen A, Haurum JS, Andersen PS. Limits for antibody affinity maturation and repertoire diversification in hypervaccinated humans. J Immunol.

2011;187(8):4229-35.

52. Benowitz NL, Kuyt F, Jacob P, 3rd, Jones RT, Osman AL. Cotinine disposition and effects. Clin Pharmacol Ther. 1983;34(5):604-11.

53. Ghosheh O, Dwoskin LP, Li WK, Crooks PA. Residence times and half-lives of nicotine metabolites in rat brain after acute peripheral administration of [2'-

(14)C]nicotine. Drug Metab Dispos. 1999;27(12):1448-55.

54. Court JA, Perry EK, Spurden D, Lloyd S, Gillespie JI, Whiting P, et al.

Comparison of the binding of nicotinic agonists to receptors from human and rat

191 cerebral cortex and from chick brain (alpha 4 beta 2) transfected into mouse fibroblasts with ion channel activity. Brain Res. 1994;667(1):118-22.

55. Syvanen S, Lindhe O, Palner M, Kornum BR, Rahman O, Langstrom B, et al.

Species differences in blood-brain barrier transport of three positron emission tomography radioligands with emphasis on P-glycoprotein transport. Drug Metab

Dispos. 2009;37(3):635-43.

56. Nyback H, Halldin C, Ahlin A, Curvall M, Eriksson L. PET studies of the uptake of

(S)- and (R)-[11C]nicotine in the human brain: difficulties in visualizing specific receptor binding in vivo. Psychopharmacology (Berl). 1994;115(1-2):31-6.

192

Chapter 5: Multi-epitope Insert Modulates Solubility-based and Chromatographic

Purification of Human Papilloma Virus 16 L1-based Vaccine Without Inhibiting

Virus-like Particle Assembly

Kyle Saylor1, Alison Waldman1,2, Frank Gillam1,3, Chenming Zhang1

1Department of Biological Systems Engineering, Virginia Tech, Blacksburg, VA

2Department of Chemical Engineering, North Carolina State University, Raleigh, NC

3Process Development Department, Locus Bioscience, Morrisville, NC

*Corresponding Author: Chenming (Mike) Zhang

302D, HABB1, 1230 Washington St., S.W.

Blacksburg, VA 24061

Voice: (540)-231-7601

Fax: (540)-231-3199

Email: [email protected]

This manuscript has been submitted for publication in the Journal of Chromatography A.

193

Abstract

The separation of heterogeneous protein mixtures has always been characterized by a trade-off between purity and yield. One way this issue has been addressed in the past is by recombinantly modifying protein to improve separations. Such modifications are mostly employed in the form of tags used specifically for affinity chromatography, though it is also possible to make changes to a protein that will have a sizeable impact on its hydrophobicity and charge/charge distribution. As such, it should also be possible to use protein tags to modulate phase separations and protein-resin binding kinetics when performing ion exchange chromatography. Here, we employed a three-step purification scheme on E. coli expressed, His-tagged, human papilloma virus 16 L1- based recombinant proteins (rHPV 16 L1) that consisted of an inclusion body (IB) wash step, a diethylaminoethyl (DEAE) anion exchange chromatography step, and an immobilized metal affinity chromatography polishing step. Purification of the wild type rHPV 16 L1 protein (WT) was characterized by substantial losses during the IB wash but relatively high yield over the DEAE column. In contrast, purification of modified rHPV

16 L1, a chimeric version of the WT protein that had the last 34 amino acids replaced with an MHC class II multi-epitope insert derived from tetanus toxin and diphtheria toxin

(WTΔC34-2TEp), was characterized by little to no losses in the IB wash but had a relatively low yield over the DEAE column. Since the fate of these proteins was to be used in vaccine formulations, it is important to note that the modifications made to the

WTΔC34-2TEp protein had little to no effect on its ability to assemble into virus-like particles. These results demonstrate that modifications of wild type protein via the recombinant insertion of immunofunctional polypeptides can modulate both phase-

194 based separation and charge-based chromatographic processes. Additionally, incorporation of the specific, multi-epitope tag used in this study may prove to be beneficial in recombinant HPV vaccine development due to its potential to improve phase separation yield and vaccine immunogenicity without inhibiting VLP formation.

Highlights

Multi-epitope tag modulates liquid-solid phase separation and weak anion exchange chromatography of recombinant human papilloma virus 16 L1 vaccine.

Tag improves overall yield and does not abrogate VLP formation under optimized conditions.

Three step purification process for wild-type and tagged protein results in > 98% purity.

Keywords protein purification, protein tag, vaccine, liquid-solid phase separation, anion exchange chromatography

195

5.1. Introduction

Conjugate vaccines, particularly the unconventional variants targeting drugs of abuse, have shown promise in past preclinical and clinical trials. Differences observed in vaccine response between subjects, however, have been a major issue (1). Evidently, the design of more broadly immunogenic vaccines is required before clinical efficacy can be achieved in the majority of those receiving these vaccines (2). Immunologically, many approaches exist that could potentially broaden the effectiveness of these vaccines. Two such approaches are 1) the activation of existing memory helper T cell pools using cognate antigen with masked B cell epitopes and 2) the use of universal

MHC class II epitopes that have the capacity to activate the adaptive immune system in the majority of vaccine recipients (3). The current study was born from a desire to explore both of these approaches.

Recombinant human papilloma virus 16 L1 protein (rHPV 16 L1), an immunogen widely used in vaccine formulations, was chosen as the carrier in order to address the first approach. Since most humans will have existing immunity to the wild type (WT) rHPV

16 L1 protein either through natural infection or immunization, initial response to rHPV

16 L1 vaccine should be accelerated by the effects of immunological memory (4, 5).

Many additional properties associated with rHPV 16 L1, however, also make it an excellent carrier protein. First, the protein spontaneously assembles into virus-like particles (VLPs) upon renaturation (6). These VLPs resemble live virus and present many of the same antigenic markers as their parent capsids. They are safe, highly immunogenic, and a novel option when considering their application as carrier proteins

(7). Second, the protein is naturally rich in conjugation sites and the repetitive surface

196 geometry of rHPV 16 L1 VLPs allows for the effective presentation of epitopes that are fixed, either genetically or chemically, to immunodominant regions on VLP surface (8,

9). This, in turn, assists with the crosslinking of B cell receptors (BCRs) and the subsequent T cell independent activation of B cells (10). Third, the protein has been shown to accommodate a wide variety of modifications without influencing the ability of subunits to form VLPs (11). Fourth, rHPV 16 L1 has successfully completed multiple clinical trials and its use has already been approved by the FDA (12). Last, the recombinant expression and purification of rHPV 16 L1 is a well explored topic. As such, ample literature exists to both guide purification efforts and compare results (13-16).

Next, two universal helper T cell epitopes were incorporated within the primary sequence of a truncated version of the WT rHPV 16 L1 protein (WTΔC34-2TEp) in order to address the second approach. If genetic variability in the human leukocyte antigen (HLA) complex is at least partially responsible for differences in vaccine performance between individual human subjects (as research suggests), then these epitopes should further potentiate adaptive immune response to vaccine by targeting a large portion of the population’s MHC class II molecules (17). In addition to modulating immunogenicity, however, a modification of this nature was likely to have some impact on purification outcomes. Since downstream processes constitute a large percentage of protein-based vaccine (PBV) manufacturing costs, it is paramount to study (from the process economics point of view) how modification of the WT protein might impact the overall yield and/or final purity (18). In fact, it seemed possible that the 2TEp modification could be exploited for purification purposes if the right conditions were discovered.

197

When investigating this possibility, the multi-epitope modification was found to indeed have a sizeable impact on protein purification outcomes when comparing WT and

WTΔC34-2TEp versions. Here, we describe the exploitation of this tag as a means of modulating inclusion body (IB) wash separations and resin binding kinetics when performing anion exchange chromatography (AEX). While the insert protein appeared to have a detrimental impact on the AEX chromatography step that could not be explained by theoretical pI and charge distribution, the tag was found to have an overall positive effect on enrichment when compared to the WT protein. It is also important to note that the tag did not negatively influence the ability of the rHPV 16 L1 proteins to form VLPs upon refolding and assembly. This work represents the first steps towards the engineering of a multi-epitope tag for use in rHPV 16 L1 vaccine production that can improve both purification outcome and immunogenicity.

5.2. Materials and Methods

5.2.1. Genetic Engineering

DNA coding for WT rHPV 16 L1, optimized for expression in E. coli (Integrated DNA

Technologies (IDT), Coralville, IA), was purchased and amplified using polymerase chain reaction (PCR) such that a plasmid-specific, overlapping sequence was incorporated at the 5’-end that could be used to insert the gene into a pET-28a expression vector (Novagen, Madison, WI). This PCR-mediated insertion was facilitated by spliced-overlap extension (SOE) method. In brief, small, equimolar amounts of gene and plasmid were subjected to 10 steps of PCR in order to ‘stich’ them together. The reaction was then stopped, forward and reverse primers for the gene-plasmid complex were added, and then the reaction was allowed to continue for an additional 10 cycles.

198

Gel electrophoresis on 1% agarose gel spiked with SYBR safe stain (Invitrogen,

Waltham, MA) was used to confirm successful reaction and then DNA was gel purified and blunt-end ligated to create functional, circular plasmid. T7 Express Competent

BL21(DE3) E. coli cells (New England Biolabs (NEB), Ipswitch, MA) were transformed with the plasmid and transformation success was confirmed by plating culture on antibiotic-spiked agar plates. Colony check PCR with T7 primers (IDT, Coralville, IA) was used to check if gene insertion had been successful, and if appropriate size insert was detected on agarose gel, and Sanger sequencing was used to confirm insert integrity. The 2TEp tag, which was designed using universal helper T cell epitopes derived from diphtheria toxin (aa 271-290) and tetanus toxin (aa 947-967) that were flanked by cathepsin S cleavage sites (GGVVRGG), was also purchased (IDT) and amplified via PCR (19, 20). Plasmid from the first successfully transformed and sequenced WT cell line was then isolated using a GeneJet Plasmid Miniprep Kit

(Thermofisher, Waltham, MA) and 2TEp tag was inserted using the same methods outlined above. WT gene was truncated 102 nucleotides (34 amino acids) at the end upon 2TEp insertion (WTΔC34-2TEp) in order to better facilitate downstream VLP assembly (9). All PCR reactions and gel purifications, excluding colony check reactions, were followed by PCR cleanup using a GeneJet PCR Purification Kit (Thermofisher). A schematic illustrating the primary structure for the WT and WTΔC34-2TEp proteins can be found in Figure 5.1 and the DNA and amino acid sequences for both proteins can be found in the Appendix (Figures B.1 and B.2).

5.2.2. Protein Expression

199

Starter culture was grown overnight at room temperature (RT) and used to inoculate sterile, kanamycin-spiked 2xYT media that had been equilibrated to 37 °C and shaking

(150 RPM) conditions. Once expression culture had reached an OD600 of 0.6-0.7, the temperature of the incubator was reduced to 28 °C, expression was induced using 1 mM Isopropyl β-D-1-thiogalactopyranosid (IPTG, Research Products International (RPI),

Mount Prospect, IL), and overexpression of protein of interest (POI) was allowed to continue overnight. Cells were collected via centrifugation (6000 g for 6 min) and stored at -20 °C until processing.

5.2.3. Cell Processing

Wet cell pellet was resuspended in lysis buffer at 1:10 (grams to milliliters) ratio and sonicated on ice using a probe sonicator to facilitate cell lysis. Soluble fraction was removed via centrifugation (10000 g for 30 min) and collected for later analysis.

Inclusion bodies (IBs), which contained the POI, were then subjected to a wash step using buffer that was absent of sarkosyl detergent but contained ethylenediaminetetraacetic acid (EDTA) and a high concentration of urea (both 6 M and

8 M concentrations were investigated). The soluble fraction from the wash step was removed via centrifugation (10000 g for 30 min) and collected for later analysis. Urea was included in the wash step because preliminary experiments showed that the WT and WTΔC34-2TEp proteins were relatively stable in high-concentration urea solutions whereas contaminating proteins were not. This contrast allowed for the removal of large amounts of contaminating proteins relative to POI losses during the step (data not shown). Washed IBs were then rinsed using lysis buffer to remove residual EDTA, isolated via centrifugation, and solubilized overnight at RT using solubilization buffer.

200

Solubilized inclusion bodies were clarified via centrifugation (10000 g for 30 min) and stored at 4 °C until purification could be completed. All buffer volumes, except that used for the solubilization step (which was at 4x the lysis buffer volume), were the same as was used in the initial lysis step. Buffer recipes were as follows: Lysis buffer, 40 mM

Na2HPO4 + 150 mM NaCl at pH 7.8; Wash buffer, 20 mM Na2HPO4 + 500 mM NaCl +

6 M urea + 10 mM EDTA at pH 6.4; Solubilization buffer, 20 mM Na2HPO4 + 50 mM

NaCl + variable urea + 0.9% (w/v) sarkosyl + 10 mM DTT at pH 8.2.

5.2.4. Protein Purification

Purification of POI was achieved via an anion exchange (AEX) chromatography step using 20.11 mL (1.6 cm diameter × 10 cm height) DEAE Sepharose FF resin and an immobilized metal chromatography polishing step using 20.11 mL (1.6 cm diameter ×

10 cm height) IMAC Sepharose 6 FF resin on two separate XK 16/20 columns (all from

GE Healthcare, Marlborough, MA). The chromatographic systems used for the two separations were an AKTA Purifier and AKTA Explorer, respectively (GE Healthcare). In brief, solubilized IB solution (and DEAE equilibration buffer) was adjusted to target pH and then passed through the DEAE column. The flow through (FT) was collected for later analysis and the eluate, which contained POI in an enriched state, was stored at 4

°C until it could be passed through the IMAC column. DEAE eluate samples that were resultant of a single cell processing step were pooled prior to the IMAC polishing step.

In the polishing step, IMAC FT was collected for later analysis and the eluate was stored at 4 °C until refolding and assembly could be conducted. All samples were equilibrated to RT prior to loading. Loading conditions for the DEAE and IMAC columns were 25 mL load (~4-6 mg POI) and 50 mL load (~5-10 mg POI), respectively.

201

Chromatographic processes utilized a volumetric flow rate of 5 mL/min when sample was on the column (i.e. for loading, flow-through, and elution steps) and a 10 mL/min flowrate for all remaining steps. Buffer recipes were as follows: DEAE equilibration buffer, 20 mM Na2HPO4 + 50 mM NaCl + 6 M urea + 0.9% (w/v) sarkosyl at variable pH; DEAE elution buffer, 20 mM Na2HPO4 + 500 mM NaCl + 6 M urea + 0.9% (w/v) sarkosyl at pH 7.8; IMAC equilibration buffer, 20 mM Na2HPO4 + 500 mM NaCl + 6 M urea + 0.9% (w/v) sarkosyl + 10 mM imidazole at pH 7.8; IMAC elution buffer, 20 mM

Na2HPO4 + 500 mM NaCl + 6 M urea + 0.9% (w/v) sarkosyl + 300 mM imidazole at pH

7.8.

5.2.5. Protein Refolding and VLP Assembly

Following the IMAC polishing step, purified POI was concentrated to 1 mg/mL using

Amicon Ultra-15 10kDa cutoff tubes (MilliporeSigma, Burlington, MA) and dialyzed stepwise against multiple buffers to facilitate protein refolding and assembly. In brief, a series of seven buffer exchanges was used in order to eliminate denaturants and detergents while providing an optimal pH, a reducing environment, and a suitable amount of buffer and stabilizing ions. In the first step, urea and imidazole were removed and the sarkosyl concentration was cut in half by dialyzing IMAC eluate against Buffer

A. The next two steps in the refolding process incrementally reduced the sarkosyl concentration by sequentially dialyzing against Buffers B and C while maintaining all other chemical concentrations. The fourth step involved dialysis against dithiothreitol

(DTT)-spiked PBS (Buffer D) and the final three steps involved dialysis against PBS

(Buffer E) to insure complete removal of all other chemical species. Final samples were centrifuged in order to remove aggregates, sterile filtered using 30 mm, 0.1 μm syringe

202 filters (Celltreat, Pepperell, MA), and stored at 4 °C until quantitative analysis could be performed. Buffer recipes were as follows: Buffer A, 20 mM Na2HPO4 + 500 mM NaCl

+ 0.45% (w/v) sarkosyl + 10 mM DTT at pH 7.8; Buffer B, 20 mM Na2HPO4 + 500 mM

NaCl + 0.1% (w/v) sarkosyl + 10 mM DTT at pH 7.8; Buffer C, 20 mM Na2HPO4 + 500 mM NaCl + 0.01% (w/v) sarkosyl + 10 mM DTT at pH 7.8; Buffer D, 10 mM Na2HPO4 +

137 mM NaCl + 2 mM KCl + 2 mM KH2PO4 + 10 mM DTT at pH 7.8; Buffer E, 10 mM

Na2HPO4 + 137 mM NaCl + 2 mM KCl + 2 mM KH2PO4 at pH 7.2.

5.2.6. Quantitative Analyses

5.2.6.1. Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis (SDS-PAGE)

Gel electrophoresis of protein samples were run under reducing condition. In brief, between 1 and 3.5 μL of sample was mixed with appropriate amounts of SDS reagent, reducing reagent, and ultrapure water and then loaded onto a 15-well, NuPAGE 4-12%

Bis-Tris Protein Gel along with Precision Plus Protein Unstained Standard (Bio-Rad

Laboratories, Hercules, CA). Gels were run at 200 V for 35 min, rinsed three times with ultrapure water, stained with SimpleBlue SafeStain, and then rinsed twice using fresh ultrapure water for one hour. Gels were then imaged using a ChemiDoc Imaging

System (Bio-Rad) and quantified using Image Lab software v6.0.1 b34 (Bio-Rad). All

SDS-PAGE materials, except those specifically noted, were purchased from Invitrogen.

5.2.6.2. Western Blot

The presence of POI in solubilized IB samples was measured by Western Blot assay. In brief, samples were processed following methods outlined for reducing SDS-PAGE analysis, with the only difference being that the protein ladder was replaced with

203

Precision Plus Protein WesternC Protein Standard and that the gel was not stained with

Coomassie blue post-electrophoresis. Instead, protein on the gel was transferred to a nitrocellulose membrane (NCM) using a Bio-Rad Turbo-Blot system and processed via a standard immunodetection assay that used 1:3000 diluted 6x-His Tag (anti-His)

Monoclonal Antibody (HIS.H8) as the primary antibody (Invitrogen), 1:5000 diluted HRP conjugated Rabbit Anti-mouse IgG Superclonal as the secondary antibody (Invitrogen),

1:10000 diluted Precision Protein StrepTactin-HRP conjugate protein standard label mixed with the secondary antibody, and Clarity Western ECL Blotting Substrates. NCM was then imaged using a ChemiDoc Imaging System and quantified using Image Lab software. All Western Blot materials, equipment, and software, except those specifically noted, were made by Bio-Rad.

5.2.6.3. Protein Quantification

Total protein concentrations in heterogeneous protein samples were determined using both A280 readings (1 Abs = 1 mg/mL with baseline correction) on a NanoDrop One spectrophotometer (Thermofisher) and bicinchoninic acid (BCA) assay (G-Biosciences,

St. Louis, MO) results. BCA assay was performed following manufacturer’s instructions.

When necessary, all interfering chemicals were removed from sample via dialysis prior to quantification. BCA assay was also used to determine purified WT and WTΔC34-

2TEp protein concentrations. These concentrations were then used to calculate A280 extinction coefficients for the two proteins so that quantification could thereafter be performed via simple spectrophotometer reading.

Multiple online tools were used to perform in silico protein quantification. First,

ProtParam tool was used to calculate the theoretical extinction coefficient, the

204 theoretical PI, the molecular weight, and the relative stability (both in vivo and in vitro) for both the WT and WTΔC34-2TEp proteins and their N-terminal regions (21). Second,

Kyte and Doolittle’s model for hydropathicity was used to estimate the hydrophobic properties of the proteins and their N-terminal regions (22). Scores were assigned using a 15-residue rolling average that was normalized by dividing all values by the highest value achieved. These two tools were accessed via the ExPASy website (23). Last, the charge function included in the EMBOSS analysis software package was used to estimate the charge distribution of the proteins and their N-terminal regions (24).

Charge scores were assigned using a 15-residue rolling average, but unlike with the hydropathy scores, they were not normalized.

5.2.6.4. Endotoxin Quantification

Endotoxin testing was accomplished by using a PierceTM LAL Chromogenic Endotoxin

Quantitation Kit (Thermofisher) and following manufacturer’s instructions. Samples were only tested one time after they had been successfully purified, refolded, and assembled.

5.2.6.5. Morphological Characterization

The visible morphology of the VLPs was characterized by using transmission electron microscopy (TEM). In brief, sample was loaded onto a formvar coated copper grid and then stained using 1% phosphotungstic acid (PTA). Excess liquid was quickly removed after sample application and PTA staining and the grids were immediately visualized using a JEM 1400 transmission electron microscope (JEOL, Tokyo, Japan) post- staining. When performing the TEM imagining, an rHPV protein (having had its inter- capsomeric disulfides removed) was also run for comparison purposes.

205

5.2.7. Statistical Analysis

Statistical analyses were conducted on data when sample size and the nature of the assay/calculation permitted. In these cases, data are expressed as the mean ± standard deviation. Comparisons were made across all groups using one-way ANOVA and

Tukey’s HSD test. Significance was determined using an α value of 0.05.

5.2.8. Purification System Scoring

Multiple comparisons were made between separation processes by using scoring systems designed in-house. First, comparisons between the WT and WTΔC34-2TEp IB wash and DEAE purification outcomes that took into consideration both the ER and yield were made by simply taking the product of the two values. Second, overall ER and overall yield scores were calculated by simply taking the product of ER and yield values between all possible separation condition combinations. Third, an overall system performance score was calculated by taking the product of all ERs and yields for all possible separation condition combinations.

5.3. Results

5.3.1. Genetic Engineering and Expression

Agarose gel electrophoresis following SOE PCR confirmed successful insertion of genes into pET-28a plasmid, and successful transformation of E. coli cells was confirmed when culture that was plated on antibiotic-spiked agar produced healthy colonies. PCR using T7 primers confirmed the presence of gene inserts within pET-28a plasmid, and after picking colonies that had appropriate size insert and isolating their plasmids, Sanger sequencing was able to confirm that the correct nucleotide sequence

206 had been inserted. Protein expression was confirmed by reducing SDS-PAGE (Figure

B.3, A) and Western Blot (Figure B.3, B) analyses on solubilized IBs. Overnight expression for both WT and WTΔC34-2TEp cell lines resulted in ~10 grams of wet cell paste per liter culture with ~8-9% of this mass being recovered total protein and ~1-2% being POI. These rough values were determined using extractions from one set of expressions (data not shown).

5.3.2. Purification

POI losses accrued during cell lysis amounted to ~24% for WT and ~20% WTΔC34-

2TEp, though it is unclear whether these losses should be attributed to soluble expression of POI or the solubilization of IBs by lysis buffer. Additionally, there were minimal losses observed during the rinse step (data not shown). In the IB wash step, the WT protein suffered substantial losses when a higher concentration of urea was used. The WTΔC34-2TEp protein, however, was largely insoluble in both the lower and the higher concentration urea wash buffers and as a result it had a higher yield over this step when comparing the two proteins (Figure 5.2, A). The enrichment ratio (ER) for this step was improved by modification of the WT protein, but these improvements were only trends and lacked statistical significance (Figure 5.2, B). Subsequent anion exchange chromatography using the DEAE Sepharose FF resin was performed successfully, though there were multiple peaks observed in both the flow through and eluate. This result, when combined with pre- and post-DEAE separation SDS-PAGE results (data not shown), indicated that both inter-protein and intra-protein (amongst the

POIs) charge and/or charge distribution heterogeneity might be present. Representative chromatograms for all of the proteins and purification conditions explored on the DEAE

207 column can be found in Figure 5.3. Both the WT and WTΔC34-2TEp protein suffered substantial losses over the DEAE column (Figure 5.4, A). However, an ER of near 2 was achieved for each protein once loading pH had been optimized (Figure 5.4, B).

Finally, the IMAC polishing step was able to achieve target purity of >98%, as was confirmed using SDS-PAGE analysis (Figure B.3, C). There were minimal losses in the

IMAC FT for both proteins, but the final yields and chromatograms suggested that a substantial amount of protein was being lost on the IMAC column (Figure B.4).

Comparisons between WT and WTΔC34-2TEp separations using in-house scoring systems provided interesting results. When comparing the IB wash step between the proteins, it appeared that the insert improved overall results (Figure 5.2, C) while detrimentally impacting the DEAE step (Figure 5.3, C), even when using optimized conditions. Overall yield and ER values, when considering all possible separation condition combinations, indicated that 2TEp insert improved yield (Figure 5.5, A) but detrimentally impacted ER (Figure 5.5, B). Finally, the overall system performance when considering both the IB wash and DEAE chromatography steps indicated comparable separation when conditions had been optimized for the WT and WTΔC34-2TEp proteins

(Figure 5.5, C).

5.3.3. Final Protein Quantification

LAL testing results were negative in final protein samples, indicating complete removal of endotoxin (data not shown). Extinction coefficients (0.1% @ 280 nm) for the WT and

WTΔC34-2TEp proteins, on the other hand, were experimentally determined to be

11.82 and 12.37, values that were 0.38 and 1.29 percent off the theoretical extinction coefficients calculated using the ProtParam tool, respectively. Finally, charge

208 distribution and hydropathicity analyses revealed that the N-terminal modification of WT protein with 2TEp insert resulted in both regional and overall decreases in hydrophobicity and charge (Figure 5.6).

5.3.4. Refolding and Assembly

TEM images of WT and WTΔC34-2TEp, when compared to capsomeric rHPV 16 L1 protein, indicated that VLP assembly had been successful (Figure 5.7, A). Upon further analysis of the TEM images, the WTΔC34-2TEp VLPs (34.09 ± 5.08 nm) were revealed to be larger than the WT VLPs (22.82 ± 5.24 nm) (Figure 5.7, B).

5.4. Discussion

Many of the approaches for vaccine development described here were strongly guided by the accomplishments reported in past studies. Lai and Middelberg showed that recombinant expression and two-step chromatographic separation of E. coli produced

HPV 16 L1 protein was possible (15). Müller et al. proved that C-terminal modifications of HPV 16 L1 were possible under certain conditions (9). Specifically, it was shown that the replacement of the protein’s last 34 residues with an insert of up to 60 residues was possible without substantially abrogating capsomere and capsid formation, albeit it is likely that insert properties in addition to size are important for successful refolding and assembly of the protein. Finally, Li et al. demonstrated that replacing interior, di-sulfide- forming cysteines with glycine residues could prevent the formation of HPV 16 L1 VLPs

(25). This discovery made it possible to compare HPV 16 L1’s capsomeric morphology to that of the VLPs formed by the WT and WTΔC34-2TEp proteins. Together, the breakthroughs published by all of these researchers helped guide the design and production of both WT and WTΔC34-2TEp VLPs in this study.

209

Overexpression of protein in the form of IBs, to certain degrees, can be both beneficial and detrimental. Expression levels can be exceptionally high and the removal of soluble proteins can provide a convenient, highly efficacious first purification step. The final product, however, can sometimes be heavily aggregated and difficult to refold into morphologically and/or physiologically active protein (26). Processing of the WT and

WTΔC34-2TEp protein IBs followed this trend. Expression levels were high, IB solubilization required high levels of detergent and denaturant, and while removal of soluble proteins during cell lysis resulted in some losses for both POIs, a huge portion of contaminating proteins was also removed prior to IB solubilization. The nature of the proteins and their intended uses, however, made IB expression an optimal choice.

Since the final goal for the proteins was simply to form VLP scaffolding for hapten- based vaccines, the success of the process mostly hinged on whether or not VLPs could be produced in an efficient, economical fashion.

Ultimately, this goal was achieved with IB-produced proteins using the optimized purification processes and the stepwise removal of denaturant and detergents described in the methods sections, though the final VLPs had morphologies that differed from some descriptions in literature. There, two different types of HPV L1 VLPs have been reported. The first morphology consists of hollow particles that are ~40-60 nm in diameter and has shells with visible capsomeric content (27-29). The second morphology consists of what appears to be ‘crude,’ mostly solid particles that are ~15-

40 nm in diameter (30-33). The difference in these morphologies may be linked to the formation of IBs, as the WT and WTΔC34-2TEp VLPs described here shared morphology that was more similar to the latter description. Additionally, soluble

210 expression of rHPV 16 L1 in E. coli has been shown to result in the formation of VLPs with the former description and electron micrographs of IBs in rHPV L1-expressing

HeLa cells show what appear to be the presence of ~30 nm solid, VLP species (27, 34).

It is important to note that both morphologies have been proven to be highly immunogenic, though the difference between the two likely indicates that the VLPs are, at least to some degree, presenting different B cell epitopes. This shouldn’t be a problem, however, since the intended use of these proteins is as carriers in conjugate vaccine formulations.

Cell processing and purification of the POIs was required prior to protein refolding and

VLP assembly. First, optimal pH, ionic strength, detergent concentration, and sonication conditions were determined for the cell lysis step. It was found that on-off sonication performed on ice was able to efficiently disrupt cells without harming the POIs. In addition, pH and ionic strength were found to mostly influence the amount of contaminating protein that could be removed over the step. On the other hand, the use of sarkosyl and/or Triton X-100 detergents, even in small quantities, resulted in sizeable

POI losses in the soluble lysate. For this reason, detergent was completely removed from the cell lysis process.

The 2TEp insert altered the outcome of all three of the purification steps employed to achieve a target purity of >98%. First, the tag was found to modulate the IB wash

(liquid-solid phase separation) step in a positive way by increasing both yield and enrichment under optimized conditions. This effect could most likely be attributed to the regional change in hydropathicity score (0.184), since the overall change (0.012) was rather small. If this was in fact the case, this result indicates that regional properties of a

211 protein may play a more important role than overall properties when considering the outcome of separation processes. Overall, the different solubilities displayed by the WT and WTΔC34-2TEp proteins when placed in solutions containing high urea concentrations follows convention as urea is known to destabilize hydrophobic interactions (35).

Results from the DEAE weak anion exchange chromatography step may also support the importance of regional protein properties when considering separation process outcomes. Since independent regions within denatured proteins should be able to interact with stationary phase independently, it is likely that they have at least some impact on separation outcomes. Consequently, if expressed POI consisted of perfectly homogeneous charges and charge distributions, this could explain why both cation exchange and anion exchange chromatography have been successfully employed for the separation of HPV L1 proteins in the past (36, 37). Alternatively, it is also possible that the charge and/or charge distribution heterogeneity of POI species suggested by the multi-peak DEAE eluate profiles and the pre- and post-separation SDS-PAGE results can explain the ability of the HPV 16 L1 protein to purify on both anionic and cationic resins. Whatever the case, the 2TEp tag resulted in regional and overall, theoretical charge distribution shifts of 0.216 and -0.015, respectively. Ultimately, the change in electrostatic properties caused the WTΔC34-2TEp protein to bind weakly to the column and as such sizeable losses in the flow through were experienced. Though the abrogation of anion exchange yield by decreasing POI charge doesn’t follow conventional thinking, this effect can most likely be attributed to the extreme shift in regional theoretical charge as opposed to the modest shift in overall theoretical charge.

212

The 2TEp tag also caused the overall, theoretical pI to shift from 8.65 to 6.65, a value that was dangerously close (and actually spanned) the two pH values that were evaluated in this study. Fortunately, no precipitation of the WTΔC34-2TEp was observed when working at these pH levels. The DEAE resin-binding kinetics of the WT and WTΔC34-2TEp proteins also contradicted convention when considering the pI of each protein and the working pH values used for the separations. Theoretically, the WT protein should have had an overall positive charge for all purification conditions tested and therefore would have had little to no affinity for the positively charged DEAE resin.

In addition, the WTΔC34-2TEp protein should have bound more strongly to the DEAE resin and produced a greater yield than the WT protein under these conditions. A number of factors could explain these discrepancies. First, the native and theoretical pIs of our proteins are likely different. The lack of precipitation of the WTΔC34-2TEp protein when working at the target pH values supports this rational. It is possible that the IB solubilization process did not completely denature POI, resulting in remaining secondary/tertiary structure that could alter which residues were solvent exposed. The presence of dimers on the reducing SDS-PAGE gel would support this explanation.

Second, the use of anionic detergent in the EQ and solubilization buffers is likely affecting the performance of the separation. IEX chromatography is especially sensitive to changes in ionic content over the column due to competition for charged resin.

Purification performance would most likely be modulated if sarkosyl were removed from the process, but the detergent was necessary in order to fully solubilize the IBs. Third, it is possible that multimodal characteristics manifested within the separation and/or that the proteins could be interacting with each other on the column.

213

Target purity was achieved with the IMAC polishing step, but considerable losses of the

WT protein over the step were unexpected and disappointing due to either precipitation and/or irreversible binding on the column. The loss was likely due to precipitation on the column, as the WT protein (and to a lesser extent the WTΔC34-2TEp protein) proved to be susceptible to precipitation upon concentration (data not shown). Since the column was run well below the recommended dynamic binding capacity of 40 mg POI per mL resin defined for the IMAC Sepharose Fast Flow resin when the issues occurred, this result indicates that stability of the proteins on the column may become a crippling bottleneck for the process in the future.

This investigation has generated multiple takeaways. First, it has demonstrated that slight modifications can be made to a protein in order to modulate its solubility and affinity for ion exchange resins. Specifically, a method has been identified to simultaneously improve antigenic coverage and solubility-based purification of IB- produced HPV 16 L1 protein via the incorporation of the 2TEp tag. Importantly, the modification didn’t abrogate the ability of the protein to assemble into VLPs. While the tag did detrimentally impact the AEX chromatography step, it may have made up for these losses during the IMAC step by preventing precipitation and/or irreversible binding on the column. Second, this research suggests that regional protein characteristics may have more impact on solubility-based and ion exchange-based separation processes than overall characteristics. Of course, this idea has already been proven for affinity- based chromatographic separations such as those that use specific tag-resin binding dynamics. More research should be done in the future in order to ascertain whether or not this phenomenon can be further exploited for separation purposes. Finally, the

214 presented results suggest that ‘convention’ should be taken with a grain of salt when planning ion exchange chromatography processes, as anion exchange binding behavior was the exact opposite of what we were expecting.

5.5. Conclusions

Here, we describe the successful design, expression, purification, and assembly of WT and chimeric rHPV 16 L1 VLPs aimed at improving the population coverage and efficacy of future conjugate vaccines. Interestingly, the 2TEp modification that was made to the WT protein also significantly modulated the conditions needed to achieve target purity while simultaneously maximizing yield. In fact, the effects the tag had on the IB wash step can be exploited in order to improve overall step performance.

Unfortunately, however, the tag had an overall detrimental impact on the AEX chromatography step, even after optimizing purification conditions, due to reduced recovery. When considering both yield and enrichment and taking IB wash and AEX chromatography performances together, the tag appeared to slightly abrogate purification. Overall efficacies of the optimized processes were comparable once conditions had been optimized, however, and even improved when performance over the IMAC column was considered. Final results from the purifications suggested that 1) proteins can be slightly modified in order to significantly modulate their solubility and affinity for ion exchange resins, 2) regional properties of protein can have a larger impact on separation process outcome than overall properties, and that 3) conventional wisdom based on theorical protein properties should be applied to ion exchange chromatography applications sparingly. Future work should focus on the evaluation of

WT and WTΔC34-2TEp VLP immunogenicity and further investigate the ion exchange

215 chromatography step. Specifically, CEX chromatography should be performed using a range of pH values and the best results should be compared with those obtained using

AEX chromatography.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgements

This work was partially supported by the National Institute on Drug Abuse

(U01DA036850), the American Association of Immunologists Careers in Immunology

Fellowship Program, and the department of Biological Systems Engineering at Virginia

Tech.

216

Tables and Figures

Figure 5.1. Primary structure for the WT and WTΔC34-2TEp proteins.

Primary structure for the A) WT and B) WTΔC34-2TEp proteins. Black boxes indicate modification location, dark gray boxes indicate surface exposed loop regions, and light gray boxes indicate buried loop regions. Loops are identified based on their position relative to beta strands (i.e. loop BC lies between beta strands B and C).

217

Figure 5.2. IB wash step yield, enrichment ratio, and overall score.

IB wash step A) yield, B) enrichment ratio, and C) overall score based on the mean of these two values (ERWash*YWash). Bars represent mean values while whiskers and letters represent STD and statistical groupings within each individual plot, respectively. Data that has been grouped differently is statistically different.

218

Figure 5.3. Representative DEAE chromatograms.

Peaks correspond to flow through (~180 mL), eluate (~275 mL), and CIP (~360 mL).

Yield (according to A280 integration) is improved in runs containing WT C-terminal sequence and greater pH. Flow-through and eluate peak profiles indicate the presence of multiple species. Purifications were performed on an AKTA Purifier system using an

XK 16/20 column packed with 20.11 mL (1.6 cm x 10 cm) DEAE Sepharose FF resin, a loading volume of 25 mL (~4-6 mg POI), and a volumetric flow rate of 5 mL/min.

219

Figure 5.4. DEAE chromatography step yield, enrichment ratio, and overall score.

DEAE chromatography step A) yield, B) enrichment ratio, and C) overall score based on the mean of these two values (ERDEAE* YDEAE). Bars represent mean values while whiskers and letters represent STD and statistical groupings within each individual plot, respectively. Data that has been grouped differently is statistically different.

220

Figure 5.5. Overall separation performance for WT and WTΔC34-2TEp proteins.

Overall A) enrichment ratio, B) yield, and C) separation performance for the WT and

WTΔC34-2TEp proteins when considering all possible separation condition combinations. Scoring was achieved by normalizing the product of the enrichment ratios

(ERDEAE*ERWash), yields (YDEAE*YWash), and enrichment ratios and yields

(ERDEAE*YDEAE*ERWash*YWash), respectively.

221

Figure 5.6. Overall and regional hydrophobic and ionic characterization of the WT and WTΔC34-2TEp proteins.

Overall and regional A) hydrophobic and B) ionic characterization of the WT and

WTΔC34-2TEp proteins. The 2TEp insert resulted in large regional decreases and more moderate overall decreases in hydrophobicity and charge. Score values displayed in the tables represent overall and insert-specific averages. The insert also increased protein molecular weight (MW) and shifted the regional pI from 12.33 (WT) to 9.98 (2TEp) and the overall pI from 8.65 (WT) to 6.65 (2TEp). The ‘inserts’ defined for the WT protein comprise the last 34, naturally occurring C-terminal residues and do not indicate modification of the protein.

222

Figure 5.7. TEM images of WT and WTΔC34-2TEp VLPs.

The WT and WTΔC34-2TEp A) TEM images and B) size distributions clearly show the presence of VLPs when compared with that of the capsomeric rHPV 16 L1 protein

(Cap).

223

References

1. Hu Y, Zhao Z, Saylor K, Zhang C. Nicotine Vaccines: The Past, the Present, and the Future. Neuroscience of nicotine : mechanisms and treatment. 1st edition. ed.

Cambridge, MA: Elsevier; 2019.

2. Saylor K, Zhang C. A simple physiologically based pharmacokinetic model evaluating the effect of anti-nicotine antibodies on nicotine disposition in the brains of rats and humans. Toxicol Appl Pharmacol. 2016;307:150-64.

3. Saylor K, Gillam F, Lohneis T, Zhang C. Designs of Antigen Structure and

Composition for Improved Protein-Based Vaccine Efficacy. Front Immunol.

2020;11:283.

4. Fraser CC, Altreuter DH, Ilyinskii P, Pittet L, LaMothe RA, Keegan M, et al.

Generation of a universal CD4 memory T cell recall peptide effective in humans, mice and non-human primates. Vaccine. 2014;32(24):2896-903.

5. Pinto LA, Edwards J, Castle PE, Harro CD, Lowy DR, Schiller JT, et al. Cellular immune responses to human papillomavirus (HPV)-16 L1 in healthy volunteers immunized with recombinant HPV-16 L1 virus-like particles. J Infect Dis.

2003;188(2):327-38.

6. Zhang W, Carmichael J, Ferguson J, Inglis S, Ashrafian H, Stanley M.

Expression of human papillomavirus type 16 L1 protein in Escherichia coli: denaturation, renaturation, and self-assembly of virus-like particles in vitro. Virology.

1998;243(2):423-31.

7. Jennings GT, Bachmann MF. The coming of age of virus-like particle vaccines.

Biol Chem. 2008;389(5):521-36.

224

8. Gillam F, Zhang J, Zhang C. Hepatitis B core antigen based novel vaccine against porcine epidemic diarrhea virus. J Virol Methods. 2018;253:61-9.

9. Muller M, Zhou J, Reed TD, Rittmuller C, Burger A, Gabelsberger J, et al.

Chimeric papillomavirus-like particles. Virology. 1997;234(1):93-111.

10. Mohsen MO, Zha L, Cabral-Miranda G, Bachmann MF. Major findings and recent advances in virus-like particle (VLP)-based vaccines. Semin Immunol. 2017;34:123-32.

11. Sadeyen JR, Tourne S, Shkreli M, Sizaret PY, Coursaget P. Insertion of a foreign sequence on capsid surface loops of human papillomavirus type 16 virus-like particles reduces their capacity to induce neutralizing antibodies and delineates a conformational neutralizing epitope. Virology. 2003;309(1):32-40.

12. Lexchin J, Arya N, Singh S. Gardasil(R) - The New HPV Vaccine: The Right

Product, the Right Time? A Commentary. Healthc Policy. 2010;5(4):26-36.

13. Schadlich L, Senger T, Kirschning CJ, Muller M, Gissmann L. Refining HPV 16

L1 purification from E. coli: reducing endotoxin contaminations and their impact on immunogenicity. Vaccine. 2009;27(10):1511-22.

14. Chen XS, Casini G, Harrison SC, Garcea RL. Papillomavirus capsid protein expression in Escherichia coli: purification and assembly of HPV11 and HPV16 L1. J

Mol Biol. 2001;307(1):173-82.

15. Lai WB, Middelberg AP. The production of human papillomavirus type 16 L1 vaccine product from Escherichia coli inclusion bodies. Bioprocess Biosyst Eng.

2002;25(2):121-8.

225

16. Bian T, Wang Y, Lu Z, Ye Z, Zhao L, Ren J, et al. Human papillomavirus type 16

L1E7 chimeric capsomeres have prophylactic and therapeutic efficacy against papillomavirus in mice. Mol Cancer Ther. 2008;7(5):1329-35.

17. Posteraro B, Pastorino R, Di Giannantonio P, Ianuale C, Amore R, Ricciardi W, et al. The link between genetic variation and variability in vaccine responses: systematic review and meta-analyses. Vaccine. 2014;32(15):1661-9.

18. Saraswat M, Musante L, Ravida A, Shortt B, Byrne B, Holthofer H. Preparative purification of recombinant proteins: current status and future trends. Biomed Res Int.

2013;2013:312709.

19. Diethelm-Okita BM, Okita DK, Banaszak L, Conti-Fine BM. Universal epitopes for human CD4+ cells on tetanus and diphtheria toxins. J Infect Dis. 2000;181(3):1001-9.

20. Bromme D, Bonneau PR, Lachance P, Storer AC. Engineering the S2 subsite specificity of human cathepsin S to a cathepsin L- and cathepsin B-like specificity. J Biol

Chem. 1994;269(48):30238-42.

21. Wilkins MR, Gasteiger E, Bairoch A, Sanchez JC, Williams KL, Appel RD, et al.

Protein identification and analysis tools in the ExPASy server. Methods Mol Biol.

1999;112:531-52.

22. Kyte J, Doolittle RF. A simple method for displaying the hydropathic character of a protein. J Mol Biol. 1982;157(1):105-32.

23. Gasteiger E, Gattiker A, Hoogland C, Ivanyi I, Appel RD, Bairoch A. ExPASy:

The proteomics server for in-depth protein knowledge and analysis. Nucleic Acids Res.

2003;31(13):3784-8.

226

24. Rice P, Longden I, Bleasby A. EMBOSS: the European Molecular Biology Open

Software Suite. Trends Genet. 2000;16(6):276-7.

25. Li M, Beard P, Estes PA, Lyon MK, Garcea RL. Intercapsomeric disulfide bonds in papillomavirus assembly and disassembly. J Virol. 1998;72(3):2160-7.

26. Yamaguchi H, Miyazaki M. Refolding techniques for recovering biologically active recombinant proteins from inclusion bodies. Biomolecules. 2014;4(1):235-51.

27. Wei M, Wang D, Li Z, Song S, Kong X, Mo X, et al. N-terminal truncations on L1 proteins of human papillomaviruses promote their soluble expression in Escherichia coli and self-assembly in vitro. Emerg Microbes Infect. 2018;7(1):160.

28. Kirnbauer R, Taub J, Greenstone H, Roden R, Durst M, Gissmann L, et al.

Efficient self-assembly of human papillomavirus type 16 L1 and L1-L2 into virus-like particles. J Virol. 1993;67(12):6929-36.

29. Deschuyteneer M, Elouahabi A, Plainchamp D, Plisnier M, Soete D, Corazza Y, et al. Molecular and structural characterization of the L1 virus-like particles that are used as vaccine antigens in Cervarix, the AS04-adjuvanted HPV-16 and -18 cervical cancer vaccine. Hum Vaccin. 2010;6(5):407-19.

30. Zahin M, Joh J, Khanal S, Husk A, Mason H, Warzecha H, et al. Scalable

Production of HPV16 L1 Protein and VLPs from Tobacco Leaves. PLoS One.

2016;11(8):e0160995.

31. Pineo CB, Hitzeroth, II, Rybicki EP. Immunogenic assessment of plant-produced human papillomavirus type 16 L1/L2 chimaeras. Plant Biotechnol J. 2013;11(8):964-75.

227

32. Chabeda A, van Zyl AR, Rybicki EP, Hitzeroth, II. Substitution of Human

Papillomavirus Type 16 L2 Neutralizing Epitopes Into L1 Surface Loops: The Effect on

Virus-Like Particle Assembly and Immunogenicity. Front Plant Sci. 2019;10:779.

33. Bazan SB, de Alencar Muniz Chaves A, Aires KA, Cianciarullo AM, Garcea RL,

Ho PL. Expression and characterization of HPV-16 L1 capsid protein in Pichia pastoris.

Arch Virol. 2009;154(10):1609-17.

34. Xiao CY, Fu BB, Li ZY, Mushtaq G, Kamal MA, Li JH, et al. Observations on the expression of human papillomavirus major capsid protein in HeLa cells. Cancer Cell Int.

2015;15:53.

35. Hua L, Zhou R, Thirumalai D, Berne BJ. Urea denaturation by stronger dispersion interactions with proteins than water implies a 2-stage unfolding. Proc Natl

Acad Sci U S A. 2008;105(44):16928-33.

36. Kim HJ, Kim SY, Lim SJ, Kim JY, Lee SJ, Kim HJ. One-step chromatographic purification of human papillomavirus type 16 L1 protein from Saccharomyces cerevisiae.

Protein Expr Purif. 2010;70(1):68-74.

37. Sun B, Zhao D, Zhang X, Gu T, Yu X, Sun S, et al. Development a scalable production process for truncated human papillomavirus type-6 L1 protein using WAVE

Bioreactor and hollow fiber membrane. Appl Microbiol Biotechnol. 2016;100(3):1231-40.

228

Chapter 6: Computational Mining of MHC class II Epitopes for the Development of

a Universal Carrier Protein

Kyle Saylor1, Ben Donnan1, Chenming Zhang1

1Department of Biological Systems Engineering, Virginia Tech, Blacksburg, VA

*Corresponding Author: Chenming (Mike) Zhang

302D, HABB1, 1230 Washington St., S.W.

Blacksburg, VA 24061

Voice: (540)-231-7601

Fax: (540)-231-3199

Email: [email protected]

This manuscript has been submitted for publication in the journal npj Vaccines.

229

Abstract

The human leukocyte antigen (HLA) gene complex, one of the most diverse gene complexes found in the human genome, largely dictates how our immune systems recognize pathogens. Specifically, HLA genetic variability has been linked to vaccine effectiveness in humans and it is highly likely that it has played some role in the shortcomings of the numerous human vaccines that have failed clinical trials. This variability is largely impossible to evaluate in animal models, however, as their immune systems generally 1) lack the diversity of the HLA complex and 2) express major histocompatibility complex (MHC) molecules that differ in specificity when compared to human MHC. The need for vaccines that can effectively engage the majority of human

MHC molecules paired with the current capabilities of MHC prediction software, the failings of standard animal models as correlates for success, and the ever-present pressure to adopt alternatives to in vivo studies leads to a situation where in silico design and evaluation of vaccine candidates becomes an acceptable alternative to traditional routes of vaccine assessment. Here, we describe the use of HLA population frequency data and MHC epitope prediction software to facilitate the in silico mining of universal helper T cell epitopes and the subsequent deign of a universal human immunogen using these predictions. This research highlights a novel approach to using in silico prediction software and data processing to direct vaccine development efforts.

Highlights

Common immunogens were successfully mined for universal HLA-DQ, HLA-DR, IAd, and IEd epitopes using MHC class II epitope prediction software, and when applicable,

HLA population frequency data.

230

Chimeric, universal antigens were constructed for each MHC class II isotype by concatenating the 25 highest scoring epitopes from post-prediction analyses with interspacing cathepsin S-sensitive sequences.

This work represents the first ever attempt to engineer chimeric, in silico-derived, universal antigen for use in conjugate vaccine formulations.

Keywords

Vaccine, immunogenetics, epitope prediction, universal immunogen, chimeric antigen

231

6.1. Introduction

The human leucocyte antigen (HLA) system, located on the short arm of chromosome

6, contains many genes that are essential to the initiation and maintenance of our immune systems. Specifically, the genes that code for major histocompatibility complex

(MHC) class I molecules (HLA-A, HLA-B, and HLA-C) and MHC class II molecules

(HLA-DP, HLA-DM, HLA-DO, HLA-DQ, and HLA-DR) reside here. These molecules activate cytotoxic T cells (TC cells, via MHC I pathway) or helper T cells (TH cells, via

MHC II pathway) when abnormal intracellular or extracellular polypeptide species, respectively, are encountered and presented on the surface of any nucleated cell (in the case of MHC class I pathway) or specifically on antigen presenting cells (APCs, in the case of MHC class II pathway). In fact, the majority of adaptive immune system effector functions must be initiated by an MHC pathway at some point in time (1).

It is easy to understand how the evolutionary pressures placed on these pathways have culminated in one of the most diverse gene systems in the human genome. While this grand level of diversity is extremely beneficial when considering species survival, it becomes less beneficial on an individual basis when designing and implementing immunotherapies (2). HLA allele variation has been implicated in malignancies, infections, and immunotherapeutic outcomes in many independent studies (3). Some examples include the association of a specific allele with the occurrence of cancer, the association of HLA-DRB1 heterozygosity with better outcomes in viral infections, and the association of HLA class I homozygosity with checkpoint inhibitor inefficacy in cancer therapies (4-7). Particularly in the case of vaccines, significant associations between HLA haplotype and vaccine outcome have been observed. Causation in these

232 studies, however, has not been established and it is important to note that other genes have also been linked to vaccine inefficacy (8-11). Nonetheless, a new age of vaccine design has been born around the targeting of MHC molecules with epitopes that have been derived either experimentally or computationally using models based on pooled in vivo and in vitro data (12, 13).

The concept behind this targeted approach to vaccine design is simple. In any vaccine, it is possible that an antigen (or antigens) will lack epitopes specific for host MHC molecules. Without the loading of processed epitope onto MHC and the subsequent presentation of the peptide-MHC complex to T cell receptors (TCRs), there can be no activation of TH and/or TC cells. Consequently, the adaptive immune response to a vaccine in such situations will be severely impaired (14). However, as the knowledgebase concerning these MHC activation systems continues to expand and mature, it is becoming more and more feasible and prudent to use all of the information we have to insure (in the case of vaccines) or prevent (in the case of biologics) the activation of T cells and the resultant adaptive immune response.

Many converging factors support the implementation of such an approach in the field of vaccinology. First, considerable variability in subjects’ response to a particular vaccine is a common occurrence and has been a deciding factor in the ultimate failure of some clinical trials (15). The inevitability of acquiring variable outcomes when dealing with variable systems, however, can possibly be avoided by applying a more personalized approach to vaccination, particularly one that targets a large number of MHC molecules.

Second, interspecies differences in HLA genotype and MHC specificity present as serious confounding elements when attempting to use animal studies as corollaries to

233 success in future clinical trials (16). Consequently, considering the MHC epitope content of antigen(s) prior to human studies may provide more insight into the likely outcome of clinical trials. Third, ample amounts of easily accessible software and data now exist online that can support an in silico approach to vaccine design. In fact, others have already begun the modulation of antigen immunogenicity based on these resources (17-

19). Last, there is an ever-present need to minimize the use of in vivo studies and adopt alternative vaccine quantitative methods. While it is unlikely that we will ever truly break free from using animal models as corollaries for success, in silico design and/or evaluation of vaccine candidates prior to human studies provides a plausible and acceptable alternative to in vivo testing.

In this study, we mined common immunogen primary sequences for HLA-DR and HLA-

DQ epitopes using HLA population frequency data and MHC class II epitope prediction software. IAd and IEd murine isotypes were also analyzed in order to make inter- method and inter-species prediction output comparisons. MHC class II epitopes were chosen for analysis due to their role as facilitators of humoral immune responses via the activation of TH cells. This activation is crucial to antibody production and as such plays an important role in the success of vaccines that are dependent on antibody effector functions, such as vaccines against drugs of abuse (20, 21). Using the output from these predictions, we conceptualize and design ‘universal’ chimeric antigens (UCAs) that can be used to target mouse models with IAd and/or IEd allotypes, and using HLA-

DQB1 and HLA-DRB1 allele frequency data, 99% of the US population based on HLA-

DQ and HLA-DR isotypes. To the best of our knowledge, we are the first to attempt the design of UCAs using HLA population frequencies and MHC class II epitope

234 predictions. By making slight modifications to the approach, our UCAs could also be used to target certain demographics and/or individual subjects.

6.2. Methods

6.2.1. HLA-DQB1 Allele Frequency Analysis

Human HLA-DQB1 and HLA-DRB1 allele frequency data with race/ethnic associations were acquired from the National Bone Marrow Transplant (NBMW) / Be The Match bioinformatics database (22). Data were sorted based on overall population frequency, and the most common beta alleles accounting for at least 99% of the database test population were selected as targets for MHC class II epitope predictions. Since only population frequency data for beta chains were available, alpha-beta pairing was done post-hoc and did not represent population frequencies for complete MHC class II molecules (which consist of one alpha chain and one beta chain).

6.2.2. Source Immunogen Selection

A thorough literature review was performed in order to identify the most common immunogens used in vaccine formulations. Some of these have been approved for use in human vaccines by the FDA, such as diphtheria toxoid (DT), tetanus toxoid (TT), and the HPV 16 L1 protein (HPV). Others, such as keyhole limpet hemocyanin 1 and keyhole limpet hemocyanin 2 (KLH1 and KLH2) and bovine serum albumin (BSA) are typically only used in proof-of-concept, animal studies. Human serum albumin (HSA) and mouse serum albumin (MSA) were selected as benchmark proteins for the analysis of the other selected immunogens. In total, the 15 most common immunogens encountered during the literature review were chosen as MHC class II epitope sources.

235

In addition to the six immunogens already mentioned, the additional nine were the Q- beta capsid protein (QB), Pseudomonas aeruginosa exoprotein A (EPA), cholera toxin subunit B (CTB), heat labile enterotoxin B (LTB), outer membrane protein C (OMPC),

Influenza A hemogglutinin (HA), hepatitis B core antigen (HBC), MS2 capsid protein

(MS2), and hepatitis C core antigen (HCC). Sequences were obtained from the UniProt website (23). A summary of these immunogens can be found in Table 6.1.

6.2.3. MHC Class II Epitope Predictions

MHC class II epitope predictions were performed locally using predication tools downloaded from the Immune Epitope Database and Analysis Resource (IEDB.org, v2.22.1) (12). Specifically, predictions were run for each immunogen-allotype pairing using neural network-based NetMHCIIPan 3.2 method (DQ, DR, and IAd predictions) and the stabilization matrix alignment-based SMM-align method (IAd and IEd predictions) (24, 25). Both of these prediction methods have been shown to have high accuracy in the past (26, 27). IAd and IEd predictions were included in the analysis in order to 1) compare between prediction methods (NetMHCIIPan vs. SMM-align), 2) compare between species (human and mouse), and 3) set up for in vivo evaluation of this immunogen design approach. An additional parameter, peptide length, was specified as 15-20 amino acids for the predictions. This range of epitope lengths (which falls within the 9-23 residue range of lengths defined for most MHC class II epitopes) was chosen in order to introduce diversity into the prediction, thus helping to eliminate any output bias that might be due to input of a single epitope length (28).

6.2.4. Epitope Scoring and Anchor Residue Identification

236

Epitope scoring was achieved by completing an unweighted summation (no core residue bias) of transformed percentile rankings (100-x) for each amino acid within each predicted epitope. The final sums were normalized by dividing by the average residue score for the endogenous, HSA protein, and an ‘immunogenicity threshold’ was established at one for these normalized values. Alternately, MSA was used as the benchmark when establishing immunogenicity thresholds for the IAd and IEd predictions. Anchor residues were identified in a similar fashion. The only differences between the two scoring mechanisms were that a weighted summation (four to one core residue to non-core residue weighting) and an ‘anchor residue’ threshold value of four were used for anchor residue identification purposes. Both the ‘unweighted, normalized, and cumulative’ (UNC) and ‘weighted, normalized, and cumulative’ (WNC) methods resulted in proteins being assessed based on the 13 HLA-DQB1 and 40 HLA-DRB1 alleles specified in the frequency analysis, even though predictions for the HLA-DQ epitopes also required the specification of an alpha chain. As a result, predictions made for HLA-DQ epitopes may not have targeted the most frequent alpha-beta pairs, though

99% of the sample population should have been targeted nonetheless. The mean and standard deviation for summed residue immunogenicity values were calculated for each protein using UNC and WNC results, respectively. These values were then plotted against residue number. All prediction output analyses were performed and all plots were generated using Matlab (ver. R2019a).

6.2.5. Unweighted Epitope Score Analyses, Ranking, and Excision

UNC outputs were subjected to a moving mean analysis (MMA, n=13) of modified prediction values that incorporated both the mean (positive, indicated likely epitope

237 region) and the standard deviation (negative, indicated intra-isotype discrepancies between prediction outputs) of the residue scores. The rational for the prediction value modification, which was achieved by subtracting 3x the standard deviation from the mean for each residue, was that regions within the analyzed immunogens most likely to display both immunogenicity AND promiscuity as MHC II epitopes were of most interest.

Local maximums were identified within UNC MMAs and excised with 12 flanking residues on each side. These 25 amino acids long ‘epitope candidates’ were then ranked based on the summation of the scores of their composing residues. Additional data was also recorded, such as the epicenter residue location, the amino acid sequence, the local maximum value, and the number(s) and location(s) of the anchor residues present within the excised epitope candidate. Epitope scoring and excision was performed using Matlab (ver. R2019a) and ranking was performed via sorting in

Microsoft Excel (ver. 1808).

6.2.6. Conception and Analysis of Isotype-specific, Universal Immunogens

Chimeric proteins designed to maximize immunogenicity for HLA-DR, HLA-DQ, IAd, and IEd allotypes were constructed by concatenating the twenty highest scoring epitopes from post-prediction UNC MMAs with interspacing di-glycine-lysine linkers (for hapten attachment, KGGKGGK) flanked by cathepsin S-sensitive sequences (to facilitate processing by APCs, GGVVRGG) (29, 30). Host-derived epitopes and those that lacked predicted anchor residues, however, were removed from the ranking pool.

Using a similar approach, non-antigenic, chimeric proteins (UCnAs) were engineered by replacing the highest scoring epitopes with those that had the lowest UNC MMA scores.

For comparative purposes, a random protein sequence that was the same length as the

238 antigenic and non-antigenic chimeric proteins was also generated in Matlab and analyzed in parallel with the UCAs and UCnAs. All three of these proteins (the antigenic, non-antigenic, and random) were then processed using the epitope prediction, scoring, and analysis methods previously described. Results from this step were used to generate line charts based on means (and standard deviations, if available) for all UCAs, UCnAs, and the random protein and heat maps for final HLA-DR and HLA-DQ isotype-specific outputs in Matlab (ver. R2019a).

6.2.7. Comparing Prediction Methods and Outputs

Differences between method/isotype-specific prediction outputs were quantified using three methods. The first approach involved taking the difference between mean residue scores (averaged over all of the allotypes within the isotype) and then subsequently finding the overall mean and standard deviation of these values. This method was performed for all method/isotype combinations. The second method involved taking the absolute value of the difference between mean residue scores (averaged over all of the allotypes within the isotype) and then subsequently finding the overall mean and standard deviation of these values. This method was also performed for all method/isotype combinations. The third method involved finding the overall mean and standard deviation for all normalized, isotype-specific residues scores. Statistical comparisons were made, when possible, in SAS JMP Pro 14 using Tukey’s HSD method.

6.3. Results

6.3.1. Input Data Collection and Setup

239

Common immunogens were successfully identified in the literature and their sequences were obtained from the UniProt website. Additionally, population frequency data for

HLA-DQB1 and HLA-DRB1 alleles were sorted and target alleles were successfully identified. These frequencies, both cumulative and for each race/ethnicity, are shown in

Figure 6.1, A and Figure 6.1, B for HLA-DQB1 and HLA-DRB1 alleles, respectively. In total, the top 13 HLA-DQB1 alleles were necessary in order to achieve 99% total population coverage. With 28 possible HLA-DQA1 chains, the total number of allotypes that were needed for the HLA-DQ prediction input was therefore 364. On the other hand, the top 40 HLA-DRB1 alleles were necessary in order to achieve 99% total population coverage. With only one possible HLA-DRA1 chain, the total number of allotypes that were needed for the HLA-DR prediction input remained at 40. Since IAd and IEd are isotypes that are represented by single allotype each, their epitope prediction runs only required a single input.

6.3.2. MHC Class II Epitope Predictions, Scoring, and Analysis

In total, 6,919 epitope predictions were necessary in order to achieve target population coverage. The breakdown for this total was 6,188 HLA-DQ predictions, 680 HLA-DR predictions, and 17 IAd (NetMHCIIPan), IAd (SMM), and IEd predictions each. All of these epitope predictions were completed successfully, though there were sizeable variations in prediction run times due to differences in protein length and allotype input requirements. For example, the largest job, the DQ/KLH2 epitope prediction, took multiple hours to finish and provided 7,435,428 independent outputs. On the other hand, the smallest jobs, the IAd/CTB and IEd/CTB epitope predictions, took less than a minute to finish and provided only 525 independent outputs.

240

Epitope scoring (using UNC prediction outcomes) and anchor residue identification

(using WNC prediction outcomes) were also completed successfully. Line plots (and dot plots) of the unweighted (and weighted) results from this step for each immunogen/isotype pairing can be found in Figure C.1 and Figure C.2 in Supplemental

Materials. Plots that combined the UNC results of all isotypes within a single plot for each immunogen can be found in Figure C.3 in Supplemental Materials. An example of the prediction and analysis outputs, using DT as the model immunogen, can be found in

Figure 6.2. According to both raw and processed prediction output, all of the immunogens analyzed contained at least one MHC class II epitope. Additionally, the

WNC analysis results revealed that each immunogen contained multiple predicted anchor residues. Inter-species, inter-isotype, and inter-method differences were all observed when comparing UNC and WNC outputs for each immunogen. A summary of the prediction, scoring, and analysis results can be found in Table 6.2. Within the table,

‘hits’ indicate the number of outputs generated by the prediction software and ‘epitopes’ indicate the number of protein regions determined to be immunogenic post-UNC processing.

As expected, MMA of the UNC results indicated that there were many discrepancies between HLA-DQ and HLA-DR isotype-specific epitope predictions. There were also stretches within the target immunogens’ primary sequence that were shown to have a considerable degree of intra-isotype (and sometimes even inter-isotype) promiscuity.

For example, multiple regions of diphtheria toxin were found to be highly promiscuous and as such were included in the isotype-specific UCA design and analysis process.

Among these, the highest rated HLA-DQ 25-mer epitope was found centered at residue

241

366. An example of the final epitope identification and excision results (those for the

HLA-DQ isotype) can be found in Table 6.3. All other results (i.e. those for the HLA-DR,

IAd NetMHC, IAd SMM, and IEd SMM isotypes/methods) can be found in Table C.1 in

Supplemental Materials.

6.3.3. Isotype-specific UCA Design and Analysis Results

The top twenty epitopes identified for each HLA isotype were successfully used to generate an isotype-specific UCA. Isotype-specific universal chimeric non-antigens

(UCnAs) were also successfully generated for each HLA isotype using the twenty least immunogenic 25-mer amino acid stretches found in the UNC MMA results. Analysis of the UNC and WNC results for the UCAs and UnCAs, in addition to the random sequence protein of equal size, indicated that promising epitopes (when considering predicted immunogenicity) had been successfully identified within each immunogen for all of the isotypes included in the study. Finally, analysis of the isotype-specific HLA-DQ and HLA-DR UNC results using heat maps revealed considerable intra-isotype overlap between epitope predictions. The HLA-DQ and HLA-DR results from these analyses can be found in Figure 6.3, A and Figure 6.3, B, respectively. For the IAdNetMHC,

IAdSMM, and IEdSMM results, please refer to Figure A.4 in Supplemental Materials.

6.3.4. Inter-method Comparisons

When directly comparing method/isotype-specific outputs by taking the mean of the differences in mean residue scores, a difference could not be detected between the

DQB1 and IAd NetMHC outputs. These results can be seen in Figure 6.4, A.

Comparisons using the absolute value of the difference between mean residue scores,

242 however, indicated that all of the method/isotype-specific outputs yielded different results on an individual residue basis. These results can be seen in Figure 6.4, B.

Interestingly, the comparison between NetMHCIIpan and SMM-align methods (both were used to predict epitopes for the IAd isotype) yielded significantly different results, indicating that, at least when using the methodologies outlined here, the prediction outputs provided statistically different results on an individual residue basis. Additionally, the overall comparison between the UNC means indicated that all of the methods produced significantly different results. These results can be found in Figure 6.4, C.

When further analyzing these results, it was interesting to note that NetMHCIIPan predictions yielded UNC means greater than one while SMM-align predictions yielded

UNC means less than one.

6.4. Discussion

The goal of this project was to create a chimeric, human immunogen based on the concatenation of conjugation sites, cathepsin cleavage sites, and computationally mined

MHC class II epitopes that were predicted to 1) be highly immunogenic, 2) be highly promiscuous, and 3) contain at least one anchor residue. The idea was that if you included enough of these epitopes within the primary structure of an engineered immunogen, you would be guaranteed to illicit a sizeable adaptive immune response in the majority of those that would receive the vaccine. We dubbed the approach “casting a wide net,” and to the best of our knowledge, this strategy for potentiating conjugate vaccine immunogenicity has not been previously explored.

Any MHC epitope-based vaccine designed for human use would be impossible to satisfactorily evaluate outside of clinical trials (even when considering the plethora of

243 humanized mouse models available for immunological research) (31, 32). As such, we decided to target mouse HLA isotypes within the study as well. In this way, the approach outlined for the development of a universal human immunogen could be evaluated in mice first. The development of an MHC epitope-rich, mouse immunogen and its pre-clinical evaluation would effectively allow the assessment of epitope prediction approaches to targeted conjugate vaccine development. Crucially, however, the isogenic nature of mice (both wild-type and humanized) does not allow for the assessment of targeting genetically diverse HLA populations with universal epitopes.

For this purpose, human studies, or numerous time-consuming binding assays, would still be necessary.

When running the initial epitope predictions, we were uncertain at first on whether to mine common immunogens or randomly generated proteins for their epitope content. In the end, common immunogens were chosen due to their prior success in vaccine formulations. This success suggested that ample numbers of MHC class II epitopes were present in their primary structures and that the epitope prediction step would likely also be a success. An approach that used random proteins would also have worked, but it would have required more computational power and time in order to yield the same number of high-potential, predicted epitope candidates. There was some concern regarding whether the use of epitopes specific for existing memory T cell pools would culminate in potentiation or regulation of immune response (33). In the event that issues such as regulation are encountered during in vivo vaccine assessment, however, it is important to note that novel epitopes mined from randomly generated proteins could always be employed in place of common immunogen-derived epitopes.

244

The output data files from many of the HLA-DQ and HLA-DR predictions were immense and as such they were not reviewed in full prior to processing. Review of the first 200 lines of output for all prediction assignments, however, indicated acceptable data integrity. It was therefore assumed that any unreviewed output data would have similar, acceptable integrity. There were many additional caveats associated with the epitope and anchor residue scoring systems. First, we interpreted overlapping epitope predictions (even those that had the same core residues) as a positive outcome that should be incorporated in the final epitope analysis process. This interpretation is contrary to some of the recommendations received from IEDB staff that supported collecting only the top prediction output established for each unique core. Second, it was hypothesized that core residues that appeared more often within the prediction output would play a more sizeable role in epitope-MHC interaction. As such, we generated code that could elucidate the presence of potential anchor residues (refer to the WNC analysis). While MHC anchor residues are an established phenomenon, there were no previous studies that could validate our approach to their identification (34).

Third, the chimeric antigen validation system presented here used nearly identical methodologies to that of the epitope ranking and anchor residue identification systems.

As such, further epitope validation (both experimental and computational) should be performed in the future in order to better evaluate the immunogenicity and promiscuity assessments established using these methods.

Many interesting observations were made when analyzing the prediction outputs. UNC results indicated that most immunogens already contained regions with high immunogenicity scores. After observing this, we were worried that our “casting a wide

245 net” approach to conjugate vaccine deigns had already been invalidated. If broadly effective immunogens already existed, why would one need to be engineered? Further analysis of the HLA-DQ and HLA-DR UNC results, however, revealed that the majority of these regions had highly variable immunogenicity scores across different isotypes.

This result indicated that, while all of the immunogens would most likely elicit a sizeable immune response in some people, responses would be variable, and variable vaccine responses are generally poor corollaries for success. In fact, as a side note, it would be interesting to see if these discrepancies correlated with past conjugate vaccine clinical failures, such as the Phase III trial for NicVAX that used the EPA immunogen, the

Phase II trial for AngQb that used the QB immunogen, or the Phase II trial for TA-NIC that used the CTB immunogen (15, 35).

When compared, the WNC analysis did not appear to completely coincide with the UNC analysis. That is, some anchor residues seemed to appear in a random manner throughout the length of each protein. This indicated that, while important, the anchor residues only told a piece of the story when it came to epitope identification. We found that some of the most immunogenic epitopes predicted for the IAd and IEd isotypes were from the MSA protein. Since these epitopes are unlikely to be initiating a self- reactive immune response towards in MSA in mice, this result indicates that epitope predictions include motifs that are likely to be cognate for regulatory T cells (TReg cells). If this is in fact the case, more work will need to be done in the future in order to accommodate our prediction assessments with a means of discriminating between epitopes cognate for TH cells and TReg cells. Finally, analysis of the UNC and WNC results for the UCAs and UnCAs, in addition to the random sequence protein of equal

246 size, indicated that promising epitopes (when considering predicted immunogenicity) had been successfully identified within each immunogen for all of the isotypes included in the study.

The comparison of top scoring epitopes and statistical comparisons between prediction methods were able to identify similarities and discrepancies between various sub- groups. Comparison of NetMHCIIPan and SMM-align methods when considering IAd isotype predictions revealed differences on an individual residue basis and discrepancies between top-ranked epitopes. As such, any strategy using the “cast a wide net” approach to in silico vaccine design may want to incorporate predicted epitopes from various methods, if available. The inter-species differences observed when comparing both individual residue scores and top-ranking epitopes illuminate the inadequacy of using animal models to evaluate epitope-based vaccines designed specifically for humans. Alternatively, the inter-isotype similarities encountered in the

UNC MMA may presage difficulties in establishing specific allotype/epitope effects on immunogenicity if human studies are ever conducted.

The tools used and some aspects of the approach described here are not novel (i.e. predicting epitopes using algorithms developed by others and the string of beads approach to vaccine design). What is novel and brings value to this research, however, is the application of population dynamics and cumulative assessment of predictions to the variable immunogenicity plagued by many conjugate vaccines in the past. In the future, this approach to epitope identification could also be used for more than just the development of a universal, chimeric immunogen. For example, the development of demographic-specific carrier proteins, individually personalized carrier proteins, and

247 pathogen-specific, universal immunogens are all within the realm of possibility. More validation work should be done, however, before pursuing these lofty goals.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgements

We would like to thank the Immune Epitope Database for the use of their resources and for inviting and funding our attendance of the 2019 IEDB User Workshop. This work was partially supported by the American Association of Immunologists Careers in

Immunology Fellowship Program, the National Institute on Drug Abuse (U01DA036850), and the Biological Systems Engineering Department at Virginia Tech.

248

Tables and Figures

Table 6.1. Common immunogen information.

249

Table 6.2. Overview of prediction, scoring, and analysis results.

250

Table 6.3. HLA-DQ epitope ranking and excision results.

251

Figure 6.1. HLA population frequencies.

(A) HLA-DQB1 and (B) HLA-DRB1 cumulative (line graphs with race/ethnicity data) and individual allele (bar graphs) population frequency information is displayed here.

Cumulative frequency plots display summed frequencies of sequentially ordered HLA beta alleles (greatest to smallest) vs. the number of alleles included in the sum.

Individual allele plots display HLA alleles vs. their respective overall frequency.

252

Figure 6.2. HLA-DQ DT epitope analysis results.

(A) DT epitope scoring (UNC) and anchor residue identification (WNC) results for the

HLA-DQ isotype. Scores, in line form (UNC) or dot form (WNC), are plotted against residue number. For HLA-DQ and HLA-DR results, the center line represents the mean score and the shaded area represents ±1 standard deviation. For IAd NetMHC, IAd

SMM, and IEd SMM results, lines represent the mean score. (B) DT UNC results for all isotypes are plotted together here for easier comparison. Scores are plotted in line form against residue number. For HLA-DQ and HLA-DR results, the shaded area represents the mean score ±1 standard deviation. For IAd NetMHC, IAd SMM, and IEd SMM results, lines represent the mean score. In both parts, shaded areas representing standard deviation could not be incorporated with the IAd and IEd results due to lack of isotype diversity (these plots summarize a single immunogen/allotype prediction run).

UNC, WNC, and combined results for other isotype/immunogen combinations can be found in Appendix B.

253

Figure 6.3. Design and assessment of DQ- and DR-specific UCAs and UCnAs.

The plots and heat maps shown here summarize epitope scores for (A) HLA-DQ and

(B) HLA-DR isotypes. Plots display scores for UCAs, UCnAs, and a random protein of the same length, in line form (UNC) or dot form (WNC), plotted against residue number.

Center lines represent mean scores and shaded areas represent ±1 standard deviation.

Heat maps display HLA allotype (y-axis) vs. residue number (x-axis) for UCAs, UCnAs, and a random protein of the same length. Lighter regions and darker regions within the heat map represent lower and higher immunogenicity scores, respectively.

254

Figure 6.4. Comparing between prediction methods.

(A) Direct comparison of differences in mean, individual residue scores between methods/isotypes. Differences were summed, means, standard deviations, and standard errors were calculated, and comparisons were made. A difference between

HLA-DQ (NetMHC) and IAd NetMHC results could not be detected using a confidence level of 0.95. (B) Direct comparison of the absolute value of the differences in individual residue scores between methods/isotypes. Differences were summed, means and standard deviations (black bars), and standard errors (red bars) were calculated, and comparisons were made. Distance of mean values from zero indicates differences in mean, individual residue scores between all methods/isotypes. (C) Direct comparison of overall mean residue scores of methods/isotypes. Differences were summed, means and standard deviations (black bars), and standard errors (red bars) were calculated, and comparisons were made. All mean values were statistically different.

255

256

References

1. Dendrou CA, Petersen J, Rossjohn J, Fugger L. HLA variation and disease. Nat

Rev Immunol. 2018;18(5):325-39.

2. Chowell D, Krishna C, Pierini F, Makarov V, Rizvi NA, Kuo F, et al. Evolutionary divergence of HLA class I genotype impacts efficacy of cancer immunotherapy. Nat

Med. 2019;25(11):1715-20.

3. Blackwell JM, Jamieson SE, Burgner D. HLA and infectious diseases. Clin

Microbiol Rev. 2009;22(2):370-85, Table of Contents.

4. Chowell D, Morris LGT, Grigg CM, Weber JK, Samstein RM, Makarov V, et al.

Patient HLA class I genotype influences cancer response to checkpoint blockade immunotherapy. Science. 2018;359(6375):582-7.

5. Hraber P, Kuiken C, Yusim K. Evidence for human leukocyte antigen heterozygote advantage against hepatitis C virus infection. Hepatology.

2007;46(6):1713-21.

6. Thursz MR, Thomas HC, Greenwood BM, Hill AV. Heterozygote advantage for

HLA class-II type in hepatitis B virus infection. Nat Genet. 1997;17(1):11-2.

7. Madeleine MM, Johnson LG, Smith AG, Hansen JA, Nisperos BB, Li S, et al.

Comprehensive analysis of HLA-A, HLA-B, HLA-C, HLA-DRB1, and HLA-DQB1 loci and squamous cell cervical cancer risk. Cancer Res. 2008;68(9):3532-9.

8. Yao Y, Yang H, Shi L, Liu S, Li C, Chen J, et al. HLA Class II Genes HLA-DRB1,

HLA-DPB1, and HLA-DQB1 Are Associated With the Antibody Response to Inactivated

Japanese Encephalitis Vaccine. Front Immunol. 2019;10:428.

257

9. Moss AJ, Gaughran FP, Karasu A, Gilbert AS, Mann AJ, Gelder CM, et al.

Correlation between human leukocyte antigen class II alleles and HAI titers detected post-influenza vaccination. PLoS One. 2013;8(8):e71376.

10. Haralambieva IH, Ovsyannikova IG, Pankratz VS, Kennedy RB, Jacobson RM,

Poland GA. The genetic basis for interindividual immune response variation to measles vaccine: new understanding and new vaccine approaches. Expert Rev Vaccines.

2013;12(1):57-70.

11. Gartland AJ, Li S, McNevin J, Tomaras GD, Gottardo R, Janes H, et al. Analysis of HLA A*02 association with vaccine efficacy in the RV144 HIV-1 vaccine trial. J Virol.

2014;88(15):8242-55.

12. Vita R, Mahajan S, Overton JA, Dhanda SK, Martini S, Cantrell JR, et al. The

Immune Epitope Database (IEDB): 2018 update. Nucleic Acids Res. 2019;47(D1):D339-

D43.

13. Moore TV, Nishimura MI. Improved MHC II epitope prediction - a step towards personalized medicine. Nat Rev Clin Oncol. 2020;17(2):71-2.

14. Saylor K, Gillam F, Lohneis T, Zhang C. Designs of Antigen Structure and

Composition for Improved Protein-Based Vaccine Efficacy. Front Immunol.

2020;11:283.

15. Hu Y, Zhao Z, Saylor K, Zhang C. Nicotine Vaccines: The Past, the Present, and the Future. Neuroscience of nicotine : mechanisms and treatment. 1st edition. ed.

Cambridge, MA: Elsevier; 2019.

258

16. Saylor K, Zhang C. A simple physiologically based pharmacokinetic model evaluating the effect of anti-nicotine antibodies on nicotine disposition in the brains of rats and humans. Toxicol Appl Pharmacol. 2016;307:150-64.

17. Moise L, De Groot A, Marcello A, Tassone R, Martin W, Cousens L. Building better biotherapeutics and vaccines by design: EpiVax, Inc., an immunology company.

R I Med J (2013). 2013;96(2):19-21.

18. Michel-Todo L, Reche PA, Bigey P, Pinazo MJ, Gascon J, Alonso-Padilla J. In silico Design of an Epitope-Based Vaccine Ensemble for Chagas Disease. Front

Immunol. 2019;10:2698.

19. Grifoni A, Sidney J, Zhang Y, Scheuermann RH, Peters B, Sette A. A Sequence

Homology and Bioinformatic Approach Can Predict Candidate Targets for Immune

Responses to SARS-CoV-2. Cell Host Microbe. 2020;27(4):671-80 e2.

20. Hu Y, Zhao Z, Harmon T, Pentel PR, Ehrich M, Zhang C. Paradox of PEGylation in fabricating hybrid nanoparticle-based nicotine vaccines. Biomaterials. 2018;182:72-

81.

21. Zhao Z, Hu Y, Harmon T, Pentel PR, Ehrich M, Zhang C. Hybrid nanoparticle- based nicotine nanovaccines: Boosting the immunological efficacy by conjugation of potent carrier proteins. Nanomedicine. 2018;14(5):1655-65.

22. Gragert L, Madbouly A, Freeman J, Maiers M. Six-locus high resolution HLA haplotype frequencies derived from mixed-resolution DNA typing for the entire US donor registry. Hum Immunol. 2013;74(10):1313-20.

23. UniProt C. UniProt: a worldwide hub of protein knowledge. Nucleic Acids Res.

2019;47(D1):D506-D15.

259

24. Andreatta M, Karosiene E, Rasmussen M, Stryhn A, Buus S, Nielsen M.

Accurate pan-specific prediction of peptide-MHC class II binding affinity with improved binding core identification. Immunogenetics. 2015;67(11-12):641-50.

25. Nielsen M, Lundegaard C, Lund O. Prediction of MHC class II binding affinity using SMM-align, a novel stabilization matrix alignment method. BMC Bioinformatics.

2007;8:238.

26. Wang P, Sidney J, Dow C, Mothe B, Sette A, Peters B. A systematic assessment of MHC class II peptide binding predictions and evaluation of a consensus approach.

PLoS Comput Biol. 2008;4(4):e1000048.

27. Jensen KK, Andreatta M, Marcatili P, Buus S, Greenbaum JA, Yan Z, et al.

Improved methods for predicting peptide binding affinity to MHC class II molecules.

Immunology. 2018;154(3):394-406.

28. Paul S, Karosiene E, Dhanda SK, Jurtz V, Edwards L, Nielsen M, et al.

Determination of a Predictive Cleavage Motif for Eluted Major Histocompatibility

Complex Class II Ligands. Front Immunol. 2018;9:1795.

29. Bromme D, Bonneau PR, Lachance P, Storer AC. Engineering the S2 subsite specificity of human cathepsin S to a cathepsin L- and cathepsin B-like specificity. J Biol

Chem. 1994;269(48):30238-42.

30. Watari E, Dietzschold B, Szokan G, Heber-Katz E. A synthetic peptide induces long-term protection from lethal infection with herpes simplex virus 2. J Exp Med.

1987;165(2):459-70.

31. Walsh NC, Kenney LL, Jangalwe S, Aryee KE, Greiner DL, Brehm MA, et al.

Humanized Mouse Models of Clinical Disease. Annu Rev Pathol. 2017;12:187-215.

260

32. Janeway C. Immunobiology : the immune system in health and disease. 6th ed.

New York: Garland Science; 2005. xxiii, 823 p. p.

33. Joosten SA, Ottenhoff TH. Human CD4 and CD8 regulatory T cells in infectious diseases and vaccination. Hum Immunol. 2008;69(11):760-70.

34. Ferrante A, Gorski J. Cooperativity of hydrophobic anchor interactions: evidence for epitope selection by MHC class II as a folding process. J Immunol.

2007;178(11):7181-9.

35. Brown MJ. Success and failure of vaccines against renin-angiotensin system components. Nat Rev Cardiol. 2009;6(10):639-47.

261

Chapter 7: General Conclusions

Tobacco use is one of the most harmful habits in which mankind has ever partaken. It has devastated world health for decades and caused direct and indirect economic damages in the US alone measuring in the trillions of USD. Existing pharmacotherapies that are used to assist those trying to quit have been able to demonstrate success, but this has come at a price; they are only marginally effective and many have been associated with serious side effects. Alternatively, nicotine vaccines are a promising, relatively new approach to helping tobacco users quit. They complement existing treatment options well, are associated with minimal side effects, only require three to four treatments, and have shown promise in early clinical trials.

This promise, unfortunately, is still waiting to be realized, as all of the past nicotine vaccines to undergo clinical trials ultimately failed. Many groups that believe the concept has capacity for success, however, have begun work on next-generation lipid nanoparticle, polymer nanoparticle, DNA-scaffold, and hybrid nanoparticle nicotine vaccines. These groups have almost completely abandoned the protein-based approach that early vaccines used. It is likely that this is a mistake, as proteins offer many key immunological advantages when used as carriers in conjugate vaccine formulations.

Proteins are the de facto antigen recognized by the immune system. They can function as PRR, BCR, and TCR agonists. When using VLP or controlled aggregate proteins, they can take on the particulate nature vaunted by nanoparticles. In fact, they are arguably the most common nanoparticle used in research. They are compatible with a wide variety of biochemical conjugation techniques and their chemical and structural

262 properties are well understood. They are cheap to manufacture, at least when compared to alternatives, and they have demonstrated success in many prophylactic vaccine formulations. Before casting the protein away in search of more complex carrier molecules, it seemed more reasonable to first investigate all of the possible reasons why previous protein-based nicotine vaccines were so successful in some but ineffective in others.

The first steps in this process consisted of squeezing at much information as possible out of the data available in literature and then applying the results towards the validation of the nicotine vaccine concept. If the data pointed towards perpetual mediocrity and/or failure, either a completely new approach (i.e. something along the lines of the next-generation vaccines) or a new project would be warranted. The PBPK model was conceived for this purpose. Beyond simply validating itself as a reasonable mechanism for assessing nicotine vaccines, this model was able to confirm that protein- based nicotine vaccines did, in fact, have potential as a tobacco use cessation aid. A handful of factors, however, would need to be addressed before any nicotine vaccine would be able to make to the market. The first obvious issue was antibody concentration. The model predicted that increases in antibody concentration would improve nicotine vaccine efficacy. The first non-obvious issue was antibody affinity.

Though concerns about generating antibodies with too high of an affinity have been raised, the model was able to establish that current nicotine vaccines have room for improvement when it comes to affinity and the subsequent (predicted) improvement it provides to anti-nicotine antibodies and their capacity to prevent nicotine from reaching the brain. The second non-obvious issue was cotinine. The bodies of smokers are

263 saturated with cotinine, a major metabolite of nicotine. The model predicted that if anti- nicotine antibodies had even the slighted affinity for cotinine, the metabolite would compete with nicotine for antibody binding and severely impair their ability to block nicotine from reaching the brain. Ultimately, the model confirmed that if the nicotine binding characteristics of vaccine elicited antibodies were to be improved to levels observed in animal studies, the vaccine would be a success.

The second step of the process was to apply all of this new information, in combination with a passable understanding of the underpinnings of immunology, towards the development of improved nicotine vaccines. Two things stood out from the circumstances of the failed clinical trials; the use of VLPs and the use of toxoid proteins coincided with some degree of success. Those developing nanoparticle-based vaccines most likely saw these events as related; standalone protein-based vaccines simple did not work when applied to the nicotine vaccine concept. When looking at the approaches from a different angle, however, it seemed more plausible that their partial success was not related. That is, they each showed promise for different reasons. The Qb VLPs

(particulate but minimally diverse) likely owed their partial success to their structure, whereas exoprotein A and the other toxoid carriers (small but diverse) likely could attribute their partial success to multiple MHC class II epitopes present in their primary structure. A hybrid approach that combined the strengths these two approaches seemed very promising.

The first attempt to achieve this hybrid approach used the HPV 16 L1 protein (in the form of a VLP) as a carrier. Previous studies had shown that the HPV 16 L1 protein was exceptionally immunogenic and could accommodate large modifications at its C-

264 terminal end. Additionally, the lysine-dense protein provided ample opportunity for hapten conjugation and exhibited many other characteristics that made it an exceptional carrier candidate. Two proteins, a WT control and a version C-terminally modified with two universal T cell epitopes, were conceived. Concerns about the economics of a process are generally a post-development endeavor. In this case, however, they were initiated alongside the development efforts due to the impact the 2TEp modification had on purification outcomes. It was found that the 2TEp insert imparted a sizeable effect on the electrochemical characteristics of the protein. This effect ultimately impacted every purification step that was employed to purify the proteins, but overall the insert had a positive impact on purification outcome. This result is very promising, as the development of dual-function protein tags for purification purposes would significantly impact the vaccine production industry. Any improvement in immunogenicity as a result of the 2TEp insert, however, remains to be evaluated due to extenuating circumstances

(COVID-19).

The second hybrid approach focused more on constructing an immunogen that, beyond a reasonable doubt, would contain MHC class II epitopes cognate to 99% of the most frequently occurring HLA-DR and HLA-DQ MHC class II molecules found within the human phenome. The intent was still to create a particulate protein molecule, but it would likely happen in a much more disorganized way when compared to the HPV 16

L1 VLP. The use of MHC class II epitope prediction software in combination with HLA population frequency to mine common immunogens for their MHC class II epitope content produced the epitope candidates. The best candidates were then strung together with interspacing lysine-dense sequences and protease cleavage sequences

265 to create a universal carrier protein. Little thought was given to the ability of the protein to fold in an organized fashion due to its intended purpose as a conjugate vaccine (its fate was to be covered in hapten and as such all potentially dangerous antigenic determinants would be masked). Unfortunately, it would be impossible to assess the in vivo efficacy of such an immunogen in an animal model due to differences between

MHC class II molecule specificities. For that reason, immunogens specific for mice were also engineered. These remain to be fabricated, however, so it is uncertain when in vivo experiments will be able to prove or disprove that the approach works.

The work here highlights novel approaches to the assessment, design, and production of conjugate nicotine vaccines. The PBPK model adds another tool to the researcher’s toolbox when assessing the potential of a vaccine. The VLP technology marries what is likely to be the most positive aspects of VLP- and toxoid-based nicotine vaccines and then utilizes the modification for yet another purpose, improving purification outcomes. Finally, the design of the universal antigen is the first ever attempt to create a multi-species, chimeric, universal antigen derived from computationally mined MHC class II epitopes. Future work should focus on the validation of these studies. More data could be generated to train the PBPK model to make more accurate predictions, especially at higher nicotine doses. Additionally, the in vivo assessment of the two immunogens designed here is the inevitable next step for those projects.

266

Appendix A: Supplementary Material for Chapter 4

Nicotine material balance (refer to Figure 4.1)

Equation A.1. Tissue compartment(s) material balance (non-lungs, non-liver, non-brain, non-skin).

푑퐶 (푉 ) ∗ 푡𝑖푠푠푢푒,푁𝑖푐 푡𝑖푠푠푢푒,푁𝑖푐 푑푡

퐶푡𝑖푠푠푢푒,푁𝑖푐 = 퐶푂 ∗ 퐵퐹퐹푡𝑖푠푠푢푒,푁𝑖푐 ∗ (퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐 − ) − 퐴푏퐴푁𝑖푐 + 퐴푏퐷푁𝑖푐 푃퐶푡𝑖푠푠푢푒,푁𝑖푐

where 퐶푡𝑖푠푠푢푒,푁𝑖푐(푡 = 0) = 0 휇푀.

Equation A.2. Lungs compartment material balance.

푑퐶푙푢푛푔푠,푁𝑖푐 (푉 ) ∗ 푙푢푛푔푠,푁𝑖푐 푑푡

퐶푙푢푛푔푠,푁𝑖푐 = 퐶푂 ∗ 퐵퐹퐹푙푢푛푔푠,푁𝑖푐 ∗ ( − 퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐) − 퐴푏퐴푁𝑖푐 + 퐴푏퐷푁𝑖푐 푃퐶푙푢푛푔푠,푁𝑖푐

where 퐶푙푢푛푔,푁𝑖푐(푡 = 0) = 퐷푂푆퐸 when nicotine is dosed via smoking and 퐶푙푢푛푔,푁𝑖푐(푡 =

0) = 0 휇푀 in all other simulations.

Equation A.3. Brain compartment material balance.

푑퐶 (푉 ) ∗ 푏푟푎𝑖푛,푁𝑖푐 푏푟푎𝑖푛,푁𝑖푐 푑푡

퐶푏푟푎𝑖푛,푁𝑖푐 = 퐶푂 ∗ 퐵퐹퐹푏푟푎𝑖푛,푁𝑖푐 ∗ (퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐 − ) − 퐴푏퐴푁𝑖푐 + 퐴푏퐷푁𝑖푐 푃퐶푏푟푎𝑖푛,푁𝑖푐

− 퐵퐷 + 퐵퐴

where 퐶푏푟푎𝑖푛,푁𝑖푐(푡 = 0) = 0 휇푀.

267

Equation A.4. Liver compartment material balance.

푑퐶 (푉 ) ∗ 푙𝑖푣푒푟,푁𝑖푐 푙𝑖푣푒푟,푁𝑖푐 푑푡

퐶푙𝑖푣푒푟,푁𝑖푐 = 퐶푂 ∗ 퐵퐹퐹푙𝑖푣푒푟,푁𝑖푐 ∗ (퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐 − ) − 퐴푏퐴푁𝑖푐 + 퐴푏퐷푁𝑖푐 푃퐶푙𝑖푣푒푟,푁𝑖푐

where 퐶푙𝑖푣푒푟,푁𝑖푐(푡 = 0) = 0 휇푀.

Equation A.5. Skin compartment material balance.

푑퐶푠푘𝑖푛,푁𝑖푐 퐶푠푘𝑖푛,푁𝑖푐 (푉푠푘𝑖푛,푁𝑖푐) ∗ = 퐶푂 ∗ 퐵퐹퐹푠푘𝑖푛,푁𝑖푐 ∗ (퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐 − ) 푑푡 푃퐶푠푘𝑖푛,푁𝑖푐

where 퐶푠푘𝑖푛,푁𝑖푐(푡 = 0) = 0 휇푀.

Equation A.6. Arterial blood compartment material balance.

푑퐶 (푉 ) ∗ 푎푟푡푒푟𝑖푎푙,푁𝑖푐 푎푟푡푒푟𝑖푎푙,푁𝑖푐 푑푡

= 퐶푂 ∗ 퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐 ∗ (퐵퐹퐹푙푢푛푔푠,푁𝑖푐 − ∑ 퐵퐹퐹푡𝑖푠푠푢푒,푁𝑖푐 − 퐵퐹퐹푏푟푎𝑖푛,푁𝑖푐

− 퐵퐹퐹푙𝑖푣푒푟,푁𝑖푐 − 퐵퐹퐹푠푘𝑖푛,푁𝑖푐) − 푀퐸푇푁𝑖푐→푂푡ℎ푒푟 − 푀퐸푇푁𝑖푐→퐶표푡 − 퐸퐿퐼푀푁𝑖푐

− 퐴푏퐴푁𝑖푐 + 퐴푏퐷푁𝑖푐

where 퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐(푡 = 0) = 0 휇푀.

Equation A.7. Venous blood compartment material balance.

268

푑퐶 (푉 ) ∗ 푣푒푛표푢푠,푁𝑖푐 푣푒푛표푢푠,푁𝑖푐 푑푡

퐶푙푢푛푔푠,푁𝑖푐 퐶푡𝑖푠푠푢푒,푁𝑖푐 = 퐶푂 ∗ (−퐵퐹퐹푙푢푛푔푠,푁𝑖푐 ∗ + ∑ 퐵퐹퐹푡𝑖푠푠푢푒,푁𝑖푐 ∗ 푃퐶푙푢푛푔푠,푁𝑖푐 푃퐶푡𝑖푠푠푢푒,푁𝑖푐

퐶푏푟푎𝑖푛,푁𝑖푐 퐶푙𝑖푣푒푟,푁𝑖푐 + 퐵퐹퐹푏푟푎𝑖푛,푁𝑖푐 ∗ + 퐵퐹퐹푙𝑖푣푒푟,푁𝑖푐 ∗ + 퐵퐹퐹푠푘𝑖푛,푁𝑖푐 푃퐶푏푟푎𝑖푛,푁𝑖푐 푃퐶푙𝑖푣푒푟,푁𝑖푐

퐶푠푘𝑖푛,푁𝑖푐 ∗ ) − 퐴푏퐴푁𝑖푐 + 퐴푏퐷푁𝑖푐 푃퐶푠푘𝑖푛,푁𝑖푐

where 퐶푣푒푛표푢푠,푁𝑖푐(푡 = 0) = 퐷푂푆퐸 when nicotine is dosed via intravenous infusion and

퐶푣푒푛표푢푠,푁𝑖푐(푡 = 0) = 0 휇푀 in all other simulations.

269

Cotinine material balance (refer to Figure 4.1)

Equation A.8. Tissue compartment(s) material balance (non-lungs, non-liver, non-brain, non-skin).

푑퐶 (푉 ) ∗ 푡𝑖푠푠푢푒,퐶표푡 푡𝑖푠푠푢푒,퐶표푡 푑푡

퐶푡𝑖푠푠푢푒,퐶표푡 = 퐶푂 ∗ 퐵퐹퐹푡𝑖푠푠푢푒,퐶표푡 ∗ (퐶푎푟푡푒푟𝑖푎푙,퐶표푡 − ) − 퐴푏퐴퐶표푡 + 퐴푏퐷퐶표푡 푃퐶푡𝑖푠푠푢푒,퐶표푡

where 퐶푡𝑖푠푠푢푒,퐶표푡(푡 = 0) = 1.2 휇푀 for smokers and 퐶푡𝑖푠푠푢푒,퐶표푡(푡 = 0) = 0 휇푀 for non- smokers.

Equation A.9. Lungs compartment material balance.

푑퐶푙푢푛푔푠,퐶표푡 (푉 ) ∗ 푙푢푛푔푠,퐶표푡 푑푡

퐶푙푢푛푔푠,퐶표푡 = 퐶푂 ∗ 퐵퐹퐹푙푢푛푔푠,퐶표푡 ∗ ( − 퐶푎푟푡푒푟𝑖푎푙,퐶표푡) − 퐴푏퐴퐶표푡 + 퐴푏퐷퐶표푡 푃퐶푙푢푛푔푠,퐶표푡

where 퐶푙푢푛푔푠,퐶표푡(푡 = 0) = 1.2 휇푀 in smokers and 퐶푙푢푛푔푠,퐶표푡(푡 = 0) = 0 휇푀 in non- smokers.

Equation A.10. Brain compartment material balance.

푑퐶 (푉 ) ∗ 푏푟푎𝑖푛,퐶표푡 푏푟푎𝑖푛,퐶표푡 푑푡

퐶푏푟푎𝑖푛,퐶표푡 = 퐶푂 ∗ 퐵퐹퐹푏푟푎𝑖푛,퐶표푡 ∗ (퐶푎푟푡푒푟𝑖푎푙,퐶표푡 − ) − 퐴푏퐴퐶표푡 + 퐴푏퐷퐶표푡 푃퐶푏푟푎𝑖푛,퐶표푡

where 퐶푏푟푎𝑖푛,퐶표푡(푡 = 0) = 1.2 휇푀 in smokers and 퐶푏푟푎𝑖푛,퐶표푡(푡 = 0) = 0 휇푀 in non- smokers.

270

Equation A.11. Liver compartment material balance.

푑퐶 (푉 ) ∗ 푙𝑖푣푒푟,퐶표푡 푙𝑖푣푒푟,퐶표푡 푑푡

퐶푙𝑖푣푒푟,퐶표푡 = 퐶푂 ∗ 퐵퐹퐹푙𝑖푣푒푟,퐶표푡 ∗ (퐶푎푟푡푒푟𝑖푎푙,퐶표푡 − ) − 퐴푏퐴퐶표푡 + 퐴푏퐷퐶표푡 푃퐶푙𝑖푣푒푟,퐶표푡

+ 푀퐸푇푁𝑖푐→퐶표푡

where 퐶푙𝑖푣푒푟,퐶표푡(푡 = 0) = 1.2 휇푀 in smokers and 퐶푙𝑖푣푒푟,퐶표푡(푡 = 0) = 0 in non-smokers.

Equation A.12. Skin compartment material balance.

푑퐶푠푘𝑖푛,퐶표푡 퐶푠푘𝑖푛,퐶표푡 (푉푠푘𝑖푛,퐶표푡) ∗ = 퐶푂 ∗ 퐵퐹퐹푠푘𝑖푛,퐶표푡 ∗ (퐶푎푟푡푒푟𝑖푎푙,퐶표푡 − ) 푑푡 푃퐶푠푘𝑖푛,퐶표푡

where 퐶푠푘𝑖푛,퐶표푡(푡 = 0) = 1.2 휇푀 in smokers and 퐶푠푘𝑖푛,퐶표푡(푡 = 0) = 0 휇푀 in non-smokers.

Equation A.13. Arterial blood compartment material balance.

푑퐶 (푉 ) ∗ 푎푟푡푒푟𝑖푎푙,퐶표푡 푎푟푡푒푟𝑖푎푙,퐶표푡 푑푡

= 퐶푂 ∗ 퐶푎푟푡푒푟𝑖푎푙,퐶표푡 ∗ (퐵퐹퐹푙푢푛푔푠,퐶표푡 − ∑ 퐵퐹퐹푡𝑖푠푠푢푒,퐶표푡 − 퐵퐹퐹푏푟푎𝑖푛,퐶표푡

− 퐵퐹퐹푙𝑖푣푒푟,퐶표푡 − 퐵퐹퐹푠푘𝑖푛,퐶표푡) − 푀퐸푇퐶표푡→푂푡ℎ푒푟 − 퐸퐿퐼푀퐶표푡 − 퐴푏퐴퐶표푡

+ 퐴푏퐷퐶표푡

where 퐶푎푟푡푒푟𝑖푎푙,퐶표푡(푡 = 0) = 1.2 휇푀 in smokers and 퐶푎푟푡푒푟𝑖푎푙,퐶표푡(푡 = 0) = 0 휇푀 in non- smokers.

Equation A.14. Venous blood compartment material balance.

271

푑퐶 (푉 ) ∗ 푣푒푛표푢푠,퐶표푡 푣푒푛표푢푠,퐶표푡 푑푡

퐶푙푢푛푔푠,퐶표푡 퐶푡𝑖푠푠푢푒,퐶표푡 = 퐶푂 ∗ (−퐵퐹퐹푙푢푛푔푠,퐶표푡 ∗ + ∑ 퐵퐹퐹푡𝑖푠푠푢푒,퐶표푡 ∗ 푃퐶푙푢푛푔푠,퐶표푡 푃퐶푡𝑖푠푠푢푒,퐶표푡

퐶푏푟푎𝑖푛,퐶표푡 퐶푙𝑖푣푒푟,퐶표푡 + 퐵퐹퐹푏푟푎𝑖푛,퐶표푡 ∗ + 퐵퐹퐹푙𝑖푣푒푟,퐶표푡 ∗ + 퐵퐹퐹푠푘𝑖푛,퐶표푡 푃퐶푏푟푎𝑖푛,퐶표푡 푃퐶푙𝑖푣푒푟,퐶표푡

퐶푠푘𝑖푛,퐶표푡 ∗ ) − 퐴푏퐴퐶표푡 + 퐴푏퐷퐶표푡 푃퐶푠푘𝑖푛,퐶표푡

where 퐶푣푒푛표푢푠,퐶표푡(푡 = 0) = 1.2 휇푀 in smokers and 퐶푣푒푛표푢푠,퐶표푡(푡 = 0) = 0 휇푀 in non- smokers.

272

Additional equations

Drug-antibody mass action kinetics:

Equation A.15. Nicotine-antibody association constant.

푘표푛,푁𝑖푐 = 퐾퐴,푁𝑖푐 ∗ 푘표푓푓,푁𝑖푐

Equation A.16. Nicotine-antibody dissociation constant.

−1 퐸푠푡푖푚푎푡푒푑 푎푠 푘표푓푓,푁𝑖푐 = 5.5푒(−4) 푠푒푐

Equation A.17. Nicotine-antibody association kinetics.

2 퐴푏퐴푁𝑖푐 = (푇퐹 ∗ 퐶 푣푒푛표푢푠,퐴푏) ∗ 퐶푡𝑖푠푠푢푒,푁𝑖푐 ∗ 푉푡𝑖푠푠푢푒 ∗ 푘표푛,푁𝑖푐

Equation A.18. Nicotine-antibody dissociation kinetics.

퐴푏퐷푁𝑖푐 = 퐶푡𝑖푠푠푢푒,푁𝑖푐퐴푏 ∗ 푉푡𝑖푠푠푢푒 ∗ 푘표푓푓,푁𝑖푐

Equation A.19. Cotinine-antibody association constant.

푘표푛,퐶표푡 = 퐾퐴,푁𝑖푐 ∗ 퐶푅퐹퐶표푡 ∗ 푘표푓푓,퐶표푡

Equation A.20. Cotinine-antibody dissociation constant.

퐸푠푡푖푚푎푡푒푑 푎푠 푘표푓푓,퐶표푡 = 10 ∗ 푘표푓푓,푁𝑖푐

Equation A.21. Cotinine-antibody association kinetics.

2 퐴푏퐴퐶표푡 = (푇퐹 ∗ 퐶푣푒푛표푢푠,퐴푏) ∗ 퐶푡𝑖푠푠푢푒,퐶표푡 ∗ 푉푡𝑖푠푠푢푒 ∗ 푘표푛,퐶표푡

273

Equation A.22. Cotinine-antibody dissociation kinetics.

퐴푏퐷퐶표푡 = 퐶푡𝑖푠푠푢푒,퐶표푡퐴푏 ∗ 푉푡𝑖푠푠푢푒 ∗ 푘표푓푓,퐶표푡

where KA,Nic is an nicotine-antibody association constant obtained from literature, TF is a tissue factor comparing compartment/tissue antibody concentrations with venous antibody concentration, and CRFCot is a cross-reactivity factor of antibodies comparing nicotine and cotinine antibody IC50 values for various nicotine haptens.

퐶푡푖푠푠푢푒,퐴푏 퐼퐶50,퐶표푡 푇퐹 = 퐶푅퐹퐶표푡 = 퐶푣푒푛표푢푠,퐴푏 퐼퐶50,푁푖푐

Brain nicotine retention kinetics:

Equation A.23. Brain nicotine dispersion kinetics.

퐵퐷 = 푘푑𝑖푠 ∗ 퐶푏푟푎𝑖푛,푁𝑖푐 ∗ 푉푏푟푎𝑖푛

Equation A.24. Brain nicotine agglomeration kinetics.

퐵퐴 = 푘푎푔푔 ∗ 퐶푏푟푎𝑖푛,푁𝑖푐 ∗ 푉푏푟푎𝑖푛

where kdis and kagg are kinetic parameters calibrated using time course nicotine concentrations in the blood and brains of rats.

Nicotine clearance and metabolism:

Equation A.25. Nicotine hepatic clearance (other).

푀퐸푇푁𝑖푐→푂푡ℎ푒푟 = (1 − 푁푁𝑖푐→퐶표푡) ∗ 퐻퐶푁𝑖푐 ∗ 퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐

Equation A.26. Nicotine hepatic clearance to cotinine.

274

푀퐸푇푁𝑖푐→퐶표푡 = 푁푁𝑖푐→퐶표푡 ∗ 퐻퐶푁𝑖푐 ∗ 퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐

Equation A.27. Nicotine renal clearance.

퐸퐿퐼푀푁𝑖푐 = 푅퐸푁𝑖푐 ∗ 퐶푎푟푡푒푟𝑖푎푙,푁𝑖푐

where NNic→Cot is the proportion of nicotine that is metabolized to cotinine, HCNic is the rate of nicotine metabolism as measured by changes in blood concentration, and RENic is the rate of nicotine renal clearance as measured by changes in blood concentration.

Cotinine clearance and metabolism:

Equation A.28. Cotinine hepatic clearance.

푀퐸푇퐶표푡→푂푡ℎ푒푟 = 퐻퐶퐶표푡 ∗ 퐶푎푟푡푒푟𝑖푎푙,퐶표푡

Equation A.29. Cotinine renal clearance.

퐸퐿퐼푀퐶표푡 = 푅퐸퐶표푡 ∗ 퐶푎푟푡푒푟𝑖푎푙,퐶표푡

where HCCot is the rate of cotinine metabolism as measured by changes in blood concentration and RECot is the rate of cotinine renal clearance as measured by changes in blood concentration.

Other definitions:

Tissue/compartment drug concentration: C, tissue/compartment volume: V, tissue/compartment blood flow factor: BFF, tissue/compartment (venous) blood:tissue partition coefficient: PC, cardiac output: CO.

275

Appendix B: Supplementary Material for Chapter 5

Figure B.1. DNA Sequences for the WT and WTΔC34-2TEp proteins.

DNA sequences for the A) WT and B) WTΔC34-2TEp proteins. Codon usage was optimized for overexpression in E. coli. Lowercase letters indicate location of the His-tag and highlighted text indicates the ΔC34 portion of the WT protein and the 2TEp tag of the WTΔC34-2TEp protein.

276

Figure B.2. Amino acid sequences for the WT and WTΔC34-2TEp proteins.

Amino acid sequences for the A) WT and B) WTΔC34-2TEp proteins. Lowercase letters indicate location of the His-tag and highlighted text indicates the ΔC34 portion of the WT protein and the 2TEp tag of the WTΔC34-2TEp protein.

277

Figure B.3. Solubilized IB purity (SDS-PAGE), solubilized IB content (Western

Blot), and final purity (SDS-PAGE).

Qualitative protein assays assessing A) solubilized IB purity (SDS-PAGE), B) solubilized IB content (Western Blot), and C) final purified protein sample purity (SDS-

PAGE). A protein ladder, the WT protein, and the WTΔC34-2TEp protein were run in lanes 1-3 in all images, respectively. Additionally, all images contain lanes that were ran on the same gel but were edited to remove interspacing lanes that assessed proteins not relevant to this study.

278

Figure B.4. Recoveries over all separation steps.

Recoveries for all of the separation steps employed for the purification of the WT and

WTΔC34-2TEp proteins.

279

Appendix C: Supplementary Material for Chapter 6

Table C.1. Epitope ranking and excision results for HLR-DR, IAd (NetMHC), IAd

(SMM), and IEd (SMM) predictions.

Table 1A Epitope Ranking and Excision Results for HLA-DR, IAd (NetMHC), IAd (SMM), and IEd (SMM) Predictions HLA-DR Epitope Ranking and Excision Results IEd (SMM) Epitope Ranking and Excision Results Protein Epicenter Residues Peak Score Cummulative Score Anchor(s) Anchor Location(s) Protein Epicenter Residues Peak Score Cummulative Score Anchor(s) Anchor Location(s) TT 97 DRFLQTMVKLFNRIKNNVAGEALLD 1.23931673 29.25771788 2 87;95 BSA 228 SARQRLRCASIQKFGERALKAWSVA 1.332061912 33.02945609 2 216;233 TT 834 TQSKNILMQYIKANSKFIGITELKK 1.222141744 29.1903913 3 829;838;844 KLH1 845 ASVIREHARVKFDKVPRSRLIRKNV 1.353263143 32.34579772 3 835;841;847 KLH1 659 HHSNVDRLWAVWQALQMRRHKPYRA 1.197818092 29.12825143 3 655;663;669 HBC 160 VVRRRDRGRSPRRRTPSPRRRRSPS 1.360402139 32.18982758 2 148;164 HA 227 PNIGSRPWVRGLSSRISIYWTIVKP 1.210621385 29.09241572 1 223 KLH2 2326 FVDKIWAVWQALQKKRKRPYHKADC 1.328297575 32.11747997 2 2323;2331 KLH2 661 FYLHHSNVDRLWAIWQALQIRRGKS 1.223016625 29.00620111 1 668 KLH2 98 FPLWHRLYVVQLERALIRKKATISI 1.326421238 32.10244565 2 97;110 TT 1245 GIPLYKKMEAVKLRDLKTYSVQLKL 1.186510887 28.89341984 2 1237;1245 KLH1 662 NVDRLWAVWQALQMRRHKPYRAHCA 1.309714166 32.01090693 2 658;664 TT 800 KAMININIFMRESSRSFLVNQMINE 1.217803608 28.6009159 2 795;805 HSA 235 KCASLQKFGERAFKAWAVARLSQRF 1.302186782 31.91042495 2 228;234 KLH2 2288 QFEVVHNAIHYLVGGRQVYALSSQH 1.203015084 28.58346358 3 2277;2284;2293 HCV 44 YLLPRRGPRLGVRATRKTSERSQPR 1.30205986 31.57044545 2 38;50 KLH2 3227 IHDLRNQPRVFAGFVLSGIYTSANV 1.177785644 28.5308818 2 3228;3233 HSA 175 LYEIARRHPYFYAPELLFFAKRYKA 1.29242438 31.47336957 1 178 KLH2 98 FPLWHRLYVVQLERALIRKKATISI 1.1920528 28.36946577 1 95 KLH1 91 TFPHWHRAYVVHMERALQTKRRTSG 1.309083941 31.42064282 3 79;87;93 KLH2 2178 VFPHWHRLYTLQMDMALLSHGSAVA 1.210229007 28.31381105 2 2170;2176 KLH1 2992 HGMSIFPHWHRLHTIQFERALKKHG 1.313433603 31.38426934 2 2984;2994 KLH1 2446 LGGTKEMAWAYNRLFKYDITHALHD 1.210726064 28.06563606 2 2440;2448 KLH2 1072 FFLHRSNTDRLWAIWQALQKYRGKP 1.274854281 31.37656082 1 1071 TT 1033 DKFNAYLANKWVFITITNDRLSSAN 1.174559512 28.01734886 2 1023;1033 HCV 104 GWLLSPRGSRPSWGPTDPRRRSRNL 1.30116621 31.28739018 1 106 KLH2 1066 TTYDPIFFLHRSNTDRLWAIWQALQ 1.168893874 27.8811394 1 1071 KLH1 1482 HATTDRIWAIWQDLQRFRKRPYREA 1.311259876 31.24866808 2 1481;1483 KLH1 1619 QTEMSFVFDRLYKLDITKALKKNGV 1.19328119 27.86061865 2 1617;1626 KLH2 3143 SGLDRLWIIWQELQKLRKKPYNAAK 1.299696022 31.23869591 2 3141;3147 HPV 257 SLFFYLRREQMFVRHLFNRAGAVGE 1.175486958 27.69445425 2 247;257 DT 483 SKTHISVNGRKIRMRCRAIDGDVTF 1.286389612 31.21354495 2 479;491 KLH1 1900 LFYIHHSQTDRIWAIWQSLQRFRGL 1.171112455 27.64838307 2 1891;1909 TT 719 FLEKRYEKWIEVYKLVKAKWLGTVN 1.283462003 31.14980806 2 712;717 TT 1093 NQYVSIDKFRIFCKALNPKEIEKLY 1.168460118 27.57953502 3 1083;1089;1101 KLH2 2048 PWQFDRLYKYDITKTLKDMKLRYDD 1.287711245 31.03143146 2 2041;2049 KLH2 2731 YLHHSNTDRIWAIWQALQKYRGFQY 1.159519542 27.56618247 1 2741 KLH2 665 HSNVDRLWAIWQALQIRRGKSYKAH 1.27735389 30.85495671 1 663 EPA 327 PVQRLVALYLAARLSWNQVDQVIRN 1.184180915 27.4596692 2 316;337 KLH1 1362 GSSVAVPYWDWTKRIEHLPHLISDA 1.261139701 30.84725123 2 1355;1374 KLH1 1946 GAPYNLNDHTHDFSKPEDTFDYQKF 0.748233492 19.57327539 0 - KLH1 1850 FIEQALLALEQTNYCDFEVQFEIMH 0.805286186 21.13511302 0 - OMPC 79 VTDQLTGYGQWEYQIQGNSAENENN 0.746625129 19.46010703 0 - KLH1 606 FNQILYAFEQEDYCDFEVQFEITHN 0.809839164 21.09270234 0 - MSA 276 KVNKECCHGDLLECADDRAELAKYM 0.729899304 19.38433828 0 - TT 921 YPDAQLVPGINGKAIHLVNNESSEV 0.841571438 21.05658874 0 - HSA 459 LVEVSRNLGKVGSKCCKHPEAKRMP 0.728592458 19.37562962 0 - KLH1 346 FVYVCIPDDNDRNDDHCEKAGDFFV 0.809661059 21.02383337 0 - KLH1 973 YVDPEDGVEKHNPWFDGHIDTVDKT 0.765266065 19.27735761 0 - HBC 81 ATWVGNNLEDPASRDLVVNYVNTNV 0.801846208 20.96414189 0 - HSA 129 KQEPERNECFLQHKDDNPNLPRLVR 0.75158902 19.26585451 0 - EPA 92 ALKLAIDNALSITSDGLTIRLEGGV 0.804823272 20.83764558 0 - KLH2 2025 ICRPDQSCQEAGYFSVLGGSSEMPW 0.718838408 19.26389585 0 - DT 447 FQGESGHDIKITAENTPLPIAGVLL 0.796239873 20.808548 0 - HSA 215 LPKLDELRDEGKASSAKQRLKCASL 0.722119949 19.20756537 0 - EPA 350 RNALASPGSGGDLGEAIREQPEQAR 0.798368095 20.76401054 0 - TT 72 NPPSSLIEGASEYYDPNYLRTDSDK 0.70618693 19.11316373 0 - KLH2 822 VAVDDGFSITVEITDVDGSPPSADL 0.790669247 20.67848405 0 - MSA 131 EPERNECFLQHKDDNPSLPPFERPE 0.760163504 19.0952844 0 - KLH1 3084 QALEEDNYCDFEVQYEILHNEVHAL 0.803185943 20.67596798 0 - OMPC 204 DALRQNGDGVGGSITYDYEGFGIGG 0.744765495 18.78368165 0 - DT 290 TALEHPELSELKTVTGTNPVFAGAN 0.805693283 20.6354513 0 - OMPC 53 LHYFSDNKDVDGDQTYMRLGFKGET 0.723260251 18.72866504 0 - KLH2 557 PELATSETYLDPVTGETKNNPFHHA 0.797069557 20.619233 0 - CRM197 62 PGYVDSIQKGIQKPKSGTQGNYDDD 0.718225232 18.60111301 0 - KLH2 45 SDEVLALEKALDDLQQDDSNQGYQA 0.794696154 20.55714007 0 - HSA 329 VENDEMPADLPSLAADFVESKDVCK 0.692475165 18.3838731 0 - TT 900 NLDINNDIISDISGFNSSVITYPDA 0.79837486 20.50010347 0 - KLH1 290 TPADLFDYCELHYDYDTLNLNGMTP 0.697513438 18.09725768 0 - KLH2 1851 VYNNWFCNQALYALEQENYCDFEIQ 0.801927935 20.38373884 0 - HSA 281 CCHGDLLECADDRADLAKYICENQD 0.691499166 17.99986701 0 - TT 146 SVSFNLLEQDPSGATTKSAMLTNLI 0.792081255 20.22939822 0 - MSA 332 DTMPADLPAIAADFVEDQEVCKNYA 0.652064588 17.72814104 0 - KLH2 2261 QDETGTSVLLDQTLLALEQTDFCDF 0.791057028 20.17475965 0 - BSA 331 DAIPENLPPLTADFAEDKDVCKNYQ 0.655413512 17.57601042 0 - KLH2 2969 DQGPNGYESIAGYHGYPFLCPEHGE 0.792779863 20.11803453 0 - BSA 284 DLLECADDRADLAKYICDNQDTISS 0.637836037 16.82007469 0 - TT 671 GNFIGALETTGVVLLLEYIPEITLP 0.790422713 19.95316646 0 - HCV 53 LGVRATRKTSERSQPRGRRQPIPKA 0.62332696 16.45283751 0 - DT 329 ETADNLEKTTAALSILPGIGSVMGI 0.790319881 19.91495592 0 - IAd (NetMHC) Epitope Ranking and Excision Results IAd (SMM) Epitope Ranking and Excision Results Protein Epicenter Residues Peak Score Cummulative Score Anchor(s) Anchor Location(s) Protein Epicenter Residues Peak Score Cummulative Score Anchor(s) Anchor Location(s) EPA 194 ESNEMQPTLAISHAGVSVVMAQAQP 1.409804768 34.01314649 2 186;199 KLH1 665 RLWAVWQALQMRRHKPYRAHCAISL 1.371749286 32.97323085 2 663;669 DT 363 EEIVAQSIALSSLMVAQAIPLVGEL 1.414018877 33.77794908 2 354;363 BSA 233 LRCASIQKFGERALKAWSVARLSQK 1.338946313 32.72370101 3 221;231;241 KLH1 1716 MKADHSSDGFQAIASFHALPPLCPS 1.365889195 33.73092348 3 1713;1719;1725 KLH1 1701 LDKRQQLSLVKALESMKADHSSDGF 1.338741132 32.70288625 3 1692;1698;1713 KLH1 1706 QLSLVKALESMKADHSSDGFQAIAS 1.367952011 33.58693783 3 1695;1698;1713 HSA 236 CASLQKFGERAFKAWAVARLSQRFP 1.326023092 32.59990087 2 232;245 KLH1 2485 FFIKVSVTAVNGTVLPASILHAPTI 1.417647854 33.48495437 3 2475;2486;2491 KLH2 2118 RMELSELTERDLASLKSAMRSLQAD 1.350780344 32.42032216 2 2114;2121 HA 226 IPNIGSRPWVRGLSSRISIYWTIVK 1.358056104 33.41123696 3 223;232;237 KLH2 108 QLERALIRKKATISIPYWDWTSELT 1.324465055 32.33067359 2 98;112 KLH1 669 VWQALQMRRHKPYRAHCAISLEHMH 1.353115008 33.27124066 3 663;669;679 KLH2 866 RKIRKAVDSLTVEEQTSLRRAMADL 1.34836376 32.22074784 2 856;871 EPA 534 ALLRVYVPRSSLPGFYRTGLTLAAP 1.349545911 33.15874102 3 523;537;541 KLH2 668 VDRLWAIWQALQIRRGKSYKAHCAS 1.347320496 32.00104725 3 657;668;674 BSA 234 RCASIQKFGERALKAWSVARLSQKF 1.36036163 33.14449496 1 231 EPA 194 ESNEMQPTLAISHAGVSVVMAQAQP 1.289190046 31.83720292 2 183;199 KLH2 2127 RDLASLKSAMRSLQADDGVNGYQAI 1.341204875 33.03828056 4 2117;2121;2136;2139 KLH2 1715 LRKALKNMQADDSPDGYQAIASFHA 1.298134058 31.82225546 2 1704;1719 KLH1 2125 SLKYALSSLQADTSADGFAAIASFH 1.354318685 32.87397558 3 2115;2130;2133 KLH1 2119 SERDIGSLKYALSSLQADTSADGFA 1.306740421 31.70407818 2 2115;2124 KLH2 2180 PHWHRLYTLQMDMALLSHGSAVAIP 1.365181713 32.87088235 3 2170;2176;2182 EPA 170 YTIEMGDELLAKLARDATFFVRAHE 1.289773198 31.55505498 3 160;166;178 KLH1 1039 EIAHNYIHALVGGAQPYGMASLRYT 1.370551578 32.68116114 3 1032;1042;1050 EPA 328 VQRLVALYLAARLSWNQVDQVIRNA 1.308365497 31.27271755 3 317;328;337 BSA 218 ETMREKVLASSARQRLRCASIQKFG 1.332662075 32.67508356 3 207;212;221 KLH2 1912 SQTDRIWAIWQALQEHRGLSGKEAH 1.278892824 31.19900037 2 1903;1918 DT 303 VTGTNPVFAGANYAAWAVNVAQVID 1.361640349 32.48626935 3 291;296;303 DT 362 TEEIVAQSIALSSLMVAQAIPLVGE 1.323594427 31.04185145 3 353;356;364 KLH1 2546 LKDAMRAVMADHGPNGYQAIAAFHG 1.342122753 32.42328523 2 2535;2550 KLH1 34 SVEHLTQEETLDLQAALRELQMDSS 1.256176641 31.01310002 2 26;38 KLH2 2494 KVEVHGVNKTALPSSAIPAPTIIYS 1.377469923 32.41908105 1 2492 KLH2 2602 WHRLFVKQMEDALAAHGAHIGIPYW 1.272617067 30.93104394 1 2596 TT 685 LLEYIPEITLPVIAALSIAESSTQK 1.357344446 32.4075778 2 681;685 KLH2 1273 GSHQADEYREAVTSASHIRKNIRDL 1.279025872 30.82299857 2 1268;1278 DT 332 DNLEKTTAALSILPGIGSVMGIADG 1.34879055 32.38855308 2 322;326 KLH1 2596 WHRLYTKQMEDALTAHGARVGLPYW 1.268922188 30.56329479 1 2590 KLH2 97 SFPLWHRLYVVQLERALIRKKATIS 1.334380256 32.32716208 2 95;104 KLH2 2182 WHRLYTLQMDMALLSHGSAVAIPYW 1.294685021 30.54971096 1 2176 BSA 337 LPPLTADFAEDKDVCKNYQEAKDAF 0.889187704 22.8777891 0 - HA 292 ECITPNGSIPNDKPFQNVNKITYGA 0.786352203 21.01383677 0 - MSA 274 LTKVNKECCHGDLLECADDRAELAK 0.849823595 22.76363978 0 - KLH2 1864 LEQENYCDFEIQFEILHNGIHSWVG 0.801795897 20.97133481 0 - KLH1 1520 FDKSDNNDEATKTHATPHDGFEYQN 0.86364277 22.72315947 0 - TT 1275 ASLGLVGTHNGQIGNDPNRDILIAS 0.805922858 20.77778389 0 - KLH1 2650 GVDTTRSPRDKLFNDPERGSESFFY 0.877124425 22.70816549 0 - KLH1 1318 KFHGSPGLCQLNGNPISCCVHGMPT 0.790628922 20.62799786 0 - KLH1 1818 SKIEFEGENVHTKRDINRDRLFQGS 0.877098613 22.59232446 0 - KLH2 2917 PTIEHHGGDHHGGDTSGHDHSERHD 0.787566592 20.57258527 0 - KLH2 1956 NLNKRTQEFSKPEDTFDYHRFGYEY 0.840257121 22.48992087 0 - KLH1 601 HHTDLFNQILYAFEQEDYCDFEVQF 0.787973017 20.56181923 0 - KLH1 1950 NLNDHTHDFSKPEDTFDYQKFGYIY 0.83961758 22.44470451 0 - TT 400 LLDDTIYNDTEGFNIESKDLKSEYK 0.782916771 20.53887951 0 - TT 769 DYEYKIYSGPDKEQIADEINNLKNK 0.841666655 22.28520861 0 - TT 869 TPIPFSYSKNLDCWVDNEEDIDVIL 0.789534822 20.47109494 0 - HSA 409 VFDEFKPLVEEPQNLIKQNCELFEQ 0.864720853 22.27577681 0 - TT 1184 GLKFIIKRYTPNNEIDSFVKSGDFI 0.809036349 20.43260043 0 - OMPC 70 RLGFKGETQVTDQLTGYGQWEYQIQ 0.878512477 22.22582967 0 - KLH1 970 HEKYVDPEDGVEKHNPWFDGHIDTV 0.78096418 20.42501522 0 - TT 62 YEFGTKPEDFNPPSSLIEGASEYYD 0.870111783 22.07546296 0 - EPA 424 DALLERNYPTGAEFLGDGGDISFST 0.780642097 20.32065703 0 - LTB 84 EVPGSQHIDSQKKAIERMKDTLRIT 0.83587423 21.96807784 0 - KLH1 1804 DPETGRDIPNPFIGSKIEFEGENVH 0.781784085 20.29800737 0 - KLH1 282 IPLTNEHSTPADLFDYCELHYDYDT 0.855398816 21.79912036 0 - KLH1 1127 SVPFNVFDYKTNFNYEYDTLEFNGL 0.782864396 20.26835735 0 - HSA 278 HTECCHGDLLECADDRADLAKYICE 0.828451991 21.35123461 0 - HPV 190 VAVNPGDCPPLELINTVIQDGDMVD 0.791701951 20.20160431 0 - HSA 123 MADCCAKQEPERNECFLQHKDDNPN 0.817674015 21.33776979 0 - TT 61 RYEFGTKPEDFNPPSSLIEGASEYY 0.785574116 20.15318307 0 - KLH2 2917 PTIEHHGGDHHGGDTSGHDHSERHD 0.81613226 21.17486224 0 - TT 347 DSNGQYIVNEDKFQILYNSIMYGFT 0.784136523 20.13840859 0 - CRM197 73 QKPKSGTQGNYDDDWKGFYSTDNKY 0.825560025 21.02226348 0 - KLH2 1605 QGGEQNCKTKAGSFTILGGETEMPF 0.78347929 20.05169614 0 - BSA 278 KECCHGDLLECADDRADLAKYICDN 0.820000146 20.9142337 0 - TT 1211 YVSYNNNEHIVGYPKDGNAFNNLDR 0.782712112 19.93961497 0 - MSA 124 ADCCTKQEPERNECFLQHKDDNPSL 0.809000415 20.67652449 0 - KLH1 346 FVYVCIPDDNDRNDDHCEKAGDFFV 0.774852208 19.76451267 0 - BSA 125 DCCEKQEPERNECFLSHKDDSPDLP 0.809703984 20.51146298 0 - OMPC 199 TNNGRDALRQNGDGVGGSITYDYEG 0.775561605 19.62668446 0 -

280

Figure C.1. MHC epitope analysis results for immunogens / benchmarks 1-10.

Epitope scoring (UNC) and anchor residue identification (WNC) results for individual isotypes / prediction methods are presented side-by-side. Scores, in line form (UNC) or dot form (WNC), are plotted against residue number. For HLA-DQ and HLA-DR results, the center line represents the mean score and the shaded area represents ±1 standard deviation. For IAd NetMHC, IAd SMM, and IEd SMM results, lines represent the mean score.

281

282

Figure C.2. MHC epitope analysis results for immunogens / benchmarks 11-17.

Epitope scoring (UNC) and anchor residue identification (WNC) results for individual isotypes / prediction methods are presented together. Scores, in line form (UNC) or dot form (WNC), are plotted against residue number. For HLA-DQ and HLA-DR results, the center line represents the mean score and the shaded area represents ±1 standard deviation. For IAd NetMHC, IAd SMM, and IEd SMM results, lines represent the mean score.

283

284

Figure C.3. Combined MHC epitope analysis results for all immunogens / benchmarks.

UNC results for all isotypes / prediction methods are plotted together for easier comparison. Scores are plotted in line form against residue number. For HLA-DQ and

HLA-DR results, the shaded area represents the mean score ±1 standard deviation. For

IAd NetMHC, IAd SMM, and IEd SMM results, lines represent the mean score. Shaded areas representing standard deviation could not be incorporated with the IAd and IEd results due to lack of isotype diversity (these plots summarize a single immunogen / allotype prediction run).

285

286

Figure C.4. Design and assessment of IAd- and IEd-specific UCAs and UCnAs.

Plots display scores for UCAs, UCnAs, and a random protein of the same length, in line form (UNC) or dot form (WNC), plotted against residue number.

287