A Cocktail of Thermally Stable, Chemically Synthesized Capture Agents for the Efficient Detection of Anti-gp41 Antibodies from Human Sera and Techniques

Citation

Pfeilsticker, Jessica A. (2014) A Cocktail of Thermally Stable, Chemically Synthesized Capture Agents for the Efficient Detection of Anti-gp41 Antibodies from Human Sera and Techniques. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/0FT2-EF04. https://resolver.caltech.edu/CaltechTHESIS:05262014-170142267

Abstract

This thesis reports on a method to improve in vitro diagnostic assays that detect immune response, with specific application to HIV-1. The inherent polyclonal diversity of the humoral immune response was addressed by using sequential in situ click chemistry to develop a cocktail of peptide-based capture agents, the components of which were raised against different, representative anti-HIV antibodies that bind to a conserved epitope of the HIV-1 envelope protein gp41. The cocktail was used to detect anti-HIV-1 antibodies from a panel of sera collected from HIV-positive patients, with improved signal-to-noise ratio relative to the gold standard commercial recombinant protein antigen. The capture agents were stable when stored as a powder for two months at temperatures close to 60°C.

Item Type:

Thesis (Dissertation (Ph.D.))

Subject Keywords:

Capture agent, PCC agent, In situ click chemistry, library screening

Degree Grantor:

California Institute of Technology

Division:

Chemistry and Chemical Engineering

Major Option:

Chemistry

Thesis Availability:

Public (worldwide access)

Research Advisor(s):

Heath, James R.

Thesis Committee:

Grubbs, Robert H. (chair)
Clemons, William M.
Okumura, Mitchio
Heath, James R.

Defense Date:

2 December 2013

Funders:

Funding Agency	Grant Number
Bill and Melinda Gates Foundation	UNSPECIFIED
Institute for Collaborative Biotechnologies	W911NF-09-0001
Beckman Insitute pilot program "Spin-Polarized Molecules for Structural and Systems Biology"	UNSPECIFIED

Record Number:

CaltechTHESIS:05262014-170142267

Persistent URL:

https://resolver.caltech.edu/CaltechTHESIS:05262014-170142267

DOI:

10.7907/0FT2-EF04

Related URLs:

URL	URL Type	Description
http://dx.doi.org/10.1371/journal.pone.0076224	DOI	Article adapted for ch. 2-3
http://dx.doi.org/10.1016/j.jmr.2010.04.010	DOI	Article adapted for ch. 4

Default Usage Policy:

No commercial reproduction, distribution, display or performance rights in this work are provided.

ID Code:

8395

Collection:

CaltechTHESIS

Deposited By:

Jessica Pfeilsticker

Deposited On:

29 May 2014 21:58

Last Modified:

04 Oct 2019 00:05

Thesis Files

Preview

PDF (Full Thesis) - Final Version
See Usage Policy.
13MB

Preview

PDF (Introductory material) - Final Version
See Usage Policy.
142kB

Preview

PDF (Ch. 1) - Final Version
See Usage Policy.
66kB

Preview

PDF (Ch. 2) - Final Version
See Usage Policy.
1MB

Preview

PDF (Ch. 3) - Final Version
See Usage Policy.
273kB

Preview

PDF (Ch. 4) - Final Version
See Usage Policy.
2MB

Preview

PDF (Appendix A) - Final Version
See Usage Policy.
2MB

Preview

PDF (Appendix B) - Final Version
See Usage Policy.
537kB

Preview

PDF (Appendix C) - Final Version
See Usage Policy.
492kB

Preview

PDF (Appendix D) - Final Version
See Usage Policy.
3MB

Preview

PDF (Appendix E) - Final Version
See Usage Policy.
4MB

Repository Staff Only: item control page

A COCKTAIL OF THERMALLY STABLE, CHEMICALLY SYNTHESIZED CAPTURE AGENTS FOR THE EFFICIENT DETECTION OF ANTI-‐GP41 ANTIBODIES FROM HUMAN SERA AND TECHNIQUES Thesis by Jessica A. Pfeilsticker In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy California Institute of Technology Pasadena, California 2014 (Defended December 2, 2013) ii © 2014 Jessica A. Pfeilsticker All Rights Reserved iii ACKNOWLEDGEMENTS I begin by thanking my advisor Professor James Heath for his incredible guidance and talent. His enthusiasm for science is unrivaled, and his uncanny ability to absorb ideas and tie them together in creative and interesting ways has been a wonderful model for how to conduct research. I have enjoyed my time in his lab immensely, and am grateful to have learned from him for the last three and a half years. I also want to thank Professor Daniel Weitekamp, from whom I learned many fundamental scientific principles. He helped me transition as a young graduate student from someone who takes classes and obtains grades to someone who thinks about and conducts research. For that, and for sharing with me some of his incredible ideas and vast intellect during my first three years at Caltech, I am thankful. I want thank my committee members Professor Robert Grubbs, Professor Mitchio Okumura, and Professor William Clemons for their guidance and insight. More than once I sought their advice on how to navigate graduate school, and always received encouragement and enthusiasm in return. Next, I thank my family. They originally inspired an interest in the natural world and the wish to explore it when I was a child, and then supported me as I pursued my education in science. To Dad and Penny, Mom and Ed, thank you for being my parents and offering love and advice. John and Jason, thanks for being the coolest brothers anyone could ask for. You both have incredible futures in front of you, and I know that iv your deep thoughtfulness and curiosity will ensure a never ending supply of fun and interest as you pursue your own paths. I want to thank my lab mates for their instrumental role in my graduate experience. Valerie Norton, Jason Ollerenshaw, Eduard Chekmenev, Jan Höevener, and Daniel Rowlands in the Weitekamp group; Aiko Umeda, Kaycie Deyle, Blake Farrow, Ryan Henning, Alex Sutherland, Joey Varghese, JingXin Liang, Arundhati Nag, Samir Das, Marilena Dimotsantou, Ann Chung, Steve Millward, Errika Romero, and Connie Hsueh in the Heath group; and Heather Agnew, Bert Lai, Rosemary Rohde, and Suresh Pitram at InDi Molecular. You all kept me sane, and offered much needed scientific and personal feedback and assistance during our time as friends and colleagues. Thank you for being my biggest allies; this experience was made so much richer because you all were in it. I also need to thank the many members of Caltech staff who make research possible. Agnes Tong, thank you for all that you do to keep chemistry running smoothly. David VanderVelde, thank you for your enthusiasm and interest in sharing your knowledge of NMR. Mona Shahgholi, Jie Zhou, Felicia Rusnak, Darryl Willick, Joe Drew, Cora Carriedo, Steve Gould, Paul Carroad, Art Seiden, Mike Roy, Richard Gerhart, Jose Mendez, Engracia Alvarez, and countless others, thank you all for making the department exist and function. All of you have personally touched my time here in some way, and I wouldn’t have been able to do this without you. Finally, it is imperative that I thank the many friends who make up my support network. Lisa Sherer, thank you for being my pseudo-‐sister for the last 13 years. Caitlin v Murphy, Lisa Mauger, Tucker Jones, Nick Stadie, Luke Boosey, and Patrick Sanan, thank you for being an awesome group of people who bonded so quickly and tightly when we all first arrived at Caltech as wide-‐eyed first years. Keenan Crane, thank you for being so thoughtful and kind, and for helping me through all manners of crisis, both personal and professional. Cole DeForest, Robin DeForest, Anya Demianenko, and Michael McCoy, thank you for kindness, compassion, wine and foie gras. Young In Oh and Jean Li, thank you for mutual support in the face of adversities. Sang Tae Park, thank you for unconditionally having my back and being someone I could always rely on for personal and technical advice. Maggie Buckles, Tony Roy, and Kellie Roy, thank you for being smart, kind, wonderful people who gave me balance and kept me sane. Jon-‐Michael Sanchez, thank you for keeping life exciting, and for becoming someone with whom I could explore new ways to express myself. I love all of you, and thank you so much for being in my life. vi ABSTRACT This thesis reports on a method to improve in vitro diagnostic assays that detect immune response, with specific application to HIV-‐1. The inherent polyclonal diversity of the humoral immune response was addressed by using sequential in situ click chemistry to develop a cocktail of peptide-‐based capture agents, the components of which were raised against different, representative anti-‐HIV antibodies that bind to a conserved epitope of the HIV-‐1 envelope protein gp41. The cocktail was used to detect anti-‐HIV-‐1 antibodies from a panel of sera collected from HIV-‐positive patients, with improved signal-‐to-‐noise ratio relative to the gold standard commercial recombinant protein antigen. The capture agents were stable when stored as a powder for two months at temperatures close to 60oC. vii TABLE OF CONTENTS LIST OF FIGURES, SCHEMES, AND TABLES………………………………………………………….……xiii CHAPTER 1: Introduction…………………………………………………………………………………….….1 1.1 PCC Agent Cocktail for HIV-‐1 Diagnostics……………………………………………….2 1.2 References……………………………………………………………………………………………..5 CHAPTER2: Developing a PCC Agent Cocktail for the Detection of Anti-‐HIV Antibodies………………………………………………………………………………………………………………7 2.1 Introduction……………………………………………………………………………………………8 2.2 In Situ Click…………………………………………………………………………………………….9 2.3 Anchor Selection…………………………………………………………………………………….10 2.4 Screening……………………………………………………………………………………………….11 2.5 Biligand Characterization……………………………………………………………………….16 2.6 Materials and Methods…………………………………………………………………………..21 2.6.1 Anchor Synthesis……………………………………………………………………..21 2.6.2 OBOC Screens………………………………………………………………………...22 viii 2.6.3 Biligand Synthesis……………………………………………………………………24 2.6.4 Assays……………………………………………………………………………………..27 2.7 Conclusion……………………………………………………………………………………………..29 2.8 Acknowledgements………………………………………………………………………………..29 2.9 References…………………………………………………………………………………………….31 CHAPTER 3: PCC Agent Based Assay for the Detection of Anti-‐HIV Antibodies from Human Sera……………………………………………………………………………………………………………33 3.1 Introduction……………………………………………………………………………………………34 3.2 Patient Sample Assay……………………………………………………………………………..34 3.3 PATH Sample Assay………………………………………………………………………………..37 3.4 Stability Assay…………………………………………………………………………………….….39 3.5 Materials and Methods……………………….…………………………………………………41 3.5.1 Patient Serum ELISA………………………………………………………………..41 3.5.2 PATH Samples ELISA………………………………………………………………..41 3.5.3 Stability Assay………………………………………………………………………….42 3.6 Conclusion……………………………………………………………………………………………..43 ix 3.7 Acknowledgements………………………………………………………………………………..44 3.8 References…………………………………………………………………………………………….45 CHAPTER 4: A Selective 15N-‐to-‐1H Polarization Transfer Sequence for More Sensitive Detection of 15N-‐Choline………………………………………………………………………………...…….46 4.1 Abstract…….……………………………………………………………………………………...…..47 4.2 Introduction……………………………………………………………………………………...…..48 4.3 Theory…………………………………………………………………………………………..…...….49 4.4 Experimental…………………………………………………………………………………….…..56 4.5 Results and Discussion……………………………………………………………………….…..61 4.6 Conclusion………………………………………………………………………………………….….68 4.7 Acknowledgements…………………………………………………………………………….….68 4.8 References…………………………………………………………………………………………....69 APPENDIX A: Understanding PCC Agent Binding to Single Point Mutation E17K of Akt1 Pleckstrin Homology Domain through Molecular Dynamics…………………………………71 A.1 Introduction…………………………………………………………………………………………..72 x A.2 Peptide Capture Agent Against E17K Mutation of Akt1 PHD…………….……73 A.3 Construction of E17K and WT Systems………………………………….……………….73 A.3.1 yleaf-‐tosyl-‐biotin……………………………………………………………………..73 A.3.2 Ligand/Protein Complexes………………………………………………………74 A.4 Molecular Dynamics………………………………………………………………………………78 A.5 Binding Energy……………………………………………………………………………………….80 A.6 Conclusion……………………………………………………………………………………………..85 A.7 Acknowledgements………………………………………………………………………………..85 A.8 References…………………………………………………………………………………………….86 APPENDIX B: Peptide-‐Based General Antibody Detection Agent……………………….…87 B.1 Introduction……………………………………………………………………………………………88 B.2 Wells Peptide…………………………………………………………………………………………88 B.3 Biligand Screen………………………………………………………………………………………89 B.4 1,4 Triazole Linked Biligand Characterization…………………………………………92 B.5 1,5 Triazole Linked Biligand Characterization…………………………………………92 B.6 Click Cyclized Wells………………………………………………………………………………..97 xi B.7 Materials and Methods………………………………………………………………………….100 B.7.1 Anchor Synthesis……………………………………………………………………..100 B.7.2 OBOC screen…………………………………………………………………………..100 B.7.3 Biligand Synthesis……………………………………………………………………102 B.7.4 Click Cyclized Wells Synthesis………………………………………………….103 B.7.5 Assays……………………………………………………………………………………..104 B.8 Conclusion……………………………………………………………………………………………..106 B.9 Acknowledgements………………………………………………………………………………..106 B.10 References…………………………………………………………………………………………..107 APPENDIX C: Algorithm for Peptide Clustering……………………………………………………..108 C.1 Introduction…………………………………………………………………………………………..109 C.2 Persistence Clustering……………………………………………………………………………109 C.3 Algorithm Description……………………………………………………………………………111 C.4 Output……………………………………………………………………………………………………111 C.5 Function Code………………………………………………………………………………………..117 C.6 Conclusion……………………………………………………………………………………………..123 xii C.7 Acknowledgements………………………………………………………………………………..123 C.8 References…………………………………………………………………………………………….124 APPENDIX D: A cocktail of thermally stable, chemically synthesized capture agents for the efficient detection of anti-‐gp41 antibodies from human sera (PloS one 2013, 8, e76224)……………………………………………………………………………………………………………………125 APPENDIX E: A Selective 15N-‐to-‐1H Polarization Transfer Sequence for More Sensitive Detection of 15N-‐Choline (Journal of Magnetic Resonance 2010, 205, 125-‐129)……..131 xiii LIST OF FIGURES, SCHEMES, AND TABLES FIGURES, Chapter 2: Figure 2.1. Differential detection of 3D6 and 4B3 by anchor ligands……………………….12 Figure 2.2. Structures of peptide ligands in PCC Agent cocktail……………………………….17 Figure 2.3. Apparent affinity of A21 and biligands directed against 3D6 as determined by SPR………………………………………………………………………………………….….……18 Figure 2.4. Apparent affinity of A22 and biligand directed against 4B3 as determined by SPR……………………………………………………………………………….………….……..19 Figure 2.5. Performance of PCC agent cocktail to detect 3D6 and 4B3 from human serum…………………………………………………………………………………………………….……20 FIGURES, Chapter 3: Figure 3.1. Patient sample ELISA……………………………………………………………………………..36 Figure 3.2. PATH sample ELISA………………………………………………………………………………..38 Figure 3.3. Stability assay………………………………………………………………………………………..40 xiv FIGURES, Chapter 4: Figure 4.1. A refocused INEPT pulse sequence for polarization transfer from 15N to 1H for detection………………………………….............................................................51 Figure 4.2. The structure of 15N-‐choline…………………………………………………………………..52 Figure 4.3. Selective refocused INEPT pulse sequence for coherent polarization transfer from 15N to methyl 1H in 15N-‐choline………………………………………………………….57 Figure 4.4. Numerical simulation of the selective 15N-‐to-‐1H INEPT pulse sequence acting on a simplified spin system……………………………………………………………60 Figure 4.5. Comparison of 1H spectrum using selective 15N-‐to-‐1H INEPT sequence to 15N spectrum……………………..…………………………………………………………………62 Figure 4.6. Measurement of the 15N longitudinal relaxation time of 15N-‐choline in D2O solution using a modified version of selective 15N-‐to-‐1H INEPT sequence……..63 Figure 4.7. Amplitude of the methyl 1H signal observed in 15N-‐to-‐1H INEPT spectra of 15N-‐choline using selective and non selective pulses…………………...............65 FIGURES, Appendix A: Figure A.1. Structure of yleaf-‐tosyl-‐biotin………………………………………………………………..75 xv Figure A.2. Top three ZDOCK predicted conformations……………………………………………76 Figure A.3. Selected ligand conformation complexed with the mutant and wild type PH domains……………………………………………………………………………………………..77 Figure A.4. Rmsd plots for the MD trajectories and restarted trajectories of the E17K and WT complexes…………………………………………………………………………………………79 Figure A.5. Energy landscapes obtained from Rosetta ab initio structure prediction simulations on Rosetta@home..................................................................81 Figure A.6. Energy landscape for E17K complex………………………………………………………82 Figure A.7. Energy landscape for WT complex…………………………………………………………83 Figure A.8. Structures of the E17K and WT reference states……………………………………84 FIGURES, Appendix B: Figure B.1. Structures of peptide biligands identified in screen against IgG Fc………..93 Figure B.2. Apparent affinity of Wells and biligands directed against Fc as determined by sandwich ELISA………………………………………………………………………………..94 Figure B.3. Structure of 1,5 triazole linked biligand………………………………………………….95 xvi Figure B.4. Apparent affinity of Wells and 1,5 triazole linked WA directed against Fc as determined by sandwich ELISA…………………………………………………………..96 Figure B.5. Structure of Cu(I) click cyclized Wells peptide………………………………………..98 Figure B.6. Performance of click cylized Wells peptide variants compared to Wells peptide tested by sandwich ELISA………………………………………………………………….99 FIGURES, Appendix C: Figure C.1. Discs formed around points with growing radius ε………………………………..110 Figure C.2. Illustrative example of how lifetimes of clusters are computed……………..112 Figure C.3. Projection of the peptides onto the top two eigenvectors taken from the diagonalization of the covariance matrix. ………………………………………114 Figure C.4. Plot of eigenvectors vs. their associated eigenvalues…………………………….115 Figure C.5. Visual representation of the components of the top four eigenvectors…………………………………………………………………………………………………………...116 xvii SCHEMES, Chapter 2: Scheme 2.1. Screening strategy for selecting capture agents against anti-‐HIV antibodies 3D6 and 4B3……………………………………………………………………………14 SCHEMES, Appendix B: Scheme B.1. Screening strategy for selecting capture agents against human IgG Fc…………………………………………………………………………………………………………..90 TABLES, Chapter 2: Table 2.1. Biligand screen results…………………………………………………………………………….15 TABLES, Appendix B: Table B.1. Biligand screen results…………………………………………………………………………….91 TABLES, Appendix C: Table C.1. List of clusters and their associated lifetimes calculated for the hits resulting from the A22/4B3 screen………………………………………………………………………….113 1 Chapter 1 Introduction 2 1.1 PCC AGENT COCKTAIL FOR HIV-‐1 DIAGNOSTICS Accurate diagnoses are critical for effective healthcare. Misdiagnoses carry large personal and societal costs, especially in cases where the disease is transmissible or carries a large social stigma. HIV is one such disease, and HIV misdiagnoses can lead to psychosocial difficulties and have dangerous public health implications [1]. According to the World Health Organization (WHO), roughly 40 million people were living with HIV as of the end of 2004, and some regions have adult prevalence as high as 15-‐39%. Many countries with high incidence of HIV are resource limited and rely on easy-‐to-‐use rapid detection tests for diagnosis; however, the performance of these tests is variable. A study of the WHO HIV two-‐test diagnostic algorithm, where two positive rapid detection tests are used alone for diagnosis, showed an unacceptably high rate of false-‐positive results (10.5%), likely due to non-‐selective cross reactivity [2]. HIV diagnoses are often made indirectly through the detection of anti-‐HIV antibodies [3]. The current gold standard detection reagents are biologically-‐produced, and are expensive, susceptible to thermal and biological degradation, and have batch to batch performance variability; all of which can adversely affect the accuracy of diagnostic assays [4,5,6]. Typically in HIV diagnostic assays, multiple immunogenic and conserved antigens from HIV are expressed as regions of a single chimeric protein. That chimeric protein is then used to capture specific antibodies from the body fluid (e.g. blood, saliva or urine) of potentially infected patients; a positive assay result implies infection. However, the antigen has been shown to have poor signal to noise [7], 3 possibly related to the off-‐target reactivity responsible for false negative results seen in rapid detection tests. Robust detection reagents are needed that recognize relevant biomarkers with high sensitivity and selectivity, and that perform well in resource limited areas where controlled laboratory environments are not readily available. Protein Catalyzed Capture agents (PCC agents) are peptide-‐based detection agents that rival the affinity and selectivity of biological reagents, and that are chemically, physically, and biologically robust. As synthetic reagents, they are also highly modular and can be easily altered for incorporation into different assay platforms. Previous work has identified a number of PCC agents that detect multiple disease relevant biomarker proteins through the process of iterative in situ click screening [8,9,10], many of which were engineered to be specifically epitope targeted, and some of which have inhibiting or activating functionalities. The application of PCC agent technology to anti-‐HIV antibody detection is not as straightforward as targeting a single biomarker; the polyclonal diversity of antibodies across patient populations can translate into large variations in assay performance from patient to patient. To selectively capture the diversity of anti-‐HIV antibodies while minimizing serological cross-‐reactivity, a cocktail of PCC agents was developed against multiple representative anti-‐HIV antibodies. Chapter 2 reports on the use of iterative in situ click chemistry [8,11] to prepare a cocktail of chemically synthesized PCC agents that is designed to sample the polyclonal diversity of an antibody-‐based immune response. Three PCC agents, derived from the 4 same epitope on HIV-‐1 envelope protein gp41, were raised against two representative anti-‐HIV antibodies to create a cocktail of capture agents that sample the variations between the representative antibodies, thus increasing the reagent’s potential to detect an anti-‐HIV immune response across multiple patients. Chapter 3 then describes a PCC agent-‐based assay developed to detect human antibodies that bind a conserved region of the HIV-‐1 envelope glycoprotein gp41, using the PCC Agent cocktail described in Chapter 2. The performance is compared against the gold standard chimeric protein antigen using sera collected from a cohort of HIV-‐1-‐ positive human subjects are tested, plus controls. The assay was also used to detect anti-‐HIV antibodies from a validation panel of HIV-‐1 specific IgG containing plasma samples that were rated HIGH, MED, LOW, or NEG in accordance with the titer of anti-‐ HIV antibodies. Additionally, the thermal stability of the capture agent cocktail is reported, with an eye towards point-‐of-‐care HIV diagnostics assays that are needed in environments where refrigeration chains may not exist. Chapters 2 and 3 were taken in part from PloS one 2013, 8, e76224 (see Appendix D). 5 1.2 REFERENCES 1. Bhattacharya R, Barton S, Catalan J (2008) When good news is bad news: psychological impact of false positive diagnosis of HIV. AIDS care 20: 560-‐564. 2. Klarkowski DB, Wazome JM, Lokuge KM, Shanks L, Mills CF, et al. (2009) The evaluation of a rapid in situ HIV confirmation test in a programme with a high failure rate of the WHO HIV two-‐test diagnostic algorithm. PloS one 4: e4351. 3. Fanales-‐Belasio E, Raimondo M, Suligoi B, Butto S (2010) HIV virology and pathogenetic mechanisms of infection: a brief overview. Ann Ist Super Sanita 46: 5-‐14. 4. Daskalakis D (2011) HIV diagnostic testing: evolving technology and testing strategies. Top Antivir Med 19: 18-‐22. 5. Ginocchio CC, Carman WF, Arens MQ (2011) Update on HIV diagnostic testing algorithms. Editors' introduction. J Clin Virol 52 Suppl 1: S1. 6. Alvarez M, Chueca N, Guillot V, Bernal Mdel C, Garcia F (2012) Improving Clinical Laboratory Efficiency: Introduction of Systems for the Diagnosis and Monitoring of HIV Infection. Open Virol J 6: 135-‐143. 7. Tiwari RP, Jain A, Khan Z, Kumar P, Bhrigu V, et al. (2012) Designing of novel antigenic peptides cocktail for the detection of antibodies to HIV-‐1/2 by ELISA. Journal of immunological methods. 8. Agnew HD, Rohde RD, Millward SW, Nag A, Yeo WS, et al. (2009) Iterative in situ click chemistry creates antibody-‐like protein-‐capture agents. Angew Chem Int Ed Engl 48: 4944-‐4948. 9. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in Situ Click Chemistry Assembles a Branched Capture Agent and Allosteric Inhibitor for Akt1. Journal of the American Chemical Society 133: 18280-‐18288. 6 10. Arundhati Nag SD, Mary Beth Yu, Kaycie M. Deyle, Steven W. Millward, James R. Heath (2013) A Chemical Epitope-‐Targeting Strategy for Protein Capture Agents: The Serine 474 Epitope of the Kinase Akt2. Angewandte Chemie International Edition In press. 11. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in situ click chemistry assembles a branched capture agent and allosteric inhibitor for Akt1. J Am Chem Soc 133: 18280-‐18288. 7 Chapter 2 Developing a PCC Agent Cocktail for the Detection of Anti-‐HIV Antibodies 8 2.1 INTRODUCTION Detecting the immune response to an infectious agent can provide a useful in vitro diagnostic surrogate relative to direct pathogen detection [1]. Such assays are commonly used for detecting HIV infection because of its characteristic immunopathology [2]. Direct detection of HIV viral RNA and p24 antigen is only effective at an early stage of infection, approximately 2-‐6 weeks of initial exposure [3,4]. Antibodies against HIV envelope proteins emerge in patients’ blood around 3-‐4 weeks of infection [2,5] as the viral RNA and p24 levels decline as a result of immunocomplex formation [6]. The high serum level of anti-‐HIV IgG is maintained throughout the course of clinical latency (2-‐20+ years), during which time viral antigens are under detection limits until the onset of acquired immunodeficiency syndrome (AIDS) [2,5]. Typically in these assays, immunogenic and conserved antigens from the HIV are expressed as regions of a single chimeric protein. That chimeric protein is then used to capture specific antibodies from the body fluid (e.g. blood, saliva or urine) of potentially infected patients; a positive assay result implies infection. However, the polyclonal diversity of antibodies across a patient population can translate into large variations in assay performance from patient to patient. In addition, the chimeric recombinant proteins are biological reagents, and so may have limitations related to shelf life and batch-‐to-‐ batch variability. These limitations can adversely influence the performance of a diagnostic test [3,5,7], especially one that is deployed in harsh physical environments. 9 Previous work in the Heath group has demonstrated the ability to selectively target biologically relevant proteins with PCC Agents [8,9]. Targeting anti-‐HIV antibodies requires selective detection of anti-‐HIV antibodies over other IgG and serum proteins while maintaining sensitivity to variations in immune response across patients. This chapter reports on the discovery, synthesis, and characterization of a cocktail of capture agents that selectively target different anti-‐HIV antibodies to increase the target capture probability while minimizing cross reactivity with other proteins in sera. 2.2 IN SITU CLICK The development of a PCC Agent against a protein target utilizes the target itself to promote the 1,3-‐dipolar cycloaddition between an acetylene and an azide group to form a triazole linkage (the in situ ‘click’ reaction) [10]. The protein effectively plays the role of an extremely selective, but much less efficient, variant of the Cu(I) catalyst that is commonly used for such couplings [11,12]. For the present work, the two reacting species are peptides – one peptide (the anchor) is a chemically modified variant of a conserved, immunogenic epitope on the HIV-‐1 gp41 protein, and the second peptide is selected via an in situ click screen from a large (106 element) one-‐bead-‐one-‐compound (OBOC) [13] peptide library. The protein targets are human monoclonal antibodies raised against variants of the gp41 epitope represented by the anchor peptide. 10 2.3 ANCHOR SELECTION The PCC Agents developed here were designed to capture antibodies that are selective for residues 600-‐612 (IWCGSGKLICTTA) of gp41. Previous studies have shown that a large fraction of HIV-‐1-‐positive patients develop antibodies against this epitope [14,15]. Our strategy for sampling the polyclonal diversity of such antibodies was to develop PCC Agents that exhibited both differential, as well as similar avidities for human monoclonal antibodies (mAbs) raised against different parts of this epitope. To select the anchor peptides, we modified the polypeptide fragment corresponding to residues 600-‐612 of gp41 with artificial amino acids at multiple locations, and tested the ability of these modified cyclic peptides to detect two different monoclonal anti-‐gp41 antibodies 3D6 and 4B3 (Polymun, Klosterneuburg, Austria). The 3D6 mAb was raised against the epitope SGKLIC, whereas the 4B3 mAb was raised against SGKLICTTA. One anchor peptide, A21, was synthesized by adding a propargyl glycine (Pra) residue at the C-‐terminus of the residues 600-‐612 of gp41. This anchor peptide was also N-‐terminally tagged with a polyethylene glycol (PEG) oligomer bridge and a biotin label. For a second anchor peptide (A22), Leu-‐607 was substituted with Pra. A22 also included an N-‐ terminal PEG-‐biotin label. A21 equally detected 3D6 and 4B3 with an estimated dissociation constant (Kd) of 1-‐50 nM, while A22 differentially detected 3D6 (Kd >10 µM) and 4B3 (Kd = 1-‐50 nM) (Figure 2.1). A21 and A22 were then separately developed into PCC agents against 3D6 and 4B3, respectively. 11 2.4 SCREENING The in situ click screens are illustrated in Scheme 2.1. The target IgG is incubated with an excess of the selected anchor peptide and a large OBOC library at 4oC overnight (see Materials and Methods). The OBOC library is synthesized on TentaGel resin (Rapp Polymere, Tuebingen, Germany), and is a comprehensive library of 5-‐mers with a 6th amino acid at the N-‐terminus presenting an azide functionality. To help ensure chemical and biochemical stability, the OBOC library is comprised of non-‐natural (D) stereoisomers of the 20 natural amino acids, excluding cysteine and methionine. The in situ screen is designed to identify a secondary (2o) peptide that, when coupled to the anchor, forms a biligand with increased selectivity and/or affinity for the target IgG. 12 Figure 2.1. Differential detection of 3D6 and 4B3 by anchor ligands. Relative affinities of A21 and A22 for 3D6 and 4B3 were determined by sandwich ELISA. Biotinylated anchor ligands A21 and A22 were immobilized on streptavidin-‐coated 96-‐well plated at a concentration of 100 nM, and incubated with the solutions of target anti-‐HIV antibodies 3D6 and 4B3 at 100 nM in TBS. Captured antibody was detected by peroxidase-‐ conjugated anti-‐human IgG antibody. 13 The screen proceeds stepwise. In the first step (not shown in Scheme 2.1) the OBOC library is cleared of beads that exhibit non-‐specific binding to alkaline phosphatase-‐conjugated streptavidin (SA-‐AP), which is used as a detection reagent in a later step. Step 2 is a target screen, and so is designed to detect the presence of the bound IgG target to specific beads, and defines possible hits. The step 3 screen is designed to remove those beads from the pool of potential hits that also exhibit binding to off-‐target serum proteins. Step 4 is called a product screen, and is unique to sequential in situ click screens [16]. This screen is designed to detect for the presence of in situ clicked reaction products, which are those hit beads containing the triazole-‐linked anchor peptide. Typically, Step 2 yields a few hundred hits (~0.05% of the OBOC library). Step 3 reduces that pool by a factor of 2 or 3 to about 100 hits, and Step 4 further reduces the number of hits to around 10. This is a manageable number, meaning that each hit can be separately synthesized as a biligand using Cu(I) catalyzed click chemistry to couple the anchor and 2o peptides. A complete list of the hits for the A21/3D6 and A22/4B3 screens is given in Table 2.1. 14 Scheme 2.1. Screening strategy for selecting capture agents against anti-‐HIV antibodies 3D6 and 4B3. The flow chart represents the use of the A21 and A22 cyclic peptides as anchor ligands for separate in situ click screens against a large OBOC azide-‐presenting peptide library. 15 Table 2.1. Biligand screen results. List of pentapeptide “hits” from OBOC biligand screens performed with A21/3D6 and A22/4B3. The selected secondary ligands corresponding to (i), (ii), and (iii) are shown in red, blue, and green respectively (shown in Figure 2.2). 16 2.5 BILIGAND CHARACTERIZATION The performance of these biligands is then characterized using immunoprecipitation (pulldown) assays detected by Western blotting from spiked serum samples for specificity (data not shown), and sandwich ELISAs and surface plasmon resonance (SPR) assays for affinity estimations (Figure 2.3-‐2.4). This approach yielded two equivalently performing biligands against 3D6, and one biligand against 4B3. These three PCC agents (Figure 2.2) were combined, in equal parts, to form a capture agent cocktail. The cocktail slightly outperformed both the standard commercial chimeric antigen and A21, when tested against healthy human serum spiked with both 3D6 and 4B3 (Figure 2.5). A21 is the equivalent of the original antigenic epitope of gp41. 17 Figure 2.2. Structures of peptide ligands in PCC Agent cocktail. Acetylene-‐presenting anchor peptides (black) were derived from the immunogenic epitope of HIV-‐1 gp41 (residues 600-‐612). A22-‐nindG (i) and A21-‐hnpfk (ii) were evolved from the original epitope appended with Pra at the C-‐terminus whereas A22-‐eihny (iii) utilizes the “substituted” anchor where residue Leu-‐607 is replaced with Pra. Secondary ligand branches (colored) were identified from the in situ click screen of a 5-‐mer OBOC library presenting an azide functionality. Biligands (i) and (ii) were raised against the target anti-‐HIV antibody 3D6, and the biligand (iii) was raised against the antibody 4B3. 18 Figure 2.3. Apparent affinity of A21 and biligands directed against 3D6 as determined by SPR. A. Sensorgram and 1st order Hill fit to affinity data for A21. B. Sensorgram and 1st order Hill fit to affinity data for A21-‐nidnG (i). C. Sensorgram and 1st order Hill fit to affinity data for A21-‐hnpfk (ii). 19 Figure 2.4. Apparent affinity of A22 and biligand directed against 4B3 as determined by SPR. A. Sensorgram and 1st order Hill fit to affinity data for A22. B. Sensorgram and 1st order Hill fit to affinity data for A22-‐eihny (iii). 20 Figure 2.5. Performance of PCC agent cocktail to detect 3D6 and 4B3 from human serum. Comparative performance of the PCC Agent cocktail versus the original gp41 epitope A21 and the commercial chimeric protein antigen was tested by a sandwich ELISA. Target antibodies 3D6 and 4B3 (4 nM each) were both spiked into diluted, HIV-‐ free human serum (1% v/v in TBS), and captured antibody was detected by peroxidase-‐ conjugated anti-‐human IgG antibody. 21 2.6 MATERIALS AND METHODS 2.6.1 Anchor Synthesis The anchor peptides A21 and A22 were synthesized C → N on 200-‐300 mgs of Rink amide resin (AnaSpec, Fremont, CA) with a Titan peptide synthesizer using standard Fmoc chemistry [17]. The synthesis used Fmoc and side chain protected L amino acids (Aaptec, Louisville, KY), Fmoc protected propargylglycine (Pra) (Aaptec, Louisville, KY), PEG5 (EMD Millipore,Germany), and biotin (Sigma-‐Aldrich, St. Louis, MO). The finished peptides were side-‐chain deprotected and cleaved from resin in a mixture of 95% v/v trifluoroacetic acid (TFA) (Sigma-‐Aldrich, St. Louis, MO), 5% v/v dH2O, and 5% triethylsilane (TES) (Sigma-‐Aldrich, St. Louis, MO), and precipitated in diethyl either. Cu(Phen)3 stock solution was prepared by mixing 150 mM Cu(II) sulfate pentahydrate (Sigma-‐Aldrich, St. Louis, MO) and 500 mM 1,10-‐phenanthroline (Sigma-‐Aldrich, St. Louis, MO) in 4:1 H2O:Ethanol. The ether-‐precipitated ligands were cyclized by stirring the cleaved peptides dissolved in ~5 mL acetonitrile (ACN) overnight with 1 mL of the Cu(Phen)3 stock solution. The cyclization mixtures were lyophilized, and the final peptide products were purified using reverse phase (RP) HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column (Phenomenex, Torrance, CA). MALDI-‐TOF MS: A21: Expected mass [M+H]+ = 2006.96, Observed mass [M+H]+ = 2006.24. A22: Expected mass [M+H]+ = 1892.89, Observed mass [M+H]+ = 1892.85 22 2.6.2 OBOC Screens All screens used naïve one-‐bead-‐one-‐compound (OBOC) D-‐pentapeptide libraries on TentaGel resin (TG) (Rapp Polymere, Tuebingen, Germany) of the form NH2-‐Az4-‐ X1X2X3X4X5-‐TG, excluding cysteine and methionine, where Az4 is L-‐azidolycine. Libraries were synthesized on a Titan peptide synthesizer (Aaptec, Louisville, KY) using a split/mix method, and couplings were done using standard Fmoc SPPS chemistry in N-‐ methylpyrrolidone (NMP). A21 with 3D6 Screening against 3D6 was carried out in four main steps: pre-‐clear, target screen, anti-‐screen, and product screen. Step 1 preclear: 400 mg of OBOC library was blocked in 5% w/v dry milk in Tris buffered saline (TBS) (25 mM Tris-‐HCl, 150 mM NaCl, 10 mM MgCl2, pH 7.6) for 2 hr at room temperature, and then incubated with 1:10,000 phosphatase-‐conjugated streptavidin (SA-‐AP, Promega, Madison, WI) in 0.5% w/v dry milk in TBS for 1 hr. The beads were washed in high salt buffer (25 mM Tris-‐HCl, 750 mM NaCl, 10 mM MgCl2, pH 7.6) for 1 hr, and then developed with BCIP/NBT (Promega, Madison, WI) in BCIP buffer (100 mM Tris-‐HCl, 150 mM NaCl, 1 mM MgCl2, pH 9). After 20 minutes, the reaction was quenched with conc. HCl, and the purple beads were discarded. The remaining library was decolorized in NMP, dried with methanol (MeOH) and dichloromethane (DCM), and then swelled and blocked in 5% w/v dry milk in TBS for 2 hr at room temperature. Step 2 target screen: 4 mL of a solution containing 11.4 µM A21 and 147 nM 3D6 in 0.5% w/v dry milk in TBS was added to the precleared OBOC 23 library, and the in situ click reaction was allowed to proceed overnight at 4°C. Bound target was probed with 1:10,000 AP-‐conjugated rabbit anti-‐human IgG Fc (Thermo Scientific, Rockford, IL) in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The beads were then washed in high salt buffer for 1 hr at room temperature, and developed with BCIP/NBT in BCIP buffer. The reaction was quenched after 20 minutes with conc. HCl, and the purple beads were retained and decolorized in NMP and dried with MeOH and DCM. The hit beads were then stripped of all bound protein by washing with 7.5 M Guanidine-‐HCl (pH 2) for 2 hr at room temperature, rinsed 10 times with water, and then blocked in 5% w/v dry milk in TBS for 2 hr at room temperature. Step 3 anti-‐screen: A solution of 1% v/v human serum (Omega Scientific, Tarzana, CA) in TBS was incubated with the hit beads from Step 2 for 1 hr at room temperature. Off-‐target human antibodies bound to the beads were probed with 1:10,000 AP-‐conjugated rabbit anti-‐human IgG Fc in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The beads were then washed in high salt buffer for 1 hr, and developed with BCIP/NBT in BCIP buffer. The reaction was quenched after 20 minutes with conc. HCl, and the purple beads were discarded. The remaining beads were decolorized in NMP and dried with MeOH and DCM. The anti-‐screened beads were swelled in water and washed with Guanidine-‐HCl for 2 hr at room temperature, rinsed 10 times with water, and then blocked in 5% w/v dry milk in TBS for 2 hr at room temperature. Step 4 product screen: The beads were incubated for 1 hr at room temperature with 1:10,000 SA-‐AP in 0.5% w/v dry milk in TBS, washed for 1 hr with high salt buffer, and then developed with BCIP/NBT in BCIP buffer for 20 minutes. The reaction was quenched with conc. HCl, and 24 the purple beads were retained and sequenced using a Procise CLC Edman Degradation Protein Sequencing System (Applied Biosystems, Kingston, RI). A22 with 4B3 Screening against 4B3 was carried out similarly to that described for 3D6, with two exceptions. In Step 2, the in situ click reaction solution contained 10 µM A22 and 430 nM 4B3 in 0.5% w/v dry milk in TBS. In Step 3, the beads were anti-‐screened against 0.1% v/v human serum in TBS spiked with 430 nM 3D6. 2.6.3 Biligand Synthesis Pra-‐OtBu The C-‐terminally protected Pra molecule used in the synthesis of (i) and (ii) was made by refluxing Fmoc-‐L-‐Pra-‐OH (Aaptec, Louisville, KY) with tert-‐butyl-‐ tricholroacetamide (Sigma-‐Aldrich, St. Louis, MO) at 50°C for 3 hr in DCM. The formed product was separated from unreacted species on a silica flash column in DCM. Peptides (i) and (ii) The secondary ligands including N-‐terminal Az4 for (i) and (ii) were synthesized C → N on 200-‐300 mgs of Rink amide resin with a Titan peptide synthesizer using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected D amino acids and L-‐Az4 (Aaptec, Louisville, KY). Fmoc-‐ and C-‐terminally protected Pra (Pra-‐OtBu) was clicked on-‐bead to the azide side chain of Az4 by adding equal molar amounts of 25 Pra-‐OtBu, Cu(I) iodide (Sigma-‐Aldrich, St. Louis, MO), and L-‐ascorbic acid (Sigma-‐Aldrich, St. Louis, MO), in 2x excess to the azide on bead. The click mixture was agitated in 20% v/v piperidine in NMP for 3 hr at room temperature, and the copper was chelated by repeated washing with 10% w/v sodium diethylduthiocarbamate trihydrate (Sigma-‐ Aldrich, St. Louis, MO) and 10% v/v N,N-‐diisopropylethylamine (DIEA ) (Sigma-‐Aldrich, St. Louis, MO) in NMP until the beads were clear. The Fmoc deprotected Pra was then the starting point for the remainder of the anchor component of each ligand, including N-‐terminal PEG5 and biotin. The anchor components were synthesized using standard Fmoc chemistry with protected L amino acids. The finished peptides were side-‐chain deprotected and cleaved from resin in 95:5:5 TFA:dH2O:TES, then precipitated in diethyl either. The ligands were cyclized by stirring the cleaved peptides dissolved in ~5 mL ACN overnight with 1 mL of the Cu(phen)3 stock solution. The cyclization mixtures were lyophilized, and the final peptide products were purified using RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. MALDI-‐TOF MS: (i): Expected mass [M+H]+ = 2733.30, Observed mass [M+H]+ = 2733.64. (ii): Expected mass [M+H]+ = 2843.40, Observed mass [M+H]+ = 2843.60. Peptide (iii) A22 was synthesized, cyclized, and purified as described above. The secondary ligand component including N-‐terminal Az4 for (iii) was synthesized C → N on Rink amide resin with a Titan peptide synthesizer using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected D amino acids and L-‐Az4. Resulting 26 secondary peptide was side-‐chain deprotected and cleaved from resin in 95:5:5 TFA:dH2O:TES, precipitated in diethyl either, and purified using RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. The purified secondary peptide was clicked to A22 in solution by mixing with 2x molar excess of A22, 10x Cu(I) iodide, 50x L-‐ascorbic acid, and 4x Tris[(1-‐benzyl-‐1H-‐1,2,3-‐triazol-‐4-‐yl)methyl]amine (TBTA) (Sigma-‐Aldrich, St. Louis, MO) in 5:1 NMP:dH2O and stirring overnight at room temperature. The clicked biligand product was purified from the reaction mixture by RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. Expected mass [M+H]+ = 2720.31, observed mass [M+H]+ = 2720.5. Scaling up (iii) The secondary ligands including N-‐terminally Boc-‐protected Az4 for (iii) was synthesized C → N on Rink amide resin with a Titan peptide synthesizer using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected D amino acids and L-‐Az4 (AnaSpec, Fremont, CA). Truncated A22 (Pra-‐ICTTA) was synthesized C → N on Sieber amide resin (ChemPep, Wellington, FL) with a Liberty 1 microwave peptide synthesizer (CEM, Matthews, NC) using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected L amino acids and Fmoc-‐L-‐Pra. Fully protected truncated A22 was cleaved from the resin with 1% v/v TFA in DCM, and neutralized with DIEA. Solvent was removed using a rotary vacuum evaporator and the protected peptide was purified by RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. Resulting truncated A22 was clicked on-‐bead to the secondary peptide through the azide side 27 chain of Boc-‐Az4 by adding equal molar amount of the protected truncated A22, 2x molar excess of Cu(I) iodide, and 2x excess of L-‐ascorbic acid to the azide on bead. The click mixture was agitated in 20% v/v piperidine in NMP overnight at room temperature, and then the copper was chelated by repeated washing with 10% w/v sodium diethyldithiocarbamate trihydrate and 10% v/v DIEA in NMP until the beads were clear. The Fmoc deprotected Pra was then the starting point for the remainder of the anchor component, including N-‐terminal PEG5 and biotin (Biotin-‐PEG5-‐IWGCSGK). The anchor components were synthesized using standard Fmoc chemistry with protected L amino acids. The finished peptide was N-‐terminal and side-‐chain deprotected and cleaved from resin in 95:5:5 TFA:dH2O:TES, then precipitated in diethyl either. The ligand was cyclized by stirring the cleaved peptide dissolved in ~5 mL ACN overnight with 1 mL of the Cu(phen)3 stock solution. The cyclization mixtures were lyophilized, and the final peptide products were purified using RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. 2.6.4 Assays Anchor binding ELISA Biotinylated anchor peptide was immobilized on streptavidin-‐coated 96-‐well plates (Thermo Scientific Pierce, Rockford, IL) at a concentration of 100 nM in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the anchor peptides were incubated with either 100 nM 3D6 or 4B3 in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with horseradish peroxidase (HRP)-‐ 28 conjugated mouse monoclonal antibody to human IgG Fc (Abcam, Cambridge, MA), diluted 1:10,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate (KPL, Gaithersburg, MD), then quenched with 1 M H2SO4 and read at 450 nm. Surface Plasmon Resonance SPR experiments were performed on a Biacore T200 instrument. Biotinylated peptides were immobilized on streptavidin-‐coated sensor chips (GE Healthcare Biosciences, Pittsburg, PA) at R.U. values ranging from 7-‐12. 4B3 or 3D6 was flowed at various concentrations in 1x HBS-‐EP buffer (GE Healthcare Biosciences, Pittsburgh, PA), and affinity curves were fit using the default settings of the Biacore evaluation software. Cocktail binding ELISA Recombinant multi-‐epitope chimeric HIV antigen (“chimera”) (BioLink International, Lisle, IL) was chemically biotinylated using ChromaLink Biotin Labeling Kit (Solulink, San Diego, CA) according to the manufacturer’s instructions, using 10x molar excess of the ChromaLink biotinylation reagent to the buffer-‐exchanged protein. Streptavidin-‐coated 96-‐well plates were saturated with the biotinylated biligand cocktail or chimera using 1 uM solutions in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the capture reagents were incubated with either 1% v/v human serum or a mixture of 4 nM each 3D6 and 4B3 spiked in 1% v/v human serum in TBS at room temperature for 1 hr. Bound antibody was probed with HRP-‐conjugated mouse 29 monoclonal antibody to human IgG Fc, diluted 1:15,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate, then quenched with 1 M H2SO4 and read at 450 nm (A450). The A450 for each sample is normalized against the A450 for the serum control, to yield a measurement of the signal-‐to-‐noise ratio of the assay. 2.7 CONCLUSION Described in this chapter is a method for developing a PCC Agent cocktail to capture the diversity of human antibodies generated in response to an infectious agent, with specific application to HIV. The components of the cocktail were raised against 4B3 and 3D6, two representative anti-‐HIV antibodies, and individual and collective binding of the capture agents to the target antibodies was characterized. The cocktail displayed a moderately enhanced performance compared to the original gp 41 epitope (A21) and the gold standard chimeric antigen when used to detect 4B3 and 3D6 from human serum. 2.8 ACKNOWLEDGEMENTS This work was done in collaboration with Aiko Umeda, Blake Farrow, Connie L. Hseuh, and Kaycie M. Deyle. It was funded primarily by a grant from the Bill and Melinda Gates Foundation. Additional funding for the development of screening approaches, and for certain capture agent characterization methods, was provided by 30 the Institute for Collaborative Biotechnologies through grant W911NF-‐09-‐0001 from the U.S. Army Research Office. The content of the information does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred. 31 2.9 REFERENCES 1. (1996) In: Baron S, editor. Medical Microbiology. 4th ed. Galveston (TX). 2. Fanales-‐Belasio E, Raimondo M, Suligoi B, Butto S (2010) HIV virology and pathogenetic mechanisms of infection: a brief overview. Ann Ist Super Sanita 46: 5-‐14. 3. Alvarez M, Chueca N, Guillot V, Bernal Mdel C, Garcia F (2012) Improving Clinical Laboratory Efficiency: Introduction of Systems for the Diagnosis and Monitoring of HIV Infection. Open Virol J 6: 135-‐143. 4. Tebourski F, Slim A, Elgaaied A (2004) The significance of combining World Health Organization and Center for Disease Control criteria to resolve indeterminate human immunodeficiency virus type-‐1 Western blot results. Diagn Microbiol Infect Dis 48: 59-‐ 61. 5. Daskalakis D (2011) HIV diagnostic testing: evolving technology and testing strategies. Top Antivir Med 19: 18-‐22. 6. Butto S, Suligoi B, Fanales-‐Belasio E, Raimondo M (2010) Laboratory diagnostics for HIV infection. Ann Ist Super Sanita 46: 24-‐33. 7. Ginocchio CC, Carman WF, Arens MQ (2011) Update on HIV diagnostic testing algorithms. Editors' introduction. J Clin Virol 52 Suppl 1: S1. 8. Agnew HD, Rohde RD, Millward SW, Nag A, Yeo WS, et al. (2009) Iterative in situ click chemistry creates antibody-‐like protein-‐capture agents. Angew Chem Int Ed Engl 48: 4944-‐4948. 9. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in Situ Click Chemistry Assembles a Branched Capture Agent and Allosteric Inhibitor for Akt1. Journal of the American Chemical Society 133: 18280-‐18288. 32 10. Huisgen R (1984) 1,3-‐Dipolar cycloaddition – introduction, survey, mechanism. In: Padwa A, editor. 1,3-‐Dipolar Cycloaddition Chemistry Wiley. pp. 1-‐176. 11. Thirumurugan P, Matosiuk D, Jozwiak K (2013) Click Chemistry for Drug Development and Diverse Chemical-‐Biology Applications. Chem Rev. 12. Finn MG, Fokin VV (2010) Click chemistry: function follows form. Chem Soc Rev 39: 1231-‐ 1232. 13. Lam KS, Lebl M, Krchnak V (1997) The "One-‐Bead-‐One-‐Compound" Combinatorial Library Method. Chem Rev 97: 411-‐448. 14. Du AP, Limal D, Semetey V, Dali H, Jolivet M, et al. (2002) Structural and immunological characterisation of heteroclitic peptide analogues corresponding to the 600-‐612 region of the HIV envelope gp41 glycoprotein. J Mol Biol 323: 503-‐521. 15. Xu JY, Gorny MK, Palker T, Karwowska S, Zolla-‐Pazner S (1991) Epitope mapping of two immunodominant domains of gp41, the transmembrane protein of human immunodeficiency virus type 1, using ten human monoclonal antibodies. J Virol 65: 4832-‐4838. 16. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in situ click chemistry assembles a branched capture agent and allosteric inhibitor for Akt1. J Am Chem Soc 133: 18280-‐18288. 17. Fields GB, Noble RL (1990) Solid phase peptide synthesis utilizing 9-‐fluorenylmethoxycarbonyl amino acids. International journal of peptide and protein research 35: 161-‐214. 33 Chapter 3 PCC Agent based assay for the detection of anti-‐HIV antibodies from human sera 34 3.1 INTRODUCTION HIV diagnostic assays typically use chimeric proteins comprised of immunogenic and conserved antigens from the HIV to capture specific anti-‐HIV antibodies from the body fluid of potentially infected patients. However, the biologically produced chimeric recombinant proteins are susceptible to thermal, chemical, and biological degradation, as well as batch to batch variability. These limitations can adversely influence the performance of a diagnostic test [1,2,3], especially when used outside of controlled laboratory settings. This chapter explores the ability of the PCC Agent cocktail described in Chapter 2 to detect anti-‐HIV antibodies from human sera as compared to that of a standard recombinant chimeric antigen like those typically used in anti-‐HIV antibody detection assays. The chimeric protein used contained a fragment of HIV-‐1 gp41 (residues 546-‐ 692), the “O” group HIV-‐1 gp41 immunodominant region (residues 580-‐623), and a fragment of HIV-‐2 gp36 (residues 591-‐617). A good performance of this chimeric antigen has been reported elsewhere [4]. Two sets of samples were assayed, one clinical and one commercial. In both cases the PCC Agent cocktail showed increased average signal to noise compared to the chimeric antigen. 3.2 PATIENT SAMPLE ASSAY The PCC Agent cocktail and the standard antigen were co-‐evaluated against a panel of clinical samples using sandwich ELISAs. Serum samples were collected from nine HIV-‐1-‐positive patients in Southern California. For the comparison assays, 35 streptavidin-‐coated 96-‐well plates were saturated with the PCC Agent cocktail or chemically biotinylated chimera in triplicates. Serum samples were diluted to 1% v/v in Tris-‐buffered saline (TBS) supplemented with 0.1% w/v bovine serum albumin (BSA), and incubated in the wells for 1 hr at room temperature. Unbound proteins were washed off and captured IgG was detected with a peroxidase-‐conjugated mouse monoclonal anti-‐human IgG-‐Fc antibody. The comparison assay results for the patient serum samples along with a healthy control are shown in Figure 3.1. The signal-‐to-‐noise (S/N) ratio in these assays is defined as the measured ELISA signal for a given patient sample, divided by that for the healthy control. The PCC Agent cocktail performed at least as well, and typically much better, than the standard chimeric protein antigen. The average S/N improvement by the cocktail PCC agent over the chimeric antigen was a factor of 2.5. 36 Figure 3.1. Comparative performance of the PCC Agent cocktail versus the commercial chimeric protein, using sandwich ELISAs to detect anti-‐HIV-‐1 IgGs from a panel of sera samples collected from nine HIV-‐positive patients. The absorbance at 450 nm (A450) for each sample is normalized against the A450 for the healthy control, to yield a measurement of the signal-‐to-‐noise ratio of the assay. The PCC Agent cocktail, which is designed to capture a subset of anti-‐gp41 IgGs, exhibits superior performance for every sample, even though the chimeric protein is designed to capture antibodies against multiple HIV-‐1-‐associated epitopes (those containing fragments of HIV-‐1 gp41, “O” group HIV-‐1 gp41 immunodominant region, and HIV-‐2 gp39). For the assays, the PCC agent cocktail and the biotinylated chimeric antigen were immobilized on a streptavidin-‐ coated 96-‐well plate and incubated with diluted patient serum (1% v/v). Captured anti-‐ HIV antibodies were detected using peroxidase-‐conjugated anti-‐human IgG antibody. 37 3.3 PATH SAMPLE ASSAY The PCC Agent cocktail and the standard antigen were also co-‐evaluated against a panel of commercial samples. Serum samples from individual patients containing high, medium, or low titers of anti-‐HIV antibodies were obtained from SeraCare, through PATH of POC diagnostics program funded by the Bill and Melinda Gates Foundation. For the comparison assays, streptavidin-‐coated 96-‐well plates were saturated with the PCC Agent cocktail or chemically biotinylated chimera in triplicates. Serum samples were diluted to 1% v/v in Tris-‐buffered saline (TBS) supplemented with 0.1% w/v bovine serum albumin (BSA), and incubated in the wells for 1 hr at room temperature. Unbound proteins were washed off and captured IgG was detected with a peroxidase-‐ conjugated mouse monoclonal anti-‐human IgG-‐Fc antibody. The comparison assay results for the serum samples are shown in Figure 3.2. The signal-‐to-‐noise (S/N) ratio in these assays is defined as the measured ELISA signal for a given high, medium, or low titer sample, divided by that for the respective healthy control. The PCC Agent cocktail generally outperformed the standard chimeric protein antigen, with the exception of the low titer samples. The reasons for this are not entirely clear. However, the average S/N improvement by the cocktail PCC agent over the chimeric antigen was a factor of 2.8. 38 Figure 3.2. Comparative performance of the PCC Agent cocktail versus the commercial chimeric protein, using sandwich ELISAs to detect anti-‐HIV-‐1 IgGs from a panel of HIV-‐1 specific IgG containing plasma samples provided by PATH. The samples were rated HIGH, MED, or LOW in accordance with the titer of anti-‐HIV antibodies, and pooled HIV negative sera samples were used as a control (NEG). The absorbance at 450 nm (A450) for each sample is normalized against the A450 for the healthy control, to yield a measurement of the signal-‐to-‐noise ratio of the assay. The PCC Agent cocktail, which is designed to capture a subset of anti-‐gp41 IgGs, exhibits superior performance for every sample, even though the chimeric protein is designed to capture antibodies against multiple HIV-‐1-‐associated epitopes (those containing fragments of HIV-‐1 gp41, “O” group HIV-‐1 gp41 immunodominant region, and HIV-‐2 gp39). For the assays, the PCC agent cocktail and the biotinylated chimeric antigen were immobilized on a streptavidin-‐ coated 96-‐well plate and incubated with diluted patient serum (1% v/v). Captured anti-‐ HIV antibodies were detected using peroxidase-‐conjugated anti-‐human IgG antibody. 39 3.4 STABILITY ASSAY We then tested the PCC Agents for thermal stability. The PCC Agent cocktail component (iii) were synthesized at a large scale for an academic setting (~7 mg, Figure 3.3) and the lyophilized samples were stored under N2 at 25°C, 37°C, or 57°C for 58 days. The samples were then analyzed by HPLC to determine the presence of any degradation product. The traces of the peptide at each temperature are nearly identical, indicating little to no degeneration at these temperatures (Figure 3.3). The performance of these stored PCC Agents was then also tested in an ELISA, with no detectable loss of performance (Figure 3.3). 40 Figure 3.3. Samples of (iii) were stored as a powder (inset photo), under N2 at temperatures up to 57oC for ~2 months, and resolved by analytical HPLC to determine the presence of any degradation product. The HPLC traces reveal that the fingerprint of the PCC Agent is unchanged. The inset shows that the assay performance of the PCC Agent is also unaffected. 41 3.5 MATERIALS AND METHODS 3.5.1 Patient serum ELISA Peripheral blood was obtained from HIV-‐1 infected patients at the University of California, Los Angeles (UCLA) Medical Center between July and August of 2012. Sera were stored at -‐80°C before subsequent analysis. Streptavidin-‐coated 96-‐well plates were saturated with the biotinylated biligand cocktail or chimera using 1 uM solutions in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the capture reagents were incubated with patient serum samples diluted to 1% v/v in TBS containing 0.1% w/v bovine serum albumin (BSA). Additionally, both cocktail and chimera capture reagents were incubated with 1% v/v commercial healthy human serum in TBS. Bound antibody was probed with HRP-‐conjugated mouse monoclonal antibody to human IgG Fc, diluted 1:10,000 in TBS with 0.1% w/v BSA. The colorimetric assay was developed with TMB substrate, then quenched with 1 M H2SO4 and read at 450 nm (A450). The A450 for each sample is normalized against the A450 for the healthy control, to yield a measurement of the signal-‐to-‐noise ratio of the assay. 3.5.2 PATH samples ELISA Serum samples were obtained from SeraCare, through PATH of POC diagnostics program funded by the Bill and Melinda Gates foundation. Five groups of four samples were tested, where each group contained high, medium, and low titer samples, along with a negative sample that did not contain anti-‐HIV antibodies. Sera were stored at -‐ 80°C before subsequent analysis. Streptavidin-‐coated 96-‐well plates were saturated 42 with the biotinylated biligand cocktail or chimera using 1 uM solutions in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the capture reagents were incubated with patient serum samples diluted to 1% v/v in TBS containing 0.1% w/v bovine serum albumin (BSA). Additionally, both cocktail and chimera capture reagents were incubated with 1% v/v commercial healthy human serum in TBS. Bound antibody was probed with HRP-‐conjugated mouse monoclonal antibody to human IgG Fc, diluted 1:10,000 in TBS with 0.1% w/v BSA. The colorimetric assay was developed with TMB substrate, then quenched with 1 M H2SO4 and read at 450 nm (A450). The A450 for the samples in each group are normalized against the A450 for the negative sample in that group. 3.5.3 Stability assay Small amounts of (iii) were stored under N2 at 25°C, 37°C, or 57°C for 58 days. Samples were diluted in 1:1 ACN/H2O with 0.1% v/v TFA to an equal concentration determined by measuring the absorbance values at 280 nm using NanoDrop spectrophotometer (Thermo Scientific, Waltham, MA) and resolved by analytical RP-‐ HPLC. The remaining samples were lyophilized and used in a single point sandwich ELISA. Streptavidin-‐coated 96-‐well plates were saturated with the samples of (iii) recovered from the storage experiments using 1 uM solutions in TBS. The plates were blocked with 5% w/v dry milk in TBS, then incubated with either 1% v/v human serum or 10 nM 4B3 spiked in 1% v/v human serum in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with HRP-‐conjugated mouse 43 monoclonal antibody to human IgG Fc diluted 1:15,000 in 0.5% w/v dry milk in TBS. The colorimetric assay was developed with TMB substrate, then quenched with 1 M H2SO4 and read at 450 nm. 3.6 CONCLUSION This thesis demonstrated the successful application of this method for HIV-‐1 diagnostics by producing a cocktail of three PCC Agents that detected the presence of anti-‐HIV antibodies in clinical samples with a significantly enhanced signal-‐to-‐noise relative to the standard, recombinant protein-‐based chimeric antigen. In a recent report, an antigenic peptide cocktail comprised of synthetic peptides derived directly from gp120/V3-‐I (HIV-‐1 Indian isolate), gp41 (HIV-‐1), and gp36 (HIV-‐2), as well as the recombinant protein rp24 (HIV-‐1) was shown to also provide superior performance relative to the chimeric antigen [5]. This points to the possibility that expanding the current approach by developing multiple cocktails of PCC Agents, each directed against a distinct HIV epitope, would likely provide superior performance to that reported here. The strategy presented provides a promising approach for developing assays for detecting the immune response to other infectious agents, especially where challenges associated with the polyclonal nature of a humoral immune response can compromise assay sensitivity. 44 3.7 ACKNOWLEDGEMENTS This work was done in collaboration with Aiko Umeda, Jocelyn T. Kim, and Bert T. Lai. It was funded primarily by a grant from the Bill and Melinda Gates Foundation. Additional funding for the development of screening approaches, and for certain capture agent characterization methods, was provided by the Institute for Collaborative Biotechnologies through grant W911NF-‐09-‐0001 from the U.S. Army Research Office. The content of the information does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred. 45 3.8 REFERENCES 1. Daskalakis D (2011) HIV diagnostic testing: evolving technology and testing strategies. Top Antivir Med 19: 18-‐22. 2. Ginocchio CC, Carman WF, Arens MQ (2011) Update on HIV diagnostic testing algorithms. Editors' introduction. J Clin Virol 52 Suppl 1: S1. 3. Alvarez M, Chueca N, Guillot V, Bernal Mdel C, Garcia F (2012) Improving Clinical Laboratory Efficiency: Introduction of Systems for the Diagnosis and Monitoring of HIV Infection. Open Virol J 6: 135-‐143. 4. Chin CD, Laksanasopin T, Cheung YK, Steinmiller D, Linder V, et al. (2011) Microfluidics-‐based diagnostics of infectious diseases in the developing world. Nat Med 17: 1015-‐1019. 5. Tiwari RP, Jain A, Khan Z, Kumar P, Bhrigu V, et al. (2012) Designing of novel antigenic peptides cocktail for the detection of antibodies to HIV-‐1/2 by ELISA. Journal of immunological methods. 46 Chapter 4 A Selective 15N-‐to-‐1H Polarization Transfer Sequence for More Sensitive Detection of 15N-‐Choline 47 4.1 ABSTRACT This chapter discusses selective polarization transfer from 15N to methyl 1H spins in 15N-‐choline. The sensitivity and information content of heteronuclear nuclear magnetic resonance is frequently optimized by transferring spin order of spectroscopic interest to the isotope of highest detection sensitivity prior to observation. This strategy was extended to 15N-‐choline using the scalar couplings to transfer polarization from 15N to choline’s nine methyl 1H spins in high field. A theoretical analysis of a sequence using nonselective pulses showed that the optimal efficiency of this transfer is decreased by 62% as the result of competing 15N-‐1H couplings involving choline’s four methylene protons. A frequency-‐selective pulse was therefore incorporated to support evolution of only the 15N-‐methyl 1H coupling during the transfer period. This sequence provided a 52% sensitivity enhancement over the nonselective version in in vitro experiments on a sample of t hermally polarized 15N-‐choline in D2O. Further, the 15N T1 of choline in D2O was measured to be 217 ± 38 s, the 15N-‐methyl 1H coupling constant was found to be 0.817 ± 0.001 Hz, and the larger of choline’s two 15N-‐methylene 1H coupling constants was found to be 3.64 ± 0.01 Hz. Possible improvements and applications to in vivo experiments using long-‐lived hyperpolarized heteronuclear spin order are discussed. 48 4.2 INTRODUCTION It is now possible to routinely generate samples of certain small biomolecules with nuclear polarizations on the order of ten percent using PASADENA (parahydrogen and synthesis allow dramatically enhanced nuclear alignment) [1] or DNP (dynamic nuclear polarization) [2; 3]. These technologies show promise for clinical applications based on the in vivo characterization of metabolic processes on the seconds timescale. Experiments of this nature, rely on storage of the polarization on heteronuclei with long population relaxation times, such as carbonyl 13C or quaternary 15N nuclei, in order to minimize relaxation losses during sample delivery and to provide time for transport, binding, and metabolism. However, the sensitivity enhancement afforded by this strategy is partly lost if the final signal is also detected on a low-‐𝛾 heteronucleus. Chekmenev et al. [4] have proposed that optimal sensitivity can be obtained by storing polarization on a slowly-‐relaxing heteronuclear spin and then using an INEPT[5] sequence to coherently transfer magnetization to nearby 1H nuclei for detection, and have demonstrated the method with two hyperpolarized reagents, 1-‐13C-‐succinate-‐d2 and 2,2,3,3-‐tetrafluoropropyl 1-‐13C-‐propionate-‐d3 (TFPP). Those experiments were designed assuming that the relevant 1H and 13C sites were 3-‐spin ABX systems. Here we demonstrate an extension of this polarization transfer strategy to hyperpolarized 15N-‐choline, an A9BB’CC’X system of 14 spin ½ nuclei. Choline is a useful biomarker, showing significantly altered uptake and metabolism in diseased brain tissue [6] and in malignant tumor cells in the breast and prostate [7; 8]. Hyperpolarization of 49 15 N-‐choline to (4.6 ± 1)% has been reported after two hours of DNP at 1.4 K [3]. The 15N longitudinal relaxation time of T1 = 120 s in blood [3] is promising for metabolic studies. We have investigated two INEPT-‐based pulse sequences for 15N-‐to-‐1H polarization transfer in 15N-‐choline. The first is a nonselective refocused INEPT sequence similar to that used by Chekmenev et al. to transfer polarization from 13C to multiple 1H sites in succinate and TFPP [4]. While this sequence is effective, a product operator analysis shows that significantly greater sensitivity can be obtained by using a selective sequence targeting choline’s nine degenerate methyl 1H spins. We therefore also present a selective refocused INEPT transfer that uses a REBURP pulse to support coherence transfer from 15N to choline’s methyl 1H spins while suppressing competing transfer pathways to methylene 1H nuclei. These pulse sequences are experimentally demonstrated in vitro using a thermally polarized sample of 15N-‐choline in aqueous solution. 4.3 THEORY At a given spin polarization and probe efficiency, the sensitivity of detection in an NMR experiment scales as the square of the gyromagnetic ratio of the nuclide being observed. In the case of 15N hyperpolarization, therefore, a sensitivity enhancement of (γ 1 H / γ 15 N )2 = 97 can be obtained by transferring to 1H for detection rather than directly observing a 15N signal, assuming unity transfer of spin angular momentum, as is closely approached within an isolated 15N-‐1H spin pair with a resolved scalar coupling. Such a transfer can be achieved using refocused INEPT (Figure 4.1), the standard pulse 50 sequence for polarization transfer between spins with a direct scalar coupling. The effect of this sequence when applied to 15N-‐choline can be determined using a product operator analysis. In the following discussion the relevant 15N-‐1H scalar coupling constants will be labeled J m , J a , and J b as indicated in Figure 4.2. Couplings between the methyl and methylene 1H spins will be unimportant and are neglected. 51 Figure 4.1. A refocused INEPT pulse sequence for polarization transfer from 15N to 1H for detection. Narrow (wide) rectangles denote radiofrequency pulses with a tip angle of 𝜋/2 (𝜋). All pulses are applied along the x axis except where indicated. 52 Figure 4.2. The structure of 15N-‐choline. Labels indicate the relevant 15N-‐1H scalar couplings. 53 First, consider the effect of the sequence on the spin system of Figure 4.2, while initially neglecting the effects of the 15N-‐methylene 1H couplings J a and J b . The system’s magnetization is proportional to the dimensionless 15N spin angular momentum, with initial value normalized to ρ0 = S z . This is converted to in-‐phase 15N single-‐quantum coherence by the first 15N (𝜋/2)x pulse to obtain ρ1 = − S y . This coherence evolves under the 15N-‐methyl 1H scalar couplings during the INEPT transfer period τ 1 to produce terms that are antiphase with respect to the methyl 1H spins. The simultaneous 𝜋 pulses in the middle of τ 1 refocus chemical shift evolution, leaving only the effects of couplings between 15N and each of the nine methyl 1H spins. At the end of τ 1 the system’s state is given by: 9 ρ 2 = − S y cos9 (π J mτ 1 ) + ∑ 2S x I iz cos8 (π J mτ 1 ) sin(π J mτ 1 ) i =1 8 9 7 2 + ∑ ∑ 4S y I iz I jz cos (π J mτ 1 ) sin (π J mτ 1 ) + HO. i =1 j =i +1 Here, HO indicates higher-‐order terms, specifically single-‐quantum 15N coherence that is antiphase with respect to three or more 1H spins. After the transfer delay, simultaneous 1 H (𝜋/2)x and 15N (𝜋/2)y pulses are applied: 9 ρ3 = − S y cos9 (π 2 J N −CH τ 1 ) + ∑ 2 S z I iy cos8 (π 2 J N −CH τ 1 ) sin(π 2 J N −CH τ 1 ) 3 8 9 3 i =1 7 2 2 3 2 + ∑ ∑ 4 S y I iy I jy cos (π J N −CH3τ 1 ) sin (π J N −CH3τ 1 ) + HQ. i =1 j =i +1 The pulses convert the terms HO to high-‐order multiple-‐quantum coherences, which have been abbreviated as HQ . At this point, 15N coherences and multiple-‐quantum 1H 54 coherences, which will not lead to observable signals at the end of the experiment, can be dropped from the density operator. This leaves only single-‐quantum 1H coherences that are antiphase with respect to 15N: 9 ρ3 = ∑ 2S z I iy cos8 (π J mτ 1 ) sin(π J mτ 1 ) .Ny → i =1 Nycos πJτ cos πJτ cos πJτ + 2NxH3zcos πJτ cos πJτ sin πJτ + 2NxH2zcos πJτ sin πJτ cos πJτ + 4NxH2zH3zcos πJτ sin πJτ sin πJτ + 2NxH1zsin πJτ cos πJτ cos πJτ + 4NxH1zH2zcos πJτ sin πJτ sin πJτ + 4NxH1zH2zsin πJτ sin πJτ cos πJτ + 8NxH1zH2zH3zsin πJτ sin πJτ sin πJτ The next part of the sequence is a refocusing period of duration τ 2 = 1/ (2 J m ) during which the antiphase single-‐quantum 1H terms evolve into in-‐phase 1H coherence: 9 ρ4 = −∑ I ix cos8 (π J mτ 1 ) sin(π J mτ 1 ) . i =1 The detected 1H signal is proportional to the total magnetization of the nine methyl protons, so the dependence of the initial signal amplitude on τ 1 is: S sel = Asel 9cos8 (π J mτ 1 ) sin(π J mτ 1 ) + C . (1) This expression includes a hardware-‐dependent proportionality constant Asel and a constant C accounting for any DC offset in the observed signal. Equation 1 is similar to the expression derived by Doddrell et al. for a coherence transfer in the opposite direction, from n equivalent 1H spins to a heteronucleus [9]. The value of τ 1 that maximizes Ssel is given by: 55 τ1 = 1 ⎛ 1 ⎞ tan −1 ⎜ ⎟ . π Jm ⎝ 2 2 ⎠ For the value J m = 0.82 Hz measured experimentally for 15N-‐choline, the optimal transfer delay is τ1 = 0.132 s, yielding Ssel = 1.873 Asel + C . The transfer function of the angular momentum from 15N to 1H or vice versa (Eq. 1) can exceed unity. The conserved quantity is the sum of the squares of the coefficients of the normalized product operators expressing the density operator. Contributions to the observable signal from choline’s nine methyl 1H spins combine linearly and constructively at the optimized value of τ 1 . The above calculation is readily generalized to include the effects of scalar coupling between the 15N spin and the four methylene protons ( J a and J b in Fig. 2), yielding an analogous relative signal function: Snon = Anon 9cos8 (π J mτ 1 ) sin(π J mτ 1 )cos 2 (π J aτ 1 )cos 2 (π J bτ 1 ) + C (2) A proportionality constant Anon and a DC offset constant C have again been included. Maximizing this function numerically for the best known values of the three coupling constants for choline, J m = 0.82 Hz, J a = −0.57 Hz, J b = 3.64 Hz, gives an optimal transfer delay τ1 = 0.233 s, which yields a signal of Snon = 0.721 Anon + C . Comparing the maximum values of Ssel and Snon , it is apparent that the 15N-‐ methylene 1H couplings have the effect of reducing the observable 1H signal by 62% in the idealized case of an experiment for which Asel = Anon and C = 0 . It would clearly be advantageous to suppress these couplings. This can be achieved by replacing the 1H p 56 pulse in the middle of τ 1 with a selective pulse tailored to invert the methyl 1H spins without affecting the methylene 1H spins. 4.4 EXPERIMENTAL Data was collected on a Varian UNITYINOVA spectrometer operating at a 1H resonance frequency of 500 MHz. The sample was prepared by dissolving 20 mg 15N-‐ choline chloride (ISOTEC, Miamisburg, Ohio) in 700 µl D2O. Refocused INEPT experiments were performed using the pulse sequence shown in Figure 4.3, which implements the selective 15N-‐methyl 1H transfer described in the theory section using a 6.695 ms 1H REBURP pulse [10] centered at the methyl 1H frequency. A time τ reb = 6.4 ms, equal to the effective evolution time of the 15N-‐methyl 1H coupling during the REBURP pulse, was subtracted from the transfer delay as indicated in Fig. 3. For nonselective INEPT experiments, the REBURP was replaced with a nonselective 𝜋 pulse and τ reb was set to zero. 57 Figure 4.3. Selective refocused INEPT pulse sequence for coherent polarization transfer from 15N to methyl 1H in 15N-‐choline. Narrow (wide) rectangles denote radiofrequency pulses with a tip angle of 𝜋/2 (𝜋), applied with B1 field strengths of 23.9 kHz for 1H and 15.6 kHz for 15N. The shaped 1H pulse is a 6.695 ms REBURP [10] centered on the methyl 1H frequency. A four step phase cycle (𝜑1 = x, x, x, x; 𝜑2 = x, y, x, y; 𝜑rec = x, y, x, y) was used to suppress signals originating from 1H magnetization and to prevent carry over of signal from one transient to the next. A recycle delay d1 of 60 s (~ 0.25 T1 ) was used. Since it is impractical to make this delay long enough to allow full relaxation of the choline 15N spin, a purge element consisting of a 15N 𝜋/2 pulse sandwiched by gradients g0 and g1 is used to eliminate 15N magnetization at the start of d1 , ensuring a consistent starting magnetization for all experiments. A 1H purge element comprising a 1H 𝜋/2 pulse and gradient g2 is used at the start of each transient. Gradients g3 and g4 are used to suppress coherence transfer pathways created by imperfect 𝜋 pulses. 58 To allow efficient testing in the absence of hyperpolarization, the sequence includes a 15N purge element before the recycle delay d1 to ensure that the system reaches a consistent state at the start of each transient. This measure is necessary because it is impractical to allow full relaxation of the choline 15N spin after each transient owing to its very long T1 . The sequence includes a 15N 𝜋/2 pulse immediately prior to acquisition in order to purge any 1H coherence that remains antiphase with respect to 15N after the refocusing period. In the absence of this pulse, intermittent phase aberrations were observed in the methyl 1H resonance. For 15N T1 measurements by inversion-‐recovery, the pulse sequence of Figure 4.3 was modified by adding a 15N 𝜋 pulse and a variable recovery delay immediately before the 1H purge pulse. For these experiments, the 15N purge element before d1 was removed and d1 was increased to 600 s. Data analysis was performed in MATLAB (The MathWorks, Natick, MA) using scripts developed in house. Spectral data was quantified by fitting the methyl 1H resonance to a complex Lorentzian lineshape function, except for inversion-‐recovery data, which was quantified by integration. A downhill simplex algorithm was used for least-‐squares data fitting, and uncertainties in the extracted parameters were estimated using a bootstrap Monte Carlo method [11]. The value of τ reb , which is subtracted from the INEPT transfer delay in order to account for evolution of the 15N-‐methyl 1H scalar coupling during the long REBURP pulse, was determined using a numerical simulation of the pulse sequence. Calculations 59 were performed in GAMMA v4.1.2 [12] in Hilbert space using a spin system of practical size, consisting of one 15N and six methyl 1H spins, with J m = 0.82 Hz. The 1H resonance frequency was set to 500 MHz, and the carrier frequencies were set on resonance with the 15N and methyl 1H spins. The REBURP and the nonselective pulses were simulated with finite widths corresponding to the experimental values, and relaxation and phase cycling were neglected. The pulse sequence shown in Figure 4.3 was simulated, and the methyl 1H signal amplitude was calculated, for each of a series of values of the INEPT transfer delay (τ1 − τ reb ) . The signal data, shown in Figure 4.4, was then subjected to least-‐squares fit to a version of Eq. 1 appropriate for a system containing six methyl 1H spins and including τ reb as an adjustable parameter, resulting in the value of τ reb = 6.4 ms used experimentally. 60 Figure 4.4. Numerical simulation of the selective 15N-‐to-‐1H INEPT pulse sequence acting on a simplified spin system. Circles are simulated methyl 1H signal amplitudes for different values of the INEPT transfer delay (τ1 − τ reb ) . The line was calculated using a modified form of Eq. 1 with parameter values from a least-‐squares fit to the data. This fit was used to determine τ reb , the effective evolution time of the 15N-‐methyl 1H coupling during the sequence’s REBURP pulse. 61 4.5 RESULTS AND DISSCUSSION Figure 4.5 shows a spectrum of 15N-‐choline recorded using the selective 15N-‐to-‐ 1 H INEPT sequence of Figure 4.3, alongside a 15N-‐detected spectrum of the same sample recorded in a similar measurement time. It is not meaningful to compare these spectra quantitatively because they were both recorded using a probe that is optimized for direct heteronuclear observation, a factor which diminishes the sensitivity advantage of 1 H detection. However, even under these unfavorable circumstances the signal-‐to-‐noise ratio is clearly much higher in the selective INEPT spectrum (Figure 4.5a) than in the 15N-‐ detected spectrum (Figure 4.5b). A modified version of the selective 15N-‐to-‐1H INEPT experiment was used to measure the 15N longitudinal relaxation time of 15N-‐choline in D2O solution, and a value of T1 = (217 ± 38) s was obtained (Figure 4.6). This is comparable to the value T1 = (285 ± 12) s previously reported by Gabellieri et al. [3] for 15N-‐choline in 90% H2O / 10% D2O. 62 Figure 4.5. (a) 1H spectrum of 15N-‐choline recorded using the selective 15N-‐to-‐1H INEPT sequence of Fig. 3. Only the methyl resonance is shown. (b) Single-‐pulse 15N spectrum of 15N-‐choline, recorded with tip angle 𝜋/2. Spectra (a) and (b) were each collected on the same Varian AutoX Dual Broadband probe, taking the sum of four transients with a recycle delay of 60 s. An enhancement factor of 11.0 greater sensitivity when using selective 15N-‐to-‐1H INEPT sequence was estimated from Monte Carlo fits of (a) and (b) to Lorentzian lineshapes. The peak amplitudes were used as “signals” and the standard deviations as “noise” to construct pseudo “SNR” values for each spectrum to take into account differences in the linewidths of 1H and 15N peaks. The ratio calculated enhancement factor. " SNR "1H " SNR "15 N gave the 63 Figure 4.6. Measurement of the 15N longitudinal relaxation time of 15N-‐choline in D2O solution using a modified version of the pulse sequence shown in Figure 3. The experiment was repeated for 13 values of the inversion-‐recovery delay T = 0.001, 15, 30, 45, …, 180 s. Crosses are experimental methyl 1H signal amplitudes obtained by integrating the spectral data. The line was calculated using the indicated equation and parameter values from a least-‐squares fit to the data. 64 A series of 15N-‐choline 15N-‐to-‐1H INEPT spectra were recorded using both selective and nonselective versions of the pulse sequence of Figure 3 with values of the transfer delay τ 1 ranging from 6.4 ms to 700 ms. The methyl 1H signal amplitudes observed in these experiments are plotted in Figure 7. The selective INEPT sequence allows 15N-‐choline to be detected with 52% greater sensitivity than the nonselective sequence, comparing the maximum signal amplitude observed in each case. In the Theory section, the amplitude of the methyl 1H signal was predicted as a function of the transfer delay for the selective (Eq. 1) and nonselective (Eq. 2) experiments using a product operator analysis. By fitting the observed signal amplitudes to Eqs. 1 and 2, the optimal length of τ 1 for each experiment can be determined, as well as precise values of choline’s 15N-‐1H coupling constants. A simultaneous least-‐squares fit to both datasets yields parameter values Asel = 0.632 ± 0.001, Anon = 1.000 ± 0.005 , J m = (0.817 ± 0.001) Hz, J b = (3.64 ± 0.01) Hz, and C = 0.000 ± 0.001 . These values for the coupling constants agree well with ones derived from previously published 1H-‐14N values: J m = 0.80 Hz and J b = 3.61 Hz [13]. The value of the smaller 15N-‐methylene 1H coupling constant was found to have a negligible effect on the fitting process, so this parameter was fixed at a reasonable value of J a = −0.57 Hz. 65 Figure 4.7. Amplitude of the methyl 1H signal observed in 15N-‐to-‐1H INEPT spectra of 15N-‐ choline using selective (circles) and nonselective (triangles) versions of the pulse sequence shown in Figure 3, as a function of the INEPT transfer period τ 1 . The lines were calculated using Eq. 1 (solid line) and Eq. 2 (dashed line) with parameter values from a least-‐squares fit to the data. 66 Note that, for experiments using the selective and nonselective versions of the pulse sequence of Figure 3 on the same spectrometer hardware and sample, the constants C in Eqs. 1 and 2 should be the same, but Asel may be smaller than Anon due to signal losses during the selective sequence’s REBURP pulse. Comparing the values of Asel and Anon from the least squares fit it appears that 36% of the signal is lost in this way. This can be attributed in part to miscalibration of the REBURP pulse. A three pulse sequence (𝜋/2 -‐ REBURP -‐ 𝜋/2) was used to calibrate the REBURP pulse by seeking pulse width and power parameters that gave the best null signal. It would be more appropriate to choose a calibration sequence that uses the REBURP pulse for inversion, as it is used in the selective INEPT sequence, rather than for refocusing. Also, a sequence with a larger number of pulses would be preferable, so that the cumulative effects of B1 inhomogeneity would be comparable in the calibration sequence and the selective INEPT sequence. We have presented a pulse sequence that suppresses the effects of the four methylene protons on the INEPT transfer from 15N to methyl 1H in 15N-‐choline. Note that the same result can be achieved by selective deuteration of 15N-‐choline. This approach might provide the additional benefit of increasing the molecule’s 15N longitudinal relaxation time and would be an option for DNP experiments. However, the present selective recoupling strategy is clearly advantageous for 15N-‐choline hyperpolarization using PASADENA. In these experiments, two of the methylene 1H spins derive from a parahydrogen molecule which reacts with an unsaturated precursor to initiate the hyperpolarization process. Although deuteration of the alkene precursor 67 can remove two of the methylene 1H spins, those which supply the spin order cannot be eliminated. The success of this selective coherence transfer strategy in 15N-‐choline has clear implications for studies of other hyperpolarized small biomolecules. By directing the spin order to a subset of the possible detection spins, sensitivity is improved. The recoupling strategy demonstrated here accomplishes this by way of a frequency selective π pulse, which simplifies, and often shortens, the polarization transfer step. Similar selective INEPT sequences could be used to improve the sensitivity of the 13C-‐to-‐ 1 H polarization transfer experiments demonstrated by Chekmenev et al. in hyperpolarized TFPP [4] by directing the spin order to one of the resolved 1H sites. The selective 15N-‐choline transfer presented here also has potentially useful applications to thermally polarized in vivo 15N spectroscopy. Pulse sequences can be devised to transfer methyl 1H polarization to 15N for an encoding period and then back to methyl 1H for detection. Such sequences would benefit from choline’s long 15N T1 while making use of the strong thermal polarization of the nine methyl 1H spins. This strategy could be used to resolve species such as choline and phosphocholine by 15N-‐1H correlation spectroscopy, or even to study the metabolic interconversion of choline and related compounds using a sequence with a longitudinal 15N period for chemical exchange. 68 4.6 CONCLUSION As part of a strategy to improve the sensivity of in vivo experiments on hyperpolarized small biomolecules, we have investigated INEPT pulse sequences for the transfer of magnetization from 15N to the methyl 1H spins in 15N-‐choline. Product operator analysis of a simple refocused INEPT sequence shows that the optimized efficiency of this transfer is diminished by 62% by the action of competing couplings between 15N and choline’s methylene 1H spins. A selective INEPT sequence, using a 1H REBURP pulse to suppress undesired couplings during the transfer period, was devised and tested in vitro on a thermally polarized 15N-‐choline sample. It was demonstrated that the selective INEPT sequence leads to a 52% stronger methyl 1H signal than the nonselective version. 4.7 ACKNOWLEGEMENTS This work was done in collaboration with Jason E. Ollerenshaw, Valerie A. Norton, and Daniel P. Weitekamp. It was supported by the Beckman Institute pilot program, “Spin-‐ Polarized Molecules for Structural and Systems Biology.” JEO was supported by a Postdoctoral Fellowship from the Natural Sciences and Engineering Research Council of Canada. 69 4.8 REFERENCES [1] E.Y. Chekmenev, J. Hovener, V.A. Norton, K. Harris, L.S. Batchelder, P. Bhattacharya, B.D. Ross, and D.P. Weitekamp, PASADENA Hyperpolarization of Succinic Acid for MRI and NMR Spectroscopy. Journal of the American Chemical Society 130 (2008) 4212-‐4213. [2] K. Golman, Metabolic imaging by hyperpolarized 13C magnetic resonance imaging for in vivo tumor diagnosis. Cancer Research 66 (2006) 10855. [3] C. Gabellieri, S. Reynolds, A. Lavie, G.S. Payne, M.O. Leach, and T.R. Eykyn, Therapeutic target metabolism observed using hyperpolarized 15N choline. J. Am. Chem. Soc. 130 (2008) 4598-‐4599. [4] E.Y. Chekmenev, V.A. Norton, D.P. Weitekamp, and P. Bhattacharya, Hyperpolarized 1H NMR employing low g nucleus for spin polarization storage. J. Am. Chem. Soc. 131 (2009) 3164-‐3165. [5] R.F. Gareth A. Morris, Enhancement of Nuclear Magnetic Resonance Signals by Polarization Transfer. Journal of the American Chemical Society 101 (1979) 763-‐762. [6] Y. Boulanger, M. Labelle, and A. Khiat, Role of phospholipase A2 on the variations of the choline signal intensity observed by 1H magnetic resonance spectroscopy in brain diseases. Brain Res. Rev. 33 (2000) 380-‐389. [7] R. Katz-‐Brull, D. Seger, D. Rivenson-‐Segal, E. Rushkin, and H. Degani, Metabolic markers of breast cancer: enhanced choline metabolism and reduced choline-‐ether-‐phospholipid synthesis. Cancer Res. 62 (2002) 1966-‐1970. [8] E. Sutinen, M. Nurmi, A. Roivainen, M. Varpula, T. Tolvanen, P. Lehikoinen, and H. Minn, Kinetics of [11C]choline uptake in prostate cancer: a PET study. Eur. J. Nucl. Med. Mol. Imaging 31 (2004) 317-‐324. 70 [9] D.M. Doddrell, D.T. Pegg, W. Brooks, and M.R. Bendall, Enhancement of 29Si or 119Sn NMR signals in the compounds M(CH3)nCl4-‐n (M = Si or Sn, R = 4, 3, 2) using proton polarization transfer. Dependence of the enhancement on the number of scalar coupled protons. J. Am. Chem. Soc. 103 (1981) 727-‐728. [10] H. Geen, and R. Freeman, Band-‐selective radiofrequency pulses. J. Magn. Reson. 93 (1991) 93-‐141. [11] W.H. Press, S.A. Teukolsky, W.T. Vetterling, and B.P. Flannery, Numerical Recipes in C: The Art of Scientific Computing 2nd edition, Cambridge University Press, Cambridge, UK, 1992. [12] S.A. Smith, T.O. Levante, B.H. Meier, and R.R. Ernst, Computer simulations in magnetic resonance. An object-‐oriented programming approach. J. Magn. Reson. A106 (1994) 75-‐ 105. [13] V. Govindaraju, K. Young, and A.A. Maudsley, Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed. 13 (2000). 71 Appendix A Understanding PCC Agent Binding to Single Point Mutation E17K of Akt1 Pleckstrin Homology Domain through Molecular Dynamics 72 A.1 INTRODUCTION The Akt kinase is a critical molecular router that mediates cell growth, apoptosis, and translation [1], and Akt overexpression and/or hyperactivation has been observed in many cancer types [2]. The Akt1 activation mechanism includes binding to the PIP3 lipid on the cell membrane with a domain called the Pleckstrin Homology Domain (PHD). Mutations in the PHD of Akt1 that increase its affinity for binding with PIP3 will cause the upregulation of downstream pathways, thereby promoting the formation of cancerous cells [3]. Recent experiments performed on mice have revealed that a single amino acid point mutation to the PHD of Akt1, called the E17K mutation, was sufficient to cause cancer [4]. This same mutation has been found in certain human ovarian, colorectal and breast cancers. The E17K mutation exchanges a negatively charged glutamate for a positively charged lysine, causing the formation of an additional hydrogen bond between the PHD and PIP3 on the cell membrane. This induces a conformational change that causes Akt1(E17K) to bind to the plasma membrane four times stronger than it would in its wild-‐type form [4]. This strengthened binding is believed to be the underlying mechanism for the creation of a type of cancerous cells in mice. It is hypothesized that blocking the binding between PIP3 and Akt1(E17K) could reduce or terminate the growth of cancerous cells derived from the E17K mutation in humans, thus serving as a less invasive and less toxic means of chemotherapy. Previous work has shown that peptide capture agents can be raised against Akt that are epitope targeted and/or inhibitory [5,6,7]. Specifically, a 5mer peptide capture agent (yleaf) has 73 been developed to bind to the E17K mutation of the PH domain of Akt1 [7]. Molecular dynamics (MD) trajectories of the peptide/protein complex are constructed for the anchor with both the mutant and wild type (WT) Akt1 PH domains, and used to calculate the free energy of binding for each system. A.2 PEPTIDE CAPTURE AGENT AGAINST E17K MUTATION OF AKT1 PHD Work in the Heath group has yielded a 5mer anchor peptide that differentiates the E17K mutated Akt1 PHD from the wild type domain [7]. The epitope targeted one-‐ bead-‐one-‐compound screening method used to identify the ligand provides an anchor point for the C-‐terminus location of peptide binding. Additionally, biotin labeling experiments done using a variant of the peptide with an N-‐terminal tosyl-‐biotin labeling arm indicate a general region of N-‐terminal peptide interaction with the target around the residue Y26 [8]. These data were used in choosing the initial configuration of the yleaf-‐tosyl-‐biotin ligand in relation to either the E17K or WT target. A.3 CONSTRUCTION OF E17K AND WT SYSTEMS A.3.1 yleaf-‐tosyl-‐biotin The N-‐terminal tosyl-‐biotin labeling arm was constructed in ChemDraw, then imported into Maestro. The bond lengths were manually doubled, and then the B3LYP/6-‐31G** optimized structure and Mulliken charges were obtained using Jaguar. Unknown bond length, bond angle, and torsion parameters for the tosyl group were 74 optimized using B3LYP/6-‐31G** in Jaguar. The yleaf peptide was grown onto the minimized tosyl structure using (D) amino acids to form the complete ligand. yleaf-‐ tosyl-‐biotin was equilibrated in explicit water at 300 K for 1 ns to obtain the ligand conformation (Figure A.1) used to construct the ligand/protein complexes. A.3.2 Ligand/Protein Complexes Crystal structures for E17K (2UZR) and wild type (1UNP) Akt1 PH domains were obtained from the Protein Data Bank. ZDOCK was used to determine an initial relative configuration of the ligand shown in Figure A.1 to the E17K PHD. The anchor peptide was initially screened against the 32mer fragment corresponding to residues 1-‐32 of the E17K PHD, and ZDOCK predictions were restricted to only include relative conformations that allowed for interaction of the ligand with this fragment. The selected ligand configuration was also used for the wild type PH domain. Figure A.2 shows the top three ZDOCK output structures, and Figure A.3 shows the chosen structure for both E17K and WT systems. 75 Figure A.1. A. ChemBioDraw structure of yleaf-‐tosyl-‐biotin. B. Equilibrated structure of yleaf-‐tosyl-‐biotin used to construct WT and E17K complexes. 76 Figure A.2. Top three ZDOCK predicted conformations of yleaf-‐tosyl-‐biotin/E17K PH domain complex. Interaction was constrained to occur only with the first 32 residues of the protein (highlighted) that represent the fragment against which the peptide ligand was raised. The two amino acids of interest from experimental binding assays, E17K and Y26, are shown as sticks. 77 Figure A.3. Selected ligand conformation complexed with the mutant (blue) and wild type (purple) PH domains. This structure was chosen because the ligand most closely spans the two amino acids of interest (E17 or E17K, and Y26). 78 A.4 MOLECULAR DYNAMICS The systems were neutralized by adding Cl− or Na+ counterions as necessary, and were fully solvated in TIP3P water boxes. Each system was first subjected to a minimization of 10,000 steps. Then the solvent molecules, ligand, and binding fragment were equilibrated for 200 ps at 300 K while the remaining protein coordinates were fixed. The full systems were then minimized for 5,000 steps. Finally, both systems were equilibrated for a total of 5 ns at 300 K. The MD simulations were carried out with the NAMD 2.6 program [9]. Plots of the root mean squared deviation (rmsd) from the initial configuration of the trajectories and restarted trajectories from both systems are shown in Figure A.4. 79 Figure A.4. Rmsd plots for the MD trajectories and restarted trajectories of the E17K and WT complexes. 80 A.5 BINDING ENERGY To obtain a good comparison of free binding energy between the two systems, energy landscapes were constructed for E17K and WT by plotting the free energy of binding of conformations representing local macrostates vs. their rmsd from the lowest energy conformation, similar to the approach taken by [10]. If the reference structure is truly the lowest energy, the landscape will be smoothly funnel shaped. If there is an alternative lowest energy structure, the landscape will be non-‐funnel shaped. Figure A.5 shows examples of this type of landscape, where each dot represents one of 200,000–400,000 independent Rosetta ab initio structure prediction simulations [10]. For the purpose of this work, the rmsd of each trajectory is like a partition function, where each point represents a microstate. Plateaus in the rmsd represent macrostates, and given an infinite amount of simulation time, the system will spend a Boltzmann weighted percentage of time in each macrostate. Binding energies were calculated for a number of 100 ps plateaus in each trajectory using a Generalized Born implicit solvent model. After identifying the conformation with the lowest calculated binding energy REF, the energies of the other conformations were plotted against the rmsd of those conformations relative to REF. The energy vs. rmsd plot for the E17K and WT systems are shown in Figures A.6 and A.7, and the structure of the chosen reference structures are shown in Figure A.8. Significantly fewer data points were able to obtained for the E17K and WT systems than the ab initio structure prediction simulations, but the plots suggest the funnel shape described by [10], indicating that the correct lowest energy reference structure had been selected. 81 Figure A.5. Energy landscapes obtained from Rosetta ab initio structure prediction simulations on Rosetta@home. Red points represent the lowest-‐energy structures obtained in independent Monte Carlo structure prediction trajectories starting from an extended chain for each sequence; the y axis shows the Rosetta all-‐atom energy and the x axis shows the Cα root mean squared deviation from the design model. Green points represent the lowest-‐energy structures obtained in trajectories starting from the design model. The bottom figure shows a landscape that is funnel shaped, and the top shows one that is non funnel shaped. Adapted from [10]. 82 Figure A.6. Energy landscape for E17K complex. The y-‐axis is the Generalized Born calculation of binding energy, and the x-‐axis is the rmsd relative to the lowest energy structure. Each point is a plateau in the trajectory rmsd, representing a macrostate of the system. The binding energy of the reference state calculated without including entropy contribution is -‐67.95 kcal/mol. 83 Figure A.7. Energy landscape for WT complex. The y-‐axis is the Generalized Born calculation of binding energy, and the x-‐axis is the rmsd relative to the lowest energy structure. Each point is a plateau in the trajectory rmsd, representing a macrostate of the system. The binding energy of the reference state calculated without including entropy contribution is -‐65.28 kcal/mol. 84 Figure A.8. Structures of the E17K and WT reference states. There are distinct differences in the conformation of the peptide around the E17K mutation cite between the two systems. 85 A.6 CONCLUSION This appendix explored the use of MD calculations to understand the selectivity of the peptide ligand yleaf-‐tosyl-‐biotin for the E17K mutant of Akt1 PH domain over the wild type domain. The structure of the ligand was constructed and minimized, then complexed with both the E17K and wild type structures of Akt1 PHD. Trajectories were run, and the energies of macrostates of the system were calculated and plotted against their rmsd relative to a reference state to determine if the lowest energy state had been identified. The binding energy of the E17K reference state was found to be 2.67 kcal/mol lower than that of the WT reference state. This corresponds to a factor of ~90 difference in Kd, which is within the error of experimental measurements. A.7 ACKNOWLEDGEMENTS This work was done in collaboration with William Goddard III, Kaycie Deyle, and Blake Farrow. 86 A.8 REFERENCES 1. Vivanco I, Sawyers CL (2002) The phosphatidylinositol 3-‐kinase–AKT pathway in human cancer. Nature Reviews Cancer 2: 489-‐501. 2. Altomare DA, Testa JR (2005) Perturbations of the AKT signaling pathway in human cancer. Oncogene 24: 7455-‐7464. 3. Rebecchi M, Scarlata S (1998) Pleckstrin homology domains: a common fold with diverse functions. Annual review of biophysics and biomolecular structure 27: 503-‐528. 4. Carpten JD, Faber AL, Horn C, Donoho GP, Briggs SL, et al. (2007) A transforming mutation in the pleckstrin homology domain of AKT1 in cancer. Nature 448: 439-‐444. 5. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in Situ Click Chemistry Assembles a Branched Capture Agent and Allosteric Inhibitor for Akt1. Journal of the American Chemical Society 133: 18280-‐18288. 6. Nag A, Das S, Yu MB, Deyle KM, Millward SW, Heath JR (2013) A Chemical Epitope-‐Targeting Strategy for Protein Capture Agents: The Serine 474 Epitope of the Kinase Akt2. Angewandte Chemie International Edition In press. 7. Deyle K et al., Submitted. 8. Tamura T, Tsukiji S, Hamachi I (2012) Native FKBP12 Engineering by Ligand-‐Directed Tosyl Chemistry: Labeling Properties and Application to Photo-‐Cross-‐Linking of Protein Complexes in Vitro and in Living Cells. Journal of the American Chemical Society 134: 2216-‐2226. 9. Phillips JC, Braun R, Wang W, Gumbart J, Tajkhorshid E, et al. (2005) Scalable molecular dynamics with NAMD. Journal of computational chemistry 26: 1781-‐1802. 10. Koga N, Tatsumi-‐Koga R, Liu G, Xiao R, Acton TB, et al. (2012) Principles for designing ideal protein structures. Nature 491: 222-‐227. 87 Appendix B Peptide-‐Based General Antibody Detection Agent 88 B.1 INTRODUCTION Detecting the immune response to an infectious agent can provide useful in vitro diagnostic information. Assays of this nature require a general antibody detection agent, often a conjugated anti-‐Fc antibody or protein substrate, i.e., Protein A. However, most biological reagents have inherent variability, which can adversely influence the performance of the diagnostic test. Peptide capture agents have been shown to be highly robust and inexpensive to produce, and to have similar affinities and selectivities as antibodies. DeLano et al. reported the discovery of a cyclic 13-‐amino acid peptide (Fc-‐III) that bound the “consensus region” of IgG Fc by phage display affinity screening [1]. We explored possible improvements to the Fc-‐III peptide in terms of affinity and selectivity by employing in situ click screening to obtain a biligand, and in terms of stability by replacing the disulfide bridge of Fc-‐III, with the goal of using the modified peptide as a detection agent in diagnostic assays. B.2 WELLS PEPTIDE Using a cyclic phage display library of the form XiCXjCXk (where C is cysteine, X is a random amino acid, and i + j + k = 18), Delano et al. found the 13-‐residue Fc binding sequence Fc-‐III (DCAWHLGELVWCT) [1,2]. When synthesized, this peptide was found to inhibit the binding of Protein A Z-‐domain to Fc with a Ki of 25 nM, which is 200 fold greater affinity than an earlier, longer consensus sequence found by the group. The increased affinity with shorter length implied that the final sequence very efficiently 89 interacts with the Fc binding pocket. We added a propargyl glycine (Pra) residue at the N-‐terminus of Fc-‐III and tested the ability of this modified cyclic peptide to bind IgG Fc. The anchor peptide, monikered Wells peptide, was also C-‐terminally tagged with a polyethylene glycol (PEG) oligomer bridge and a biotin label. B.3 BILIGAND SCREEN The in situ click screen is illustrated in Scheme 2.1. The target IgG Fc is incubated with an excess of the selected anchor peptide and a large OBOC library at 4oC for 6 hrs (see Materials and Methods). The OBOC library is synthesized on TentaGel resin (Rapp Polymere, Tuebingen, Germany), and is a comprehensive library of 5-‐mers with a 6th amino acid at the C-‐terminus presenting an azide functionality. The OBOC library is comprised of non-‐natural (D) stereoisomers of the natural amino acids, excluding cysteine and methionine. The in situ screen is designed to identify a secondary (2o) peptide that, when coupled to the anchor, forms a biligand with increased selectivity and/or affinity for the target IgG Fc. In Step 1 of the screen, the OBOC library is cleared of beads that exhibit non-‐ specific binding to Wells peptide and alkaline phosphatase-‐conjugated streptavidin (SA-‐ AP), which is used as a detection reagent in the next step. Step 2 is called a product screen, and is unique to sequential in situ click screens [3]. This screen detects the presence of in situ clicked reaction products, which are those hit beads containing the triazole-‐linked anchor peptide. A complete list of screen hits is given in Table B.1. 90 Scheme B.1. Screening strategy for selecting capture agents against human IgG Fc. The flow chart represents the use of the cyclic Wells peptide as an anchor ligand for in situ click screening against a large OBOC azide-‐presenting peptide library. 91 Table B.1. Biligand screen results. List of pentapeptide “hits” from OBOC biligand screens performed with Wells peptide and human IgG Fc. The selected secondary ligands corresponding to WA and WG are shown in blue and red respectively (shown in Figure B.1). 92 B.4 1,4 TRIAZOLE LINKED BILIGAND CHARACTERIZATION The performance of these biligands was then characterized using an in-‐house developed an on-‐bead ELISA assay (data not shown). Sandwich ELISAs were used for affinity estimations of the two best performing biligands (Figure B.1) from the on-‐bead experiments. The affinity increase of the biligands over Wells is marginal. Hill fits to the ELISA data give Kd values of 18 nM for WA, 21 nM for WG, and 25 nM for Wells (Figure B.2). B.5 1,5 TRIAZOLE LINKED BILIGAND CHARACTERIZATION The in situ click process does not yield a stereospecific triazole product, and it is impossible to measure which click product variant the protein “catalyzed” during a screen. Therefore it can be useful to test the 1,5 triazole linked versions of potential biligands. Here, the biligand WA was synthesized using a 1,5 triazole small molecule linker and its performance compared to that of Wells by sandwich ELISA (Figures B.2 -‐ B.3). The difference in affinity between the 1,5 triazole linked biligand and Wells was negligible. 93 Figure B.1. Structures of peptide biligands identified in screen against IgG Fc. Acetylene-‐ presenting anchor peptides (black) were derived from Fc-‐III. Wells-‐ahriG (WA) and Wells-‐Gehni (WG) were evolved from the original peptide appended with Pra at the N-‐ terminus, and secondary ligand branches (colored) were identified from the in situ click screen of a 5-‐mer OBOC library presenting an azide functionality. 94 Figure B.2. Apparent affinity of Wells and biligands directed against Fc as determined by sandwich ELISA. Lines represent 1st order Hill fits to the affinity data. Wells: Kd = 25 nM, WA: Kd = 18 nM, WG: Kd = 21 nM. 95 Figure B.3. Structure of 1,5 triazole linked biligand. Biligand peptide WA was synthesized with a 1,5 triazole linker. 96 Figure B.4. Apparent affinity of Wells and 1,5 triazole linked WA directed against Fc as determined by sandwich ELISA. Lines represent 1st order Hill fits to the affinity data. 97 B.6 CLICK CYCLIZED WELLS To enhance the stability of the Wells anchor peptide, we attempted to replace the disulphide bridge with a Cu(I) catalyzed click triazole. (D) or (L) azidoalanine (Az1) was substituted for Cys2, and Pra was substituted for Cys12, then the peptide was cyclized using Cu(I) catalyzed click chemistry (Figure B.5). Single point ELISAs showed that this approach to cyclization removed all affinity to IgG Fc (Figure B.6). This could be due to the change in confirmation of the peptide cycle when constrained by an alternative cyclization moiety. The use of both (D) and (L) Az1 was an attempt to circumvent that, but was unsuccessful. 98 Figure B.5. Structure of Cu(I) click cyclized Wells peptide. The disulfide bond in Wells peptide was replaced with a Cu(I) catalyzed triazole formed from substitutions of Az1 and Pra for Cys2 and Cys12 respectively. 99 Figure B.6. Performance of click cylized Wells peptide variants compared to Wells peptide tested by sandwich ELISA. Fc (500 nM) was spiked into TBS, and captured antibody was detected by peroxidase-‐conjugated anti-‐human IgG antibody. 100 B.7 MATERIALS AND METHODS B.7.1 Anchor Synthesis The anchor Wells peptide was synthesized C → N on 200-‐300 mgs of Biotin NovaTag resin (EMD Biosciences, San Diego, CA) with a Titan peptide synthesizer using standard Fmoc chemistry [4]. The synthesis used Fmoc and side chain protected L amino acids (Aaptec, Louisville, KY), Fmoc protected propargylglycine (Pra) (Aaptec, Louisville, KY) and PEG5 (EMD Millipore, Germany). The finished peptide was side-‐chain deprotected and cleaved from resin in a mixture of 95% v/v trifluoroacetic acid (TFA) (Sigma-‐Aldrich, St. Louis, MO), 5% v/v dH2O, and 5% triethylsilane (TES) (Sigma-‐Aldrich, St. Louis, MO), and precipitated in diethyl either. Cu(Phen)3 stock solution was prepared by mixing 150 mM Cu(II) sulfate pentahydrate (Sigma-‐Aldrich, St. Louis, MO) and 500 mM 1,10-‐phenanthroline (Sigma-‐Aldrich, St. Louis, MO) in 4:1 H2O:Ethanol. The ether-‐ precipitated ligand was cyclized by stirring the cleaved peptides dissolved in ~5 mL acetonitrile (ACN) overnight with 1 mL of the Cu(Phen)3 stock solution. The cyclization mixture was lyophilized, and the final peptide product was purified using reverse phase (RP) HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column (Phenomenex, Torrance, CA). B.7.2 OBOC Screen The screen used naïve one-‐bead-‐one-‐compound (OBOC) D-‐pentapeptide libraries on TentaGel resin (TG) (Rapp Polymere, Tuebingen, Germany) of the form NH2-‐ 101 X1X2X3X4X5-‐Az4-‐TG, excluding cysteine and methionine, where Az4 is L-‐azidolycine. Libraries were synthesized on a Titan peptide synthesizer (Aaptec, Louisville, KY) using a split/mix method, and couplings were done using standard Fmoc SPPS chemistry in N-‐ methylpyrrolidone (NMP). Screening against Fc was carried out in two main steps: pre-‐clear, and product screen. Step 1 preclear: 200 mg of OBOC library was blocked in 5% w/v dry milk in Tris buffered saline (TBS) (25 mM Tris-‐HCl, 150 mM NaCl, 10 mM MgCl2, pH 7.6) for 2 hr at room temperature, and then incubated with 1:10,000 phosphatase-‐conjugated streptavidin (SA-‐AP, Promega, Madison, WI) and 25 µM Wells peptide in 0.5% w/v dry milk in TBS for 1 hr. The beads were washed in high salt buffer (25 mM Tris-‐HCl, 750 mM NaCl, 10 mM MgCl2, pH 7.6) for 1 hr, and then developed with BCIP/NBT (Promega, Madison, WI) in BCIP buffer (100 mM Tris-‐HCl, 150 mM NaCl, 1 mM MgCl2, pH 9). After 1 hour, the reaction was quenched with conc. HCl, and the purple beads were discarded. The remaining library was decolorized in NMP, dried with methanol (MeOH) and dichloromethane (DCM), and then swelled and blocked in 5% w/v dry milk in TBS for 2 hr at room temperature. Step 2 product screen: 4 mL of a solution containing 25 µM A21 and 23 nM Fc in 0.5% w/v dry milk in TBS was added to the precleared OBOC library, and the in situ click reaction was allowed to proceed for 6 hours at 4°C. Click solution was drained and the beads were washed for 1 hr in high salt buffer. The beads were incubated for 1 hr at room temperature with 1:10,000 SA-‐AP in 0.5% w/v dry milk in TBS, washed for 1 hr with high salt buffer, and then developed with BCIP/NBT in BCIP buffer for 1 hour. The reaction was quenched with conc. HCl, and the purple beads 102 were retained and sequenced using a Procise CLC Edman Degradation Protein Sequencing System (Applied Biosystems, Kingston, RI). B.7.3 Biligand Synthesis 1,4 Triazole linked WA and WG The secondary ligands including C-‐terminal Az4 for WA and WG were synthesized C → N on 200-‐300 mgs of Rink amide resin with a Titan peptide synthesizer using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected D amino acids and L-‐Az4 (Aaptec, Louisville, KY). The ligands were side-‐chain deprotected and cleaved from resin in 95:5:5 TFA:dH2O:TES, then precipitated in diethyl either and purified using RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. The purified secondary ligands were clicked to cyclized Wells peptide in solution using 2x molar excess of secondary, 10x Cu(I) iodide, 50x L-‐ascorbic acid, and 4x Tris[(1-‐benzyl-‐1H-‐1,2,3-‐ triazol-‐4-‐yl)methyl]amine (TBTA) (Sigma-‐Aldrich, St. Louis, MO) in 5:1 NMP:dH2O and stirring overnight at room temperature. The clicked biligand products were purified from the reaction mixture by RP-‐HPLC on a C18 semi prep column. 1,5 Triazole linked WA The anchor peptide was synthesized C → N omitting Pra on 200-‐300 mgs of Biotin NovaTag resin (EMD Biosciences, San Diego, CA) with a Titan peptide synthesizer using standard Fmoc chemistry [4]. A pre-‐clicked Fmoc protected 1,5 triazole Pra-‐Az4 small molecule linker (InDi Molecular) was amide coupled to the N-‐terminal Asp of Wells 103 peptide, which after Fmoc deprotection served as the starting point for the synthesis of the secondary ligand. The finished peptide was side-‐chain deprotected and cleaved from resin in a mixture of 95% v/v trifluoroacetic acid (TFA) (Sigma-‐Aldrich, St. Louis, MO), 5% v/v dH2O, and 5% triethylsilane (TES) (Sigma-‐Aldrich, St. Louis, MO), and precipitated in diethyl either. Cu(Phen)3 stock solution was prepared by mixing 150 mM Cu(II) sulfate pentahydrate (Sigma-‐Aldrich, St. Louis, MO) and 500 mM 1,10-‐ phenanthroline (Sigma-‐Aldrich, St. Louis, MO) in 4:1 H2O:Ethanol. The ether-‐ precipitated ligand was cyclized by stirring the cleaved peptides dissolved in ~5 mL acetonitrile (ACN) overnight with 1 mL of the Cu(Phen)3 stock solution. The cyclization mixtures were lyophilized, and the final peptide product was purified using reverse phase (RP) HPLC on a C18 prep column (Phenomenex, Torrance, CA). B.7.4 Click Cyclized Wells Synthesis The anchor Wells peptide was synthesized C → N, including N-‐terminal PEG and biotin, on 200-‐300 mgs of Rink amid resin (AnaSpec, Fremont, CA) with a Titan peptide synthesizer using standard Fmoc chemistry [4]. The synthesis used Fmoc and side chain protected L amino acids (Aaptec, Louisville, KY), Fmoc protected propargylglycine (Pra) (Aaptec, Louisville, KY), Fmoc protected (D) and (L) azidoalanine (Iris Biotech, Germany), PEG5 (EMD Millipore, Germany), and biotin (Sigma-‐Aldrich, St. Louis, MO). The finished peptide was side-‐chain deprotected and cleaved from resin in a mixture of 95% v/v trifluoroacetic acid (TFA) (Sigma-‐Aldrich, St. Louis, MO), 5% v/v dH2O, and 5% 104 triethylsilane (TES) (Sigma-‐Aldrich, St. Louis, MO), and precipitated in diethyl either. A Cu(I) click cyclization solution was prepared by mixing 10x Cu(I) iodide, 50x L-‐ascorbic acid, and 4x Tris[(1-‐benzyl-‐1H-‐1,2,3-‐triazol-‐4-‐yl)methyl]amine (TBTA) (Sigma-‐Aldrich, St. Louis, MO) in 5:1 NMP:dH2O. The ether-‐precipitated ligands were cyclized by stirring the cleaved peptides dissolved in ~2 mL Cu(I) click solution. The clicked biligand products were purified from the reaction mixture by RP-‐HPLC on a C18 semi prep column. B.7.5 Assays Biligand binding ELISA Biotinylated WA, WG, and Wells peptides were immobilized on streptavidin-‐ coated 96-‐well plates (Thermo Scientific Pierce, Rockford, IL) at a concentration of 400 nM in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the anchor peptides were incubated with varying concentrations of Fc in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with horseradish peroxidase (HRP)-‐conjugated mouse monoclonal antibody to human IgG Fc (Abcam, Cambridge, MA), diluted 1:10,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate (KPL, Gaithersburg, MD), then quenched with 1 M H2SO4 and read at 450 nm. 1,5 triazole linked biligand binding ELISA Biotinylated 1,5 triazole linked WA peptide was immobilized on streptavidin-‐ coated 96-‐well plates (Thermo Scientific Pierce, Rockford, IL) at a concentration of 300 105 nM in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the anchor peptides were incubated with varying concentrations of Fc in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with horseradish peroxidase (HRP)-‐conjugated mouse monoclonal antibody to human IgG Fc (Abcam, Cambridge, MA), diluted 1:10,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate (KPL, Gaithersburg, MD), then quenched with 1 M H2SO4 and read at 450 nm. Click cyclized Wells ELISA Biotinylated Cu(I) click cyclized Wells peptide was immobilized on streptavidin-‐ coated 96-‐well plates (Thermo Scientific Pierce, Rockford, IL) at a concentration of 1 µM in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the anchor peptides were incubated with 500 nM of Fc in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with horseradish peroxidase (HRP)-‐ conjugated mouse monoclonal antibody to human IgG Fc (Abcam, Cambridge, MA), diluted 1:10,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate (KPL, Gaithersburg, MD), then quenched with 1 M H2SO4 and read at 450 nm. 106 B.8 CONCLUSION Described in this appendix is an attempt to improve the previously discovered Fc binding peptide Fc-‐III in terms of affinity, selectivity, and stability. While a number of approaches were taken, the performance of the original Wells anchor peptide was not significantly improved. Preliminary immunoprecipitation data (not shown) indicated possible selectivity enhancement, but was not verified. The lack of improvement of Wells peptide with the addition of a secondary ligand may reflect the increased dwell time that such a strong binder has with its target during a screen; there are less significant constraints on the nature of the secondary ligand that clicks to the anchor in the in situ process due to the relatively constant presence of the anchor ligand at the target binding site. B.9 ACKNOWLEDGEMENTS This work was done in collaboration with Steven Millward, Kaycie M. Deyle, Errika Romero. 107 B.10 REFERENCES 1. DeLano WL, Ultsch MH, Wells JA (2000) Convergent solutions to binding at a protein-‐protein interface. Science 287: 1279-‐1283. 2. Lowman HB, Chen YM, Skelton NJ, Mortensen DL, Tomlinson EE, et al. (1998) Molecular mimics of insulin-‐like growth factor 1 (IGF-‐1) for inhibiting IGF-‐1: IGF-‐binding protein interactions. Biochemistry 37: 8870-‐8878. 3. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in situ click chemistry assembles a branched capture agent and allosteric inhibitor for Akt1. J Am Chem Soc 133: 18280-‐18288. 4. Fields GB, Noble RL (1990) Solid phase peptide synthesis utilizing 9-‐fluorenylmethoxycarbonyl amino acids. International journal of peptide and protein research 35: 161-‐214. 108 Appendix C Algorithm for Peptide Clustering 109 C.1 INTRODUCTION The one-‐bead-‐one-‐compound in situ click screening approach often yields a number of peptide “hits.” Rather than synthesizing and vetting each individual sequence, it can be useful to group the results into clusters of peptides based on physical properties, and then to choose a representative peptide from each cluster to scale up and test. This appendix provides an example clustering algorithm that uses the concept of persistence homology to group peptides by hydrophobicity, isoelectric point, and residue weight. The grouping is amino acid order dependent. C.2 PERSISTENCE CLUSTERING Persistence states that structures that persist at multiple scales are most likely to be real. For example, if an image is sampled by using a small number of points, persistence can be used to determine properties of the original shape. The points are grouped by drawing a disc of radius ε around each point. Each collection of overlapping discs is considered a cluster. These clusters provide one possible interpretation of the original data. To obtain a more complete analysis of the data, ε is continuously varied and the clusters tracked as the discs grow around the points. As discs from one cluster begin to overlap with discs from another cluster, the two initial clusters “die” and a new merged cluster is “born.” The lifetime of a cluster is the time between a cluster’s birth and death. Thus, groups of points that are far from other points will have a long lifetime, which indicates that they are more meaningful clusters. 110 Figure C.1. Discs formed around points with growing radius ε. As the discs around points overlap, new clusters form that include points rooting the newly touching discs. Eventually all the discs will touch, and the cluster containing all points will continue for infinite “time.” Adapted from [1]. 111 C.3 ALGORITHM DESCRIPTION The MatLab function BasicParam takes as input text files that contain a list of sequences and the number n of amino acids per sequence. It then uses hydrophobicity, isoelectric point, and residue weight to construct a point in 3n dimensional for each sequence. To prevent any of the physical parameters from overly contributing to the clustering based on their absolute amounts, they are all scaled by the inverse of their average. Once each sequence is represented by a point, the Euclidian pair wise distances between the points are computed. Initially, each point is its own cluster. Then, the two points with the smallest pair wise distance are merged. The time of death of the initial point clusters, and the time of birth of the new merged cluster, is set to the distance between those two points (see Figure B.2). Clusters containing points with the next least distance continue to iteratively merge until all points in the data set exist as a single cluster. Additionally, a covariance matrix was constructed and diagonalized to find the eigenvectors and eigenvalues that characterize the sequence space. C.4 OUTPUT The function outputs a list of clusters and their associated lifetimes, along with a plot of the input sequences projected onto the top two eigenvectors. The complete list of eigenvectors is also plotted against the associated eigenvalues, and the top four eigenvectors are decomposed into their individual components as charts. The output from clustering the hits obtained in the A22/4B3 screen described in Chapter 2 is shown in Table B.1 and Figures B.3 – B.5. 112 Figure C.2. Illustrative example of how lifetimes of clusters are computed. The lifetimes of clusters “a” and “b” are 3, because they are closer to each other than any other point and that is the distance between them. 3 then becomes the birth time for cluster “ab,” and the lifetime of “ab” is 5 because it dies at time 8, when the cluster “abc” is born. 113 Table C.1. List of clusters and their associated lifetimes calculated for the hits resulting from the A22/4B3 screen. The peptides input to the algorithm were qndwk, ltsry, diksp, eihny, qpidq, kidrv, qspwl, qsdwl, eqtfd, aapsl, kqdtq, kwtkf, tgnek, kysfq, ffftk, and vgeit. 114 Figure C.3. Projection of the peptide “points” onto the top two eigenvectors taken from the diagonalization of the covariance matrix. The clusters containing more than one point with the three longest lifetimes from Table B.1 are circled. 115 Figure C.4. Plot of eigenvectors vs. their associated eigenvalues. This allows users to know how many relevant eigenvectors describe the set of input peptides. 116 Figure C.5. Visual representation of the components of the top four eigenvectors. This shows what physical characteristics are dominant when describing the list of input peptides. The order of the amino acids is preserved during clustering, so Hydrophobicity1 refers to the hydrophobicity of the N-‐terminal amino acid of the peptides. For this set, the top four eigenvectors have large hydrophobicity components, though not all from the same amino acid index. 117 C.5 FUNCTION CODE function BasicParam(filename, pepLength) %Feb 8, 2013 %Physical parameters come from the wwHydrophobicity scale, Sigma PI values, %and resudue weight. %Note: hydrophobicities are at pH 7 except for proline, which was estimated %from an InDi (or AN SD) measurement %Note: the number of traits is hard coded in, so that has to be changed %every time you mess with the parameters. %http://www.sigmaaldrich.com/life-science/metabolomics/learningcenter/amin %o-acid-reference-chart.html Hydr = 1/((41+49+55+31+100+0+8+99+23+97+74+28+50+10+14+5+13+76+97+63)/20); PI = 1/((6+5.07+2.77+3.22+5.48+5.97+7.59+6.02+9.74+5.98+5.74+5.41+6.3+5.56+1 0.76+5.68+5.6+5.96+5.89+5.66)/20); RW = 1/((71.08+103.15+115.09+129.12+147.18+57.05+137.14+113.16+128.18+113.16 +131.2+114.11+97.12+128.13+156.19+87.07+101.11+99.13+186.22+163.18)/20) ; scaling = [Hydr PI RW]; %scaling = 1; %[Hydrophobicity pI ResidueWeight] A = [41 6.00 71.08].*scaling; C = [49 5.07 103.15].*scaling; D = [-55 2.77 115.09].*scaling; E = [-31 3.22 129.12].*scaling; F = [100 5.48 147.18].*scaling; G = [0 5.97 57.05].*scaling; H = [8 7.59 137.14].*scaling; I = [99 6.02 113.16].*scaling; K = [-23 9.74 128.18].*scaling; L = [97 5.98 113.16].*scaling; M = [74 5.74 131.20].*scaling; N = [-28 5.41 114.11].*scaling; P = [50 6.30 97.12].*scaling; Q = [-10 5.56 128.13].*scaling; R = [-14 10.76 156.19].*scaling; S = [-5 5.68 87.07].*scaling; T = [13 5.60 101.11].*scaling; V = [76 5.96 99.13].*scaling; W = [97 5.89 186.22].*scaling; Y = [63 5.66 163.18].*scaling; %Read in sequences from text file function argument, and determine the 118 %numberof sequences sequences = textread(filename,'%s', -1); nseq = length(sequences); %Initialize matrix to hold each sequence vector ntrait = 3; Z=zeros(nseq,pepLength*ntrait); %For each individual sequence assign a vector based on Grantham Values for j = 1:nseq seq = sequences{j}; o = []; for i=1:length(seq) switch seq(i) case 's' o = [o S]; case 'r' o = [o R]; case 'l' o = [o L]; case 'p' o = [o P]; case 't' o = [o T]; case 'a' o = [o A]; case 'v' o = [o V]; case 'g' o = [o G]; case 'i' o = [o I]; case 'f' o = [o F]; case 'y' o = [o Y]; case 'c' o = [o C]; case 'h' o = [o H]; case 'q' o = [o Q]; case 'n' o = [o N]; case 'k' o = [o K]; case 'd' o = [o D]; case 'e' o = [o E]; case 'm' o = [o M]; case 'w' o = [o W]; 119 end end Z(j,:) = o; end %Compute pairwise distances between each "sequence point" given by %a row in the Z matrix and put in array n. Also, create a 2x#pairs matrix %cataloging pair indices to go along with distances. n=[]; Idx=[]; k=1; for i = 1:nseq for j = i+1:nseq n(k) = norm(Z(i,:)-Z(j,:)); Idx(1,k) = i; Idx(2,k) = j; k=k+1; end end %Sort the array of distances from smallest to largest and use returned %permutation to analogously sort pair indices. [N,O] = sort(n); Idx = Idx(:,O); %Create array of parents of "sequence points" p = [1:nseq]; %Create an array of cluster birth tb = zeros(1,nseq); % %Union the points with distance smaller than 200 % for k = 1:length(N) % if(N(k)>= 200) % %sprintf('stopping at distance %f\n', N(k)) % break; % end % %sprintf('Unioning %f and %f with distance %f\n',Idx(1,k),Idx(2,k),N(k)) % p = Union(Idx(1,k),Idx(2,k),p); % end % %'Find' all points to see which ones share a root after Union and put into % %array 'root' % for i = 1:nseq % j = Find(i,p); % root(i) = j; % end % [R,O_r] = sort(root); % sequences(O_r); 120 %Itteratively 'Union' clusters containing the points with the next %least distance sets = {}; ages = []; birth = []; death = []; for i = 1:length(N) %Find roots of the two sets being joined r1 = Find(Idx(1,i),p); r2 = Find(Idx(2,i),p); %Set the current time to the distance between the two merging clusters t = N(i); %Check if the two sequences are already in the same cluster if r1 ~= r2 %Save the sequences in the two merging clusters %disp('====================================='); set1 = MkSet(r1,nseq,p,sequences); set2 = MkSet(r2,nseq,p,sequences); sets(end+1) = {set1}; sets(end+1) = {set2}; %Compute and save the ages of the merging clusters age1 = t - tb(r1); age2 = t - tb(r2); ages = [ages age1]; ages = [ages age2]; %Keep running tally of birth/death times for persistence barcode birth = [birth tb(r1)]; birth = [birth tb(r2)]; death = [birth t]; death = [birth t]; %Union the two clusters p = Union(Idx(1,i),Idx(2,i),p); %Save birth time for new cluster tb(r2) = t; end end birth; death; %Sort ages from largest to smallest, accordingly rearrainge sets of %sequence names 121 [Ages,O_A] = sort(ages, 'descend'); Sets = sets(O_A); %Read out vectors of ages and sets %disp('--------------------------------------------------------') %DispSet(sets, ages); % disp('--------------------------------------------------------') DispSet(Sets, Ages); disp('--------------------------------------------------------') %Plot "persistence barcode" to see if any clustering is actually present %Do PCA analysis to see if there exist any major axes along which the %sequence data falls. m = size(Z); %Center data center = zeros(1,m(2)); for i = 1:nseq center = center + Z(i,:); end center = center/nseq; Zc = zeros(m); for i = 1:nseq Zc(i,:) = Z(i,:) - center; end %Create covarience matrix covar = zeros(m(2), m(2)); CoVar = zeros(m(2), m(2)); for i = 1:nseq covar = Zc(i,:)'*Zc(i,:); CoVar = CoVar + covar; end %Diagonalize covarience matrix to find largest eigenvalues and %corresponding eigenvectors [vec,val] = eig(CoVar); CoVar*vec - vec*val; %Plot eigenvalues and display eigenvectors associated with two largest %values. Also label eigenvectors with next two largest eigenvalues to %decompose into pie charts to examine components. Diagval = diag(val); [Val,O_val] = sort(Diagval, 'descend'); Vec = vec(:,O_val); plot(Val, '.k', 'MarkerSize',10); eig1 = Vec(:,1); eig2 = Vec(:,2); eig3 = Vec(:,3); eig4 = Vec(:,4); %Project sequence points onto the two largest eigenvectors and plot Projection(Vec(:,1), Vec(:,2), Zc, sequences, nseq); 122 %Create pie chart to examine the components of the four eigenvectors with %the largest eigenvalues labels = {'Hydrophobicity_1', 'pI_1', 'RW_1','Hydrophobicity_2', 'pI_2',... 'RW_2', 'Hydrophobicity_3', 'pI_3', 'RW_3', 'Hydrophobicity_4', 'pI_4',... 'RW_4','Hydrophobicity_5', 'pI_5', 'RW_5'}; figure pie(abs(eig1), labels) title('Eigenvector 1') figure pie(abs(eig2), labels) title('Eigenvector 2') figure pie(abs(eig3), labels) title('Eigenvector 3') figure pie(abs(eig4), labels) title('Eigenvector 4') end %-----------------------------------------------------------------------% function y = Find(x,p) y = x; while(p(y)~=y) y = p(y); end end function p = Union(x1,x2,p) r1 = Find(x1,p); r2 = Find(x2,p); p(r1) = r2; end function Set = MkSet(r1,nseq,p,seq) Set = {}; for i = 1:nseq r = Find(i,p); if r == r1 Set(end+1) = seq(i); end end end function DispSet(Sets, Ages) 123 for i = 1:length(Sets) Si = Sets{i}; str = sprintf('%f\t',Ages(i)); for j=1:length(Si) str = sprintf('%s %s\t', str, char(Si{j})); end disp(str); end end function Projection(eig1, eig2, Zc, sequences, nseq) x = []; y = []; for i = 1:nseq x = [x Zc(i,:)*eig1]; y = [y Zc(i,:)*eig2]; end figure; plot(x, y, '*') dx = .2; dy = 0.2; text(x+dx, y+dy, sequences); end C.6 CONCLUSION This appendix describes a clustering algorithm designed to help characterize groups of peptide hits from in situ click screens. The objective is to provide guidance about which peptides are useful to scale up and assay, and give information about the physical characteristics that predominately describe the differences between input peptides. C.7 ACKNOWLEGDEMENTS I would like to thank Keenan Crane for introducing me to the concept of persistent homology, and for his incredibly helpful guidance while developing this algorithm. 124 C.8 REFERENCES 1. Ghrist R (2008) Barcodes: the persistent topology of data. Bulletin of the American Mathematical Society 45: 61-‐75. 125 Appendix D A cocktail of thermally stable, chemically synthesized capture agents for the efficient detection of anti-‐gp41 antibodies from human sera Reproduced from PloS one 2013, 8, e76224 126 127 128 129 130 131 Appendix E A Selective 15N-‐to-‐1H Polarization Transfer Sequence for More Sensitive Detection of 15N-‐Choline Reproduced from Journal of Magnetic Resonance 2010, 205, 125-‐129 132 133 134 135 136 A COCKTAIL OF THERMALLY STABLE, CHEMICALLY SYNTHESIZED CAPTURE AGENTS FOR THE EFFICIENT DETECTION OF ANTI-‐GP41 ANTIBODIES FROM HUMAN SERA AND TECHNIQUES Thesis by Jessica A. Pfeilsticker In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy California Institute of Technology Pasadena, California 2014 (Defended December 2, 2013) ii © 2014 Jessica A. Pfeilsticker All Rights Reserved iii ACKNOWLEDGEMENTS I begin by thanking my advisor Professor James Heath for his incredible guidance and talent. His enthusiasm for science is unrivaled, and his uncanny ability to absorb ideas and tie them together in creative and interesting ways has been a wonderful model for how to conduct research. I have enjoyed my time in his lab immensely, and am grateful to have learned from him for the last three and a half years. I also want to thank Professor Daniel Weitekamp, from whom I learned many fundamental scientific principles. He helped me transition as a young graduate student from someone who takes classes and obtains grades to someone who thinks about and conducts research. For that, and for sharing with me some of his incredible ideas and vast intellect during my first three years at Caltech, I am thankful. I want thank my committee members Professor Robert Grubbs, Professor Mitchio Okumura, and Professor William Clemons for their guidance and insight. More than once I sought their advice on how to navigate graduate school, and always received encouragement and enthusiasm in return. Next, I thank my family. They originally inspired an interest in the natural world and the wish to explore it when I was a child, and then supported me as I pursued my education in science. To Dad and Penny, Mom and Ed, thank you for being my parents and offering love and advice. John and Jason, thanks for being the coolest brothers anyone could ask for. You both have incredible futures in front of you, and I know that iv your deep thoughtfulness and curiosity will ensure a never ending supply of fun and interest as you pursue your own paths. I want to thank my lab mates for their instrumental role in my graduate experience. Valerie Norton, Jason Ollerenshaw, Eduard Chekmenev, Jan Höevener, and Daniel Rowlands in the Weitekamp group; Aiko Umeda, Kaycie Deyle, Blake Farrow, Ryan Henning, Alex Sutherland, Joey Varghese, JingXin Liang, Arundhati Nag, Samir Das, Marilena Dimotsantou, Ann Chung, Steve Millward, Errika Romero, and Connie Hsueh in the Heath group; and Heather Agnew, Bert Lai, Rosemary Rohde, and Suresh Pitram at InDi Molecular. You all kept me sane, and offered much needed scientific and personal feedback and assistance during our time as friends and colleagues. Thank you for being my biggest allies; this experience was made so much richer because you all were in it. I also need to thank the many members of Caltech staff who make research possible. Agnes Tong, thank you for all that you do to keep chemistry running smoothly. David VanderVelde, thank you for your enthusiasm and interest in sharing your knowledge of NMR. Mona Shahgholi, Jie Zhou, Felicia Rusnak, Darryl Willick, Joe Drew, Cora Carriedo, Steve Gould, Paul Carroad, Art Seiden, Mike Roy, Richard Gerhart, Jose Mendez, Engracia Alvarez, and countless others, thank you all for making the department exist and function. All of you have personally touched my time here in some way, and I wouldn’t have been able to do this without you. Finally, it is imperative that I thank the many friends who make up my support network. Lisa Sherer, thank you for being my pseudo-‐sister for the last 13 years. Caitlin v Murphy, Lisa Mauger, Tucker Jones, Nick Stadie, Luke Boosey, and Patrick Sanan, thank you for being an awesome group of people who bonded so quickly and tightly when we all first arrived at Caltech as wide-‐eyed first years. Keenan Crane, thank you for being so thoughtful and kind, and for helping me through all manners of crisis, both personal and professional. Cole DeForest, Robin DeForest, Anya Demianenko, and Michael McCoy, thank you for kindness, compassion, wine and foie gras. Young In Oh and Jean Li, thank you for mutual support in the face of adversities. Sang Tae Park, thank you for unconditionally having my back and being someone I could always rely on for personal and technical advice. Maggie Buckles, Tony Roy, and Kellie Roy, thank you for being smart, kind, wonderful people who gave me balance and kept me sane. Jon-‐Michael Sanchez, thank you for keeping life exciting, and for becoming someone with whom I could explore new ways to express myself. I love all of you, and thank you so much for being in my life. vi ABSTRACT This thesis reports on a method to improve in vitro diagnostic assays that detect immune response, with specific application to HIV-‐1. The inherent polyclonal diversity of the humoral immune response was addressed by using sequential in situ click chemistry to develop a cocktail of peptide-‐based capture agents, the components of which were raised against different, representative anti-‐HIV antibodies that bind to a conserved epitope of the HIV-‐1 envelope protein gp41. The cocktail was used to detect anti-‐HIV-‐1 antibodies from a panel of sera collected from HIV-‐positive patients, with improved signal-‐to-‐noise ratio relative to the gold standard commercial recombinant protein antigen. The capture agents were stable when stored as a powder for two months at temperatures close to 60oC. vii TABLE OF CONTENTS LIST OF FIGURES, SCHEMES, AND TABLES………………………………………………………….……xiii CHAPTER 1: Introduction…………………………………………………………………………………….….1 1.1 PCC Agent Cocktail for HIV-‐1 Diagnostics……………………………………………….2 1.2 References……………………………………………………………………………………………..5 CHAPTER2: Developing a PCC Agent Cocktail for the Detection of Anti-‐HIV Antibodies…………………………………………………………………………………………………………….…7 2.1 Introduction……………………………………………………………………………………………8 2.2 In Situ Click……………………………………………………………………………………………..9 2.3 Anchor Selection…………………………………………………………………………………….10 2.4 Screening……………………………………………………………………………………………….11 2.5 Biligand Characterization…………………………………………………………………….….16 2.6 Materials and Methods…………………………………………………………………………..21 2.6.1 Anchor Synthesis……………………………………………………………………..21 2.6.2 OBOC Screens………………………………………………………………………....22 viii 2.6.3 Biligand Synthesis……………………………………………………………………24 2.6.4 Assays……………………………………………………………………………………..27 2.7 Conclusion……………………………………………………………………………………………..29 2.8 Acknowledgements………………………………………………………………………………..29 2.9 References………………………………………………………………………………………….….31 CHAPTER 3: PCC Agent Based Assay for the Detection of Anti-‐HIV Antibodies from Human Sera……………………………………………………………………………………………………….……33 3.1 Introduction……………………………………………………………………………………………34 3.2 Patient Sample Assay……………………………………………………………………………..34 3.3 PATH Sample Assay………………………………………………………………………………..37 3.4 Stability Assay…………………………………………………………………………………….….39 3.5 Materials and Methods……………………….…………………………………………………41 3.5.1 Patient Serum ELISA………………………………………………………………..41 3.5.2 PATH Samples ELISA………………………………………………………………..41 3.5.3 Stability Assay………………………………………………………………………….42 3.6 Conclusion……………………………………………………………………………………………..43 ix 3.7 Acknowledgements………………………………………………………………………………..44 3.8 References……………………………………………………………………….…………………….45 CHAPTER 4: A Selective 15N-‐to-‐1H Polarization Transfer Sequence for More Sensitive Detection of 15N-‐Choline…………………………………………………………………….…………...…….46 4.1 Abstract…….……………………………………………………………………………………...…..47 4.2 Introduction……………………………………………………………………………………...…..48 4.3 Theory…………………………………………………………………………………………..…...….49 4.4 Experimental………………………………………………………………….………………….…..56 4.5 Results and Discussion……………………………………………………………………….…..61 4.6 Conclusion………………………………………………………………………………………….….68 4.7 Acknowledgements…………………………………………………………………………….….68 4.8 References…………………………………………………………………………………………....69 APPENDIX A: Understanding PCC Agent Binding to Single Point Mutation E17K of Akt1 Pleckstrin Homology Domain through Molecular Dynamics…………………..………………71 A.1 Introduction…………………………………………………………………………………………..72 x A.2 Peptide Capture Agent Against E17K Mutation of Akt1 PHD…………….……73 A.3 Construction of E17K and WT Systems………………………………….……………….73 A.3.1 yleaf-‐tosyl-‐biotin……………………………………………………………………..73 A.3.2 Ligand/Protein Complexes………………………………………………………74 A.4 Molecular Dynamics………………………………………………………………………………78 A.5 Binding Energy……………………………………………………………………………………….80 A.6 Conclusion……………………………………………………………………………………………..85 A.7 Acknowledgements………………………………………………………………………………..85 A.8 References…………………………………………………………………………………………….86 APPENDIX B: Peptide-‐Based General Antibody Detection Agent…………………..…….…87 B.1 Introduction……………………………………………………………………………………………88 B.2 Wells Peptide…………………………………………………………………………………………88 B.3 Biligand Screen………………………………………………………………………………………89 B.4 1,4 Triazole Linked Biligand Characterization…………………………………………92 B.5 1,5 Triazole Linked Biligand Characterization…………………………………………92 B.6 Click Cyclized Wells………………………………………………………………………………..97 xi B.7 Materials and Methods………………………………………………………………………….100 B.7.1 Anchor Synthesis………………………………………………………..…………..100 B.7.2 OBOC screen…………………………………………………………………………..100 B.7.3 Biligand Synthesis……………………………………………………………………102 B.7.4 Click Cyclized Wells Synthesis………………………………………………….103 B.7.5 Assays……………………………………………………………………………………..104 B.8 Conclusion……………………………………………………………………………………………..106 B.9 Acknowledgements………………………………………………………………………………..106 B.10 References…………………………………………………………………………………………..107 APPENDIX C: Algorithm for Peptide Clustering…………………………………………….………..108 C.1 Introduction…………………………………………………………………………………………..109 C.2 Persistence Clustering……………………………………………………………………………109 C.3 Algorithm Description……………………………………………………………………………111 C.4 Output……………………………………………………………………………………………………111 C.5 Function Code………………………………………………………………………………………..117 C.6 Conclusion……………………………………………………………………………………………..123 xii C.7 Acknowledgements………………………………………………………………………………..123 C.8 References…………………………………………………………………………………………….124 APPENDIX D: A cocktail of thermally stable, chemically synthesized capture agents for the efficient detection of anti-‐gp41 antibodies from human sera (PloS one 2013, 8, e76224)……………………………………………………………………………………………………………………125 APPENDIX E: A Selective 15N-‐to-‐1H Polarization Transfer Sequence for More Sensitive Detection of 15N-‐Choline (Journal of Magnetic Resonance 2010, 205, 125-‐129)……..131 xiii LIST OF FIGURES, SCHEMES, AND TABLES FIGURES, Chapter 2: Figure 2.1. Differential detection of 3D6 and 4B3 by anchor ligands……………………….12 Figure 2.2. Structures of peptide ligands in PCC Agent cocktail……………………………….17 Figure 2.3. Apparent affinity of A21 and biligands directed against 3D6 as determined by SPR………………………………………………………………………………………….………18 Figure 2.4. Apparent affinity of A22 and biligand directed against 4B3 as determined by SPR……………………………………………………………………………….………….……..19 Figure 2.5. Performance of PCC agent cocktail to detect 3D6 and 4B3 from human serum…………………………………………………………………………………………………….……20 FIGURES, Chapter 3: Figure 3.1. Patient sample ELISA……………………………………………………………………………..36 Figure 3.2. PATH sample ELISA………………………………………………………………………………..38 Figure 3.3. Stability assay………………………………………………………………………………………..40 xiv FIGURES, Chapter 4: Figure 4.1. A refocused INEPT pulse sequence for polarization transfer from 15N to 1H for detection………………………………….............................................................51 Figure 4.2. The structure of 15N-‐choline…………………………………………………………………..52 Figure 4.3. Selective refocused INEPT pulse sequence for coherent polarization transfer from 15N to methyl 1H in 15N-‐choline………………………………………………………….57 Figure 4.4. Numerical simulation of the selective 15N-‐to-‐1H INEPT pulse sequence acting on a simplified spin system……………………………………………………………60 Figure 4.5. Comparison of 1H spectrum using selective 15N-‐to-‐1H INEPT sequence to 15N spectrum……………………..…………………………………………………………………62 Figure 4.6. Measurement of the 15N longitudinal relaxation time of 15N-‐choline in D2O solution using a modified version of selective 15N-‐to-‐1H INEPT sequence……..63 Figure 4.7. Amplitude of the methyl 1H signal observed in 15N-‐to-‐1H INEPT spectra of 15N-‐choline using selective and non selective pulses…………………...............65 FIGURES, Appendix A: Figure A.1. Structure of yleaf-‐tosyl-‐biotin………………………………………………………………..75 xv Figure A.2. Top three ZDOCK predicted conformations……………………………………………76 Figure A.3. Selected ligand conformation complexed with the mutant and wild type PH domains……………………………………………………………………………………………..77 Figure A.4. Rmsd plots for the MD trajectories and restarted trajectories of the E17K and WT complexes…………………………………………………………………………………………79 Figure A.5. Energy landscapes obtained from Rosetta ab initio structure prediction simulations on Rosetta@home..................................................................81 Figure A.6. Energy landscape for E17K complex………………………………………………………82 Figure A.7. Energy landscape for WT complex…………………………………………………………83 Figure A.8. Structures of the E17K and WT reference states……………………………………84 FIGURES, Appendix B: Figure B.1. Structures of peptide biligands identified in screen against IgG Fc………..93 Figure B.2. Apparent affinity of Wells and biligands directed against Fc as determined by sandwich ELISA………………………………………………………………………………..94 Figure B.3. Structure of 1,5 triazole linked biligand………………………………………………….95 xvi Figure B.4. Apparent affinity of Wells and 1,5 triazole linked WA directed against Fc as determined by sandwich ELISA…………………………………………………………..96 Figure B.5. Structure of Cu(I) click cyclized Wells peptide………………………………………..98 Figure B.6. Performance of click cylized Wells peptide variants compared to Wells peptide tested by sandwich ELISA………………………………………………………………….99 FIGURES, Appendix C: Figure C.1. Discs formed around points with growing radius ε………………………………..110 Figure C.2. Illustrative example of how lifetimes of clusters are computed……………..112 Figure C.3. Projection of the peptides onto the top two eigenvectors taken from the diagonalization of the covariance matrix. ………………………………………114 Figure C.4. Plot of eigenvectors vs. their associated eigenvalues…………………………….115 Figure C.5. Visual representation of the components of the top four eigenvectors…………………………………………………………………………………………………………...116 xvii SCHEMES, Chapter 2: Scheme 2.1. Screening strategy for selecting capture agents against anti-‐HIV antibodies 3D6 and 4B3……………………………………………………………………………14 SCHEMES, Appendix B: Scheme B.1. Screening strategy for selecting capture agents against human IgG Fc…………………………………………………………………………………………………………..90 TABLES, Chapter 2: Table 2.1. Biligand screen results…………………………………………………………………………….15 TABLES, Appendix B: Table B.1. Biligand screen results…………………………………………………………………………….91 TABLES, Appendix C: Table C.1. List of clusters and their associated lifetimes calculated for the hits resulting from the A22/4B3 screen………………………………………………………………………….113 1 Chapter 1 Introduction 2 1.1 PCC AGENT COCKTAIL FOR HIV-‐1 DIAGNOSTICS Accurate diagnoses are critical for effective healthcare. Misdiagnoses carry large personal and societal costs, especially in cases where the disease is transmissible or carries a large social stigma. HIV is one such disease, and HIV misdiagnoses can lead to psychosocial difficulties and have dangerous public health implications [1]. According to the World Health Organization (WHO), roughly 40 million people were living with HIV as of the end of 2004, and some regions have adult prevalence as high as 15-‐39%. Many countries with high incidence of HIV are resource limited and rely on easy-‐to-‐use rapid detection tests for diagnosis; however, the performance of these tests is variable. A study of the WHO HIV two-‐test diagnostic algorithm, where two positive rapid detection tests are used alone for diagnosis, showed an unacceptably high rate of false-‐positive results (10.5%), likely due to non-‐selective cross reactivity [2]. HIV diagnoses are often made indirectly through the detection of anti-‐HIV antibodies [3]. The current gold standard detection reagents are biologically-‐produced, and are expensive, susceptible to thermal and biological degradation, and have batch to batch performance variability; all of which can adversely affect the accuracy of diagnostic assays [4,5,6]. Typically in HIV diagnostic assays, multiple immunogenic and conserved antigens from HIV are expressed as regions of a single chimeric protein. That chimeric protein is then used to capture specific antibodies from the body fluid (e.g. blood, saliva or urine) of potentially infected patients; a positive assay result implies infection. However, the antigen has been shown to have poor signal to noise [7], 3 possibly related to the off-‐target reactivity responsible for false negative results seen in rapid detection tests. Robust detection reagents are needed that recognize relevant biomarkers with high sensitivity and selectivity, and that perform well in resource limited areas where controlled laboratory environments are not readily available. Protein Catalyzed Capture agents (PCC agents) are peptide-‐based detection agents that rival the affinity and selectivity of biological reagents, and that are chemically, physically, and biologically robust. As synthetic reagents, they are also highly modular and can be easily altered for incorporation into different assay platforms. Previous work has identified a number of PCC agents that detect multiple disease relevant biomarker proteins through the process of iterative in situ click screening [8,9,10], many of which were engineered to be specifically epitope targeted, and some of which have inhibiting or activating functionalities. The application of PCC agent technology to anti-‐HIV antibody detection is not as straightforward as targeting a single biomarker; the polyclonal diversity of antibodies across patient populations can translate into large variations in assay performance from patient to patient. To selectively capture the diversity of anti-‐HIV antibodies while minimizing serological cross-‐reactivity, a cocktail of PCC agents was developed against multiple representative anti-‐HIV antibodies. Chapter 2 reports on the use of iterative in situ click chemistry [8,11] to prepare a cocktail of chemically synthesized PCC agents that is designed to sample the polyclonal diversity of an antibody-‐based immune response. Three PCC agents, derived from the 4 same epitope on HIV-‐1 envelope protein gp41, were raised against two representative anti-‐HIV antibodies to create a cocktail of capture agents that sample the variations between the representative antibodies, thus increasing the reagent’s potential to detect an anti-‐HIV immune response across multiple patients. Chapter 3 then describes a PCC agent-‐based assay developed to detect human antibodies that bind a conserved region of the HIV-‐1 envelope glycoprotein gp41, using the PCC Agent cocktail described in Chapter 2. The performance is compared against the gold standard chimeric protein antigen using sera collected from a cohort of HIV-‐1-‐ positive human subjects are tested, plus controls. The assay was also used to detect anti-‐HIV antibodies from a validation panel of HIV-‐1 specific IgG containing plasma samples that were rated HIGH, MED, LOW, or NEG in accordance with the titer of anti-‐ HIV antibodies. Additionally, the thermal stability of the capture agent cocktail is reported, with an eye towards point-‐of-‐care HIV diagnostics assays that are needed in environments where refrigeration chains may not exist. Chapters 2 and 3 were taken in part from PloS one 2013, 8, e76224 (see Appendix D). 5 1.2 REFERENCES 1. Bhattacharya R, Barton S, Catalan J (2008) When good news is bad news: psychological impact of false positive diagnosis of HIV. AIDS care 20: 560-‐564. 2. Klarkowski DB, Wazome JM, Lokuge KM, Shanks L, Mills CF, et al. (2009) The evaluation of a rapid in situ HIV confirmation test in a programme with a high failure rate of the WHO HIV two-‐test diagnostic algorithm. PloS one 4: e4351. 3. Fanales-‐Belasio E, Raimondo M, Suligoi B, Butto S (2010) HIV virology and pathogenetic mechanisms of infection: a brief overview. Ann Ist Super Sanita 46: 5-‐14. 4. Daskalakis D (2011) HIV diagnostic testing: evolving technology and testing strategies. Top Antivir Med 19: 18-‐22. 5. Ginocchio CC, Carman WF, Arens MQ (2011) Update on HIV diagnostic testing algorithms. Editors' introduction. J Clin Virol 52 Suppl 1: S1. 6. Alvarez M, Chueca N, Guillot V, Bernal Mdel C, Garcia F (2012) Improving Clinical Laboratory Efficiency: Introduction of Systems for the Diagnosis and Monitoring of HIV Infection. Open Virol J 6: 135-‐143. 7. Tiwari RP, Jain A, Khan Z, Kumar P, Bhrigu V, et al. (2012) Designing of novel antigenic peptides cocktail for the detection of antibodies to HIV-‐1/2 by ELISA. Journal of immunological methods. 8. Agnew HD, Rohde RD, Millward SW, Nag A, Yeo WS, et al. (2009) Iterative in situ click chemistry creates antibody-‐like protein-‐capture agents. Angew Chem Int Ed Engl 48: 4944-‐4948. 9. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in Situ Click Chemistry Assembles a Branched Capture Agent and Allosteric Inhibitor for Akt1. Journal of the American Chemical Society 133: 18280-‐18288. 6 10. Arundhati Nag SD, Mary Beth Yu, Kaycie M. Deyle, Steven W. Millward, James R. Heath (2013) A Chemical Epitope-‐Targeting Strategy for Protein Capture Agents: The Serine 474 Epitope of the Kinase Akt2. Angewandte Chemie International Edition In press. 11. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in situ click chemistry assembles a branched capture agent and allosteric inhibitor for Akt1. J Am Chem Soc 133: 18280-‐18288. 7 Chapter 2 Developing a PCC Agent Cocktail for the Detection of Anti-‐HIV Antibodies 8 2.1 INTRODUCTION Detecting the immune response to an infectious agent can provide a useful in vitro diagnostic surrogate relative to direct pathogen detection [1]. Such assays are commonly used for detecting HIV infection because of its characteristic immunopathology [2]. Direct detection of HIV viral RNA and p24 antigen is only effective at an early stage of infection, approximately 2-‐6 weeks of initial exposure [3,4]. Antibodies against HIV envelope proteins emerge in patients’ blood around 3-‐4 weeks of infection [2,5] as the viral RNA and p24 levels decline as a result of immunocomplex formation [6]. The high serum level of anti-‐HIV IgG is maintained throughout the course of clinical latency (2-‐20+ years), during which time viral antigens are under detection limits until the onset of acquired immunodeficiency syndrome (AIDS) [2,5]. Typically in these assays, immunogenic and conserved antigens from the HIV are expressed as regions of a single chimeric protein. That chimeric protein is then used to capture specific antibodies from the body fluid (e.g. blood, saliva or urine) of potentially infected patients; a positive assay result implies infection. However, the polyclonal diversity of antibodies across a patient population can translate into large variations in assay performance from patient to patient. In addition, the chimeric recombinant proteins are biological reagents, and so may have limitations related to shelf life and batch-‐to-‐ batch variability. These limitations can adversely influence the performance of a diagnostic test [3,5,7], especially one that is deployed in harsh physical environments. 9 Previous work in the Heath group has demonstrated the ability to selectively target biologically relevant proteins with PCC Agents [8,9]. Targeting anti-‐HIV antibodies requires selective detection of anti-‐HIV antibodies over other IgG and serum proteins while maintaining sensitivity to variations in immune response across patients. This chapter reports on the discovery, synthesis, and characterization of a cocktail of capture agents that selectively target different anti-‐HIV antibodies to increase the target capture probability while minimizing cross reactivity with other proteins in sera. 2.2 IN SITU CLICK The development of a PCC Agent against a protein target utilizes the target itself to promote the 1,3-‐dipolar cycloaddition between an acetylene and an azide group to form a triazole linkage (the in situ ‘click’ reaction) [10]. The protein effectively plays the role of an extremely selective, but much less efficient, variant of the Cu(I) catalyst that is commonly used for such couplings [11,12]. For the present work, the two reacting species are peptides – one peptide (the anchor) is a chemically modified variant of a conserved, immunogenic epitope on the HIV-‐1 gp41 protein, and the second peptide is selected via an in situ click screen from a large (106 element) one-‐bead-‐one-‐compound (OBOC) [13] peptide library. The protein targets are human monoclonal antibodies raised against variants of the gp41 epitope represented by the anchor peptide. 10 2.3 ANCHOR SELECTION The PCC Agents developed here were designed to capture antibodies that are selective for residues 600-‐612 (IWCGSGKLICTTA) of gp41. Previous studies have shown that a large fraction of HIV-‐1-‐positive patients develop antibodies against this epitope [14,15]. Our strategy for sampling the polyclonal diversity of such antibodies was to develop PCC Agents that exhibited both differential, as well as similar avidities for human monoclonal antibodies (mAbs) raised against different parts of this epitope. To select the anchor peptides, we modified the polypeptide fragment corresponding to residues 600-‐612 of gp41 with artificial amino acids at multiple locations, and tested the ability of these modified cyclic peptides to detect two different monoclonal anti-‐gp41 antibodies 3D6 and 4B3 (Polymun, Klosterneuburg, Austria). The 3D6 mAb was raised against the epitope SGKLIC, whereas the 4B3 mAb was raised against SGKLICTTA. One anchor peptide, A21, was synthesized by adding a propargyl glycine (Pra) residue at the C-‐terminus of the residues 600-‐612 of gp41. This anchor peptide was also N-‐terminally tagged with a polyethylene glycol (PEG) oligomer bridge and a biotin label. For a second anchor peptide (A22), Leu-‐607 was substituted with Pra. A22 also included an N-‐ terminal PEG-‐biotin label. A21 equally detected 3D6 and 4B3 with an estimated dissociation constant (Kd) of 1-‐50 nM, while A22 differentially detected 3D6 (Kd >10 µM) and 4B3 (Kd = 1-‐50 nM) (Figure 2.1). A21 and A22 were then separately developed into PCC agents against 3D6 and 4B3, respectively. 11 2.4 SCREENING The in situ click screens are illustrated in Scheme 2.1. The target IgG is incubated with an excess of the selected anchor peptide and a large OBOC library at 4oC overnight (see Materials and Methods). The OBOC library is synthesized on TentaGel resin (Rapp Polymere, Tuebingen, Germany), and is a comprehensive library of 5-‐mers with a 6th amino acid at the N-‐terminus presenting an azide functionality. To help ensure chemical and biochemical stability, the OBOC library is comprised of non-‐natural (D) stereoisomers of the 20 natural amino acids, excluding cysteine and methionine. The in situ screen is designed to identify a secondary (2o) peptide that, when coupled to the anchor, forms a biligand with increased selectivity and/or affinity for the target IgG. 12 Figure 2.1. Differential detection of 3D6 and 4B3 by anchor ligands. Relative affinities of A21 and A22 for 3D6 and 4B3 were determined by sandwich ELISA. Biotinylated anchor ligands A21 and A22 were immobilized on streptavidin-‐coated 96-‐well plated at a concentration of 100 nM, and incubated with the solutions of target anti-‐HIV antibodies 3D6 and 4B3 at 100 nM in TBS. Captured antibody was detected by peroxidase-‐ conjugated anti-‐human IgG antibody. 13 The screen proceeds stepwise. In the first step (not shown in Scheme 2.1) the OBOC library is cleared of beads that exhibit non-‐specific binding to alkaline phosphatase-‐conjugated streptavidin (SA-‐AP), which is used as a detection reagent in a later step. Step 2 is a target screen, and so is designed to detect the presence of the bound IgG target to specific beads, and defines possible hits. The step 3 screen is designed to remove those beads from the pool of potential hits that also exhibit binding to off-‐target serum proteins. Step 4 is called a product screen, and is unique to sequential in situ click screens [16]. This screen is designed to detect for the presence of in situ clicked reaction products, which are those hit beads containing the triazole-‐linked anchor peptide. Typically, Step 2 yields a few hundred hits (~0.05% of the OBOC library). Step 3 reduces that pool by a factor of 2 or 3 to about 100 hits, and Step 4 further reduces the number of hits to around 10. This is a manageable number, meaning that each hit can be separately synthesized as a biligand using Cu(I) catalyzed click chemistry to couple the anchor and 2o peptides. A complete list of the hits for the A21/3D6 and A22/4B3 screens is given in Table 2.1. 14 Scheme 2.1. Screening strategy for selecting capture agents against anti-‐HIV antibodies 3D6 and 4B3. The flow chart represents the use of the A21 and A22 cyclic peptides as anchor ligands for separate in situ click screens against a large OBOC azide-‐presenting peptide library. 15 Table 2.1. Biligand screen results. List of pentapeptide “hits” from OBOC biligand screens performed with A21/3D6 and A22/4B3. The selected secondary ligands corresponding to (i), (ii), and (iii) are shown in red, blue, and green respectively (shown in Figure 2.2). 16 2.5 BILIGAND CHARACTERIZATION The performance of these biligands is then characterized using immunoprecipitation (pulldown) assays detected by Western blotting from spiked serum samples for specificity (data not shown), and sandwich ELISAs and surface plasmon resonance (SPR) assays for affinity estimations (Figure 2.3-‐2.4). This approach yielded two equivalently performing biligands against 3D6, and one biligand against 4B3. These three PCC agents (Figure 2.2) were combined, in equal parts, to form a capture agent cocktail. The cocktail slightly outperformed both the standard commercial chimeric antigen and A21, when tested against healthy human serum spiked with both 3D6 and 4B3 (Figure 2.5). A21 is the equivalent of the original antigenic epitope of gp41. 17 Figure 2.2. Structures of peptide ligands in PCC Agent cocktail. Acetylene-‐presenting anchor peptides (black) were derived from the immunogenic epitope of HIV-‐1 gp41 (residues 600-‐612). A22-‐nindG (i) and A21-‐hnpfk (ii) were evolved from the original epitope appended with Pra at the C-‐terminus whereas A22-‐eihny (iii) utilizes the “substituted” anchor where residue Leu-‐607 is replaced with Pra. Secondary ligand branches (colored) were identified from the in situ click screen of a 5-‐mer OBOC library presenting an azide functionality. Biligands (i) and (ii) were raised against the target anti-‐HIV antibody 3D6, and the biligand (iii) was raised against the antibody 4B3. 18 Figure 2.3. Apparent affinity of A21 and biligands directed against 3D6 as determined by SPR. A. Sensorgram and 1st order Hill fit to affinity data for A21. B. Sensorgram and 1st order Hill fit to affinity data for A21-‐nidnG (i). C. Sensorgram and 1st order Hill fit to affinity data for A21-‐hnpfk (ii). 19 Figure 2.4. Apparent affinity of A22 and biligand directed against 4B3 as determined by SPR. A. Sensorgram and 1st order Hill fit to affinity data for A22. B. Sensorgram and 1st order Hill fit to affinity data for A22-‐eihny (iii). 20 Figure 2.5. Performance of PCC agent cocktail to detect 3D6 and 4B3 from human serum. Comparative performance of the PCC Agent cocktail versus the original gp41 epitope A21 and the commercial chimeric protein antigen was tested by a sandwich ELISA. Target antibodies 3D6 and 4B3 (4 nM each) were both spiked into diluted, HIV-‐ free human serum (1% v/v in TBS), and captured antibody was detected by peroxidase-‐ conjugated anti-‐human IgG antibody. 21 2.6 MATERIALS AND METHODS 2.6.1 Anchor Synthesis The anchor peptides A21 and A22 were synthesized C → N on 200-‐300 mgs of Rink amide resin (AnaSpec, Fremont, CA) with a Titan peptide synthesizer using standard Fmoc chemistry [17]. The synthesis used Fmoc and side chain protected L amino acids (Aaptec, Louisville, KY), Fmoc protected propargylglycine (Pra) (Aaptec, Louisville, KY), PEG5 (EMD Millipore,Germany), and biotin (Sigma-‐Aldrich, St. Louis, MO). The finished peptides were side-‐chain deprotected and cleaved from resin in a mixture of 95% v/v trifluoroacetic acid (TFA) (Sigma-‐Aldrich, St. Louis, MO), 5% v/v dH2O, and 5% triethylsilane (TES) (Sigma-‐Aldrich, St. Louis, MO), and precipitated in diethyl either. Cu(Phen)3 stock solution was prepared by mixing 150 mM Cu(II) sulfate pentahydrate (Sigma-‐Aldrich, St. Louis, MO) and 500 mM 1,10-‐phenanthroline (Sigma-‐Aldrich, St. Louis, MO) in 4:1 H2O:Ethanol. The ether-‐precipitated ligands were cyclized by stirring the cleaved peptides dissolved in ~5 mL acetonitrile (ACN) overnight with 1 mL of the Cu(Phen)3 stock solution. The cyclization mixtures were lyophilized, and the final peptide products were purified using reverse phase (RP) HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column (Phenomenex, Torrance, CA). MALDI-‐TOF MS: A21: Expected mass [M+H]+ = 2006.96, Observed mass [M+H]+ = 2006.24. A22: Expected mass [M+H]+ = 1892.89, Observed mass [M+H]+ = 1892.85 22 2.6.2 OBOC Screens All screens used naïve one-‐bead-‐one-‐compound (OBOC) D-‐pentapeptide libraries on TentaGel resin (TG) (Rapp Polymere, Tuebingen, Germany) of the form NH2-‐Az4-‐ X1X2X3X4X5-‐TG, excluding cysteine and methionine, where Az4 is L-‐azidolycine. Libraries were synthesized on a Titan peptide synthesizer (Aaptec, Louisville, KY) using a split/mix method, and couplings were done using standard Fmoc SPPS chemistry in N-‐ methylpyrrolidone (NMP). A21 with 3D6 Screening against 3D6 was carried out in four main steps: pre-‐clear, target screen, anti-‐screen, and product screen. Step 1 preclear: 400 mg of OBOC library was blocked in 5% w/v dry milk in Tris buffered saline (TBS) (25 mM Tris-‐HCl, 150 mM NaCl, 10 mM MgCl2, pH 7.6) for 2 hr at room temperature, and then incubated with 1:10,000 phosphatase-‐conjugated streptavidin (SA-‐AP, Promega, Madison, WI) in 0.5% w/v dry milk in TBS for 1 hr. The beads were washed in high salt buffer (25 mM Tris-‐HCl, 750 mM NaCl, 10 mM MgCl2, pH 7.6) for 1 hr, and then developed with BCIP/NBT (Promega, Madison, WI) in BCIP buffer (100 mM Tris-‐HCl, 150 mM NaCl, 1 mM MgCl2, pH 9). After 20 minutes, the reaction was quenched with conc. HCl, and the purple beads were discarded. The remaining library was decolorized in NMP, dried with methanol (MeOH) and dichloromethane (DCM), and then swelled and blocked in 5% w/v dry milk in TBS for 2 hr at room temperature. Step 2 target screen: 4 mL of a solution containing 11.4 µM A21 and 147 nM 3D6 in 0.5% w/v dry milk in TBS was added to the precleared OBOC 23 library, and the in situ click reaction was allowed to proceed overnight at 4°C. Bound target was probed with 1:10,000 AP-‐conjugated rabbit anti-‐human IgG Fc (Thermo Scientific, Rockford, IL) in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The beads were then washed in high salt buffer for 1 hr at room temperature, and developed with BCIP/NBT in BCIP buffer. The reaction was quenched after 20 minutes with conc. HCl, and the purple beads were retained and decolorized in NMP and dried with MeOH and DCM. The hit beads were then stripped of all bound protein by washing with 7.5 M Guanidine-‐HCl (pH 2) for 2 hr at room temperature, rinsed 10 times with water, and then blocked in 5% w/v dry milk in TBS for 2 hr at room temperature. Step 3 anti-‐screen: A solution of 1% v/v human serum (Omega Scientific, Tarzana, CA) in TBS was incubated with the hit beads from Step 2 for 1 hr at room temperature. Off-‐target human antibodies bound to the beads were probed with 1:10,000 AP-‐conjugated rabbit anti-‐human IgG Fc in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The beads were then washed in high salt buffer for 1 hr, and developed with BCIP/NBT in BCIP buffer. The reaction was quenched after 20 minutes with conc. HCl, and the purple beads were discarded. The remaining beads were decolorized in NMP and dried with MeOH and DCM. The anti-‐screened beads were swelled in water and washed with Guanidine-‐HCl for 2 hr at room temperature, rinsed 10 times with water, and then blocked in 5% w/v dry milk in TBS for 2 hr at room temperature. Step 4 product screen: The beads were incubated for 1 hr at room temperature with 1:10,000 SA-‐AP in 0.5% w/v dry milk in TBS, washed for 1 hr with high salt buffer, and then developed with BCIP/NBT in BCIP buffer for 20 minutes. The reaction was quenched with conc. HCl, and 24 the purple beads were retained and sequenced using a Procise CLC Edman Degradation Protein Sequencing System (Applied Biosystems, Kingston, RI). A22 with 4B3 Screening against 4B3 was carried out similarly to that described for 3D6, with two exceptions. In Step 2, the in situ click reaction solution contained 10 µM A22 and 430 nM 4B3 in 0.5% w/v dry milk in TBS. In Step 3, the beads were anti-‐screened against 0.1% v/v human serum in TBS spiked with 430 nM 3D6. 2.6.3 Biligand Synthesis Pra-‐OtBu The C-‐terminally protected Pra molecule used in the synthesis of (i) and (ii) was made by refluxing Fmoc-‐L-‐Pra-‐OH (Aaptec, Louisville, KY) with tert-‐butyl-‐ tricholroacetamide (Sigma-‐Aldrich, St. Louis, MO) at 50°C for 3 hr in DCM. The formed product was separated from unreacted species on a silica flash column in DCM. Peptides (i) and (ii) The secondary ligands including N-‐terminal Az4 for (i) and (ii) were synthesized C → N on 200-‐300 mgs of Rink amide resin with a Titan peptide synthesizer using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected D amino acids and L-‐Az4 (Aaptec, Louisville, KY). Fmoc-‐ and C-‐terminally protected Pra (Pra-‐OtBu) was clicked on-‐bead to the azide side chain of Az4 by adding equal molar amounts of 25 Pra-‐OtBu, Cu(I) iodide (Sigma-‐Aldrich, St. Louis, MO), and L-‐ascorbic acid (Sigma-‐Aldrich, St. Louis, MO), in 2x excess to the azide on bead. The click mixture was agitated in 20% v/v piperidine in NMP for 3 hr at room temperature, and the copper was chelated by repeated washing with 10% w/v sodium diethylduthiocarbamate trihydrate (Sigma-‐ Aldrich, St. Louis, MO) and 10% v/v N,N-‐diisopropylethylamine (DIEA ) (Sigma-‐Aldrich, St. Louis, MO) in NMP until the beads were clear. The Fmoc deprotected Pra was then the starting point for the remainder of the anchor component of each ligand, including N-‐terminal PEG5 and biotin. The anchor components were synthesized using standard Fmoc chemistry with protected L amino acids. The finished peptides were side-‐chain deprotected and cleaved from resin in 95:5:5 TFA:dH2O:TES, then precipitated in diethyl either. The ligands were cyclized by stirring the cleaved peptides dissolved in ~5 mL ACN overnight with 1 mL of the Cu(phen)3 stock solution. The cyclization mixtures were lyophilized, and the final peptide products were purified using RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. MALDI-‐TOF MS: (i): Expected mass [M+H]+ = 2733.30, Observed mass [M+H]+ = 2733.64. (ii): Expected mass [M+H]+ = 2843.40, Observed mass [M+H]+ = 2843.60. Peptide (iii) A22 was synthesized, cyclized, and purified as described above. The secondary ligand component including N-‐terminal Az4 for (iii) was synthesized C → N on Rink amide resin with a Titan peptide synthesizer using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected D amino acids and L-‐Az4. Resulting 26 secondary peptide was side-‐chain deprotected and cleaved from resin in 95:5:5 TFA:dH2O:TES, precipitated in diethyl either, and purified using RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. The purified secondary peptide was clicked to A22 in solution by mixing with 2x molar excess of A22, 10x Cu(I) iodide, 50x L-‐ascorbic acid, and 4x Tris[(1-‐benzyl-‐1H-‐1,2,3-‐triazol-‐4-‐yl)methyl]amine (TBTA) (Sigma-‐Aldrich, St. Louis, MO) in 5:1 NMP:dH2O and stirring overnight at room temperature. The clicked biligand product was purified from the reaction mixture by RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. Expected mass [M+H]+ = 2720.31, observed mass [M+H]+ = 2720.5. Scaling up (iii) The secondary ligands including N-‐terminally Boc-‐protected Az4 for (iii) was synthesized C → N on Rink amide resin with a Titan peptide synthesizer using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected D amino acids and L-‐Az4 (AnaSpec, Fremont, CA). Truncated A22 (Pra-‐ICTTA) was synthesized C → N on Sieber amide resin (ChemPep, Wellington, FL) with a Liberty 1 microwave peptide synthesizer (CEM, Matthews, NC) using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected L amino acids and Fmoc-‐L-‐Pra. Fully protected truncated A22 was cleaved from the resin with 1% v/v TFA in DCM, and neutralized with DIEA. Solvent was removed using a rotary vacuum evaporator and the protected peptide was purified by RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. Resulting truncated A22 was clicked on-‐bead to the secondary peptide through the azide side 27 chain of Boc-‐Az4 by adding equal molar amount of the protected truncated A22, 2x molar excess of Cu(I) iodide, and 2x excess of L-‐ascorbic acid to the azide on bead. The click mixture was agitated in 20% v/v piperidine in NMP overnight at room temperature, and then the copper was chelated by repeated washing with 10% w/v sodium diethyldithiocarbamate trihydrate and 10% v/v DIEA in NMP until the beads were clear. The Fmoc deprotected Pra was then the starting point for the remainder of the anchor component, including N-‐terminal PEG5 and biotin (Biotin-‐PEG5-‐IWGCSGK). The anchor components were synthesized using standard Fmoc chemistry with protected L amino acids. The finished peptide was N-‐terminal and side-‐chain deprotected and cleaved from resin in 95:5:5 TFA:dH2O:TES, then precipitated in diethyl either. The ligand was cyclized by stirring the cleaved peptide dissolved in ~5 mL ACN overnight with 1 mL of the Cu(phen)3 stock solution. The cyclization mixtures were lyophilized, and the final peptide products were purified using RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. 2.6.4 Assays Anchor binding ELISA Biotinylated anchor peptide was immobilized on streptavidin-‐coated 96-‐well plates (Thermo Scientific Pierce, Rockford, IL) at a concentration of 100 nM in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the anchor peptides were incubated with either 100 nM 3D6 or 4B3 in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with horseradish peroxidase (HRP)-‐ 28 conjugated mouse monoclonal antibody to human IgG Fc (Abcam, Cambridge, MA), diluted 1:10,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate (KPL, Gaithersburg, MD), then quenched with 1 M H2SO4 and read at 450 nm. Surface Plasmon Resonance SPR experiments were performed on a Biacore T200 instrument. Biotinylated peptides were immobilized on streptavidin-‐coated sensor chips (GE Healthcare Biosciences, Pittsburg, PA) at R.U. values ranging from 7-‐12. 4B3 or 3D6 was flowed at various concentrations in 1x HBS-‐EP buffer (GE Healthcare Biosciences, Pittsburgh, PA), and affinity curves were fit using the default settings of the Biacore evaluation software. Cocktail binding ELISA Recombinant multi-‐epitope chimeric HIV antigen (“chimera”) (BioLink International, Lisle, IL) was chemically biotinylated using ChromaLink Biotin Labeling Kit (Solulink, San Diego, CA) according to the manufacturer’s instructions, using 10x molar excess of the ChromaLink biotinylation reagent to the buffer-‐exchanged protein. Streptavidin-‐coated 96-‐well plates were saturated with the biotinylated biligand cocktail or chimera using 1 uM solutions in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the capture reagents were incubated with either 1% v/v human serum or a mixture of 4 nM each 3D6 and 4B3 spiked in 1% v/v human serum in TBS at room temperature for 1 hr. Bound antibody was probed with HRP-‐conjugated mouse 29 monoclonal antibody to human IgG Fc, diluted 1:15,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate, then quenched with 1 M H2SO4 and read at 450 nm (A450). The A450 for each sample is normalized against the A450 for the serum control, to yield a measurement of the signal-‐to-‐noise ratio of the assay. 2.7 CONCLUSION Described in this chapter is a method for developing a PCC Agent cocktail to capture the diversity of human antibodies generated in response to an infectious agent, with specific application to HIV. The components of the cocktail were raised against 4B3 and 3D6, two representative anti-‐HIV antibodies, and individual and collective binding of the capture agents to the target antibodies was characterized. The cocktail displayed a moderately enhanced performance compared to the original gp 41 epitope (A21) and the gold standard chimeric antigen when used to detect 4B3 and 3D6 from human serum. 2.8 ACKNOWLEDGEMENTS This work was done in collaboration with Aiko Umeda, Blake Farrow, Connie L. Hseuh, and Kaycie M. Deyle. It was funded primarily by a grant from the Bill and Melinda Gates Foundation. Additional funding for the development of screening approaches, and for certain capture agent characterization methods, was provided by 30 the Institute for Collaborative Biotechnologies through grant W911NF-‐09-‐0001 from the U.S. Army Research Office. The content of the information does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred. 31 2.9 REFERENCES 1. (1996) In: Baron S, editor. Medical Microbiology. 4th ed. Galveston (TX). 2. Fanales-‐Belasio E, Raimondo M, Suligoi B, Butto S (2010) HIV virology and pathogenetic mechanisms of infection: a brief overview. Ann Ist Super Sanita 46: 5-‐14. 3. Alvarez M, Chueca N, Guillot V, Bernal Mdel C, Garcia F (2012) Improving Clinical Laboratory Efficiency: Introduction of Systems for the Diagnosis and Monitoring of HIV Infection. Open Virol J 6: 135-‐143. 4. Tebourski F, Slim A, Elgaaied A (2004) The significance of combining World Health Organization and Center for Disease Control criteria to resolve indeterminate human immunodeficiency virus type-‐1 Western blot results. Diagn Microbiol Infect Dis 48: 59-‐ 61. 5. Daskalakis D (2011) HIV diagnostic testing: evolving technology and testing strategies. Top Antivir Med 19: 18-‐22. 6. Butto S, Suligoi B, Fanales-‐Belasio E, Raimondo M (2010) Laboratory diagnostics for HIV infection. Ann Ist Super Sanita 46: 24-‐33. 7. Ginocchio CC, Carman WF, Arens MQ (2011) Update on HIV diagnostic testing algorithms. Editors' introduction. J Clin Virol 52 Suppl 1: S1. 8. Agnew HD, Rohde RD, Millward SW, Nag A, Yeo WS, et al. (2009) Iterative in situ click chemistry creates antibody-‐like protein-‐capture agents. Angew Chem Int Ed Engl 48: 4944-‐4948. 9. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in Situ Click Chemistry Assembles a Branched Capture Agent and Allosteric Inhibitor for Akt1. Journal of the American Chemical Society 133: 18280-‐18288. 32 10. Huisgen R (1984) 1,3-‐Dipolar cycloaddition – introduction, survey, mechanism. In: Padwa A, editor. 1,3-‐Dipolar Cycloaddition Chemistry Wiley. pp. 1-‐176. 11. Thirumurugan P, Matosiuk D, Jozwiak K (2013) Click Chemistry for Drug Development and Diverse Chemical-‐Biology Applications. Chem Rev. 12. Finn MG, Fokin VV (2010) Click chemistry: function follows form. Chem Soc Rev 39: 1231-‐ 1232. 13. Lam KS, Lebl M, Krchnak V (1997) The "One-‐Bead-‐One-‐Compound" Combinatorial Library Method. Chem Rev 97: 411-‐448. 14. Du AP, Limal D, Semetey V, Dali H, Jolivet M, et al. (2002) Structural and immunological characterisation of heteroclitic peptide analogues corresponding to the 600-‐612 region of the HIV envelope gp41 glycoprotein. J Mol Biol 323: 503-‐521. 15. Xu JY, Gorny MK, Palker T, Karwowska S, Zolla-‐Pazner S (1991) Epitope mapping of two immunodominant domains of gp41, the transmembrane protein of human immunodeficiency virus type 1, using ten human monoclonal antibodies. J Virol 65: 4832-‐4838. 16. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in situ click chemistry assembles a branched capture agent and allosteric inhibitor for Akt1. J Am Chem Soc 133: 18280-‐18288. 17. Fields GB, Noble RL (1990) Solid phase peptide synthesis utilizing 9-‐fluorenylmethoxycarbonyl amino acids. International journal of peptide and protein research 35: 161-‐214. 33 Chapter 3 PCC Agent Based Assay for the Detection of Anti-‐HIV Antibodies from Human Sera 34 3.1 INTRODUCTION HIV diagnostic assays typically use chimeric proteins comprised of immunogenic and conserved antigens from the HIV to capture specific anti-‐HIV antibodies from the body fluid of potentially infected patients. However, the biologically produced chimeric recombinant proteins are susceptible to thermal, chemical, and biological degradation, as well as batch to batch variability. These limitations can adversely influence the performance of a diagnostic test [1,2,3], especially when used outside of controlled laboratory settings. This chapter explores the ability of the PCC Agent cocktail described in Chapter 2 to detect anti-‐HIV antibodies from human sera as compared to that of a standard recombinant chimeric antigen like those typically used in anti-‐HIV antibody detection assays. The chimeric protein used contained a fragment of HIV-‐1 gp41 (residues 546-‐ 692), the “O” group HIV-‐1 gp41 immunodominant region (residues 580-‐623), and a fragment of HIV-‐2 gp36 (residues 591-‐617). A good performance of this chimeric antigen has been reported elsewhere [4]. Two sets of samples were assayed, one clinical and one commercial. In both cases the PCC Agent cocktail showed increased average signal to noise compared to the chimeric antigen. 3.2 PATIENT SAMPLE ASSAY The PCC Agent cocktail and the standard antigen were co-‐evaluated against a panel of clinical samples using sandwich ELISAs. Serum samples were collected from nine HIV-‐1-‐positive patients in Southern California. For the comparison assays, 35 streptavidin-‐coated 96-‐well plates were saturated with the PCC Agent cocktail or chemically biotinylated chimera in triplicates. Serum samples were diluted to 1% v/v in Tris-‐buffered saline (TBS) supplemented with 0.1% w/v bovine serum albumin (BSA), and incubated in the wells for 1 hr at room temperature. Unbound proteins were washed off and captured IgG was detected with a peroxidase-‐conjugated mouse monoclonal anti-‐human IgG-‐Fc antibody. The comparison assay results for the patient serum samples along with a healthy control are shown in Figure 3.1. The signal-‐to-‐noise (S/N) ratio in these assays is defined as the measured ELISA signal for a given patient sample, divided by that for the healthy control. The PCC Agent cocktail performed at least as well, and typically much better, than the standard chimeric protein antigen. The average S/N improvement by the cocktail PCC agent over the chimeric antigen was a factor of 2.5. 36 Figure 3.1. Comparative performance of the PCC Agent cocktail versus the commercial chimeric protein, using sandwich ELISAs to detect anti-‐HIV-‐1 IgGs from a panel of sera samples collected from nine HIV-‐positive patients. The absorbance at 450 nm (A450) for each sample is normalized against the A450 for the healthy control, to yield a measurement of the signal-‐to-‐noise ratio of the assay. The PCC Agent cocktail, which is designed to capture a subset of anti-‐gp41 IgGs, exhibits superior performance for every sample, even though the chimeric protein is designed to capture antibodies against multiple HIV-‐1-‐associated epitopes (those containing fragments of HIV-‐1 gp41, “O” group HIV-‐1 gp41 immunodominant region, and HIV-‐2 gp39). For the assays, the PCC agent cocktail and the biotinylated chimeric antigen were immobilized on a streptavidin-‐ coated 96-‐well plate and incubated with diluted patient serum (1% v/v). Captured anti-‐ HIV antibodies were detected using peroxidase-‐conjugated anti-‐human IgG antibody. 37 3.3 PATH SAMPLE ASSAY The PCC Agent cocktail and the standard antigen were also co-‐evaluated against a panel of commercial samples. Serum samples from individual patients containing high, medium, or low titers of anti-‐HIV antibodies were obtained from SeraCare, through PATH of POC diagnostics program funded by the Bill and Melinda Gates Foundation. For the comparison assays, streptavidin-‐coated 96-‐well plates were saturated with the PCC Agent cocktail or chemically biotinylated chimera in triplicates. Serum samples were diluted to 1% v/v in Tris-‐buffered saline (TBS) supplemented with 0.1% w/v bovine serum albumin (BSA), and incubated in the wells for 1 hr at room temperature. Unbound proteins were washed off and captured IgG was detected with a peroxidase-‐ conjugated mouse monoclonal anti-‐human IgG-‐Fc antibody. The comparison assay results for the serum samples are shown in Figure 3.2. The signal-‐to-‐noise (S/N) ratio in these assays is defined as the measured ELISA signal for a given high, medium, or low titer sample, divided by that for the respective healthy control. The PCC Agent cocktail generally outperformed the standard chimeric protein antigen, with the exception of the low titer samples. The reasons for this are not entirely clear. However, the average S/N improvement by the cocktail PCC agent over the chimeric antigen was a factor of 2.8. 38 Figure 3.2. Comparative performance of the PCC Agent cocktail versus the commercial chimeric protein, using sandwich ELISAs to detect anti-‐HIV-‐1 IgGs from a panel of HIV-‐1 specific IgG containing plasma samples provided by PATH. The samples were rated HIGH, MED, or LOW in accordance with the titer of anti-‐HIV antibodies, and pooled HIV negative sera samples were used as a control (NEG). The absorbance at 450 nm (A450) for each sample is normalized against the A450 for the healthy control, to yield a measurement of the signal-‐to-‐noise ratio of the assay. The PCC Agent cocktail, which is designed to capture a subset of anti-‐gp41 IgGs, exhibits superior performance for every sample, even though the chimeric protein is designed to capture antibodies against multiple HIV-‐1-‐associated epitopes (those containing fragments of HIV-‐1 gp41, “O” group HIV-‐1 gp41 immunodominant region, and HIV-‐2 gp39). For the assays, the PCC agent cocktail and the biotinylated chimeric antigen were immobilized on a streptavidin-‐ coated 96-‐well plate and incubated with diluted patient serum (1% v/v). Captured anti-‐ HIV antibodies were detected using peroxidase-‐conjugated anti-‐human IgG antibody. 39 3.4 STABILITY ASSAY We then tested the PCC Agents for thermal stability. The PCC Agent cocktail component (iii) were synthesized at a large scale for an academic setting (~7 mg, Figure 3.3) and the lyophilized samples were stored under N2 at 25°C, 37°C, or 57°C for 58 days. The samples were then analyzed by HPLC to determine the presence of any degradation product. The traces of the peptide at each temperature are nearly identical, indicating little to no degeneration at these temperatures (Figure 3.3). The performance of these stored PCC Agents was then also tested in an ELISA, with no detectable loss of performance (Figure 3.3). 40 Figure 3.3. Samples of (iii) were stored as a powder (inset photo), under N2 at temperatures up to 57oC for ~2 months, and resolved by analytical HPLC to determine the presence of any degradation product. The HPLC traces reveal that the fingerprint of the PCC Agent is unchanged. The inset shows that the assay performance of the PCC Agent is also unaffected. 41 3.5 MATERIALS AND METHODS 3.5.1 Patient serum ELISA Peripheral blood was obtained from HIV-‐1 infected patients at the University of California, Los Angeles (UCLA) Medical Center between July and August of 2012. Sera were stored at -‐80°C before subsequent analysis. Streptavidin-‐coated 96-‐well plates were saturated with the biotinylated biligand cocktail or chimera using 1 uM solutions in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the capture reagents were incubated with patient serum samples diluted to 1% v/v in TBS containing 0.1% w/v bovine serum albumin (BSA). Additionally, both cocktail and chimera capture reagents were incubated with 1% v/v commercial healthy human serum in TBS. Bound antibody was probed with HRP-‐conjugated mouse monoclonal antibody to human IgG Fc, diluted 1:10,000 in TBS with 0.1% w/v BSA. The colorimetric assay was developed with TMB substrate, then quenched with 1 M H2SO4 and read at 450 nm (A450). The A450 for each sample is normalized against the A450 for the healthy control, to yield a measurement of the signal-‐to-‐noise ratio of the assay. 3.5.2 PATH samples ELISA Serum samples were obtained from SeraCare, through PATH of POC diagnostics program funded by the Bill and Melinda Gates foundation. Five groups of four samples were tested, where each group contained high, medium, and low titer samples, along with a negative sample that did not contain anti-‐HIV antibodies. Sera were stored at -‐ 80°C before subsequent analysis. Streptavidin-‐coated 96-‐well plates were saturated 42 with the biotinylated biligand cocktail or chimera using 1 uM solutions in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the capture reagents were incubated with patient serum samples diluted to 1% v/v in TBS containing 0.1% w/v bovine serum albumin (BSA). Additionally, both cocktail and chimera capture reagents were incubated with 1% v/v commercial healthy human serum in TBS. Bound antibody was probed with HRP-‐conjugated mouse monoclonal antibody to human IgG Fc, diluted 1:10,000 in TBS with 0.1% w/v BSA. The colorimetric assay was developed with TMB substrate, then quenched with 1 M H2SO4 and read at 450 nm (A450). The A450 for the samples in each group are normalized against the A450 for the negative sample in that group. 3.5.3 Stability assay Small amounts of (iii) were stored under N2 at 25°C, 37°C, or 57°C for 58 days. Samples were diluted in 1:1 ACN/H2O with 0.1% v/v TFA to an equal concentration determined by measuring the absorbance values at 280 nm using NanoDrop spectrophotometer (Thermo Scientific, Waltham, MA) and resolved by analytical RP-‐ HPLC. The remaining samples were lyophilized and used in a single point sandwich ELISA. Streptavidin-‐coated 96-‐well plates were saturated with the samples of (iii) recovered from the storage experiments using 1 uM solutions in TBS. The plates were blocked with 5% w/v dry milk in TBS, then incubated with either 1% v/v human serum or 10 nM 4B3 spiked in 1% v/v human serum in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with HRP-‐conjugated mouse 43 monoclonal antibody to human IgG Fc diluted 1:15,000 in 0.5% w/v dry milk in TBS. The colorimetric assay was developed with TMB substrate, then quenched with 1 M H2SO4 and read at 450 nm. 3.6 CONCLUSION This thesis demonstrated the successful application of this method for HIV-‐1 diagnostics by producing a cocktail of three PCC Agents that detected the presence of anti-‐HIV antibodies in clinical samples with a significantly enhanced signal-‐to-‐noise relative to the standard, recombinant protein-‐based chimeric antigen. In a recent report, an antigenic peptide cocktail comprised of synthetic peptides derived directly from gp120/V3-‐I (HIV-‐1 Indian isolate), gp41 (HIV-‐1), and gp36 (HIV-‐2), as well as the recombinant protein rp24 (HIV-‐1) was shown to also provide superior performance relative to the chimeric antigen [5]. This points to the possibility that expanding the current approach by developing multiple cocktails of PCC Agents, each directed against a distinct HIV epitope, would likely provide superior performance to that reported here. The strategy presented provides a promising approach for developing assays for detecting the immune response to other infectious agents, especially where challenges associated with the polyclonal nature of a humoral immune response can compromise assay sensitivity. 44 3.7 ACKNOWLEDGEMENTS This work was done in collaboration with Aiko Umeda, Jocelyn T. Kim, and Bert T. Lai. It was funded primarily by a grant from the Bill and Melinda Gates Foundation. Additional funding for the development of screening approaches, and for certain capture agent characterization methods, was provided by the Institute for Collaborative Biotechnologies through grant W911NF-‐09-‐0001 from the U.S. Army Research Office. The content of the information does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred. 45 3.8 REFERENCES 1. Daskalakis D (2011) HIV diagnostic testing: evolving technology and testing strategies. Top Antivir Med 19: 18-‐22. 2. Ginocchio CC, Carman WF, Arens MQ (2011) Update on HIV diagnostic testing algorithms. Editors' introduction. J Clin Virol 52 Suppl 1: S1. 3. Alvarez M, Chueca N, Guillot V, Bernal Mdel C, Garcia F (2012) Improving Clinical Laboratory Efficiency: Introduction of Systems for the Diagnosis and Monitoring of HIV Infection. Open Virol J 6: 135-‐143. 4. Chin CD, Laksanasopin T, Cheung YK, Steinmiller D, Linder V, et al. (2011) Microfluidics-‐based diagnostics of infectious diseases in the developing world. Nat Med 17: 1015-‐1019. 5. Tiwari RP, Jain A, Khan Z, Kumar P, Bhrigu V, et al. (2012) Designing of novel antigenic peptides cocktail for the detection of antibodies to HIV-‐1/2 by ELISA. Journal of immunological methods. 46 Chapter 4 A Selective 15N-‐to-‐1H Polarization Transfer Sequence for More Sensitive Detection of 15N-‐Choline 47 4.1 ABSTRACT This chapter discusses selective polarization transfer from 15N to methyl 1H spins in 15N-‐choline. The sensitivity and information content of heteronuclear nuclear magnetic resonance is frequently optimized by transferring spin order of spectroscopic interest to the isotope of highest detection sensitivity prior to observation. This strategy was extended to 15N-‐choline using the scalar couplings to transfer polarization from 15N to choline’s nine methyl 1H spins in high field. A theoretical analysis of a sequence using nonselective pulses showed that the optimal efficiency of this transfer is decreased by 62% as the result of competing 15N-‐1H couplings involving choline’s four methylene protons. A frequency-‐selective pulse was therefore incorporated to support evolution of only the 15N-‐methyl 1H coupling during the transfer period. This sequence provided a 52% sensitivity enhancement over the nonselective version in in vitro experiments on a sample of t hermally polarized 15N-‐choline in D2O. Further, the 15N T1 of choline in D2O was measured to be 217 ± 38 s, the 15N-‐methyl 1H coupling constant was found to be 0.817 ± 0.001 Hz, and the larger of choline’s two 15N-‐methylene 1H coupling constants was found to be 3.64 ± 0.01 Hz. Possible improvements and applications to in vivo experiments using long-‐lived hyperpolarized heteronuclear spin order are discussed. 48 4.2 INTRODUCTION It is now possible to routinely generate samples of certain small biomolecules with nuclear polarizations on the order of ten percent using PASADENA (parahydrogen and synthesis allow dramatically enhanced nuclear alignment) [1] or DNP (dynamic nuclear polarization) [2; 3]. These technologies show promise for clinical applications based on the in vivo characterization of metabolic processes on the seconds timescale. Experiments of this nature, rely on storage of the polarization on heteronuclei with long population relaxation times, such as carbonyl 13C or quaternary 15N nuclei, in order to minimize relaxation losses during sample delivery and to provide time for transport, binding, and metabolism. However, the sensitivity enhancement afforded by this strategy is partly lost if the final signal is also detected on a low-‐𝛾 heteronucleus. Chekmenev et al. [4] have proposed that optimal sensitivity can be obtained by storing polarization on a slowly-‐relaxing heteronuclear spin and then using an INEPT[5] sequence to coherently transfer magnetization to nearby 1H nuclei for detection, and have demonstrated the method with two hyperpolarized reagents, 1-‐13C-‐succinate-‐d2 and 2,2,3,3-‐tetrafluoropropyl 1-‐13C-‐propionate-‐d3 (TFPP). Those experiments were designed assuming that the relevant 1H and 13C sites were 3-‐spin ABX systems. Here we demonstrate an extension of this polarization transfer strategy to hyperpolarized 15N-‐choline, an A9BB’CC’X system of 14 spin ½ nuclei. Choline is a useful biomarker, showing significantly altered uptake and metabolism in diseased brain tissue [6] and in malignant tumor cells in the breast and prostate [7; 8]. Hyperpolarization of 49 15 N-‐choline to (4.6 ± 1)% has been reported after two hours of DNP at 1.4 K [3]. The 15N longitudinal relaxation time of T1 = 120 s in blood [3] is promising for metabolic studies. We have investigated two INEPT-‐based pulse sequences for 15N-‐to-‐1H polarization transfer in 15N-‐choline. The first is a nonselective refocused INEPT sequence similar to that used by Chekmenev et al. to transfer polarization from 13C to multiple 1H sites in succinate and TFPP [4]. While this sequence is effective, a product operator analysis shows that significantly greater sensitivity can be obtained by using a selective sequence targeting choline’s nine degenerate methyl 1H spins. We therefore also present a selective refocused INEPT transfer that uses a REBURP pulse to support coherence transfer from 15N to choline’s methyl 1H spins while suppressing competing transfer pathways to methylene 1H nuclei. These pulse sequences are experimentally demonstrated in vitro using a thermally polarized sample of 15N-‐choline in aqueous solution. 4.3 THEORY At a given spin polarization and probe efficiency, the sensitivity of detection in an NMR experiment scales as the square of the gyromagnetic ratio of the nuclide being observed. In the case of 15N hyperpolarization, therefore, a sensitivity enhancement of (γ 1 H / γ 15 N )2 = 97 can be obtained by transferring to 1H for detection rather than directly observing a 15N signal, assuming unity transfer of spin angular momentum, as is closely 50 approached within an isolated 15N-‐1H spin pair with a resolved scalar coupling. Such a transfer can be achieved using refocused INEPT (Figure 4.1), the standard pulse sequence for polarization transfer between spins with a direct scalar coupling. The effect of this sequence when applied to 15N-‐choline can be determined using a product operator analysis. In the following discussion the relevant 15N-‐1H scalar coupling constants will be labeled J m , J a , and J b as indicated in Figure 4.2. Couplings between the methyl and methylene 1H spins will be unimportant and are neglected. 51 Figure 4.1. A refocused INEPT pulse sequence for polarization transfer from 15N to 1H for detection. Narrow (wide) rectangles denote radiofrequency pulses with a tip angle of 𝜋/2 (𝜋). All pulses are applied along the x axis except where indicated. 52 Figure 4.2. The structure of 15N-‐choline. Labels indicate the relevant 15N-‐1H scalar couplings. 53 First, consider the effect of the sequence on the spin system of Figure 4.2, while initially neglecting the effects of the 15N-‐methylene 1H couplings J a and J b . The system’s magnetization is proportional to the dimensionless 15N spin angular momentum, with initial value normalized to ρ0 = S z . This is converted to in-‐phase 15N single-‐quantum coherence by the first 15N (𝜋/2)x pulse to obtain ρ1 = − S y . This coherence evolves under the 15N-‐methyl 1H scalar couplings during the INEPT transfer period τ 1 to produce terms that are antiphase with respect to the methyl 1H spins. The simultaneous 𝜋 pulses in the middle of τ 1 refocus chemical shift evolution, leaving only the effects of couplings between 15N and each of the nine methyl 1H spins. At the end of τ 1 the system’s state is given by: 9 ρ 2 = − S y cos9 (π J mτ 1 ) + ∑ 2S x I iz cos8 (π J mτ 1 ) sin(π J mτ 1 ) i =1 8 9 7 2 + ∑ ∑ 4S y I iz I jz cos (π J mτ 1 ) sin (π J mτ 1 ) + HO. i =1 j =i +1 Here, HO indicates higher-‐order terms, specifically single-‐quantum 15N coherence that is antiphase with respect to three or more 1H spins. After the transfer delay, simultaneous 1 H (𝜋/2)x and 15N (𝜋/2)y pulses are applied: 9 ρ3 = − S y cos9 (π 2 J N −CH τ 1 ) + ∑ 2 S z I iy cos8 (π 2 J N −CH τ 1 ) sin(π 2 J N −CH τ 1 ) 3 8 9 3 i =1 7 2 2 2 + ∑ ∑ 4 S y I iy I jy cos (π J N −CH3τ 1 ) sin (π J N −CH3τ 1 ) + HQ. i =1 j =i +1 3 54 The pulses convert the terms HO to high-‐order multiple-‐quantum coherences, which have been abbreviated as HQ . At this point, 15N coherences and multiple-‐quantum 1H coherences, which will not lead to observable signals at the end of the experiment, can be dropped from the density operator. This leaves only single-‐quantum 1H coherences that are antiphase with respect to 15N: 9 ρ3 = ∑ 2S z I iy cos8 (π J mτ 1 ) sin(π J mτ 1 ) . i =1 The next part of the sequence is a refocusing period of duration τ 2 = 1/ (2 J m ) during which the antiphase single-‐quantum 1H terms evolve into in-‐phase 1H coherence: 9 ρ4 = −∑ I ix cos8 (π J mτ 1 ) sin(π J mτ 1 ) . i =1 The detected 1H signal is proportional to the total magnetization of the nine methyl protons, so the dependence of the initial signal amplitude on τ 1 is: S sel = Asel 9cos8 (π J mτ 1 ) sin(π J mτ 1 ) + C . (1) This expression includes a hardware-‐dependent proportionality constant Asel and a constant C accounting for any DC offset in the observed signal. Equation 1 is similar to the expression derived by Doddrell et al. for a coherence transfer in the opposite direction, from n equivalent 1H spins to a heteronucleus [9]. The value of τ 1 that maximizes Ssel is given by: 55 τ1 = 1 ⎛ 1 ⎞ tan −1 ⎜ ⎟ . π Jm ⎝ 2 2 ⎠ For the value J m = 0.82 Hz measured experimentally for 15N-‐choline, the optimal transfer delay is τ1 = 0.132 s, yielding Ssel = 1.873 Asel + C . The transfer function of the angular momentum from 15N to 1H or vice versa (Eq. 1) can exceed unity. The conserved quantity is the sum of the squares of the coefficients of the normalized product operators expressing the density operator. Contributions to the observable signal from choline’s nine methyl 1H spins combine linearly and constructively at the optimized value of τ 1 . The above calculation is readily generalized to include the effects of scalar coupling between the 15N spin and the four methylene protons ( J a and J b in Fig. 2), yielding an analogous relative signal function: Snon = Anon 9cos8 (π J mτ 1 ) sin(π J mτ 1 )cos 2 (π J aτ 1 )cos 2 (π J bτ 1 ) + C (2) A proportionality constant Anon and a DC offset constant C have again been included. Maximizing this function numerically for the best known values of the three coupling constants for choline, J m = 0.82 Hz, J a = −0.57 Hz, J b = 3.64 Hz, gives an optimal transfer delay τ1 = 0.233 s, which yields a signal of Snon = 0.721 Anon + C . Comparing the maximum values of Ssel and Snon , it is apparent that the 15N-‐ methylene 1H couplings have the effect of reducing the observable 1H signal by 62% in 56 the idealized case of an experiment for which Asel = Anon and C = 0 . It would clearly be advantageous to suppress these couplings. This can be achieved by replacing the 1H p pulse in the middle of τ 1 with a selective pulse tailored to invert the methyl 1H spins without affecting the methylene 1H spins. 4.4 EXPERIMENTAL Data was collected on a Varian UNITYINOVA spectrometer operating at a 1H resonance frequency of 500 MHz. The sample was prepared by dissolving 20 mg 15N-‐ choline chloride (ISOTEC, Miamisburg, Ohio) in 700 µl D2O. Refocused INEPT experiments were performed using the pulse sequence shown in Figure 4.3, which implements the selective 15N-‐methyl 1H transfer described in the theory section using a 6.695 ms 1H REBURP pulse [10] centered at the methyl 1H frequency. A time τ reb = 6.4 ms, equal to the effective evolution time of the 15N-‐methyl 1H coupling during the REBURP pulse, was subtracted from the transfer delay as indicated in Fig. 3. For nonselective INEPT experiments, the REBURP was replaced with a nonselective 𝜋 pulse and τ reb was set to zero. 57 Figure 4.3. Selective refocused INEPT pulse sequence for coherent polarization transfer from 15N to methyl 1H in 15N-‐choline. Narrow (wide) rectangles denote radiofrequency pulses with a tip angle of 𝜋/2 (𝜋), applied with B1 field strengths of 23.9 kHz for 1H and 15.6 kHz for 15N. The shaped 1H pulse is a 6.695 ms REBURP [10] centered on the methyl 1H frequency. A four step phase cycle (𝜑1 = x, x, x, x; 𝜑2 = x, y, x, y; 𝜑rec = x, y, x, y) was used to suppress signals originating from 1H magnetization and to prevent carry over of signal from one transient to the next. A recycle delay d1 of 60 s (~ 0.25 T1 ) was used. Since it is impractical to make this delay long enough to allow full relaxation of the choline 15N spin, a purge element consisting of a 15N 𝜋/2 pulse sandwiched by gradients g0 and g1 is used to eliminate 15N magnetization at the start of d1 , ensuring a consistent starting magnetization for all experiments. A 1H purge element comprising a 1H 𝜋/2 pulse and gradient g2 is used at the start of each transient. Gradients g3 and g4 are used to suppress coherence transfer pathways created by imperfect 𝜋 pulses. 58 To allow efficient testing in the absence of hyperpolarization, the sequence includes a 15N purge element before the recycle delay d1 to ensure that the system reaches a consistent state at the start of each transient. This measure is necessary because it is impractical to allow full relaxation of the choline 15N spin after each transient owing to its very long T1 . The sequence includes a 15N 𝜋/2 pulse immediately prior to acquisition in order to purge any 1H coherence that remains antiphase with respect to 15N after the refocusing period. In the absence of this pulse, intermittent phase aberrations were observed in the methyl 1H resonance. For 15N T1 measurements by inversion-‐recovery, the pulse sequence of Figure 4.3 was modified by adding a 15N 𝜋 pulse and a variable recovery delay immediately before the 1H purge pulse. For these experiments, the 15N purge element before d1 was removed and d1 was increased to 600 s. Data analysis was performed in MATLAB (The MathWorks, Natick, MA) using scripts developed in house. Spectral data was quantified by fitting the methyl 1H resonance to a complex Lorentzian lineshape function, except for inversion-‐recovery data, which was quantified by integration. A downhill simplex algorithm was used for least-‐squares data fitting, and uncertainties in the extracted parameters were estimated using a bootstrap Monte Carlo method [11]. The value of τ reb , which is subtracted from the INEPT transfer delay in order to account for evolution of the 15N-‐methyl 1H scalar coupling during the long REBURP 59 pulse, was determined using a numerical simulation of the pulse sequence. Calculations were performed in GAMMA v4.1.2 [12] in Hilbert space using a spin system of practical size, consisting of one 15N and six methyl 1H spins, with J m = 0.82 Hz. The 1H resonance frequency was set to 500 MHz, and the carrier frequencies were set on resonance with the 15N and methyl 1H spins. The REBURP and the nonselective pulses were simulated with finite widths corresponding to the experimental values, and relaxation and phase cycling were neglected. The pulse sequence shown in Figure 4.3 was simulated, and the methyl 1H signal amplitude was calculated, for each of a series of values of the INEPT transfer delay (τ1 − τ reb ) . The signal data, shown in Figure 4.4, was then subjected to least-‐squares fit to a version of Eq. 1 appropriate for a system containing six methyl 1H spins and including τ reb as an adjustable parameter, resulting in the value of τ reb = 6.4 ms used experimentally. 60 Figure 4.4. Numerical simulation of the selective 15N-‐to-‐1H INEPT pulse sequence acting on a simplified spin system. Circles are simulated methyl 1H signal amplitudes for different values of the INEPT transfer delay (τ1 − τ reb ) . The line was calculated using a modified form of Eq. 1 with parameter values from a least-‐squares fit to the data. This fit was used to determine τ reb , the effective evolution time of the 15N-‐methyl 1H coupling during the sequence’s REBURP pulse. 61 4.5 RESULTS AND DISSCUSSION Figure 4.5 shows a spectrum of 15N-‐choline recorded using the selective 15N-‐to-‐ 1 H INEPT sequence of Figure 4.3, alongside a 15N-‐detected spectrum of the same sample recorded in a similar measurement time. It is not meaningful to compare these spectra quantitatively because they were both recorded using a probe that is optimized for direct heteronuclear observation, a factor which diminishes the sensitivity advantage of 1 H detection. However, even under these unfavorable circumstances the signal-‐to-‐noise ratio is clearly much higher in the selective INEPT spectrum (Figure 4.5a) than in the 15N-‐ detected spectrum (Figure 4.5b). A modified version of the selective 15N-‐to-‐1H INEPT experiment was used to measure the 15N longitudinal relaxation time of 15N-‐choline in D2O solution, and a value of T1 = (217 ± 38) s was obtained (Figure 4.6). This is comparable to the value T1 = (285 ± 12) s previously reported by Gabellieri et al. [3] for 15N-‐choline in 90% H2O / 10% D2O. 62 Figure 4.5. (a) 1H spectrum of 15N-‐choline recorded using the selective 15N-‐to-‐1H INEPT sequence of Fig. 3. Only the methyl resonance is shown. (b) Single-‐pulse 15N spectrum of 15N-‐choline, recorded with tip angle 𝜋/2. Spectra (a) and (b) were each collected on the same Varian AutoX Dual Broadband probe, taking the sum of four transients with a recycle delay of 60 s. An enhancement factor of 11.0 greater sensitivity when using selective 15N-‐to-‐1H INEPT sequence was estimated from Monte Carlo fits of (a) and (b) to Lorentzian lineshapes. The peak amplitudes were used as “signals” and the standard deviations as “noise” to construct pseudo “SNR” values for each spectrum to take into account differences in the linewidths of 1H and 15N peaks. The ratio calculated enhancement factor. " SNR "1H " SNR "15 N gave the 63 Figure 4.6. Measurement of the 15N longitudinal relaxation time of 15N-‐choline in D2O solution using a modified version of the pulse sequence shown in Figure 3. The experiment was repeated for 13 values of the inversion-‐recovery delay T = 0.001, 15, 30, 45, …, 180 s. Crosses are experimental methyl 1H signal amplitudes obtained by integrating the spectral data. The line was calculated using the indicated equation and parameter values from a least-‐squares fit to the data. 64 A series of 15N-‐choline 15N-‐to-‐1H INEPT spectra were recorded using both selective and nonselective versions of the pulse sequence of Figure 3 with values of the transfer delay τ 1 ranging from 6.4 ms to 700 ms. The methyl 1H signal amplitudes observed in these experiments are plotted in Figure 7. The selective INEPT sequence allows 15N-‐choline to be detected with 52% greater sensitivity than the nonselective sequence, comparing the maximum signal amplitude observed in each case. In the Theory section, the amplitude of the methyl 1H signal was predicted as a function of the transfer delay for the selective (Eq. 1) and nonselective (Eq. 2) experiments using a product operator analysis. By fitting the observed signal amplitudes to Eqs. 1 and 2, the optimal length of τ 1 for each experiment can be determined, as well as precise values of choline’s 15N-‐1H coupling constants. A simultaneous least-‐squares fit to both datasets yields parameter values Asel = 0.632 ± 0.001, Anon = 1.000 ± 0.005 , J m = (0.817 ± 0.001) Hz, J b = (3.64 ± 0.01) Hz, and C = 0.000 ± 0.001 . These values for the coupling constants agree well with ones derived from previously published 1H-‐14N values: J m = 0.80 Hz and J b = 3.61 Hz [13]. The value of the smaller 15N-‐methylene 1H coupling constant was found to have a negligible effect on the fitting process, so this parameter was fixed at a reasonable value of J a = −0.57 Hz. 65 Figure 4.7. Amplitude of the methyl 1H signal observed in 15N-‐to-‐1H INEPT spectra of 15N-‐ choline using selective (circles) and nonselective (triangles) versions of the pulse sequence shown in Figure 3, as a function of the INEPT transfer period τ 1 . The lines were calculated using Eq. 1 (solid line) and Eq. 2 (dashed line) with parameter values from a least-‐squares fit to the data. 66 Note that, for experiments using the selective and nonselective versions of the pulse sequence of Figure 3 on the same spectrometer hardware and sample, the constants C in Eqs. 1 and 2 should be the same, but Asel may be smaller than Anon due to signal losses during the selective sequence’s REBURP pulse. Comparing the values of Asel and Anon from the least squares fit it appears that 36% of the signal is lost in this way. This can be attributed in part to miscalibration of the REBURP pulse. A three pulse sequence (𝜋/2 -‐ REBURP -‐ 𝜋/2) was used to calibrate the REBURP pulse by seeking pulse width and power parameters that gave the best null signal. It would be more appropriate to choose a calibration sequence that uses the REBURP pulse for inversion, as it is used in the selective INEPT sequence, rather than for refocusing. Also, a sequence with a larger number of pulses would be preferable, so that the cumulative effects of B1 inhomogeneity would be comparable in the calibration sequence and the selective INEPT sequence. We have presented a pulse sequence that suppresses the effects of the four methylene protons on the INEPT transfer from 15N to methyl 1H in 15N-‐choline. Note that the same result can be achieved by selective deuteration of 15N-‐choline. This approach might provide the additional benefit of increasing the molecule’s 15N longitudinal relaxation time and would be an option for DNP experiments. However, the present selective recoupling strategy is clearly advantageous for 15N-‐choline hyperpolarization using PASADENA. In these experiments, two of the methylene 1H spins derive from a parahydrogen molecule which reacts with an unsaturated precursor to initiate the hyperpolarization process. Although deuteration of the alkene precursor 67 can remove two of the methylene 1H spins, those which supply the spin order cannot be eliminated. The success of this selective coherence transfer strategy in 15N-‐choline has clear implications for studies of other hyperpolarized small biomolecules. By directing the spin order to a subset of the possible detection spins, sensitivity is improved. The recoupling strategy demonstrated here accomplishes this by way of a frequency selective π pulse, which simplifies, and often shortens, the polarization transfer step. Similar selective INEPT sequences could be used to improve the sensitivity of the 13C-‐to-‐ 1 H polarization transfer experiments demonstrated by Chekmenev et al. in hyperpolarized TFPP [4] by directing the spin order to one of the resolved 1H sites. The selective 15N-‐choline transfer presented here also has potentially useful applications to thermally polarized in vivo 15N spectroscopy. Pulse sequences can be devised to transfer methyl 1H polarization to 15N for an encoding period and then back to methyl 1H for detection. Such sequences would benefit from choline’s long 15N T1 while making use of the strong thermal polarization of the nine methyl 1H spins. This strategy could be used to resolve species such as choline and phosphocholine by 15N-‐1H correlation spectroscopy, or even to study the metabolic interconversion of choline and related compounds using a sequence with a longitudinal 15N period for chemical exchange. 68 4.6 CONCLUSION As part of a strategy to improve the sensivity of in vivo experiments on hyperpolarized small biomolecules, we have investigated INEPT pulse sequences for the transfer of magnetization from 15N to the methyl 1H spins in 15N-‐choline. Product operator analysis of a simple refocused INEPT sequence shows that the optimized efficiency of this transfer is diminished by 62% by the action of competing couplings between 15N and choline’s methylene 1H spins. A selective INEPT sequence, using a 1H REBURP pulse to suppress undesired couplings during the transfer period, was devised and tested in vitro on a thermally polarized 15N-‐choline sample. It was demonstrated that the selective INEPT sequence leads to a 52% stronger methyl 1H signal than the nonselective version. 4.7 ACKNOWLEGEMENTS This work was done in collaboration with Jason E. Ollerenshaw, Valerie A. Norton, and Daniel P. Weitekamp. It was supported by the Beckman Institute pilot program, “Spin-‐ Polarized Molecules for Structural and Systems Biology.” JEO was supported by a Postdoctoral Fellowship from the Natural Sciences and Engineering Research Council of Canada. 69 4.8 REFERENCES [1] E.Y. Chekmenev, J. Hovener, V.A. Norton, K. Harris, L.S. Batchelder, P. Bhattacharya, B.D. Ross, and D.P. Weitekamp, PASADENA Hyperpolarization of Succinic Acid for MRI and NMR Spectroscopy. Journal of the American Chemical Society 130 (2008) 4212-‐4213. [2] K. Golman, Metabolic imaging by hyperpolarized 13C magnetic resonance imaging for in vivo tumor diagnosis. Cancer Research 66 (2006) 10855. [3] C. Gabellieri, S. Reynolds, A. Lavie, G.S. Payne, M.O. Leach, and T.R. Eykyn, Therapeutic target metabolism observed using hyperpolarized 15N choline. J. Am. Chem. Soc. 130 (2008) 4598-‐4599. [4] E.Y. Chekmenev, V.A. Norton, D.P. Weitekamp, and P. Bhattacharya, Hyperpolarized 1H NMR employing low g nucleus for spin polarization storage. J. Am. Chem. Soc. 131 (2009) 3164-‐3165. [5] R.F. Gareth A. Morris, Enhancement of Nuclear Magnetic Resonance Signals by Polarization Transfer. Journal of the American Chemical Society 101 (1979) 763-‐762. [6] Y. Boulanger, M. Labelle, and A. Khiat, Role of phospholipase A2 on the variations of the choline signal intensity observed by 1H magnetic resonance spectroscopy in brain diseases. Brain Res. Rev. 33 (2000) 380-‐389. [7] R. Katz-‐Brull, D. Seger, D. Rivenson-‐Segal, E. Rushkin, and H. Degani, Metabolic markers of breast cancer: enhanced choline metabolism and reduced choline-‐ether-‐phospholipid synthesis. Cancer Res. 62 (2002) 1966-‐1970. [8] E. Sutinen, M. Nurmi, A. Roivainen, M. Varpula, T. Tolvanen, P. Lehikoinen, and H. Minn, Kinetics of [11C]choline uptake in prostate cancer: a PET study. Eur. J. Nucl. Med. Mol. Imaging 31 (2004) 317-‐324. 70 [9] D.M. Doddrell, D.T. Pegg, W. Brooks, and M.R. Bendall, Enhancement of 29Si or 119Sn NMR signals in the compounds M(CH3)nCl4-‐n (M = Si or Sn, R = 4, 3, 2) using proton polarization transfer. Dependence of the enhancement on the number of scalar coupled protons. J. Am. Chem. Soc. 103 (1981) 727-‐728. [10] H. Geen, and R. Freeman, Band-‐selective radiofrequency pulses. J. Magn. Reson. 93 (1991) 93-‐141. [11] W.H. Press, S.A. Teukolsky, W.T. Vetterling, and B.P. Flannery, Numerical Recipes in C: The Art of Scientific Computing 2nd edition, Cambridge University Press, Cambridge, UK, 1992. [12] S.A. Smith, T.O. Levante, B.H. Meier, and R.R. Ernst, Computer simulations in magnetic resonance. An object-‐oriented programming approach. J. Magn. Reson. A106 (1994) 75-‐ 105. [13] V. Govindaraju, K. Young, and A.A. Maudsley, Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed. 13 (2000). 71 Appendix A Understanding PCC Agent Binding to Single Point Mutation E17K of Akt1 Pleckstrin Homology Domain through Molecular Dynamics 72 A.1 INTRODUCTION The Akt kinase is a critical molecular router that mediates cell growth, apoptosis, and translation [1], and Akt overexpression and/or hyperactivation has been observed in many cancer types [2]. The Akt1 activation mechanism includes binding to the PIP3 lipid on the cell membrane with a domain called the Pleckstrin Homology Domain (PHD). Mutations in the PHD of Akt1 that increase its affinity for binding with PIP3 will cause the upregulation of downstream pathways, thereby promoting the formation of cancerous cells [3]. Recent experiments performed on mice have revealed that a single amino acid point mutation to the PHD of Akt1, called the E17K mutation, was sufficient to cause cancer [4]. This same mutation has been found in certain human ovarian, colorectal and breast cancers. The E17K mutation exchanges a negatively charged glutamate for a positively charged lysine, causing the formation of an additional hydrogen bond between the PHD and PIP3 on the cell membrane. This induces a conformational change that causes Akt1(E17K) to bind to the plasma membrane four times stronger than it would in its wild-‐type form [4]. This strengthened binding is believed to be the underlying mechanism for the creation of a type of cancerous cells in mice. It is hypothesized that blocking the binding between PIP3 and Akt1(E17K) could reduce or terminate the growth of cancerous cells derived from the E17K mutation in humans, thus serving as a less invasive and less toxic means of chemotherapy. Previous work has shown that peptide capture agents can be raised against Akt that are epitope targeted and/or inhibitory [5,6,7]. Specifically, a 5mer peptide capture agent (yleaf) has 73 been developed to bind to the E17K mutation of the PH domain of Akt1 [7]. Molecular dynamics (MD) trajectories of the peptide/protein complex are constructed for the anchor with both the mutant and wild type (WT) Akt1 PH domains, and used to calculate the free energy of binding for each system. A.2 PEPTIDE CAPTURE AGENT AGAINST E17K MUTATION OF AKT1 PHD Work in the Heath group has yielded a 5mer anchor peptide that differentiates the E17K mutated Akt1 PHD from the wild type domain [7]. The epitope targeted one-‐ bead-‐one-‐compound screening method used to identify the ligand provides an anchor point for the C-‐terminus location of peptide binding. Additionally, biotin labeling experiments done using a variant of the peptide with an N-‐terminal tosyl-‐biotin labeling arm indicate a general region of N-‐terminal peptide interaction with the target around the residue Y26 [8]. These data were used in choosing the initial configuration of the yleaf-‐tosyl-‐biotin ligand in relation to either the E17K or WT target. A.3 CONSTRUCTION OF E17K AND WT SYSTEMS A.3.1 yleaf-‐tosyl-‐biotin The N-‐terminal tosyl-‐biotin labeling arm was constructed in ChemDraw, then imported into Maestro. The bond lengths were manually doubled, and then the B3LYP/6-‐31G** optimized structure and Mulliken charges were obtained using Jaguar. Unknown bond length, bond angle, and torsion parameters for the tosyl group were 74 optimized using B3LYP/6-‐31G** in Jaguar. The yleaf peptide was grown onto the minimized tosyl structure using (D) amino acids to form the complete ligand. yleaf-‐ tosyl-‐biotin was equilibrated in explicit water at 300 K for 1 ns to obtain the ligand conformation (Figure A.1) used to construct the ligand/protein complexes. A.3.2 Ligand/Protein Complexes Crystal structures for E17K (2UZR) and wild type (1UNP) Akt1 PH domains were obtained from the Protein Data Bank. ZDOCK was used to determine an initial relative configuration of the ligand shown in Figure A.1 to the E17K PHD. The anchor peptide was initially screened against the 32mer fragment corresponding to residues 1-‐32 of the E17K PHD, and ZDOCK predictions were restricted to only include relative conformations that allowed for interaction of the ligand with this fragment. The selected ligand configuration was also used for the wild type PH domain. Figure A.2 shows the top three ZDOCK output structures, and Figure A.3 shows the chosen structure for both E17K and WT systems. 75 Figure A.1. A. ChemBioDraw structure of yleaf-‐tosyl-‐biotin. B. Equilibrated structure of yleaf-‐tosyl-‐biotin used to construct WT and E17K complexes. 76 Figure A.2. Top three ZDOCK predicted conformations of yleaf-‐tosyl-‐biotin/E17K PH domain complex. Interaction was constrained to occur only with the first 32 residues of the protein (highlighted) that represent the fragment against which the peptide ligand was raised. The two amino acids of interest from experimental binding assays, E17K and Y26, are shown as sticks. 77 Figure A.3. Selected ligand conformation complexed with the mutant (blue) and wild type (purple) PH domains. This structure was chosen because the ligand most closely spans the two amino acids of interest (E17 or E17K, and Y26). 78 A.4 MOLECULAR DYNAMICS The systems were neutralized by adding Cl− or Na+ counterions as necessary, and were fully solvated in TIP3P water boxes. Each system was first subjected to a minimization of 10,000 steps. Then the solvent molecules, ligand, and binding fragment were equilibrated for 200 ps at 300 K while the remaining protein coordinates were fixed. The full systems were then minimized for 5,000 steps. Finally, both systems were equilibrated for a total of 5 ns at 300 K. The MD simulations were carried out with the NAMD 2.6 program [9]. Plots of the root mean squared deviation (rmsd) from the initial configuration of the trajectories and restarted trajectories from both systems are shown in Figure A.4. 79 Figure A.4. Rmsd plots for the MD trajectories and restarted trajectories of the E17K and WT complexes. 80 A.5 BINDING ENERGY To obtain a good comparison of free binding energy between the two systems, energy landscapes were constructed for E17K and WT by plotting the free energy of binding of conformations representing local macrostates vs. their rmsd from the lowest energy conformation, similar to the approach taken by [10]. If the reference structure is truly the lowest energy, the landscape will be smoothly funnel shaped. If there is an alternative lowest energy structure, the landscape will be non-‐funnel shaped. Figure A.5 shows examples of this type of landscape, where each dot represents one of 200,000–400,000 independent Rosetta ab initio structure prediction simulations [10]. For the purpose of this work, the rmsd of each trajectory is like a partition function, where each point represents a microstate. Plateaus in the rmsd represent macrostates, and given an infinite amount of simulation time, the system will spend a Boltzmann weighted percentage of time in each macrostate. Binding energies were calculated for a number of 100 ps plateaus in each trajectory using a Generalized Born implicit solvent model. After identifying the conformation with the lowest calculated binding energy REF, the energies of the other conformations were plotted against the rmsd of those conformations relative to REF. The energy vs. rmsd plot for the E17K and WT systems are shown in Figures A.6 and A.7, and the structure of the chosen reference structures are shown in Figure A.8. Significantly fewer data points were able to obtained for the E17K and WT systems than the ab initio structure prediction simulations, but the plots suggest the funnel shape described by [10], indicating that the correct lowest energy reference structure had been selected. 81 Figure A.5. Energy landscapes obtained from Rosetta ab initio structure prediction simulations on Rosetta@home. Red points represent the lowest-‐energy structures obtained in independent Monte Carlo structure prediction trajectories starting from an extended chain for each sequence; the y axis shows the Rosetta all-‐atom energy and the x axis shows the Cα root mean squared deviation from the design model. Green points represent the lowest-‐energy structures obtained in trajectories starting from the design model. The bottom figure shows a landscape that is funnel shaped, and the top shows one that is non funnel shaped. Adapted from [10]. 82 Figure A.6. Energy landscape for E17K complex. The y-‐axis is the Generalized Born calculation of binding energy, and the x-‐axis is the rmsd relative to the lowest energy structure. Each point is a plateau in the trajectory rmsd, representing a macrostate of the system. The binding energy of the reference state calculated without including entropy contribution is -‐67.95 kcal/mol. 83 Figure A.7. Energy landscape for WT complex. The y-‐axis is the Generalized Born calculation of binding energy, and the x-‐axis is the rmsd relative to the lowest energy structure. Each point is a plateau in the trajectory rmsd, representing a macrostate of the system. The binding energy of the reference state calculated without including entropy contribution is -‐65.28 kcal/mol. 84 Figure A.8. Structures of the E17K and WT reference states. There are distinct differences in the conformation of the peptide around the E17K mutation cite between the two systems. 85 A.6 CONCLUSION This appendix explored the use of MD calculations to understand the selectivity of the peptide ligand yleaf-‐tosyl-‐biotin for the E17K mutant of Akt1 PH domain over the wild type domain. The structure of the ligand was constructed and minimized, then complexed with both the E17K and wild type structures of Akt1 PHD. Trajectories were run, and the energies of macrostates of the system were calculated and plotted against their rmsd relative to a reference state to determine if the lowest energy state had been identified. The binding energy of the E17K reference state was found to be 2.67 kcal/mol lower than that of the WT reference state. This corresponds to a factor of ~90 difference in Kd, which is within the error of experimental measurements. A.7 ACKNOWLEDGEMENTS This work was done in collaboration with William Goddard III, Kaycie Deyle, and Blake Farrow. 86 A.8 REFERENCES 1. Vivanco I, Sawyers CL (2002) The phosphatidylinositol 3-‐kinase–AKT pathway in human cancer. Nature Reviews Cancer 2: 489-‐501. 2. Altomare DA, Testa JR (2005) Perturbations of the AKT signaling pathway in human cancer. Oncogene 24: 7455-‐7464. 3. Rebecchi M, Scarlata S (1998) Pleckstrin homology domains: a common fold with diverse functions. Annual review of biophysics and biomolecular structure 27: 503-‐528. 4. Carpten JD, Faber AL, Horn C, Donoho GP, Briggs SL, et al. (2007) A transforming mutation in the pleckstrin homology domain of AKT1 in cancer. Nature 448: 439-‐444. 5. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in Situ Click Chemistry Assembles a Branched Capture Agent and Allosteric Inhibitor for Akt1. Journal of the American Chemical Society 133: 18280-‐18288. 6. Nag A, Das S, Yu MB, Deyle KM, Millward SW, Heath JR (2013) A Chemical Epitope-‐Targeting Strategy for Protein Capture Agents: The Serine 474 Epitope of the Kinase Akt2. Angewandte Chemie International Edition In press. 7. Deyle K et al., Submitted. 8. Tamura T, Tsukiji S, Hamachi I (2012) Native FKBP12 Engineering by Ligand-‐Directed Tosyl Chemistry: Labeling Properties and Application to Photo-‐Cross-‐Linking of Protein Complexes in Vitro and in Living Cells. Journal of the American Chemical Society 134: 2216-‐2226. 9. Phillips JC, Braun R, Wang W, Gumbart J, Tajkhorshid E, et al. (2005) Scalable molecular dynamics with NAMD. Journal of computational chemistry 26: 1781-‐1802. 10. Koga N, Tatsumi-‐Koga R, Liu G, Xiao R, Acton TB, et al. (2012) Principles for designing ideal protein structures. Nature 491: 222-‐227. 87 Appendix B Peptide-‐Based General Antibody Detection Agent 88 B.1 INTRODUCTION Detecting the immune response to an infectious agent can provide useful in vitro diagnostic information. Assays of this nature require a general antibody detection agent, often a conjugated anti-‐Fc antibody or protein substrate, i.e., Protein A. However, most biological reagents have inherent variability, which can adversely influence the performance of the diagnostic test. Peptide capture agents have been shown to be highly robust and inexpensive to produce, and to have similar affinities and selectivities as antibodies. DeLano et al. reported the discovery of a cyclic 13-‐amino acid peptide (Fc-‐III) that bound the “consensus region” of IgG Fc by phage display affinity screening [1]. We explored possible improvements to the Fc-‐III peptide in terms of affinity and selectivity by employing in situ click screening to obtain a biligand, and in terms of stability by replacing the disulfide bridge of Fc-‐III, with the goal of using the modified peptide as a detection agent in diagnostic assays. B.2 WELLS PEPTIDE Using a cyclic phage display library of the form XiCXjCXk (where C is cysteine, X is a random amino acid, and i + j + k = 18), Delano et al. found the 13-‐residue Fc binding sequence Fc-‐III (DCAWHLGELVWCT) [1,2]. When synthesized, this peptide was found to inhibit the binding of Protein A Z-‐domain to Fc with a Ki of 25 nM, which is 200 fold greater affinity than an earlier, longer consensus sequence found by the group. The increased affinity with shorter length implied that the final sequence very efficiently 89 interacts with the Fc binding pocket. We added a propargyl glycine (Pra) residue at the N-‐terminus of Fc-‐III and tested the ability of this modified cyclic peptide to bind IgG Fc. The anchor peptide, monikered Wells peptide, was also C-‐terminally tagged with a polyethylene glycol (PEG) oligomer bridge and a biotin label. B.3 BILIGAND SCREEN The in situ click screen is illustrated in Scheme 2.1. The target IgG Fc is incubated with an excess of the selected anchor peptide and a large OBOC library at 4oC for 6 hrs (see Materials and Methods). The OBOC library is synthesized on TentaGel resin (Rapp Polymere, Tuebingen, Germany), and is a comprehensive library of 5-‐mers with a 6th amino acid at the C-‐terminus presenting an azide functionality. The OBOC library is comprised of non-‐natural (D) stereoisomers of the natural amino acids, excluding cysteine and methionine. The in situ screen is designed to identify a secondary (2o) peptide that, when coupled to the anchor, forms a biligand with increased selectivity and/or affinity for the target IgG Fc. In Step 1 of the screen, the OBOC library is cleared of beads that exhibit non-‐ specific binding to Wells peptide and alkaline phosphatase-‐conjugated streptavidin (SA-‐ AP), which is used as a detection reagent in the next step. Step 2 is called a product screen, and is unique to sequential in situ click screens [3]. This screen detects the presence of in situ clicked reaction products, which are those hit beads containing the triazole-‐linked anchor peptide. A complete list of screen hits is given in Table B.1. 90 Scheme B.1. Screening strategy for selecting capture agents against human IgG Fc. The flow chart represents the use of the cyclic Wells peptide as an anchor ligand for in situ click screening against a large OBOC azide-‐presenting peptide library. 91 Table B.1. Biligand screen results. List of pentapeptide “hits” from OBOC biligand screens performed with Wells peptide and human IgG Fc. The selected secondary ligands corresponding to WA and WG are shown in blue and red respectively (shown in Figure B.1). 92 B.4 1,4 TRIAZOLE LINKED BILIGAND CHARACTERIZATION The performance of these biligands was then characterized using an in-‐house developed an on-‐bead ELISA assay (data not shown). Sandwich ELISAs were used for affinity estimations of the two best performing biligands (Figure B.1) from the on-‐bead experiments. The affinity increase of the biligands over Wells is marginal. Hill fits to the ELISA data give Kd values of 18 nM for WA, 21 nM for WG, and 25 nM for Wells (Figure B.2). B.5 1,5 TRIAZOLE LINKED BILIGAND CHARACTERIZATION The in situ click process does not yield a stereospecific triazole product, and it is impossible to measure which click product variant the protein “catalyzed” during a screen. Therefore it can be useful to test the 1,5 triazole linked versions of potential biligands. Here, the biligand WA was synthesized using a 1,5 triazole small molecule linker and its performance compared to that of Wells by sandwich ELISA (Figures B.2 -‐ B.3). The difference in affinity between the 1,5 triazole linked biligand and Wells was negligible. 93 Figure B.1. Structures of peptide biligands identified in screen against IgG Fc. Acetylene-‐ presenting anchor peptides (black) were derived from Fc-‐III. Wells-‐ahriG (WA) and Wells-‐Gehni (WG) were evolved from the original peptide appended with Pra at the N-‐ terminus, and secondary ligand branches (colored) were identified from the in situ click screen of a 5-‐mer OBOC library presenting an azide functionality. 94 Figure B.2. Apparent affinity of Wells and biligands directed against Fc as determined by sandwich ELISA. Lines represent 1st order Hill fits to the affinity data. Wells: Kd = 25 nM, WA: Kd = 18 nM, WG: Kd = 21 nM. 95 Figure B.3. Structure of 1,5 triazole linked biligand. Biligand peptide WA was synthesized with a 1,5 triazole linker. 96 Figure B.4. Apparent affinity of Wells and 1,5 triazole linked WA directed against Fc as determined by sandwich ELISA. Lines represent 1st order Hill fits to the affinity data. 97 B.6 CLICK CYCLIZED WELLS To enhance the stability of the Wells anchor peptide, we attempted to replace the disulphide bridge with a Cu(I) catalyzed click triazole. (D) or (L) azidoalanine (Az1) was substituted for Cys2, and Pra was substituted for Cys12, then the peptide was cyclized using Cu(I) catalyzed click chemistry (Figure B.5). Single point ELISAs showed that this approach to cyclization removed all affinity to IgG Fc (Figure B.6). This could be due to the change in confirmation of the peptide cycle when constrained by an alternative cyclization moiety. The use of both (D) and (L) Az1 was an attempt to circumvent that, but was unsuccessful. 98 Figure B.5. Structure of Cu(I) click cyclized Wells peptide. The disulfide bond in Wells peptide was replaced with a Cu(I) catalyzed triazole formed from substitutions of Az1 and Pra for Cys2 and Cys12 respectively. 99 Figure B.6. Performance of click cylized Wells peptide variants compared to Wells peptide tested by sandwich ELISA. Fc (500 nM) was spiked into TBS, and captured antibody was detected by peroxidase-‐conjugated anti-‐human IgG antibody. 100 B.7 MATERIALS AND METHODS B.7.1 Anchor Synthesis The anchor Wells peptide was synthesized C → N on 200-‐300 mgs of Biotin NovaTag resin (EMD Biosciences, San Diego, CA) with a Titan peptide synthesizer using standard Fmoc chemistry [4]. The synthesis used Fmoc and side chain protected L amino acids (Aaptec, Louisville, KY), Fmoc protected propargylglycine (Pra) (Aaptec, Louisville, KY) and PEG5 (EMD Millipore, Germany). The finished peptide was side-‐chain deprotected and cleaved from resin in a mixture of 95% v/v trifluoroacetic acid (TFA) (Sigma-‐Aldrich, St. Louis, MO), 5% v/v dH2O, and 5% triethylsilane (TES) (Sigma-‐Aldrich, St. Louis, MO), and precipitated in diethyl either. Cu(Phen)3 stock solution was prepared by mixing 150 mM Cu(II) sulfate pentahydrate (Sigma-‐Aldrich, St. Louis, MO) and 500 mM 1,10-‐phenanthroline (Sigma-‐Aldrich, St. Louis, MO) in 4:1 H2O:Ethanol. The ether-‐ precipitated ligand was cyclized by stirring the cleaved peptides dissolved in ~5 mL acetonitrile (ACN) overnight with 1 mL of the Cu(Phen)3 stock solution. The cyclization mixture was lyophilized, and the final peptide product was purified using reverse phase (RP) HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column (Phenomenex, Torrance, CA). B.7.2 OBOC Screen The screen used naïve one-‐bead-‐one-‐compound (OBOC) D-‐pentapeptide libraries on TentaGel resin (TG) (Rapp Polymere, Tuebingen, Germany) of the form NH2-‐ 101 X1X2X3X4X5-‐Az4-‐TG, excluding cysteine and methionine, where Az4 is L-‐azidolycine. Libraries were synthesized on a Titan peptide synthesizer (Aaptec, Louisville, KY) using a split/mix method, and couplings were done using standard Fmoc SPPS chemistry in N-‐ methylpyrrolidone (NMP). Screening against Fc was carried out in two main steps: pre-‐clear, and product screen. Step 1 preclear: 200 mg of OBOC library was blocked in 5% w/v dry milk in Tris buffered saline (TBS) (25 mM Tris-‐HCl, 150 mM NaCl, 10 mM MgCl2, pH 7.6) for 2 hr at room temperature, and then incubated with 1:10,000 phosphatase-‐conjugated streptavidin (SA-‐AP, Promega, Madison, WI) and 25 µM Wells peptide in 0.5% w/v dry milk in TBS for 1 hr. The beads were washed in high salt buffer (25 mM Tris-‐HCl, 750 mM NaCl, 10 mM MgCl2, pH 7.6) for 1 hr, and then developed with BCIP/NBT (Promega, Madison, WI) in BCIP buffer (100 mM Tris-‐HCl, 150 mM NaCl, 1 mM MgCl2, pH 9). After 1 hour, the reaction was quenched with conc. HCl, and the purple beads were discarded. The remaining library was decolorized in NMP, dried with methanol (MeOH) and dichloromethane (DCM), and then swelled and blocked in 5% w/v dry milk in TBS for 2 hr at room temperature. Step 2 product screen: 4 mL of a solution containing 25 µM A21 and 23 nM Fc in 0.5% w/v dry milk in TBS was added to the precleared OBOC library, and the in situ click reaction was allowed to proceed for 6 hours at 4°C. Click solution was drained and the beads were washed for 1 hr in high salt buffer. The beads were incubated for 1 hr at room temperature with 1:10,000 SA-‐AP in 0.5% w/v dry milk in TBS, washed for 1 hr with high salt buffer, and then developed with BCIP/NBT in BCIP buffer for 1 hour. The reaction was quenched with conc. HCl, and the purple beads 102 were retained and sequenced using a Procise CLC Edman Degradation Protein Sequencing System (Applied Biosystems, Kingston, RI). B.7.3 Biligand Synthesis 1,4 Triazole linked WA and WG The secondary ligands including C-‐terminal Az4 for WA and WG were synthesized C → N on 200-‐300 mgs of Rink amide resin with a Titan peptide synthesizer using standard Fmoc chemistry. The synthesis used Fmoc and side chain protected D amino acids and L-‐Az4 (Aaptec, Louisville, KY). The ligands were side-‐chain deprotected and cleaved from resin in 95:5:5 TFA:dH2O:TES, then precipitated in diethyl either and purified using RP-‐HPLC on a Kinetex 5µ XB-‐C18 250 x 4.6 mm prep column. The purified secondary ligands were clicked to cyclized Wells peptide in solution using 2x molar excess of secondary, 10x Cu(I) iodide, 50x L-‐ascorbic acid, and 4x Tris[(1-‐benzyl-‐1H-‐1,2,3-‐ triazol-‐4-‐yl)methyl]amine (TBTA) (Sigma-‐Aldrich, St. Louis, MO) in 5:1 NMP:dH2O and stirring overnight at room temperature. The clicked biligand products were purified from the reaction mixture by RP-‐HPLC on a C18 semi prep column. 1,5 Triazole linked WA The anchor peptide was synthesized C → N omitting Pra on 200-‐300 mgs of Biotin NovaTag resin (EMD Biosciences, San Diego, CA) with a Titan peptide synthesizer using standard Fmoc chemistry [4]. A pre-‐clicked Fmoc protected 1,5 triazole Pra-‐Az4 small molecule linker (InDi Molecular) was amide coupled to the N-‐terminal Asp of Wells 103 peptide, which after Fmoc deprotection served as the starting point for the synthesis of the secondary ligand. The finished peptide was side-‐chain deprotected and cleaved from resin in a mixture of 95% v/v trifluoroacetic acid (TFA) (Sigma-‐Aldrich, St. Louis, MO), 5% v/v dH2O, and 5% triethylsilane (TES) (Sigma-‐Aldrich, St. Louis, MO), and precipitated in diethyl either. Cu(Phen)3 stock solution was prepared by mixing 150 mM Cu(II) sulfate pentahydrate (Sigma-‐Aldrich, St. Louis, MO) and 500 mM 1,10-‐ phenanthroline (Sigma-‐Aldrich, St. Louis, MO) in 4:1 H2O:Ethanol. The ether-‐ precipitated ligand was cyclized by stirring the cleaved peptides dissolved in ~5 mL acetonitrile (ACN) overnight with 1 mL of the Cu(Phen)3 stock solution. The cyclization mixtures were lyophilized, and the final peptide product was purified using reverse phase (RP) HPLC on a C18 prep column (Phenomenex, Torrance, CA). B.7.4 Click Cyclized Wells Synthesis The anchor Wells peptide was synthesized C → N, including N-‐terminal PEG and biotin, on 200-‐300 mgs of Rink amid resin (AnaSpec, Fremont, CA) with a Titan peptide synthesizer using standard Fmoc chemistry [4]. The synthesis used Fmoc and side chain protected L amino acids (Aaptec, Louisville, KY), Fmoc protected propargylglycine (Pra) (Aaptec, Louisville, KY), Fmoc protected (D) and (L) azidoalanine (Iris Biotech, Germany), PEG5 (EMD Millipore, Germany), and biotin (Sigma-‐Aldrich, St. Louis, MO). The finished peptide was side-‐chain deprotected and cleaved from resin in a mixture of 95% v/v trifluoroacetic acid (TFA) (Sigma-‐Aldrich, St. Louis, MO), 5% v/v dH2O, and 5% 104 triethylsilane (TES) (Sigma-‐Aldrich, St. Louis, MO), and precipitated in diethyl either. A Cu(I) click cyclization solution was prepared by mixing 10x Cu(I) iodide, 50x L-‐ascorbic acid, and 4x Tris[(1-‐benzyl-‐1H-‐1,2,3-‐triazol-‐4-‐yl)methyl]amine (TBTA) (Sigma-‐Aldrich, St. Louis, MO) in 5:1 NMP:dH2O. The ether-‐precipitated ligands were cyclized by stirring the cleaved peptides dissolved in ~2 mL Cu(I) click solution. The clicked biligand products were purified from the reaction mixture by RP-‐HPLC on a C18 semi prep column. B.7.5 Assays Biligand binding ELISA Biotinylated WA, WG, and Wells peptides were immobilized on streptavidin-‐ coated 96-‐well plates (Thermo Scientific Pierce, Rockford, IL) at a concentration of 400 nM in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the anchor peptides were incubated with varying concentrations of Fc in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with horseradish peroxidase (HRP)-‐conjugated mouse monoclonal antibody to human IgG Fc (Abcam, Cambridge, MA), diluted 1:10,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate (KPL, Gaithersburg, MD), then quenched with 1 M H2SO4 and read at 450 nm. 1,5 triazole linked biligand binding ELISA Biotinylated 1,5 triazole linked WA peptide was immobilized on streptavidin-‐ coated 96-‐well plates (Thermo Scientific Pierce, Rockford, IL) at a concentration of 300 105 nM in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the anchor peptides were incubated with varying concentrations of Fc in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with horseradish peroxidase (HRP)-‐conjugated mouse monoclonal antibody to human IgG Fc (Abcam, Cambridge, MA), diluted 1:10,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate (KPL, Gaithersburg, MD), then quenched with 1 M H2SO4 and read at 450 nm. Click cyclized Wells ELISA Biotinylated Cu(I) click cyclized Wells peptide was immobilized on streptavidin-‐ coated 96-‐well plates (Thermo Scientific Pierce, Rockford, IL) at a concentration of 1 µM in TBS. The plates were blocked with 5% w/v dry milk in TBS, and then the anchor peptides were incubated with 500 nM of Fc in 0.5% w/v dry milk in TBS at room temperature for 1 hr. Bound antibody was probed with horseradish peroxidase (HRP)-‐ conjugated mouse monoclonal antibody to human IgG Fc (Abcam, Cambridge, MA), diluted 1:10,000 in 0.5% w/v dry milk in TBS for 1 hr at room temperature. The colorimetric assay was developed with TMB substrate (KPL, Gaithersburg, MD), then quenched with 1 M H2SO4 and read at 450 nm. 106 B.8 CONCLUSION Described in this appendix is an attempt to improve the previously discovered Fc binding peptide Fc-‐III in terms of affinity, selectivity, and stability. While a number of approaches were taken, the performance of the original Wells anchor peptide was not significantly improved. Preliminary immunoprecipitation data (not shown) indicated possible selectivity enhancement, but was not verified. The lack of improvement of Wells peptide with the addition of a secondary ligand may reflect the increased dwell time that such a strong binder has with its target during a screen; there are less significant constraints on the nature of the secondary ligand that clicks to the anchor in the in situ process due to the relatively constant presence of the anchor ligand at the target binding site. B.9 ACKNOWLEDGEMENTS This work was done in collaboration with Steven Millward, Kaycie M. Deyle, Errika Romero. 107 B.10 REFERENCES 1. DeLano WL, Ultsch MH, Wells JA (2000) Convergent solutions to binding at a protein-‐protein interface. Science 287: 1279-‐1283. 2. Lowman HB, Chen YM, Skelton NJ, Mortensen DL, Tomlinson EE, et al. (1998) Molecular mimics of insulin-‐like growth factor 1 (IGF-‐1) for inhibiting IGF-‐1: IGF-‐binding protein interactions. Biochemistry 37: 8870-‐8878. 3. Millward SW, Henning RK, Kwong GA, Pitram S, Agnew HD, et al. (2011) Iterative in situ click chemistry assembles a branched capture agent and allosteric inhibitor for Akt1. J Am Chem Soc 133: 18280-‐18288. 4. Fields GB, Noble RL (1990) Solid phase peptide synthesis utilizing 9-‐fluorenylmethoxycarbonyl amino acids. International journal of peptide and protein research 35: 161-‐214. 108 Appendix C Algorithm for Peptide Clustering 109 C.1 INTRODUCTION The one-‐bead-‐one-‐compound in situ click screening approach often yields a number of peptide “hits.” Rather than synthesizing and vetting each individual sequence, it can be useful to group the results into clusters of peptides based on physical properties, and then to choose a representative peptide from each cluster to scale up and test. This appendix provides an example clustering algorithm that uses the concept of persistence homology to group peptides by hydrophobicity, isoelectric point, and residue weight. The grouping is amino acid order dependent. C.2 PERSISTENCE CLUSTERING Persistence states that structures that persist at multiple scales are most likely to be real. For example, if an image is sampled by using a small number of points, persistence can be used to determine properties of the original shape. The points are grouped by drawing a disc of radius ε around each point. Each collection of overlapping discs is considered a cluster. These clusters provide one possible interpretation of the original data. To obtain a more complete analysis of the data, ε is continuously varied and the clusters tracked as the discs grow around the points. As discs from one cluster begin to overlap with discs from another cluster, the two initial clusters “die” and a new merged cluster is “born.” The lifetime of a cluster is the time between a cluster’s birth and death. Thus, groups of points that are far from other points will have a long lifetime, which indicates that they are more meaningful clusters. 110 Figure C.1. Discs formed around points with growing radius ε. As the discs around points overlap, new clusters form that include points rooting the newly touching discs. Eventually all the discs will touch, and the cluster containing all points will continue for infinite “time.” Adapted from [1]. 111 C.3 ALGORITHM DESCRIPTION The MatLab function BasicParam takes as input text files that contain a list of sequences and the number n of amino acids per sequence. It then uses hydrophobicity, isoelectric point, and residue weight to construct a point in 3n dimensional for each sequence. To prevent any of the physical parameters from overly contributing to the clustering based on their absolute amounts, they are all scaled by the inverse of their average. Once each sequence is represented by a point, the Euclidian pair wise distances between the points are computed. Initially, each point is its own cluster. Then, the two points with the smallest pair wise distance are merged. The time of death of the initial point clusters, and the time of birth of the new merged cluster, is set to the distance between those two points (see Figure B.2). Clusters containing points with the next least distance continue to iteratively merge until all points in the data set exist as a single cluster. Additionally, a covariance matrix was constructed and diagonalized to find the eigenvectors and eigenvalues that characterize the sequence space. C.4 OUTPUT The function outputs a list of clusters and their associated lifetimes, along with a plot of the input sequences projected onto the top two eigenvectors. The complete list of eigenvectors is also plotted against the associated eigenvalues, and the top four eigenvectors are decomposed into their individual components as charts. The output from clustering the hits obtained in the A22/4B3 screen described in Chapter 2 is shown in Table B.1 and Figures B.3 – B.5. 112 Figure C.2. Illustrative example of how lifetimes of clusters are computed. The lifetimes of clusters “a” and “b” are 3, because they are closer to each other than any other point and that is the distance between them. 3 then becomes the birth time for cluster “ab,” and the lifetime of “ab” is 5 because it dies at time 8, when the cluster “abc” is born. 113 Table C.1. List of clusters and their associated lifetimes calculated for the hits resulting from the A22/4B3 screen. The peptides input to the algorithm were qndwk, ltsry, diksp, eihny, qpidq, kidrv, qspwl, qsdwl, eqtfd, aapsl, kqdtq, kwtkf, tgnek, kysfq, ffftk, and vgeit. 114 Figure C.3. Projection of the peptide “points” onto the top two eigenvectors taken from the diagonalization of the covariance matrix. The clusters containing more than one point with the three longest lifetimes from Table B.1 are circled. 115 Figure C.4. Plot of eigenvectors vs. their associated eigenvalues. This allows users to know how many relevant eigenvectors describe the set of input peptides. 116 Figure C.5. Visual representation of the components of the top four eigenvectors. This shows what physical characteristics are dominant when describing the list of input peptides. The order of the amino acids is preserved during clustering, so Hydrophobicity1 refers to the hydrophobicity of the N-‐terminal amino acid of the peptides. For this set, the top four eigenvectors have large hydrophobicity components, though not all from the same amino acid index. 117 C.5 FUNCTION CODE function BasicParam(filename, pepLength) %Feb 8, 2013 %Physical parameters come from the wwHydrophobicity scale, Sigma PI values, %and resudue weight. %Note: hydrophobicities are at pH 7 except for proline, which was estimated %from an InDi (or AN SD) measurement %Note: the number of traits is hard coded in, so that has to be changed %every time you mess with the parameters. %http://www.sigmaaldrich.com/life-science/metabolomics/learningcenter/amin %o-acid-reference-chart.html Hydr = 1/((41+49+55+31+100+0+8+99+23+97+74+28+50+10+14+5+13+76+97+63)/20); PI = 1/((6+5.07+2.77+3.22+5.48+5.97+7.59+6.02+9.74+5.98+5.74+5.41+6.3+5.56+1 0.76+5.68+5.6+5.96+5.89+5.66)/20); RW = 1/((71.08+103.15+115.09+129.12+147.18+57.05+137.14+113.16+128.18+113.16 +131.2+114.11+97.12+128.13+156.19+87.07+101.11+99.13+186.22+163.18)/20) ; scaling = [Hydr PI RW]; %scaling = 1; %[Hydrophobicity pI ResidueWeight] A = [41 6.00 71.08].*scaling; C = [49 5.07 103.15].*scaling; D = [-55 2.77 115.09].*scaling; E = [-31 3.22 129.12].*scaling; F = [100 5.48 147.18].*scaling; G = [0 5.97 57.05].*scaling; H = [8 7.59 137.14].*scaling; I = [99 6.02 113.16].*scaling; K = [-23 9.74 128.18].*scaling; L = [97 5.98 113.16].*scaling; M = [74 5.74 131.20].*scaling; N = [-28 5.41 114.11].*scaling; P = [50 6.30 97.12].*scaling; Q = [-10 5.56 128.13].*scaling; R = [-14 10.76 156.19].*scaling; S = [-5 5.68 87.07].*scaling; T = [13 5.60 101.11].*scaling; V = [76 5.96 99.13].*scaling; W = [97 5.89 186.22].*scaling; Y = [63 5.66 163.18].*scaling; %Read in sequences from text file function argument, and determine the 118 %numberof sequences sequences = textread(filename,'%s', -1); nseq = length(sequences); %Initialize matrix to hold each sequence vector ntrait = 3; Z=zeros(nseq,pepLength*ntrait); %For each individual sequence assign a vector based on Grantham Values for j = 1:nseq seq = sequences{j}; o = []; for i=1:length(seq) switch seq(i) case 's' o = [o S]; case 'r' o = [o R]; case 'l' o = [o L]; case 'p' o = [o P]; case 't' o = [o T]; case 'a' o = [o A]; case 'v' o = [o V]; case 'g' o = [o G]; case 'i' o = [o I]; case 'f' o = [o F]; case 'y' o = [o Y]; case 'c' o = [o C]; case 'h' o = [o H]; case 'q' o = [o Q]; case 'n' o = [o N]; case 'k' o = [o K]; case 'd' o = [o D]; case 'e' o = [o E]; case 'm' o = [o M]; case 'w' o = [o W]; 119 end end Z(j,:) = o; end %Compute pairwise distances between each "sequence point" given by %a row in the Z matrix and put in array n. Also, create a 2x#pairs matrix %cataloging pair indices to go along with distances. n=[]; Idx=[]; k=1; for i = 1:nseq for j = i+1:nseq n(k) = norm(Z(i,:)-Z(j,:)); Idx(1,k) = i; Idx(2,k) = j; k=k+1; end end %Sort the array of distances from smallest to largest and use returned %permutation to analogously sort pair indices. [N,O] = sort(n); Idx = Idx(:,O); %Create array of parents of "sequence points" p = [1:nseq]; %Create an array of cluster birth tb = zeros(1,nseq); % %Union the points with distance smaller than 200 % for k = 1:length(N) % if(N(k)>= 200) % %sprintf('stopping at distance %f\n', N(k)) % break; % end % %sprintf('Unioning %f and %f with distance %f\n',Idx(1,k),Idx(2,k),N(k)) % p = Union(Idx(1,k),Idx(2,k),p); % end % %'Find' all points to see which ones share a root after Union and put into % %array 'root' % for i = 1:nseq % j = Find(i,p); % root(i) = j; % end % [R,O_r] = sort(root); % sequences(O_r); 120 %Itteratively 'Union' clusters containing the points with the next %least distance sets = {}; ages = []; birth = []; death = []; for i = 1:length(N) %Find roots of the two sets being joined r1 = Find(Idx(1,i),p); r2 = Find(Idx(2,i),p); %Set the current time to the distance between the two merging clusters t = N(i); %Check if the two sequences are already in the same cluster if r1 ~= r2 %Save the sequences in the two merging clusters %disp('====================================='); set1 = MkSet(r1,nseq,p,sequences); set2 = MkSet(r2,nseq,p,sequences); sets(end+1) = {set1}; sets(end+1) = {set2}; %Compute and save the ages of the merging clusters age1 = t - tb(r1); age2 = t - tb(r2); ages = [ages age1]; ages = [ages age2]; %Keep running tally of birth/death times for persistence barcode birth = [birth tb(r1)]; birth = [birth tb(r2)]; death = [birth t]; death = [birth t]; %Union the two clusters p = Union(Idx(1,i),Idx(2,i),p); %Save birth time for new cluster tb(r2) = t; end end birth; death; %Sort ages from largest to smallest, accordingly rearrainge sets of %sequence names 121 [Ages,O_A] = sort(ages, 'descend'); Sets = sets(O_A); %Read out vectors of ages and sets %disp('--------------------------------------------------------') %DispSet(sets, ages); % disp('--------------------------------------------------------') DispSet(Sets, Ages); disp('--------------------------------------------------------') %Plot "persistence barcode" to see if any clustering is actually present %Do PCA analysis to see if there exist any major axes along which the %sequence data falls. m = size(Z); %Center data center = zeros(1,m(2)); for i = 1:nseq center = center + Z(i,:); end center = center/nseq; Zc = zeros(m); for i = 1:nseq Zc(i,:) = Z(i,:) - center; end %Create covarience matrix covar = zeros(m(2), m(2)); CoVar = zeros(m(2), m(2)); for i = 1:nseq covar = Zc(i,:)'*Zc(i,:); CoVar = CoVar + covar; end %Diagonalize covarience matrix to find largest eigenvalues and %corresponding eigenvectors [vec,val] = eig(CoVar); CoVar*vec - vec*val; %Plot eigenvalues and display eigenvectors associated with two largest %values. Also label eigenvectors with next two largest eigenvalues to %decompose into pie charts to examine components. Diagval = diag(val); [Val,O_val] = sort(Diagval, 'descend'); Vec = vec(:,O_val); plot(Val, '.k', 'MarkerSize',10); eig1 = Vec(:,1); eig2 = Vec(:,2); eig3 = Vec(:,3); eig4 = Vec(:,4); %Project sequence points onto the two largest eigenvectors and plot Projection(Vec(:,1), Vec(:,2), Zc, sequences, nseq); 122 %Create pie chart to examine the components of the four eigenvectors with %the largest eigenvalues labels = {'Hydrophobicity_1', 'pI_1', 'RW_1','Hydrophobicity_2', 'pI_2',... 'RW_2', 'Hydrophobicity_3', 'pI_3', 'RW_3', 'Hydrophobicity_4', 'pI_4',... 'RW_4','Hydrophobicity_5', 'pI_5', 'RW_5'}; figure pie(abs(eig1), labels) title('Eigenvector 1') figure pie(abs(eig2), labels) title('Eigenvector 2') figure pie(abs(eig3), labels) title('Eigenvector 3') figure pie(abs(eig4), labels) title('Eigenvector 4') end %-----------------------------------------------------------------------% function y = Find(x,p) y = x; while(p(y)~=y) y = p(y); end end function p = Union(x1,x2,p) r1 = Find(x1,p); r2 = Find(x2,p); p(r1) = r2; end function Set = MkSet(r1,nseq,p,seq) Set = {}; for i = 1:nseq r = Find(i,p); if r == r1 Set(end+1) = seq(i); end end end function DispSet(Sets, Ages) 123 for i = 1:length(Sets) Si = Sets{i}; str = sprintf('%f\t',Ages(i)); for j=1:length(Si) str = sprintf('%s %s\t', str, char(Si{j})); end disp(str); end end function Projection(eig1, eig2, Zc, sequences, nseq) x = []; y = []; for i = 1:nseq x = [x Zc(i,:)*eig1]; y = [y Zc(i,:)*eig2]; end figure; plot(x, y, '*') dx = .2; dy = 0.2; text(x+dx, y+dy, sequences); end C.6 CONCLUSION This appendix describes a clustering algorithm designed to help characterize groups of peptide hits from in situ click screens. The objective is to provide guidance about which peptides are useful to scale up and assay, and give information about the physical characteristics that predominately describe the differences between input peptides. C.7 ACKNOWLEGDEMENTS I would like to thank Keenan Crane for introducing me to the concept of persistent homology, and for his incredibly helpful guidance while developing this algorithm. 124 C.8 REFERENCES 1. Ghrist R (2008) Barcodes: the persistent topology of data. Bulletin of the American Mathematical Society 45: 61-‐75. 125 Appendix D A cocktail of thermally stable, chemically synthesized capture agents for the efficient detection of anti-‐gp41 antibodies from human sera Reproduced from PloS one 2013, 8, e76224 126 127 128 129 130 131 Appendix E A Selective 15N-‐to-‐1H Polarization Transfer Sequence for More Sensitive Detection of 15N-‐Choline Reproduced from Journal of Magnetic Resonance 2010, 205, 125-‐129 132 133 134 135 136