University of Southern California Dissertations and Theses

Generation and characterization of peptide theranostics by mRNA display

(USC Thesis Other)

Generation and characterization of peptide theranostics by mRNA display

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Generation and Characterization of Peptide Theranostics by mRNA Display Farzad Jalali-Yazdi A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (CHEMICAL ENGINEERING) August 2015 2 This work is dedicated to my wife Sarah. Here’s to many more adventures. 3 Acknowledgements: First and foremost, I acknowledge and thank Professor Richard Roberts’ contribution to this work. None of what follows would be possible without his help, vision and effort. I feel extremely lucky to have Rich as my mentor. I have come to realize how deeply he cares for his students. After we initially spoke about me entering the PhD program at USC, he arranged for me to speak with four other professors whose research aligned with my interests, to make sure I did not feel pressured into only joining his lab. His thoughtfulness and support was a constant throughout my career here at USC. Initially, the transition from the industry back into academia was difficult for me. It had been some years since I had taken any academic courses. I found balancing course work and lab work difficult, and Rich allowed me to focus on my class work for the first year. This helped me excel in my courses and reduced the amount of stress I was under at the time. He has always relished at a chance to teach and educate me when my lack of chemistry knowledge became clear, and has never made me feel insignificant, inadequate or incapable. Rich manages his lab in a unique way by allowing his students to work very independently. Because of the freedoms that Rich provides to his graduate students in choosing their own projects I was able to work on problems that interested me the most. This led me to initiate two projects (chapters 6 and 7 of this work) with my former employer (BioScale Inc.), and devise and work on the project in chapter 8. Without Rich’s focus on innovation and independence, the second half of this thesis would not be possible. I thank Professor Terry Takahashi, who walked me through every step of mRNA display (generally more than once) and taught me all I know about protein cloning, expression, purification, validation, and interactions. Without Terry I would be completely lost, and this work would have taken at least twice as long to accomplish. I am also indebted to Terry due to 4 four chapters of this work (chapters 5-8) only being possible due to his selection for ultra-high affinity peptide ligands against Bcl-x L . Even to this day, I tend to discuss any ideas I have with him, and rely greatly on his help in all of my projects. I thank my dissertation committee members: Professor Noah Malmstadt, Professor Pin Wang, and Professor Scott Fraser. I feel privileged to have such outstanding scientists on my committee, and thank them for their time and advice. I also thank my collaborators Professor Pin Wang and Si Li who performed the HDM2 cell experiments (chapter 4), and Professor Ren Run, Anders Olson, and Hangfei Qi who performed the HCV core protein cell experiments (chapter 2). Without their work this dissertation would be significantly lacking. I thank Jasmine Corbin, Cynthia Wang, Makana Krulce, and Karl Heyes, my undergraduate research assistants, for their help in performing many rounds of selection, countless radiolabeled binding assays and ELISAs. Jasmine started as a technician in the lab; however she was driven, ambitious, and interested in research. She helped me initially in the HCV project by performing over 100 binding assays, which allowed us to really understand the mechanisms of action for our sequences. After her work in the HCV project, she started her own project under my supervision and worked as hard as any graduate student in the lab. Her efforts paid off in a large way and she was able to identify molecules with picomolar affinity to streptavidin (not part of this work). She was also involved in training Cynthia Wang, my second undergraduate research assistant. Jasmine’s work and dedication got her accepted into a PhD program at UC Davis, where she is currently happily performing research on plant cells, going to conferences, and spending summers in Japan! I recruited Cynthia in my third year when I realized that Jasmine would soon leave to pursue her academic goals. Cynthia initially helped me characterize a more efficient way to perform and purify ligations (a step in mRNA display). After her successful work, she started 5 her own independent projects, and is currently hard at work on writing and publishing the results of her newest project, understanding the rules of protease digestion of peptides (not included in this work). Karl and Makana are the newest recruits to the Roberts lab. I recruited them in my fifth year, when I was looking for a single undergraduate research assistant. I ended up recruiting both of them because it was impossible to decide between the two of them. I am glad that they both joined the lab. Karl’s project on using mRNA display to generate diagnostic reagents against HDM2 protein (not included in this work) has enabled us to produce the first antibody free ELISA assay, and won him the runner up prize for undergraduate research symposium at USC in the physical sciences and engineering section. Makana’s project on optimizing the un- translated region upstream of the start codon can help us make mRNA display a much more effective method of ligand generation, as well as making in vitro translation of peptides and proteins much more efficient. Recruiting, mentoring and supervising the undergraduate research assistants for the Roberts lab has truly been one of my most cherished experiences as a graduate student, and their work has made the thesis before you possible. I thank Brett Masters, Martin Latterich, Matthew Dickerson, Michael Miller, and Ed O'Brien (all at BioScale) for discussions of data and experimental design; and BioScale Inc. for providing access to the ViBE BioAnalyzer and the universal detection cartridges. Their advice and generosity made chapters 6 and 7 of this work possible. I also thank Dr. Stephen Fiacco and Amanda Hardy for making the HDM2 project (chapter 4) possible by applying for the Ideas Empowered program at USC. Their hard work and dedication lead to the project being funded by the program. Amanda’s help throughout the project especially with high throughput DNA sequencing also made it possible for us to identify our candidate peptides for testing. I thank Lan Huong Lai for her help in the high throughput affinity characterization project (chapter 8). While 6 it took us a few days simply to plan the experiment and two tries to get it right, her help in performing the experiments, PCR amplifying the pools, and discussion of the results were invaluable. I would also like to thank the rest of the members of the Roberts lab (Garret Gross, Shannon Howell, Mehmet Cetin, Aaron Nichols, John Mac, and William Evenson) who were always ready to render assistance and advice on matters related or unrelated to bench work. Our relaxing evenings spent drinking beer at various locations while discussing work or life made the lab environment a place that I wanted to be and spend time with my friends. Finally, I would like to thank my parents for their unwavering support and love throughout my life. I always knew that I wanted to perform research, and their never ending encouragement and support made it possible. They encouraged me to believe in myself, and believe that I could compete at the highest academic levels. They encouraged me to apply to the MITES program at MIT which opened many opportunities for me, including my acceptance to MIT. After MIT, while I was working at BioScale Inc., they encouraged me to apply for graduate studies, and have supported me while I have been at USC both emotionally and financially. 7 Table of Content: Chapter 1: Introduction ………………….………………………………………………… P10 Chapter 2: Generating Peptide Therapeutics against the HCV Core Nucleocapsid Protein 2.1 Introduction…………………………………………………………………… P12 2.2 Materials and Methods………………………………………………………... P15 2.3 Results and Discussion……………………………………………………….. P24 2.4 Conclusions…………………………………………………………………… P37 2.5 Figures………………………………………………………………………… P38 2.6 Appendix……………………………………………………………………… P56 Chapter 3: Serum Stable Natural Peptides Designed by mRNA Display 3.1 Introduction…………………………………………………………………… P62 3.2 Materials and Methods………………………………………………………... P64 3.3 Results and Discussion………………………………………………………... P67 3.4 Conclusions…………………………………………………………………… P73 3.5 Figures………………………………………………………………………… P74 3.6 Supplementary Figures.……………………………………………………….. P78 Chapter 4: Cell-permeable and Biologically Stable Peptide Therapeutics Aimed at Disrupting the HDM2-p53 Interaction 4.1 Introduction…………………………………………………………………… P80 4.2 Materials and Methods………………………………………………………... P84 4.3 Results and Discussion………………………………………………………... P90 4.4 Conclusions…………………………………………………………………… P95 4.5 Figures………………………………………………………………………… P96 4.6 Appendix……………………………………………………………………… P103 8 Chapter 5: Generating Ultrahigh Affinity Peptide Ligands Using a Fragment-based Approach 5.1 Introduction…………………………………………………………………… P105 5.2 Materials and Methods………………………………………………………... P107 5.3 Results and Discussion………………………………………………………... P122 5.4 Conclusions…………………………………………………………………… P129 5.5 Figures………………………………………………………………………… P130 5.6 Supplementary Figures.……………………………………………………….. P136 Chapter 6: Analysis of Proteins using Peptide Immunoreagents by an Acoustic Resonant Sensor 6.1 Introduction…………………………………………………………………… P140 6.2 Materials and Methods………………………………………………………... P143 6.3 Results and Discussion………………………………………………………... P149 6.4 Conclusions…………………………………………………………………… P159 6.5 Figures………………………………………………………………………… P161 6.6 Appendix……………………………………………………………………… P171 Chapter 7: A General, Label-Free Method for Determining K d and Ligand Concentration Simultaneously 7.1 Introduction…………………………………………………………………… P172 7.2 Materials and Methods………………………………………………………... P174 7.3 Results and Discussion………………………………………………………... P181 7.4 Conclusions…………………………………………………………………… P191 7.5 Figures………………………………………………………………………… P192 7.6 Supplementary Figures.……………………………………………………….. P200 9 Chapter 8: High-throughput Binding Kinetics Measurement by mRNA Display and Next- Generation Sequencing 8.1 Introduction…………………………………………………………………… P208 8.2 Materials and Methods………………………………………………………... P210 8.3 Results and Discussion………………………………………………………... P218 8.4 Conclusions…………………………………………………………………… P221 8.5 Figures………………………………………………………………………… P222 8.6 Supplementary Figures.……………………………………………………….. P225 Chapter 9: Conclusions…………………………………………………………………….. P227 References…………………………………………………………………………………. P229 10 Chapter 1: Introduction mRNA display, as designed and implemented by the Roberts and Szostak, 1, 2 is a powerful approach to design peptides and proteins via in vitro selection. (detailed protocol in Takahashi and Roberts, 2003). 3 This approach enables peptide and protein design by directed molecular evolution, using libraries with more than 10 13 independent sequences, in a strictly monomeric format, with precise control over both binding conditions and library sequence content. Additionally, libraries may contain unnatural amino acids, 4, 5, 6 covalent modifications including pharmacophores 7 and covalent cyclization (N-terminal to sidechain). 8 More recently, our lab has tried other post translational steps such as protease digestion 9 or heat treatment to increase the stringency of selection, and find ligands with the desired properties. In mRNA display, mRNA molecules bearing a pendant 3’ puromycin are translated in vitro to generate covalent mRNA-protein fusions. In order to perform directed protein evolution experiments, synthesis of the mRNA-protein fusions is incorporated into an in vitro genetic cycle. An initial library is created in the form of double stranded linear DNA using PCR. Diversity in the library is generated using randomized DNA cassettes 4, 7, 8, 10, 11, 12 doped cassettes, 13, 14 mutagenic PCR, 15, 16 or a combination of these methods. In vitro translation results in protein sequences covalently attached to their own mRNA. 10 These mRNA-fusion molecules will then act as a molecular Rosetta Stones, and can be recovered after affinity maturation and amplified. Traditionally, mRNA display has been used in order to generate peptide and protein ligands which can be used as therapeutic or diagnostic reagents. 17, 18 More recently, through combining mRNA display with high throughput DNA sequencing, mRNA display has also become a tool to analyze ligand characteristics or even protein structures. 19 11 We explore the effectiveness of peptides as therapeutic reagents in chapters 2-4 of this work. In Chapter 2 we show successful inhibition of hepatitis C virus particle production in infected human hepatocellular carcinoma cells. In chapter 3 we construct peptides composed entirely of natural amino acids which resist proteases and peptidases. Subsequently, in chapter 4, we develop cell permeable, biologically stable peptides that inhibit the growth of colorectal cancer cells in situ. In chapter 5, we use a fragment-based design approach to select ultrahigh affinity peptide reagents. Using this method, we select picomolar affinity peptides that outperform most clinical antibodies in terms of affinity and specificity, capable of target recognition in complex matrices. Chapters 6 and 7 explore methods to best identify and characterize the highest affinity ligands through either enzyme linked immunosorbent assays (ELISAs) or using an ultrasensitive acoustic resonant sensor. Chapter 8 describes the combination of mRNA display and next generation DNA sequencing to create a powerful tool capable of high-throughput analysis of ligand affinities. Using this technique, we obtain the dissociation constant for tens of thousands of ligands against Bcl-x L and identify the highest affinity peptide-protein interaction ever known. 12 Chapter 2: Generating Peptide Therapeutics against the HCV Core Nucleocapsid Protein 2.1 Introduction Hepatitis C virus (HCV) is a major human health concern with an estimated 170 million people infected with HCV worldwide. 20 In comparison, 40 million people are infected with HIV in the world. HCV causes liver diseases including chronic hepatitis, cirrhosis, and hepatocellular carcinoma and is responsible for approximately 22% of the new cases of liver cancer each year. 20 HCV has a positive-sense, single-stranded RNA genome of about 9.6 kilobases (kb), which codes for a single polyprotein. The polyprotein is then proteolyzed into 10 mature proteins using both host and HCV proteases. 21 These proteins consist of 3 structural proteins – core (nucleocapsid), E1, and E2 – and 7 nonstructural proteins. There are 6 major genotypes of HCV, each predominantly observed in a geographic locality (Table 1.) The different genotypes of HCV, though are not found to have major biological differences in terms of disease progression and effects, tend to respond very differently to treatment. 22 Until 2011, only approved therapeutics for HCV consisted of 24 or 48 week regiments of pegylated interferon alpha and ribavirin. These treatments had a Sustained Virological Response (SVR) of less than 50%, and were associated with severe side effects such as influenza-like symptoms, alopecia, and suicidal ideations. Since 2011, two generations of anti-HCV drugs have entered the market. The first generation drugs Boceprevir (Victrelis) and Telaprevir (Incivek) inhibit the NS3/4A protease, and can increase SVR to ~65% and ~75% respectively when administered in conjugation with IFN and ribavirin. The second generation of drugs (Sovaldi, Harvoni, Olysio, and Viekira Pak) 13 was approved by the FDA in 2013 and they target a larger array of HCV proteins. Neither the first nor the second generation therapeutics, however, targets the HCV core protein. HCV core protein has been implicated in binding the viral mRNA and transferring it to the virus particle, 23 as well as various other functions such as cell signaling, apoptosis, carcinogenesis, lipid metabolism, and interferon resistance. 21 Amino acids 16-40 of the HCV core protein are highly conserved among many genotypes of the HCV virus (greater than 95% conservation in genotypes 1a, 1b, 2, and 5; greater than 95% conservation with the exception of a Q20M mutation in genotypes 4 and 6. Table 1) 24, 25 and are essential in virus particle production. 26 The goal of this project is to find peptide ligands which can bind to the HCV core protein in vivo, and stop its function in viral particle production. We decided to find peptide binders that specifically target this epitope of the HCV core protein. To do this, we chemically synthesized amino acids 16-40 of the core protein with an N-terminal flexible linker to be used as the target for our mRNA display selection. The main concern with using a peptide segment as the target is that the epitope might have a different conformation in the protein while being unstructured as a peptide. This could lead to ligands which recognize the unstructured peptide, but not function during the course of in vivo experiments. This concern has been somewhat mitigated in our case, due to a number of antibodies derived from human or mouse sera that target this epitope of the HCV core protein and bind to peptides segments of the core protein as well as the truncated protein (amino acids 1- 120). 27, 28 The peptide seems to exhibit a helix-loop-helix structure in solution containing 40% TFE or when bound by the 19D9D6 antibody. 28, 29 The evidence suggests that this peptide segment of the core protein structurally resembles the epitope on the core protein. 14 One major advantage of our proposed approach is that the selected binders not only would be potential therapeutics, but can also be used to study the fundamental HCV biology, the mechanism of viral replication, and virus-host interactions at the scale of viral proteome. In contrast to antibodies which contain disulfide bonds, peptide ligands can be functional in vivo. We can use the binders to determine the functions and cellular locations of the HCV core protein during different stages of the viral life cycle. 15 2.2 Materials and Methods HCV Core Peptide Synthesis: Two versions of HCV core peptide (Table 1) were chemically synthesized using Fmoc solid phase peptide synthesis by a PS3 peptide synthesizer (Protein Technologies Inc.) 30 The first version of the peptide has an N-terminal serine followed by a flexible linker, amino acids 16-40 of HCV core protein, and a C-terminal amide (NH2- SGGSGGNRRPQDVKFPGGGQIVGGVYLLPRR-NH2). The second version of the peptide substitutes an N-terminal biotin in place of the N-terminal serine for immobilization onto avidin beads (biotin-GGSGGNRRPQDVKFPGGGQIVGGVYLLPRR-NH2). 200 μmol scale syntheses were performed and peptides were cleaved from resin and HPLC purified using the same process as described in Chapter 2.2. The mass of the peptides was measured by MALDI-TOF Mass Spectrometry (ABI). For both the N-terminal serine peptide (mass expected: 3,180.7 g/mol, mass found: 3,180.35 g/mol, Figure 2a) and the N-terminal biotin peptide (mass expected: 3,319.8 g/mol, mass observed: 3,319.4 g/mol) the fractions with the correct molecular weight were combined, lyophilized, and dissolved in DMSO. Peptide Immobilization on Beads: 300 nmols of S-terminal HCV core peptide in 50% DMSO (v/v) and 1X PBS was incubated with 15 mg of sodium periodate and protected from light. After 3 minutes at room temperature, the reaction was buffer exchanged using NAP-25 columns (GE Healthcare) to remove sodium periodate from the solution. The oxidized peptide was analyzed using MALDI-TOF Mass Spectrometry to confirm peptide oxidation. Oxidized peptide has a lower molecular weight than the unmodified peptide by 31 Da, which can be seen in Figure 2. The oxidized peptide was incubated with 300 μL of agarose hydrazide beads 16 (CarboLink, Thermo Scientific) or 300 μL of acrylamide hydrazide beads (UltraLink Hydrazide Resin, Thermo Scientific) for 72 hours. 100 μL of 1 M Tris-HCl (pH 8.0) was added to beads to quench the unreacted hydrazide groups on the beads. The beads were washed 3 time with 0.1% Tween 20 in selection buffer (15 mM sodium bicarbonate, 10mM potassium phosphate, 150mM potassium hydroxide, 0.5 mM magnesium chloride, pH adjusted to 7.2 with HCl) and resuspended in complex buffer (100 µg/mL yeast tRNA (Roche) + 1 mg/mL BSA (Equitech- Bio) + 0.1% Tween 20 in selection buffer). To quantitate the amount of HCV core peptide, 30 μL of beads were washed 5 times in 1X selection buffer, and were incubated with 100 μL of 0.1 M HCl for three hours. The 20 μL of 1 M Tris-HCl pH 8.0 was added to the supernatant, and the beads were discarded. The cleaved peptide was then analyzed on the HPLC along with known amounts of periodate-oxidized HCV core peptide and un-modified peptide (Figure 3.) Periodate- oxidized peptide has a longer retention time on a reverse phase C-18 column, using water/acetonitrile gradients to elute the peptide (described in detail in Chapter 2.2). This is due to the fact that the oxidized peptide has a hydrophilic amine group removed from the N-terminal, making it more hydrophobic. The amount of peptide on the beads can be calculated using the area under the curve for the standards and comparing it to the oxidized peptide cleaved from the beads. Library Construction: The anti-sense strand DNA oligonucleotide for the polypeptide library was obtained (synthesized at the Keck Oligo Facility at Yale). The library coded for 10 random amino acids followed by lysine doped at 60% and a constant C-terminal linker (5’- GCCAGATCCGCT111322322322322322322322322322322CATTGTAATTGTAAATAGTAA TTG-3’ Nucleotide 1: 81% T, 5% G, 7% A, 7% C; Nucleotide 2: 20% T, 20% G, 30%A, 30% C; 17 Nucleotide 3: 60% C and 40% G). After normalizing for the different rates of nucleotide coupling, 31 the nucleotide composition is as follows: Nucleotide 1: 85% T, 5% G, 5% A, 5% C; Nucleotide 2: 25% T, 25% G, 25% A, 25% C; Nucleotide 3: 50% G, 50% C. 4.3 pmols of the library (~3E12 diversity) was PCR amplified with a 5’ primer containing a T7 promoter and ΔTMV (Sequence in Appendix A) and a 3’ primer containing the C-terminal flexible linker “SGSGSG” (5’- TCCGCTGCCAGATCCGCT-3’). After PCR amplification, the diversity of the library was estimated to be around 1E12 (the library was roughly 40% extendable) which accounts for ~0.5% of the total theoretical library diversity. mRNA Display Selection: The DNA pool was in vitro transcribed, ligated to F30P (using the DNA splint: 5’-TTTTTTTTTTTNTCCGCTGCCAGA-3’), and Urea-PAGE purified. 4 nmols of ligated mRNA was in vitro translated with supplemented amino acids, while 160 pmols of ligated mRNA was translated using 35 S-labeled methionine and non-labeled amino acids without methionine (described in Chapter 2.2). In vitro translated samples were incubated with Oligo-dT cellulose beads (1 mg/25 μL translation reaction) in dT buffer (50mM Hepes-KOH pH 7.5, 1 M NaCl, 1 mM EDTA, 0.05% Tween20). After one hour of incubation, the beads were washed 7 times or until no significant amount of radiation was eluted from the beads using wash dT buffer. The ligated mRNA molecules were then eluted from the beads using 65 °C water. Half of the radiolabeled samples were added to the non-radiolabeled samples to track the fusion molecules throughout the selection process, while the other half was kept to assess the binding of the pool to HCV core peptide. Solutions of 1 mg/mL DSG in DMF and 4X phosphate buffer (pH = 8.0) were added to sample to achieve the final DSG concentration of 0.25 mg/mL in 1X phosphate buffer (modified protocol from Millward et. al., 2007). 32 Samples were incubated for an hour, 18 and then incubated with Oligo-dT cellulose beads (1 mg/25 μL translation reaction) in IB buffer (50mM Tris-HCl pH 8.0, 1 M NaCl, 0.2% Triton X-100). The Tris present in the IB buffer quenches the un-reacted NHS groups on DSG and inhibits covalent attachment of the binders to their target. dT-cellulose beads were washed 5 times in IB buffer, and the ligated mRNA was eluted from the beads using 65 °C water. Both samples were reverse transcribed. The binding assay samples were incubated with 1 nmol of HCV core peptide immobilized on agarose hydrazide beads or agarose hydrazide beads without target, in assay buffer at 4 °C for 1 hour, and washed 5 times. The radiolabeled counts for the flow-through, all of the washes, and the beads were measured using a scintillation counter. The selection samples were incubated with 4 nmols of immobilized target at 4 °C for 1 hour. The samples were then washed 7 times and eluted with 2 times 100 μL of 0.15% SDS solution. 15 μL of SDS-out solution (Thermo Scientific) was added to the samples and incubated for 1 hour at 4 C. The samples were filtered to eliminate precipitated SDS, diluted to 1 mL in 1X PCR buffer, and PCR amplified to the final concentration of 0.1 μM, using the same primers as pool 0, to regenerate the DNA library. For subsequent rounds, 80 pmols of ligated mRNA of the pool was in vitro translated using non-labeled methionine, and 40 pmols of ligated mRNA of the pool was translated using radiolabeled 35 S-methioine. The selection and binding assays for rounds 1-5 were performed with 300 pmols of HCV core peptide on agarose hydrazide beads. Rounds 6-9 were performed with ~700 pmols of HCV core peptide on acrylamide hydrazide beads. Sequencing: Pools 0, 6, 8 and 9 were cloned using the TOPO cloning kit (Life Technologies) following the manufacturer’s instructions. Pool 9 was also sequenced using Next Gen high- 19 throughput sequencing (illumina). The data was analyzed using in-house developed code using python and bio-python to count the occurrence of each unique sequence in the sequencing data. Radiolabeled Binding Assays: The DNA for sequences 8.02, 8.07, 9.01, 9.08, 9.10, 9.15, and 9.19 were obtained from the TOPO cloning vectors (Appendix B). The plasmids were PCR amplified with the same 5’ and 3’ primers used to generate the affinity enriched pools. Synthetic DNA oligonucleotides were ordered to construct sequences I9.01 and I9.09 (Appendix C). C- terminally and N-terminally modified 9.15 and 9.01 sequences were constructed using synthetic oligonucleotides and PCR (Appendix D). 50 pmols of each sequence was in vitro transcribed, ligated to F30P using the appropriate DNA splint, and purified. For each binding reaction, 80 pmols of ligated mRNA was in vitro translated with supplemented 35 S-methionine and dT purified using dT buffer (same protocol as above). The purified samples were then incubated in a solution containing 25% DMF, 1X phosphate buffer pH 8, for 1 hour. DSG or Biotin modified samples were incubated with the same buffer with the addition of DSG (0.25 mg/mL) or NHS- Biotin (0.25 mg/mL). The samples were then dT purified using IB buffer and RNAse treated (as described in Chapter 2.2). The samples were incubated with ~700 pmols of HCV core peptide on acrylamide hydrazide beads, or ~300 pmols of biotin-labeled HCV core peptide immobilized on streptavidin agarose beads in assay buffer for 1 hour at room temperature. The samples were washed with assay buffer 5 times. The flow through, all of the washes, and the beads were counted using a scintillation counter. Construction of ECFP-Ligand Plasmids: ECFP-ligand fusion plasmids were constructed from the ECFP-EYFP FRET sensor plasmid. 17 The selected sequences were extended by PCR 20 amplification to include an HA tag and the helical linker (Appendix E). The ligands were then agarose gel purified using the Nucleospin Gel and PCR Cleanup kits (MACHEREY-NAGEL) according to the manufacturer’s instructions. Ligands were cloned into the vector using XbaI and HindIII. MBP-Ligand Fusion Protein Expression and Purification: Ligands were PCR amplified using oligonucleotides shown in Appendix F and gel-purified (same procedure as above). The ligands and the control were cloned into pET24a vector containing the MBP protein using BamHI and NotI. BL21 (DE3) competent cells were used to express the fusion proteins, using 100 mL of auto-induction media. 33 Cells were lysed with 12 mL of Bper (Pierce) and spun down. The supernatant was loaded onto a column containing 2 mL of HisPur Ni-NTA resin (Thermo Scientific) and allowed drain gravimetrically. The column was washed 5 times with Buffer A (50mM Hepes-KOH pH 7.5 + 150 mM NaCl + 10mM imidazole), 5 times with Buffer B (50mM Hepes-KOH pH 7.5 + 1 M NaCl + 10mM immidozle) and 5 times with Buffer A again. The fusion proteins were eluted stepwise with 30, 50 and 125 mM immidozole in Buffer A. Fractions of pure protein were combined, concentrated, and desalted into 50mM Hepes-KOH pH 7.5 and 150 mM NaCl. The protein was frozen in liquid nitrogen and stored at -80 °C. MBP-Ligand Fusion Protein ELISA: Polystyrene plates were incubated with 5 pmols of streptavidin per well overnight at 4 °C. The plates were washed with wash buffer (1X PBS + 0.1% Tween 20), filled completely with 5% BSA (w/v) in 1X PBS, and incubated for 2 hours. After incubations, the plates were washed. The control wells (fusion proteins without the HCV core peptide) were incubated with assay buffer (1X PBS, 1 mg/mL BSA, 0.1% Tween 20). The 21 remaining wells were incubated with 5 pmols of biotin-labeled HCV core peptide per well and incubated for 90 minutes. The ELISA plate was washed and incubated with different concentrations of MBP fusion proteins, or the MBP control. After a 90 minute incubation, the plate was washed and incubated for 1 hour with 100 μL/well of 1:1000 dilution of HRP conjugated to anti-HA antibody (Pierce). The plate was then washed and incubated with TMB substrate for approximately 5 minutes. The reaction was stopped with equal volume of 2M sulfuric acid, and the OD450 was measured using a plate reader. Expression, Purification and loading of the Truncated HCV Core Protein on Beads: The truncated protein sequence was purchased from GENEWIZ (optimized for E.Coli expression). The sequence was PCR amplified and extended to include a C-terminal avitag for in vivo biotinylation by BirA enzyme. 34 The extended sequence was inserted into pET-24a vector using NdeI and XhoI. The vector was transformed into BL21 (DE3) competent cells containing BirA enzyme expressing plasmids. 25 mL of cells (grown overnight) were added to 1 L of media supplemented with 50 μM biotin and grown at 37 °C for 2 hours. The cells were then induced with 400 μM IPTG for 4 hours and harvested. Cells were lysed using Buffer A (100 mM sodium phosphate, 10 mM Tris-Cl, 8 M urea, pH adjusted to 8.0) and centrifuged. The supernatant was purified using Ni-NTA superflow resin on an FPLC (Bio-Rad), using a gradient from 0 mM to 400 mM imidazole (Buffer A: same as above. Buffer B: 100 mM sodium phosphate, 10 mM Tris-Cl, 8 M urea, 400mM imidazole, pH adjusted to 8.0). Fractions with pure truncated HCV core protein were combined, concentrated, and desalted into Buffer A, and kept at 4 °C for up to 1 month. 22 To load the protein onto the Streptavidin beads, 300 µL of streptavidin agarose beads were incubated with 20 nmols of protein in 3 M urea. After 2 hours of incubation at 4 °C, the beads were first washed with 50 mM Tris-HCl pH 8.0 + 150 mM NaCl + 3 M urea, and then washed with 50 mM Tris-HCl pH 8.0 + 150 mM NaCl + 0.1 % Tween 20. The beads were suspended in a solution containing 50 mM Tris-HCl pH 8.0 + 150 mM NaCl + 1 mg/mL BSA + 0.1 % Tween 20 and kept at 4 °C. HCV Cell Culture Experiments: Huh7.5.1 human liver cells line were seeded in a 24 well plate and cultured in Dulbecco's Modified Eagle Medium (DMEM, Invitrogen) supplemented with 10% of fetal bovine serum (FBS), 10mM non-essential amino acids (Invitrogen), 10mM HEPES pH 7.5, 100 units/mL penicillin, 100 mg/mL streptomycin, and 2 mM L-glutamine at 37 °C with 5% CO 2 . 26 After 24 hours, the cells were transfected in triplicate with 800 ng of ECFP control or ECFP-ligand fusion proteins plasmids, and allowed to grow for 24 hours. The cells were then treated with 800 ng/µL of G418 antibiotic. After 24 hours, the cells were infected with HCV GT2a strain JFH-1 at 0.1 multiplicity of infection. 6 hours post infection, the cells were washed 3 times with 1X PBS, and fresh media was added to the cells, and cells were allowed to grow for three days. The supernatant was then collected and used for virus titer by fluorescent focus assay, and the cell pellets were used to determine HCV RNA copy number by quantitative reverse transcription-PCR (both procedures are described in detail in Arumugaswami et. al., 2008). 26 For samples with interferon treatment, interferon alpha B2 protein (Fitzgerald) was incubated with the cells post infection at 1 U/mL. 23 Influenza Cell Culture Experiments: Huh7.5.1 cells were cultured and transfected with ECFP or ECFP-ligands constructs as described above. 24 hours after the antibiotic screen, the cells were infected with the human influenza virus for 6 hours. The cells were washed 3 times in PBS and incubated with fresh growth media for three days. The supernatant was then collected and used for a virus titer assay (Plaque Forming Assay) using A549 cells grown in DMEM, supplemented with 10% FBS, 100 units/mL penicillin, 100 mg/mL streptomycin, and 2 mM L- glutamine at 37 °C with 5% CO 2 . 24 2.3 Results and Discussion The Peptide Library: Our plan was to use a cyclic peptide library (MX 10 KSGSGSG), cyclized using a bis-NHS liker (DSG) connecting the primary amine on a lysine residue to the N- terminal amine of the peptide. One concern with using such a library was that if a shorter cyclic peptide was preferred for binding to the target, a lysine residue would be required in the X 10 random segment of the library. Having a fixed lysine residue at position 12 would create multiple possible cyclization products with varying probabilities (cyclization efficiencies are higher for smaller cycles) 35 which would reduce the fraction of correct product, thus creating major disadvantage for the smaller cyclic peptides. In order to design a robust method that allows for shorter cyclic peptides, we designed the library in such as a way that amino acid lysine at position 12 would be in ~65% of the final library, and other amino acids would appear 35% of the time. This would allow for robust shorter cycle products in 35% of the library. Selection: The peptide library underwent 9 rounds of enrichment. The number of PCR amplification cycles required to generate each affinity enriched pool, as well as the translation efficiency of each pool is shown in Figure 4. mRNA display generally requires several rounds of enrichment before target specific ligands compose a significant portion of the pool (selection convergence). The enriched pool needs to be amplified, transcribed, ligated, and translated after every round of enrichment before going through another affinity enrichment step. Consequently, enrichment will not solely benefit sequences that bind the desired target, but will also select for sequences which can PCR amplify, transcribe, ligate, or translate with higher efficiencies. This effect can be used to track the progression of selection through different rounds, since there is 25 generally increased translation efficiency for the pool undergoing selection. Figure 4 shows this effect clearly, as the translation efficiency of the pool nearly tripled when comparing pool 0 to pool 9. This can be used as a tool to ensure that the selection is progressing as designed. We also performed radiolabeled binding assays with each round of selection to look for specific binding of the pool to the target and to make sure the binding of the pool to the solid matrix is minimal. Since the area of the solid matrix covered by the selection target is much smaller than the total area of the solid matrix, it is not unusual to get high levels of binding to solid matrix after a few rounds of selection. 17 We observed this phenomenon after 6 rounds of selection against HCV core peptide on agarose hydrazide beads (Figure 5). Pool 6 shows a much higher level of binding to beads without target than binding to beads with target. To continue the selection and reduce the frequency of the peptides that bound to the agarose hydrazide matrix, we switched to acrylamide hydrazide beads using an altered spacer between the beads and the peptide. We tested the binding of pool 6 to the acrylamide beads, and measured minimal binding of the pool to the new solid matrix, while the pool showed a significant level of binding to immobilized HCV core peptide. We performed three more rounds of selection both to improve the frequency of the HCV binding peptides in the pool, and to reduce the frequency of the agarose hydrazide binding peptides. During the selection process, the radiolabeled binding assays performed on the pools were designed to closely resemble the conditions of selection itself. mRNA-peptide fusions were reverse transcribed, and then bound to immobilized HCV core peptide. Figure 6 shows the binding of pool 9 to immobilized HCV core peptide with or without the mRNA template. Removing the mRNA template from the fusion molecule increases the binding by approximately a factor of 4. These results are consistent with historical observations (unpublished observations). 26 From this point on, to better represent the behavior of the peptide in solution, all binding experiments were done by incubating the mRNA-peptide fusions molecules with RNAse enzyme, and then proceeding with the radiolabeled binding assays. Sequencing: Rounds 0, 6, 8, and 9 were sequenced via TOPO cloning (Table 2.) Round 9 was also sequenced by Next Gen sequencing (Table 3). Even though only a limited number of sequences were acquired for pool 0, the pool resembles the design closely. Approximately 75% of the sequences had a lysine at position 12, which is in line with the designed 65%. The calculated frequency of sequences without a stop codon in a stretch of 10 random amino acids is 52% using the NNS triple codons (where N means any possible nucleotide, and S stands for G and C nucleotides). Using the NNS triple codon instead of NNN to code for randomized amino acid segments reduces the probability of stop codons from 3/32 to 1/16. We observed 50% of the sequences in pool 0 without a stop codon, in line with the design. One sequence in pool 0 and one in pool 6 (shown in red) have C-termini which are not coded for by the 3’ primer. The origin of these sequences is unknown, but is likely due to recombinatorial events. The percent of sequences with stop codons is reduced throughout the selection process from ~50% in pool 0 to ~20% in pool 9, when comparing sequences with TOPO cloning. We obtained over 4.3 million sequences when we observed pool 9 composition via Next Gen sequencing (illumina). Approximately 6% of the pool was the top most frequent clone, and together with the next 9 most abundant sequences, they composed ~20% of the pool, which points to the convergence of the selection (Figure 7a). Interestingly, the number of sequences vs. copy number was related by a power function (linear on a log-log graph, Figure 7b). About one quarter of sequences from the Next Gen sequencing have a single occurrence in the sequencing 27 data, which demonstrates that there is still a significant amount of diversity left in pool 9. None of the top 30 most abundant sequences in pool 9, when analyzed using Next Gen sequencing (Table 3) have a stop codon in the random region. This can be interpreted as a sign that selection progressed as expected, since sequences with stop codons do not form fusions as easily as sequences without stop codons, and should be selected against under normal selection conditions. We were able to obtain 6 sequences out of the top 7 most abundant sequences, and 9 out of the top 30 most abundant sequences in pool 9 by random sampling of 26 sequences (TOPO cloning results, Table 2.) These results are not too surprising, since statistically ~2 copies of the top sequence should have been observed (2 copies were obtained) and ~5 out of the 26 sequences should have been from the top 10 most abundant sequences (7 sequences were obtained.) These results point out that the Next Gen sequencing results are indeed representative of the pool 9 composition, and that the selection has converged. Another indication for the progression of affinity selection is the amino acid composition for the random region. By comparing the amino acid composition of the random region for the top 200 most abundant sequences (Figure 8) against the initial composition, we see a clear bias towards specific amino acids (F, I, L, and V). Other amino acids are under-represented (C, D, E, H, P, and T). When we compare pool 9 random segment composition with the propensity of amino acids to be found at the interface of protein-protein interactions 36 , no clear pattern emerges. This could be the result of not all the amino acids in the random segment making contact with the HCV core peptide or a consequence of the peptide-peptide interaction not sufficiently resembling the same interface propensities as the corresponding protein-protein interaction. 28 Frame Shift and Loss of Lysine at Position 12: As is clear from Tables 2 and 3, there was a consistent increase in frequency of the peptides with a frame shift from the original designed frame, throughout the selection process. This is shown more clearly in Figure 9. None of the sequences obtained by TOPO cloning of pool 0 showed a frame shift mutation, whereas greater than 90% of the top 200 most abundant sequences in pool 9 are in the -1 frame. There is also a consistent increase, throughout the selection progress, in the number of sequences in the -1 frame which points to an evolutionary advantage for these sequences. To better understand the reason for the frame-shift, we analyzed the three translational frames of the constant C-terminal region (Table 4.) Mean grand average hydropathicity score enumerates the hydropathic characters of a peptide. 37 Negative scores mean the peptide is hydrophilic while positive scores mean that the peptide segment is hydrophobic. Helical propensity scale was developed to show the likelihood of a specific amino acid to be part of an alpha-helical structure. 38 The lower the helical propensity score, the higher than probability of the peptide segment adopting an alpha-helical structure. Table 4 shows that the -1 frame is the most hydrophobic and has both the highest probability of conforming to an alpha-helical structure among the possible translational frames. We further analyzed the increase in hydrophobicity and likelihood of alpha-helical structure for pool 9 by analyzing the top 200 most abundant sequences and comparing them to the other sequenced pools. We did the analysis with or without the constant C-terminal segment for hydrophobicity and helical propensity (Figures 10 and 11). There is a large consistent increase in the frequency of hydrophobic residues in the pools throughout the selection progress (Figure 10a). This trend is enlarged when the constant C-terminal segment is eliminated from analysis. 29 The increase in hydrophobic residue frequency without the constant C-terminal region mutation, points to a selective advantage for sequences with hydrophobic residues independent of the frame shift mutation. These results are not unexpected, since protein-protein interactions are usually driven by burial of hydrophobic clusters at the site of interaction. 39 Helical propensity and hydropathicity are not co-dependent variables. As an example, the 8 most likely amino acids to be found in helices are alanine, arginine, leucine, methionine, lysine, glutamine, glutamic acid, and isoleucine, which cover the spectrum of hydrophilic to hydrophobic. 38 This means that the increase in hydrophobicity and alpha-helicity can be analyzed independent of each other. There is a consistent increase in amino acids with high likelihood of helical structure when looking at the sequenced pools (Figure 11a). While the increase in the likelihood of helicity of the pool remains, it is somewhat lessened when we remove the contribution of the constant C-terminal mutation from the analysis (Figure 11b). These results are also not surprising, since the HCV core peptide is known to from a helix-loop- helix conformation, and helix-helix interactions are one of the common forms of protein-protein interactions. 40 The helical propensity of peptides from pool 9 is examined in more detail with individual sequences below. Another striking pattern that is clear from Table 3 is the low frequency of lysine amino acid at position 12. If there was a selective disadvantage for having lysine at that position, it could explain the consistent frame shift in the library, either through acquired mutations during selection or enrichment of sequences with one less nucleotide due to chemical synthesis. However, as it is shown in Appendix A, the -1 frame shift would change position 12’s anti-sense codon from 111 to T11 (where nucleotide 1 is composed of 85% T, 5% G, 5% A, and 5% C), which would change the frequency of lysine at that position from ~65% to ~72%. This means 30 that the fame shift did not happen to move the library away from lysine at position 12, and the two events are independent of each other. It is striking that even though the frame shift mutation increased the likelihood of lysine at position 12, it is observed only in 22% of the top 200 sequences in pool 9. Lower frequency of lysine at position 12 could be explained by the formation of smaller amino acid cycles, which was originally part of the design of the library. By analyzing the top 200 most frequent sequences in pool 9, however, we discovered that approximately two third of the sequences without lysine at position 12 do not contain a lysine anywhere in the random segment. This means that reduction in the frequency of lysine at position 12 is not necessarily due to the formation of smaller cyclic peptides. We tested the binding of DSG cyclized and non-cyclic pool 9 to HCV core peptide (Figure 12) to examine the effect of DSG modification on the pool, where less than half of the sequences have the ability to from cyclic peptides. DSG modified pool 9 bound to immobilized HCV core peptide approximately 4 times higher than non-DSG modified pool. The binding level to the beads without target was also significantly lower. These results are not entirely surprising, since DSG modification was performed prior to every round of selection. It stands to reason that sequences that have been enriched throughout the selection process, have been selected for in the context of DSG modification. However, the mechanism by which DSG modification helps the binding of pool 9 to HCV core peptide is not clear. The effects of DSG modification on specific clones in pool 9 are explored further below. Sequence Binding: An expected sign that selection by a naïve library has progressed as planned is the emergence of homologous families of sequences. This is due to the fact that if a large enough portion of the initial library is sampled and only a cluster of 4-6 amino acids is 31 required to form a core peptide that recognizes the target, this core can appear in several locations throughout the sequence, flanked on either end by a larger diversity of amino acids. By looking through the top 200 most abundant sequences in pool 9, we were able to recognize 6 major sequence motifs (Figure 13). Approximately 25% of the top 200 most abundant sequences contained an (S or G)WI motif, and a smaller subset had this motif flanked by tyrosine at the N- terminus and leucine at the C-terminus. Other motifs include VRFV, YNQIF, AGFV, AWFA and LIV. We chose 10 sequences based on binding motif and frequency in pool 9. The top 7 most frequent sequences (Table 3), as well as at least a single sequence from each family of binders (Figure 13) were chosen to study further. Sequences 8.02, 9.09, and I9.09 contain a lysine at position 12; sequences 9.01, 9.17 and 9.19 contain lysine residues in the random region, and sequences 8.07, 9.08, 9.15 and I9.01 do not contain any lysine residues. Helical peptide segments have a strong dipole moment due to the alignment of the peptide backbone in the alpha helix structure. 41, 42 If our peptide binders are indeed alpha helical in nature, we hypothesized that the reason for DSG modification’s improvement to binding of sequences to HCV core peptide is removing of the positive charge at the N-terminal of the peptide and stabilizing the helical structure. If this is indeed the case, modifying the N-terminal with a different modifier should have a similar effect. We tested the binding of non-modified, DSG modified, and biotin modified (using NHS-biotin to react with the primary amine at the N- terminal) sequences to HCV core peptide (Figure 14a). The binding assays were performed over three days. Since radiolabeled assays tend to have a high degree of day to day variability (mostly due to variable amount of target present, and different degrees of bead blocking) the results 32 should be compared within a single day, and not across the whole chart (the three separate day’s experiments are separated by vertical lines). DSG modified peptides have a higher level of binding than the non-DSG modified peptides in every tested sequence. These results demonstrate the importance of selection conditions and context, and corroborate previous binding experiments with DSG modified or non-modified pool 9. Sequence 8.02, which contains a single lysine at position 12, is the only sequence for which DSG cyclization leads to a higher level of binding than biotin modification. All other sequences show a higher level of binding with NHS-biotin modified mRNA-peptide fusions, even sequences 9.09 and I9.09 which have a lysine at position 12. Even though it seems like modifying the N-terminal of the peptide to remove the positive charge is helpful in binding of ligands to HCV core peptide, this experiment by itself cannot be taken as conclusive evidence of helicity of binders. By removing the positive charge at the N-terminus, and attaching a biotin molecule, we significantly increase the hydrophobicity of the resulting peptide. The increased in the binding experienced here could simply be due to the increased hydrophobicity of the sequences, a trait which was significantly favored during the selection process. To compare directly the tested sequences, we performed a set of binding assays on a single day, which looked at the best binding condition for 6 of the selected sequences (Figure 14b). Sequences 9.01 and 9.15 show a slightly higher binding level than the rest of the tested sequences, and are used more extensively in future experiments. We decided to modify the N- and C-terminus of some of our HCV core binders to further analyze the frame shift and test the structure of the peptides. We tested the binding of the C- terminally modified 9.01 sequence, where the constant region was changed from ADLAA to AAAAAAG. The new sequence would not change the sequence’s hydropathicity or helical 33 propensity by a large degree; however, it can highlight the role of leucine and aspartic acid residues in the constant C-terminal segment. Figure 15a clearly shows that the -1 frame was not favored in the selection due to the appearance of leucine and aspartic acid residues, and the selection is likely due to other reasons. Next, we tested the helicity of peptides 9.01 and 9.15 by extending them at the N-terminal by a flexible or helical linker. The helical linker (AEAAAKEA) should stabilize helix formation through the formation of salt bridges between the lysine and glutamic acid. 43 If the peptides are helical, the helical linker should increase binding of the peptide to its target by not only taking the positive charge away from the N-terminal of the binding sequence, but by extending and stabilizing the helix. Figure 15b-c shows that even though the binding of sequences to HCV core peptide is increased more with the helical linker compared to the flexible linker, neither extension matches the increase in binding achieved by biotin modification. Even though there is some evidence of helical structure in the peptide ligands, the data suggests that hydrophobic interactions seems to be the main driving force for binding. Confirmation of Binding: Even though the selection had been performed using two different solid matrices, it is essential to show that binding is not dependent on the solid matrix used. We have been unable (data not shown) to synthesize and purify the peptides via chemical synthesis, since the peptides are very hydrophobic, and hard to purify. We tested the binding of DSG modified pool 9, and sequences 9.01 and 9.15 to biotin-labeled HCV core peptide immobilized on streptavidin beads (Figure 16). Sequences 9.01 and 9.15 show significant binding to HCV core peptide immobilized on streptavidin beads, though at lower levels than peptide on acrylamide hydrazide beads. Pool 9 shows a much lower level of binding to HCV core peptide 34 immobilized on streptavidin beads. Streptavidin beads provide a much lower target density than acrylamide hydrazide beads, which can account for much of the different between the binding levels. None of the tested samples show any significant binding to beads without target. To further show the specificity and affinity of the selected sequences, we expressed the 9.15 sequence as a maltose binding protein-fusions (MBP-fusions), and performed an enzyme linked immunosorbent assay (ELISA). We designed an HA tag between the 9.15 sequence and the MBP, and we used the helical linker to connect the HA tag to the 9.15 sequence. We also expressed a version of the protein without the 9.15 ligand. The binding schematic for the assay is shown in Figure 17a. Biotin labeled HCV core peptide binds to streptavidin immobilized on the polystyrene ELISA plate. The MBP-9.15 fusion protein then binds the peptide target, and is recognized by an anti-HA antibody conjugated to horseradish peroxidase (HRP). Signal is generated HRP’s cleavage of TMB substrate. MBP-9.15 fusion protein produces a dose- dependent response, which is robustly higher than background at concentrations as low as 10 nM (Figure 17b). Elimination of the 9.15 sequence from the MBP-fusion molecule, or the HCV core peptide, results in low, dose independent signal. These results confirm that the peptide can bind to the HCV core segment without the puromycin or the DNA linker. It also confirms that the peptide being made via in vitro translation behaves similarly to E.Coli expressed peptide. So far, we have shown that the selected ligands can bind to HCV core peptide either as peptides fused to a DNA liker via puromycin, or as MBP protein fusions. Even though having peptides that bind specifically to another peptide by itself is a very interesting and exciting find, which could have many applications in synthetic biology or as in vivo research tools, we still needed to make sure the peptide ligands can also bind to the core protein of the HCV, and not just the peptide segment to be able to fulfill our goal for this project. We expressed a truncated 35 version (amino acids 1-120) of the core protein with a C-terminal avitag for in vivo biotinylation by the BirA enzyme 34 (Appendix G). After expression, we purified the protein under denaturing conditions (6 M urea). Removing the urea from the protein solution proved difficult, since the core protein would precipitate completely after buffer exchanging into a solution without high urea concentrations. To avoid this problem, we immobilized the protein onto streptavidin beads, and buffer exchanged the beads into 1X PBS + 1 mg/mL BSA + 0.1% Tween 20. We performed radiolabeled binding experiments with DSG modified 9.01 and 9.15 ligands, as well as an unrelated peptide (Figure 18). We measured a very high binding level for both 9.01 (~42%) and 9.15 (~50%) ligands, while the unrelated peptide showed medium levels of binding (~5%). A relatively high level of binding for the unrelated peptide to HCV core protein is indicative of the presence of unfolded protein on the beads surface. The ability of our peptides to bind to the core protein with an order of magnitude higher level can be evidence of their ability of bind to the core protein specifically; however, this evidence is not conclusive due to the high levels of denatured proteins on the beads surface. Behavior of HCV Core Peptide Ligands in Vivo: We made DNA plasmid constructs with Enhanced Cyan Fluorescent Protein (ECFP) 44 ligand fusions (Appendix E) for in vivo experiments using hepatocellular carcinoma cells. Huh7.5.1 cells were transfected in triplicates with ECFP or ECFP fusions constructs, screened for transfection, and infected with JFH1 (genotype 2a) strain of HCV. Extracellular virus particle production was assayed 3 days post infection (Figure 19a) by fluorescein focus assay. Both 9.08 and 9.15 sequences reduced the virus particle production to background levels (no infection), while sequences 8.07 and 9.01 reduced viral levels by ~80% and ~30% respectively. These results conclusively show that not 36 only do our ligands bind to the HCV core protein, but they also inhibit its function in viral particle production. In order to make sure the cells were viable and our peptides did not simply disrupt normal cell functions, we took a look at internal viral RNA levels. Since the HCV core protein does not play a role in viral infection, we hypothesized that the Huh7.5.1 cells should have similar copies of the HCV genome independent of ECFP-ligand transfection. If normal cell functions were inhibited due to the hydrophobic nature of our ligands, or any non-specific binding, we would expect lower copies for the HCV genome in the infected cells. Figure 17b shows that the internal mRNA levels in the infected cells (measured via RT-qPCR) are not affected by the presence of our peptide ligands. Lastly, to show the specificity of our ligands, we infected Huh7.5.1 cells, transfected with our ligands, with the human influenza virus. Figure 17c shows the extracellular virus production measured via the plaque forming assay. Sequence 9.15 reduces the extracellular influenza virus levels by ~40%, followed by sequence 8.07 (~66%), and sequences 9.01 and 9.08 (~88%). The data suggests that the 9.15 sequence is both highly effective in stopping viral particle production during HCV infection, while being minimally disruptive to normal cell functions, or orthogonal infections. Even though sequence 9.08 seems as effective as 9.15 in stopping viral particle production, it seems to be disruptive to an orthogonal infection, which can be interpreted as non-specific binding to hydrophobic sequences. 37 2.4 Conclusions Amino acids 16-40 of the HCV core peptide are highly conserved among many genotypes of the HCV virus, make up the immunodominant region of the core protein, and are crucial in virus particle production. 26, 27 Through the use of mRNA display, we were able to generate peptide binders that specifically target this epitope of the HCV core protein. This peptide target/peptide ligand pairing is inherently exciting due to its potential applications in synthetic biology, cell signaling, and other in vivo research tools. We were able to show that the peptide ligands bind to the truncated core protein, thus addressing the concerns that using a peptide segment as the target could lead to ligands which recognize the unstructured peptide, but not function during the course of in vivo experiments. Our in vivo data, using hepatocellular carcinoma cells to test the ability of protein fusions containing the peptide ligands in the C-terminal, clearly demonstrates that the 9.15 sequence is highly effective in stopping viral particle production during HCV infection, minimally disruptive to normal cell functions, and specific to HCV infection. Unfortunately, we have been unable to synthesize and purify the peptides via chemical synthesis, since the peptides are very hydrophobic and hard to purify. Since our goal was to develop potential therapeutics for chronic HCV infection, one approach toward that goal is to create transgenic pluripotent stem cells (iPSC) derived from patient’s somatic cells, each bearing an anti-HCV peptide ligand. Repopulating the decompensated and infected liver with autologous iPSC-derived hepatocytes is likely an attractive, fast, cost effective alternative for liver transplantation. Over time, the HCV resistant hepatocytes will have a growth/selection advantage over the original hepatocytes, facilitating tissue replacement. 38 2.5 Figures Core Peptide AA#: 16 20 25 30 35 Periodate-oxidizable peptide: S-GGSGG NRRP QDVKF PGGGQ IVGGV YLLPRR Biotin-labeled peptide: bt-GGSGG NRRP QDVKF PGGGQ IVGGV YLLPRR HCV genotype % of Sequences with Mutations in This Segment Most Common Mutation Geographical Location Genotype 1a 3.8% N16Y USA, Europe Genotype 1b 8.8% N16Y, R39P USA, Chile, Japan Genotype 2 6.8% D21N China Genotype 3 100% N16I, L36V, D21N, Q20K, V31Y, G32V, G26A, G33E England, Thailand, India Genotype 4 100% Q20M Egypt, Middle East, North Africa Genotype 5 0% South Africa Genotype 6 74% Q20M Asia Table 1: The HCV core protein segment used as the target in the mRNA display selection compared to the same segment in the different genotypes of HCV. The peptide segment (Bold letters) chosen for mRNA display is the dominant sequence for genotypes 1, 2 and 5. Genotypes 4 and 6 have a single glutamine to methionine mutation at position 20, while genotype 3 significantly diverges from the rest of the genotypes in this segment. (Analysis performed on 783 sequences available in the database) 24, 25 39 Figure 1: Mechanism of attaching the HCV core peptide to hydrazide beads. The peptide with an N-terminal serine residue is first oxidized by sodium periodate to form an aldehyde at the N- terminus. The aldehyde reacts with hydrazide group at the end of the solid support to immobilize the HCV core peptide on the beads using acid-labile hydrazone linkage. 40 Figure 2: Periodate-oxidization of the HCV core peptide with an N-terminal serine. (a) MALDI- TOF data show the original molecular weight of the HCV core peptide with an N-terminal serine as 3180 g/mol. (b) The molecular weight of HCV core peptide after periodate-oxidation is 31 Da lower than the non-oxidized peptide. Figure 3: Quantitation of HCV core peptide immobilized on hydrazide beads. HCV core peptide is immobilized on hydrazide beads with an acid-labile hydrazone bond. The amount of HCV core peptide on beads can be quantitated using the area under the curve and a set of standards. Non-periodate oxidized peptide has a different retention time than the periodate-oxidized peptide. 41 Figure 4: PCR cycles and translation efficiency through the selection process. The PCR cycles through the selection process stayed relatively consistent, however translation efficiency improved consistently. Figure 5: Binding of affinity-enriched pools to immobilized HCV core peptide throughout the selection process. Binding of radiolabeled pools 0-5 to HCV core peptide immobilized on agarose hydrazide beads is low and comparable to binding to beads without target. Pool 6 shows considerable binding to agarose hydrazide beads without target. Pools 6-9 bind to HCV core peptide immobilized on acrylamide hydrazide beads while showing no significant binding to beads without target. All pools were DSG modified prior to the binding assay. 42 Figure 6: Effect of removing the attached mRNA template from the fusion molecule. Binding of Radiolabeled RNAse treated pool 9 fusion molecules to immobilized HCV core peptide was approximately five fold higher than that of non-RNAse treated fusions. Neither group showed any significant binding to beads without target. The binding assays were performed with DSG modified fusion molecules. 43 Name Sequence Frequency Name Sequence Frequency Pool 0 Pool 9 HCVCP 0.00 MAAGM-GLML-K SGSGSG 1 HCVCP 8.07 MYIILLYNQIFI ADLAA 2 HCVCP 0.01 MVLMWAPSCYNK SGSGSG 1 HCVCP 8.02 MWFLVNSWIFVK ADLAA 1 HCVCP 0.02 MAHRGAYPSKMK SGSGSG 1 HCVCP 9.01 MAKLLIAWFAFI ADLAA 1 HCVCP 0.03 IHGLDRWRFA-Q SGSGSG 1 HCVCP 9.02 MF-WFPNSWTPN SGSGSG 1 HCVCP 0.04 M-RRLITYAS-K SGSGSG 1 HCVCP 9.03 MDYYVWLYAMIL SGSGSG 1 HCVCP 0.05 MT-YFGSMLVNK SGSGSG 1 HCVCP 9.04 MKMLFFVGSVVI ADLAA 1 HCVCP 0.06 MTPNWIHQTLVKSEGRIR 1 HCVCP 9.05 MYKLLPKPERL- SGSGSG 1 HCVCP 0.07 MPRGLLLLLTRI SGSGSG 1 HCVCP 9.06 MALLTIAWFAFI SGSGSG 1 Pool 6 HCVCP 9.07 MLLQMVWARQNQ RGSGSG 1 HCVCP 6.00 MFMWWY-PIWFKTDSGTE 1 HCVCP 9.08 MLRMLIVRFVLI ADLAA 1 HCVCP 6.01 MYIALYGWILVK SGSGSG 1 HCVCP 9.09 MLLYLLAGFVLK ADLAA 1 HCVCP 6.02 MVLSGRTAVRMT SGSGSG 1 HCVCP 9.10 MLVLITAGFVLK ADLAA 1 HCVCP 6.03 MSQLSRPDAVEE RIWQR 1 HCVCP 9.11 MYYVIYLPVVLI SGSGSG 1 HCVCP 6.04 MWQFHNSWVIV- SGSGSG 1 HCVCP 9.12 MNDLKWLSWILI ADLAA 1 HCVCP 6.05 MM-MSQSQLSLR SGSGSG 1 HCVCP 9.13 MAQ-LAQPDSLI SGSGSG 1 HCVCP 6.06 MLKWWYDQIWFK ADLAA 1 HCVCP 9.14 MKIYIFGWITIK ADLAA 1 Pool 8 HCVCP 9.15 MLIVMIGRILLI ADLAA 1 HCVCP 8.00 M-SSSQITM-PN SGSGSG 1 HCVCP 9.16 MYIFGAYGWILK ADLAA 1 HCVCP 8.01 MLLKWMFCWTCI SGSGSG 1 HCVCP 9.17 MHLLKVFAWFAI ADLAA 1 HCVCP 8.02 MWFLVNSWIFVK ADLAA 1 HCVCP 9.18 MLKLLYGWICAI ADLAA 1 HCVCP 8.03 M-IVYPMFPRSK ADLAA 1 HCVCP 9.19 MKFLIFMKAKIL ADLAA 1 HCVCP 8.04 M-FILMYWIIFI ADLAA 1 HCVCP 9.20 MYGVR-YQXIFK ADLAA 1 HCVCP 8.05 MKSLMLLSWIYI ADLAA 1 HCVCP 9.21 MC---PLGWTVI SGSGSG 1 HCVCP 8.06 XXRMSNAKILTK RIWQR 1 HCVCP 9.22 MFLYMFSLISLI ADLAA 1 HCVCP 8.07 MYIILLYNQIFI ADLAA 1 HCVCP 9.23 MKWFFVRFVYEL ADLAA 1 HCVCP 8.08 MLWILFGQVVMI ADLAA 1 HCVCP 9.24 MYLRSQSRIMLR SGSGSG 1 Table 2: Sequences obtained by TOPO cloning from pools 0, 6, 8, and 9. Pools 0, 6, 8, and 9 were sequenced using TOPO cloning. The sequences are shown above. The space in the sequence column distinguishes the constant C-terminal region and the N-terminal random region. Sequences in red are not from the 3’ primer, and their origin is unknown. Sequences in bold were chosen to be tested further. 44 Figure 7: The composition of enriched pool 9 based on Next Gen high-throughput sequencing. (a) The most abundant sequence in affinity enriched pool 9 accounts for approximately 6% of the pool, while approximately a quarter of the sequences were present at a single copy level. The top 10 sequences account for ~20% of the pool. (b) Sequences with greater than 1,000 copies (greater than 250 ppm of the pool) account for greater than 35% of the pool 9 composition. 45 Name Sequence Frequency HCVCP 8.07 MYIILLYNQIFI ADLAA 251,306 HCVCP 9.15 MLIVMIGRILLI ADLAA 116,559 HCVCP I9.01 MLWILFGQVVVI ADLAA 113,962 HCVCP 9.17 MHLLKVFAWFAI ADLAA 112,227 HCVCP 8.02 MWFLVNSWIFVK ADLAA 61,511 HCVCP 9.19 MKFLIFMKAKIL ADLAA 54,502 HCVCP 9.08 MLRMLIVRFVLI ADLAA 36,276 HCVCP I9.02 MKYFLNLSWIAL ADLAA 30,059 HCVCP I9.03 MMLYKISWILHI ADLAA 25,607 HCVCP I9.04 MQMYLLVQILWI ADLAA 23,249 HCVCP 9.09 MLLYLLAGFVLK ADLAA 23,113 HCVCP I9.12 MGMVLIVRFVII ADLAA 20,753 HCVCP I9.06 MLQVTLFVRFAI ADLAA 18,882 HCVCP I9.08 MIVIILTGFVFT ADLAA 15,326 HCVCP I9.07 MYFVLLAGFVYK ADLAA 14,831 HCVCP 9.10 MLVLITAGFVLK ADLAA 13,641 HCVCP I9.09 MYIFVAYGWILK ADLAA 12,816 HCVCP I9.10 MLLLYNGWIVLI SGSGSG 11,671 HCVCP I9.11 MYYFGYIATILI ADLAA 10,073 HCVCP I9.12 MLIFFNSWILLK ADLAA 9,630 HCVCP I9.13 MYWLCIAGFCLK ADLAA 9,520 HCVCP I9.14 MFFIKLKVRFVI SGSGSG 9,331 HCVCP I9.15 MTYLLIASFSLI ADLAA 9,248 HCVCP I9.16 MQFIELISWILI ADLAA 9,172 HCVCP I9.17 MLIVLFYQVWK ADLAA 8,882 HCVCP I9.18 MVYVVYMGWIFK ADLAA 7,795 HCVCP I9.19 MLVVFNVGFVFI ADLAA 7,781 HCVCP I9.20 MGLLFELSWILI ADLAA 7,423 HCVCP I9.21 MKVLLYFSWIHI ADLAA 7,046 HCVCP 9.01 MAKLLIAWFAFI ADLAA 6,530 Table 3: Top 30 most abundant sequences in affinity-enriched pool 9 based on Next Gen high- throughput sequencing. Over 4.2 million sequences were obtained, of which 1.2 million were unique. The space in the sequence column distinguishes the constant C-terminal region and the N-terminal random region. Sequences in bold were chosen to be tested further. 46 Figure 8: Amino acid composition of pools 0 and 9 compared to theoretical pool 0 composition and interface propensity of amino acids. There is a significant over-representation for amino acids phenylalanine, isoleucine, and leucine, while cysteine, aspartic acid, glutamic acid, proline, arginine, and threonine are underrepresented in pool 9. Comparing pool 9 to amino acid interface propensity 36 reveals an under-representation for cysteine, aspartic acid, glutamic acid, histidine, methionine, and proline, while amino acids isoleucine, leucine, and valine are over represented. Figure 9: Progression of frame shift mutation through the selection process for the sequenced pools. (a) None of the sequences tested for pool 0 were out of the original designed frame. The selection progression is accompanied by an increasing percent of sequences shifting to the -1 frame. 47 Frame Sequence Mean Grand Average of Hydropathicity Score Mean Helical Propensity (kcal/mol) Original Frame SGSGSG -0.6 0.75 -1 Frame ADLAA 1.1 0.18 -2 Frame RIWQR -1.8 0.34 Table 4: Properties of the constant C-terminal segment in various frames. The C-terminal segment of the library in its original frame is moderately hydrophilic (negative GRAVY score) and unlikely to adopt a helical structure (large Mean Helical Propensity value). The -1 frame is highly hydrophobic and is the most favorable frame to adopt an alpha-helical structure. Figure 10: Increase in hydrophobicity of enriched pools through the selection process. (a) There is a significant increase in hydrophobic residue frequency as the selection progresses. (b) The increase in hydrophobic residue frequency is even greater in the random segment of the pool, where the C-terminal constant region is eliminated from analysis. 48 Figure 11: Increased helical likelihood of the pools through selection. (a) There is a consistent increase in enriched pools’ disposition to helical structure, based on sequence analysis, throughout the selection process. (b) Though this disposition is lessened, it is still present when the constant C-terminal region is excluded from the analysis 49 Figure 12: The effect of DSG modification on pool 9 binding. DSG modification improves the binding of the pool 9 radiolabeled fusion molecules to immobilized HCV core peptide even though most of the sequences lack the lysine residue in the C-terminus of the library required for cyclization. 50 Figure 13: Sequence homology in the top 200 most abundant sequences from the high- throughput screen. Six major motifs were found. At least one member of each homologous family was tested further. (S/G)WIL Motif: I9.09 YIFVAYGWIL I9.77 YLIVGYGWIL I9.65 YVYYSWILLI I9.83 LLLYGWILLV I9.27 HLVYVYSWIL I9.43 AIKLIVSWIL I9.59 YTYLLVSWIL I9.16 QFIELISWIL I9.03 MLYKISWILH I9.20 GLLFELSWIL I9.72 IFLIESWILH I9.31 HVVYLFGWIL I9.12 LIFFNSWILL AGFV Motif: 9.09 LLYLLAGFVL 9.10 LVLITAGFVL I9.07 YFVLLAGFVY I9.32 YLLLYAGFVI AWFA Motif: 9.01 AKLLIAWFAF 9.17 HLLKVFAWFA I9.34 SFLFIAWFAL VRFV Motif: 9.08 LRMLIVRFVL I9.46 MMYLLIVRFV I9.12 GMVLIVRFVI I9.14 FFIKLKVRFV I9.86 LLLLDVRFVF I9.81 MVMLYVRFVV I9.87 KYMLLMFRFV I9.37 LLIIRLRFVS YNQIF Motif: 8.07 YIILLYNQIF I9.29 HLYNQIFFFI I9.23 YVVIYNQIFF I9.58 QFYNQIVLLV I9.57 IIYLFINQIF LIV Motif: 9.15 LIVMIGRILL I9.26 KYLIVLVSQV I9.77 YLIVGYGWIL I9.43 AIKLIVSWIL 51 Figure 14: Binding of selected radiolabeled peptides to HCV core peptide immobilized on acrylamide hydrazide beads. (a) Radiolabeled binding assays performed on the selected sequences without modification, with DSG modification, and with biotin modification on three separate days. Since day to day variability of radiolabeled binding experiments is large, each segment of the graph should be compared to itself. Sequence 8.02 is the only sequence which shows a higher binding levels when cyclized with DSG. All other ligands show a higher binding level when modified with biotin. None of the sequences show significant binding to beads without target. (b) Direct comparison of the best binding condition of six selected HCV core peptide ligand sequences. 52 Figure 15: Effects of N-terminal and C-terminal modifications in the binding of 9.01 and 9.15 sequences to HCV core peptide. (a) Changing the C-terminal constant region of the peptide from ADLAA to AAAAA does not affect the binding of radiolabeled 9.01 sequence to HCV core peptide. (b) The helical linker N-terminal to the 9.01 sequence increases the binding of the sequence to the target compared to the flexible linker. Both linkers fail to reach the binding level of biotin-modified 9.01 sequence. (c) The helical linker N-terminal to the 9.15 sequence increases the binding of the sequence to the target compared to the flexible linker, however, it fails to reach the binding level of un-modified or biotin-modified 9.15 sequence. 53 Figure 16: Binding of radiolabeled fusions for DSG modified sequences 9.01and 9.15 and pool 9 to HCV core peptide immobilized on streptavidin agarose or acrylamide hydrazide beads. Tested samples show binding to HCV core peptide immobilized on streptavidin agarose beads while they do not show any significant binding to beads without target. Samples show a higher binding level to HCV core peptide immobilized on acrylamide hydrazide beads than to target immobilized on streptavidin agarose beads. Figure 17: Binding of MBP-9.15 fusion protein to the HCV core peptide in ELISA. (a) biotin- labeled HCV core peptide was immobilized on streptavidin coated polystyrene plates. Ha-tagged MBP-9.15 fusion protein then links the anti-HA antibody conjugated to HRP to the ELISA plate by binding to the peptide target. Signal is generated by HRP’s cleavage of TMB substrate. (b) MBP-915 fusion protein generates robust signal over background at 20nM concentration. Concentration independent low signal is generated from either the MBP protein without the 9.15 sequence, or the MBP-915 fusion protein without immobilized HCV core peptide. 54 Figure 18: Binding of 9.01 and 9.15 sequences to truncated HCV core peptide/protein. (a) Binding of DSG modified 9.01 sequence to HCV core peptide or truncated protein. DSG modified 9.15 sequence shows a much higher level of binding to the truncated HCV core protein than to the HCV core peptide, while showing no significant binding to beads without target. (b) Binding of DSG modified 9.01 and 9.15 sequences as well as an un-related peptide to the truncated HCV core protein. Both sequences 9.01 and 9.15 show significant binding to the truncated HCV core protein. While the binding of the unrelated peptide to the truncated HCV core protein is relatively high, it is an order of magnitude lower than the HCV binding peptides. 55 Figure 19: Effect of HCV core peptide ligands on Huh7 cells infected with JFH1, or the human influenza virus. (a) Extracellular HCV production levels measured by fluorescent focus assay, produced by cells transfected with HCV core ligands fused to ECFP, and infected with HCV. Sequences 9.08 and 9.15 reduce the extracellular virus levels to background levels of signal (no HCV infection). Sequence 9.01 reduces the extracellular virus levels by ~80% and sequence 9.01 reduces the levels of ~30%. (b) HCV genome copy number in the infected cells via qRT-PCR. The transfection of HCV core ligands does not significantly alter the levels of intercellular genome copy number of HCV in infected cells. (c) Extracellular influenza virus production levels in cells, measured by plaque assay. Sequence 9.15 reduces the extracellular influenza virus levels by ~40%, followed by sequence 8.07 (~66%), and sequences 9.01 and 9.08 (~88%). 56 2.6 Appendix DNA sequences used throughout the above research A) Peptide Library HCV Core Peptide Selection Library: 5’ Primer: T7 Promoter ΔTMV TAATACGACTCACTATAGGGACAATTACTATTTACAATTACA 3’ Primer: TCCGCTGCCAGATCCGCT Anti-sense Pool Strand: GCCAGATCCGCT 111322322322322322322322322322322CAT TGTAATTGTAAATAGTAATTG 1: 81% T, 05% G, 07% A, 07% C 2: 20% T, 20% G, 30% A, 30% C 3: 00% T, 40% G, 00% A, 60% C Splint: TTTTTTTTTTTNTCCGCTGCCAGA B) Topo Cloning Sequences from TOPO Cloning: 8.02 ATGTGGTTCTTGGTGAACAGCTGGATCTTCGTAAA AGCGGATCTGGCAGCGGA 8.07 ATGTACATCATCCTGTTGTACAACCAGATCTTCAT AGCGGATCTGGCAGCGGA 9.01 ATGGCCAAGTTGCTCATCGCCTGGTTCGCGTTCAT AGCGGATCTGGCAGCGGA 9.08 ATGTTGAGGATGCTCATCGTCAGGTTCGTGCTCAT AGCGGATCTGGCAGCGGA 9.09 ATGTTGCTGTACTTGTTGGCCGGGTTCGTCTTGAA AGCGGATCTGGCAGCGGA 9.15 ATGTTGATCGTGATGATCGGCCGGATCCTCCTGAT AGCGGATCTGGCAGCGGA 9.17 ATGCACCTGTTGAAGGTGTTCGCCTGGTTCGCAAT AGCGGATCTGGCAGCGGA 9.19 ATGAAGTTCCTCATCTTCATGAAGGCCAAGATCTT AGCGGATCTGGCAGCGGA 57 C) Sequences from Next Gen Sequencing Sequences Constructed from High-Throughput Next Gen Sequencing: I9.01 Sequence: (3’ primer and splint same as pool) 5’ Primer: TAATACGACTCACTATA GGGACAATTACTATTTACAATTACA ATGTTGTGGATC Template TTTACAATTACAATGTTGTGGATCTTGTTCGGCCAGGTCGTCGTGAT AGCGGATCTGGCA I9.09 Sequence: (3’ primer and splint same as pool) 5’ Primer: TAATACGACTCACTATA GGGACAATTACTATTTACAATTACA ATGTACATCTTC Template: TTTACAATTACAATGTACATCTTCGTGGCGTACGGCTGGATCCTCAA AGCGGATCTGGCA 58 D) N- and C-terminal Modifications 9.01 C-terminally Modified: (5’ primer same as the pool, template from TOPO cloning) 3’ Primer: CTGCCTGCCGCTGCCGCTGCCGCT ATGAACGCGAACCAGGC Splint: TTTTTTTTTTTNCTGCCTGCCGCTGCCGCT 9.01 N-terminal Flexible: (splint same as the pool) 5’ Primer: TAATACGACTCACTATA GGGACAATTACTATTTACAATTACA ATGGGAGACGGCGGGGAT Template: GGAGACGGCGGGGATGGTGGA ATGGCCAAGTTGCTCATCGCCTGGTTCGCGTTCAT AGCG 3’ Primer: TCCGCTGCCAGATCCGCT ATGAACGCGAACCAGGC 9.01 N-terminal Helical: (3’ primer same as 9.01 N-terminal flexible 3’ primer, template from TOPO cloning, splint same as the pool) 5’ Primer1: GCGGAAGCCGCAGCTAAAGAGGCA ATGGCCAAGTTGCTCATCGCCTGGTTCGCGTTCAT A 5’ Primer2: TAATACGACTCACTATA GGGACAATTACTATTTACAATTACA ATGGCGGAAGCCGCAGCT 9.15 N-terminal Flexible: (5’ primer same as 9.01 N-terminal flexible 5’ primer, splint same as the pool) Template: ATGGGAGACGGCGGGGATGGTGGA ATGTTGATCGTGATGATCGGCCGGATCCTCCTG 3’ Primer: TCCGCTGCCAGATCCGCT ATCAGGAGGATCCGGCCGAT 9.15 N-terminal Helical: (3’ primer same as 9.15 N-terminal flexible 3’ primer, 5’ primer same as 9.01 Helical 5’ Primer 2, splint same as the pool) Template: GGCGGAAGCCGCAGCTAAAGAGGCA ATGTTGATCGTGATGATCGGCCGGATCCTCCTG 59 E) ECFP-Ligands Fusion Proteins for HCV Cell Culture Experiments ECFP-Ligand Construction: ECFP-Control: 5’ Primer: HindIII HA tag GACTTAGCAAGCTTGTCTGGATCT TACCCCTACGACGTGCCCGACTACGCC AGCGGCAGC Template: Helical Linker XbaI GCCAGCGGCAGCGGC GCCGAGGCCGCCGCCAAGGAGGCC ATGTAGTCTAGAACTGTGCAG 3’ Primer: CTGCACAGTTCTAGACTACAT ECFP-8.07 Sequence: (5’ primer 2 same as ECFP-Control 5’ primer, template from TOPO cloning) 5’ Primer 1: GCCAGCGGCAGC GGCGCCGAGGCCGCCGCCAAGGAGGCC ATGTACATCATCCTGTTGTAC 3’ Primer: CTGCACAGTTCTAGACTACGCTGCCAGATCCGCT AT ECFP-9.01 Sequence: (5’ primer 2 same as ECFP-Control 5’ primer, template from TOPO cloning, 3’ primer same as ECFP-8.07 Sequence 3’ primer) 5’ Primer 1: GCCAGCGGCAGC GGCGCCGAGGCCGCCGCCAAGGAGGCC ATGGCCAAGTTGCTCATCGCC ECFP-9.08 Sequence: (5’ primer 2 same as ECFP-Control 5’ primer, template from TOPO cloning, 3’ primer same as ECFP-8.07 Sequence 3’ primer) 5’ Primer 1: GCCAGCGGCAGC GGCGCCGAGGCCGCCGCCAAGGAGGCC ATGTTGAGGATGCTCATCGTC ECFP-9.15 Sequence: (5’ primer 2 same as ECFP-Control 5’ primer, template from TOPO cloning, 3’ primer same as ECFP-8.07 Sequence 3’ primer) 5’ Primer 1: GCCAGCGGCAGC GGCGCCGAGGCCGCCGCCAAGGAGGCC ATGTTGATCGTGATGATCGGC 60 F) MBP-Ligands Fusion Proteins MBP-Ligand Construction: MBP-Control (MBP-HA): 5’ Primer: BamHI HA Tag GACTTAGCGGATCCTCTGGATCTTACCCCTACGACGTGCCCGAC Template: HA Tag Helical Linker CCTACGACGTGCCCGACTACGCCAGCGGCAGCGGCGCCGAGGCCGCCGCCAAGGAGGC 3’ Primer: NotI Helical Linker CTGCACAGTGCGGCCGCGGCCTCCTTGGCGGC MBP-8.07 Sequence: (template the same as TOPO clone) 5’ Primer 1: GCCAGCGGCAGCGGCGCCGAGGCCGCCGCCAAGGAGGCC ATGTACATCATCCTGTTGTAC 5’ Primer 2: GACTTAGCGGATCCTCTGGATCTTACCCCTACGACGTGCCCGACTACGCCAGCGGCAGC 3’ Primer: CTGCACAGTGCGGCCGCCGCTGCCAGATCCGCTAT MBP-9.01 Sequence: (template the same as TOPO clone, 5’ primer 2 and 3’ primer the same as MBP-8.07 sequence) 5’ Primer 1: GCCAGCGGCAGCGGCGCCGAGGCCGCCGCCAAGGAGGCC ATGGCCAAGTTGCTCATCGCC MBP-9.08 Sequence: (template the same as TOPO clone, 5’ primer 2 and 3’ primer the same as MBP-8.07 sequence) 5’ Primer 1: GCCAGCGGCAGCGGCGCCGAGGCCGCCGCCAAGGAGGCC ATGTTGAGGATGCTCATCGTC MBP-9.15 Sequence: (5’ primer 2 and 3’ primer the same as MBP-8.07 sequence) 5’ Primer 1: GCCAGCGGCAGCGGCGCCGAGGCCGCCGCCAAGGAGGCC ATGTTGATCGTGATGATCGGC Template: ATGTTGATCGTGATGATCGGCCGTATCCTCCTGAT AGCGGATCTGGCAGCGGA 61 G) Truncated HCV Core Protein Truncated HCV Core Protein Sequence: Template: (ordered from GENEWIZ) ATGAGCACCAACCCGAAGCCTCAGCGCAAGACCAAGCGCAACACCAATCGCCGCCCGCAGGACGTGAAGTT TCCGGGTGGCGGTCAGATCGTGGGTGGTGTGTATCTGCTGCCGCGCCGTGGTCCTCGTTTAGGCGTTCGCGCCACCC GCAAGACCAGCGAACGTAGTCAGCCGCGCGGTCGTCGTCAACCGATTCCGAAAGCACGCCAGCCTGAGGGTCGTGCA TGGGCACAGCCTGGTTACCCGTGGCCGCTGTACGGCAATGAGGGCATGGGCTGGGCAGGCTGGTTACTGAGCCCTCG CGGTAGTCGTCCGAGTTGGGGTCCGACCGATCCTCGTCGTCGTAGCCGCAACCTGGGC 5’ Primer: NdeI GACTTAGCACATATGAGCACCAACCCGAAGCCTCAGCGCAAGACCA 3’ Primer 1: ATCGTTCAGGCCCGAGCCCGAGCCGCCCAGGTTGCGGCTACGACGACGAGGATCGGTCGG 3’ Primer 2: XhoI Avi Tag CTGCACAGTCTCGAGGCCACCAGTGTCCTCGTGCCATTCGATTTTCTGAGCTTCGAAAATATCGTTCAGGC C 62 Chapter 3: Serum Stable Natural Peptides Designed by mRNA Display This work has been adapted from the following publication: Howell SM, Fiacco SV, Takahashi TT, Jalali-Yazdi F, Millward SW, Hu B, Wang P, Roberts RW. Serum stable natural peptides designed by mRNA display. Sci Rep 4, 6008 (2014). 3.1 Introduction The major limitation for using peptides as affinity reagents, probes, and therapeutics is their inherent instability in biological environments. Although the amide backbone is chemically stable, peptides are readily broken down in a matter of seconds in the digestive tract, in blood, plasma, serum, and inside cells due to the presence of proteases 45 . Because of this instability, many routes have been devised to chemically alter or modify natural peptides such as 1) including the addition of N-methylation to the backbone 46 , 2) insertion of β-amino acids 47, 48 , 3) changing the location of the side chain (e.g., peptoids), 49, 50 or 4) by covalent cyclization via insertion of chemical bridges. 51, 52 Left unanswered by these studies is the question of how much a functional peptide with natural amino acids could be stabilized by sequence optimization alone. Previously, we demonstrated a route to create high-diversity cyclic peptide libraries via mRNA display 1, 53 and used this approach to isolate a high affinity binder termed cycGiBP to the signaling protein Gαi1-GDP. 8 cycGiBP showed a very high affinity (K d = 2.1 nM) and ~3-fold increase in protease resistance as compared with the corresponding linear sequence. One observation in that work was that only one of the three possible cleavage sites predominated when cycGiBP was subjected to chymotrypsin. Using the standard substrate notation for proteases (P 3 -P 2 -P 1 -P 1 '-P 2 '-P 3 '; P's 63 represent the amino acid identity at a position and the scissile bond is located between P 1 and P 1 ' residue), chymotrypsin has a strong P 1 preference for W > Y > F >> L 54 . However, cycGiBP shows cleavage at P 1 = Y 5 , but not at P 1 = W 4 or F 7 . 8 We wondered if there were other members in that library that had improved protease resistance while retaining binding function. To address this issue, we performed a dual selection for chymotrypsin resistance and binding function on a library previously only sieved for binding function. Those experiments resulted in a several highly protease resistant peptides and indicate that sequence optimization can improve hydrolytic stability by as much as 400-fold compared to peptides sequences isolated without this selective pressure. 64 3.2 Materials and Methods E. Coli Expression of G i1-GDP and mRNA Display: G i1-GDP with a C-terminal BirA tag was expressed and purified as previously described 14 . mRNA display selection targeting G i1-GDP was performed starting at round 7 of our previous work and performed as described 8 with the modification that the cDNA fusions were digested with 2 mg of immobilized chymotrypsin (Sigma-Aldrich) per 10 6 cpm of fusions at room temperature for 15 minutes in 50 mM Sodium Phosphate buffer (pH = 8.0). The chymotrypsin beads were removed by centrifuge tube filters before the selection step. Peptide Synthesis: R6A (MSQTKRLDDQLYWWEYL), Biotin-labeled R6A (Bio- MSQTKRLDDQLYWWEYL), cycPRP-1 (MITWIDFISPSK), cycPRP-2 (MTWFEYLSGSK), cycPRP-3 (MTWFEFLSSTSK), and cycGiBP (MITWYEFVAGTK) were synthesized on Rink Amide AM Resin LL (Novabiochem) and cyclized using DSG as described by Millward et al 8 . After the reaction the cyclized peptides were purified via C 18 HPLC and the mass confirmed by MALDI-TOF MS. Binding Constant Determination: Binding constants were determined relative to the R6A peptide [K d = 60 nM] 55 by equilibrium competition using 35 [S]-Met radiolabeled G i1-GDP and biotinylated R6A as previously described 53 and the data analyzed using GraphPad Prism 5.0. Protease Resistance: Peptides (250 nmol of peptide in DMSO) were added to 50 mM sodium phosphate buffer (pH 8.0) with a final DMSO concentration of 2% (v/v). Sixty units of 65 immobilized chymotrypsin (Sigma Aldrich) were added and allowed to incubate at room temperature. Aliquots were taken at various time points and subsequently filtered. The aliquots were then injected onto a C 18 reverse phase column and separated by a gradient elution from 15 to 90% B in 25 minutes. Solvent A consisted of 0.1% (v/v) TFA in water and solvent B contained CH 3 CN with 0.1% (v/v) TFA. The area under the starting material peak was quantitated using the 32 KaratGold Software package (Beckman). The plotted values represent the mean of two experimental values, and the error bars represent the standard error of the mean. The graph was generated by fitting the data to a one phase exponential decay equation (GraphPad Prism 5.0). K m and V max Determination: The peptides were prepared and characterized as described above in the Protease resistance experiment. Only a single 2 minute time point with varying concentrations of the peptides (0, 5, 11, 22, 65, and 260 µM) was analyzed using Michaelis- Menten enzyme kinetics regression equation (GraphPad Prism 5.0). Human Serum Digests: Lyophilized human serum (Thermo Scientific) was reconstituted by adding 2 mL ddH 2 O to each 5 mg vial from the manufacturer. Each reaction contained 250 nmol of peptide, 50 µL of sodium phosphate buffer (pH 8.0), and 10% DMSO (v/v). One milliliter of reconstituted serum was added to each sample and incubated at 37 ˚C. For each time point, 100 µL aliquots were taken from the reaction and quenched in 300 µL of acetonitrile. These quenched samples were centrifuged to separate precipitate, and the supernatant was diluted in water to 1.5 mL. Samples were then analyzed via HPLC as described above. 66 Circular Dichroism Spectroscopy (CD): Far UV-CD spectra were obtained using a Jasco J810 spectropolarimeter (provided by the USC NanoBiophysics Core Facility) equipped with a Peltier device. The peptides (25–100 μM) were prepared in 10 mM phosphate buffer at pH 7 and placed in a 1 mm path length cuvette. Thereafter, CD spectra were recorded in the range of 195– 240 nm. Five spectra were acquired, and averaged. Spectra were baseline corrected by subtracting blank spectra of the corresponding solutions without peptide and ellipticities were converted to mean residue molar ellipticities in degrees cm 2 dmol -1 .Hepatitis C virus (HCV) is a major human health concern with an estimated 170 million people 67 3.3 Results and Discussion Previously, we created a trillion-member mRNA display cyclic peptide library with the form MXXXXXXXXXK (termed MX 10 K) and used seven rounds of selection to isolate cyclic peptides that bind to the signaling protein Gαi1-GDP 8 . The best binder from that selection was cyclic GiBP (cycGiBP), a specific, 12-residue cyclic peptide with a K d = 2.1 nM. Additionally, we found that cycGiBP was ~3-fold more resistant to chymotrypsin as compared to a linear version of the peptide 8 . In an effort to see if protease-resistant, natural-sequence peptides could be found, we took the Pool 7 library from our original experiment (Fig. 1a) and subjected it to a two-step selection protocol (Figure 1b). After cyclization with DSG, the library of mRNA peptide fusions was first subjected to degradation by immobilized chymotrypsin for 15 minutes at room temperature and then selected for binding to the target of interest, here Gαi1-GDP. The 15 minute digestion should have resulted in loss of the majority of cycGiBP, providing selection pressure to reveal more protease resistant sequences. After three rounds of selection, representative clones from Pool 10 were sequenced (Figure 1c). We then chose three sequences from pool 10 for further characterization. We term these molecules cycPRPs for Cyclic Protease Resistant Peptides (cycPRP-1, cycPRP-2, and cycPRP- 3). The cycPRP sequence consensus in Pool 10 is relatively similar to the consensus sequence seen in Pool 7 and also in other Gαi1-GDP binding peptides. The core sequence of the cycPRPs (TWIDFI, TWFEYL, TWFEFL) is very similar to cycGiBP (TWYEFV) 53 , R6A (YWWEYL) 55 , KB-752 (TWYDFL) 56 , the GSP peptide (TVWEFL) 14 , and the AR6-05 peptides (YWWEFL) 57 selected by both our lab and others. The fact that the sequence consensus is similar suggests that 68 the cycPRP peptides bind to Gαi1-GDP and proteolysis did not destroy binding. Additionally, all previously selected peptides bind near switch 2 of Gαi1-GDP, and based on this observation, we hypothesize that the cycPRPs bind to the same site on Gαi1-GDP. To verify that the cycPRPs bind to Gαi1-GDP as well as determine binding constants, the cycPRPs were initially assessed by in vitro pull-down with immobilized Gαi1-GDP. These data indicated that all three molecules were functional, with cycPRP-3 being the best binder followed by cycPRP-1 and cycPRP-2 (data not shown). Consistent with this view, we determined the K d values of cycPRP-3 to be 8.8 nM and cycPRP-1 to be 88 nM for Gαi1-GDP. This binding is excellent compared to KB-752 (K d = 3.9 µM), and on par with to cycGiBP (K d = 2.1 nM), raising the possibility that the dual selection optimized the product of protease resistance and binding affinity, rather than one or the other. The fact that the sequences retained a large fraction of aromatic residues in the core motif was somewhat surprising given the strong preference for chymotrypsin to cleave the backbone adjacent to tryptophan, tyrosine, or phenylalanine. 57, 58 Indeed, our initial prediction was that the chymotrypsin selection would result in sequences lacking aromatic residues that still retained binding function, which clearly did not occur. Based on the primary sequence, cycPRP-1, cycPRP-2 and cycPRP-3, each have 2-3 chymotrypsin cleavage sites (Figure 2), essentially the same as peptides isolated originally without protease challenge 8 . Initial HPLC/MALDI analysis of chymotrypsin-cleaved peptides revealed that only one cut site was predominant in each of the cycPRP sequences, similar to the result observed originally for cycGiBP. Interestingly, the major cut site varies, depending on the sequence—occurring between Y-E in cycGiBP, W-I in cycPRP- 1, and W-F in cycPRP-2 and cycPRP-3 (Figure 2). 69 We next determined the half-life of the cycPRPs as substrates for chymotrypsin, under conditions similar to the chymotrypsin digestion during the selection (Table 1). The linear GiBP peptide provides a reference point to gauge improvements in protease resistance in this series— and we can use this fixed point to address the effects of protease challenge, cyclization, as well as comparing variations in the primary sequence. Regarding the overall effect of the selection, we observed a 100-200 fold improvement in linear PRPs resistance and a 100-300 fold increase of cycPRPs resistance to chymotrypsin, as compared with linear GiBP. The improved half-life of the PRPs (t 1/2 = 29 to 99 min) is also consistent with the peptides being able to withstand the 15 minute chymotrypsin digestion largely intact. Regarding cyclization, it is clear that cyclization improves chymotrypsin resistance for cycGiBP, but for the protease resistant peptides, these gains are either quite modest (2-fold for cycPRP-1) or non-existent (cycPRP-2, cycPRP-3). The data thus indicate that cyclization is not the primary driver of chymotrypsin resistance in peptides selected using this protease challenge. There are no obvious variations in the primary sequence that would indicate the cycPRPs have improved chymotrypsin resistance as compared with cycGiBP. Put another way, there is no sequence model of which we are aware that predicts the cycPRPs should be more protease resistant than cycGiBP. Amongst the cycPRPs, the number of aromatic amino acids is almost identical to cycGiBP and there are no systematic alterations in the number of beta-branched amino acids, prolines, glycines, or other specific residues near or adjacent to the cut sites. Indeed, at least one statistical model predicts the W-F dipeptide seen at the P 1 -P 1 ' position in cycPRP-2 and cycPRP-3 would be particularly susceptible to cleavage 57 . Previous kinetic work with model substrates indicates that W at P 1 is a markedly better substrate than Y or F (rank order P 1 ; W > Y 70 > F) 54 . In this view, cycGiBP could be seen as anomalous, since the primary cleavage site occurs between Y and E, even though there is a W-Y dipeptide step present. Our work here differs from model kinetic studies in that each of our substrates contains multiple possible cleavage sites. On the other hand, exhaustive analysis and alignment of many model substrates might be expected to reveal the importance of residues at other positions in the chain to the overall cleavage rate. Schellenberger et al 58 ., constructed exactly such a QSAR model based on available k cat /K m data and found that k cat /K m depends on the residue identity from P 3 to P 2 '. Including residues outside this window did not improve the predictive power of the model, and the primary determinant at the P 2 ' position was whether the residue was proline or not. While this model has several caveats (it treats the effect each position as independent and additive), the overall observations are relevant here. Most notable is the predicted variation in cleavage based on residue identity at each position. For P 3 and P 2 , the model predicts variation in k cat /K m of 40-fold (A > R > G, D > P, K) and 10-fold (L,V > P > A > G) respectively. For P 1 , the model predicts a relatively small variation in Y, W, and F (2-fold) and for P 1 ' predicts a 20-fold variation (F, A > L, V, G). In order to quantitatively compare our peptides with previous model of chymotrypsin substrates, we determined k cat /K m values for cycGiBP, cycPRP-1, cycPRP-2, cycPRP-3 (Table 2). The k cat /K m value for cycGiBP is 1.1 x 10 5 sec -1 M -1 which is in line with model amide bonded substrates, 58, 59 supporting the view that cycGiBP acts as normal substrate for the enzyme. Interestingly, the increased protease resistance of the PRP peptides originates from changes in both k cat and K m —with substrates showing 8 – 16 fold decreases in binding and 8 – 50 fold decreases in catalytic cleavage rates. cycPRP-2 and cycPRP-1 have the smallest k cat /K m values, reduced 420-fold and 350-fold as compared to cycGiBP. This reduction is remarkable 71 because it substantially exceeds what would have been expected if all the aromatic residues had been converted to leucine 57 . cycPRP-3 shows a 110-fold reduction in k cat /K m compared to cycGiBP, but is also the highest affinity binder. This data also argues that the selection optimized the product of the binding affinity and the reduction in k cat /K m , such that survival represents a compromise between these selected traits. Previous observations have demonstrated that, at least with proteins, protease resistance can result from differences in protein structure 60, 61 . While peptides are generally unstructured in solution, a possible mechanism that could reduce the cleavage rate of the PRPs would be if the peptides had significant levels of secondary structure. In this model, only the unfolded form of the peptide would be available for binding and cleavage by the enzyme. This model would not account for changes in k cat , but could contribute to changes in K m by decreasing the fraction of free peptide available. To test this, we took the CD spectrum of both the linear and cyclic versions of GiBP, cycPRP-1, cycPRP-2, and cycPRP-3 in phosphate buffer at 10 °C (Figure 3) and at 60 °C (see supplemental information). The reduced temperature was used in order to maximize structure formation, as peptide folding transitions are often quite broad due to their modest enthalpies of formation. 43, 62 The largest structure difference between linear and cyclic molecules is seen for cycGiBP (Figure 3a) and very small differences are seen for cycPRP-1, cycPRP-2, and cycPRP-3 (Figure 3b-d). These spectra do not conform simply to established model helix, sheet, or coil spectra. 63 The lack of a minima near λ = 222 nm indicates the peptides are not helical and the overall spectra look most similar to the anti-parallel beta sheet structure. 63 However, the spectra also are similar to unstructured peptides and show little changes upon heating the samples to 60 °C (see supplemental information). The CD spectra and lack of 72 temperature dependence are most consistent with both the linear and cyclic peptides lacking regular secondary structure, although they do not eliminate non-canonical folded forms. The lack of peptide structure and the relatively small importance of cyclization imply that the majority of the protease resistance results from sequence effects that alter the efficiency of these peptides as chymotrypsin substrates. This could be because the peptides are optimized to be poor chymotrypsin substrates or because of more general changes that make the peptides poor protease substrates. To test this, we incubated cycGiBP, cycPRP-1 and cycPRP-3 in human serum containing active proteases and peptidases and examined the half-life of the molecules (Table 3). Serum generally degrades peptides very quickly 64 due to the presence of multiple proteases with differing specificity (e.g., thrombin, plasmin, and kallikrein). In line with that view, linear GiBP is undetectable at the first time point after incubation (t 1/2 < 1 min) and cycGiBP shows a half-life of ~20 minutes. Surprisingly, cyclization of the PRPs gives similar dramatic improvements in stability against serum digestion with increases of 20- to 60-fold for cycPRP-1 and cycPRP-3 respectively. This difference is consistent with serum aminopeptidase, which can digest linear peptides, but whose activity is blocked when the peptides are cyclized 65 The fact that the cyclic PRPs show dramatically improved stability against digestion by both chymotrypsin and the proteases present in serum, even though the peptides were never exposed to serum proteases during the selection, argues that the peptides have been optimized in a general way to resist proteolytic cleavage. This optimization is somewhat idiosyncratic, as the most stable serum peptide (cycPRP-3) is the least stable of the series against chymotrypsin alone. 73 3.4 Conclusions Our two-step selection protocol (mRNA display using protease challenge and binding selection) demonstrates that it is possible to isolate highly functional natural peptides that have the ability to bind a target of interest and that are resistant to proteases. This procedure results in stability increases ranging from 100- to 400-fold. Mechanistically, the protease resistance results from both a lower rate of cleavage (k cat ) and a weaker interaction with the enzyme (K m ). Surprisingly, the protease resistant peptides do not contain dramatic sequence changes compared with non-protease resistant molecules. Only a few changes are needed to increase proteolytic stability, with 4-6 relatively conservative amino acid changes (out of 10 positions) needed to convert a peptide susceptible to protease into one resistant to protease. These changes are not predictable with existing heuristics/models, do not remove key residues that are substrates for the protease (here W, Y, and F), and do not appear to increase the ordered structure of the molecules. Rather, the changes appear to affect endopeptidase cleavage generally, perhaps through local conformational biases. Nonetheless, these results are exciting because they provide a route to create protease-resistant homologs of biologically-active peptides for use as reagents, diagnostics and therapeutics. 74 3.5 Figures Figure 1: mRNA Display selection for chymotrypsin resistance. (a) Pool 7 of the cyclic peptide library (MX 10 K) targeting Gαi1-GDP was used as the starting point for the selection. (b) In rounds 8-10, the library was cyclized and subjected to chymotrypsin degradation and binding selection. (c) Representative clones from Pool 10 were sequenced (see supplemental information) and peptides cycPRP-1, cycPRP-2, and cycPRP-3 were further characterized. 75 Figure 2: Theoretical (red) and observed (blue) chymotrypsin digest sites of peptides that bind Gαi1-GDP. cycGIBP is from Pool 7 of the original selection for function only, while the three cycPRP peptides (cycPRP1-3) are from Pool 10 of the protease/binding selection. Red arrows denote theoretical digest sites and blue arrows show the actual digest sited as determined by MALDI-TOF. 76 Figure 3. CD spectra of linear and cyclic versions of GiBP, PRP-1, PRP-2, and PRP-3. All spectra were taken at 10°C in 10 mM phosphate buffer pH 7.4 with linear peptides are shown in blue and cyclic peptides in red. (a) GIBP. (b) PRP-1. (c) PRP-2. (d) PRP-3. 77 Peptide Linear Half-Life (Minutes) Cyclic Half-Life (Minutes) GiBP 0.3 3.1 PRP-1 47 99 PRP-2 58 57 PRP-3 32 29 Table 1: Chymotrypsin resistance of linear and cyclic Peptide K m (µM) V max (µM min -1 ) k cat /K m (M -1 sec -1 ) Fold Improvement cycGiBP 1.4 37 1.1 x 10 5 1 cycPRP-1 11 0.8 3.1 x 10 2 350 cycPRP-2 22 1.4 2.6 x 10 2 420 cycPRP-3 18 4.2 9.9 x 10 2 110 Table 2: K m and V max enhancements for cycGiBP and cycPRPs Peptide Linear Half-Life (Minutes) Cyclic Half-Life (Minutes) GiBP <0.02 0.3 PRP-1 0.4 7.7 PRP-3 0.4 26 Table 3: Stability of linear and cyclic peptides in human serum 78 3.6 Supplementary Figures Figure S1: Evolution of sequences in response to the addition of chymotrypsin as a selective pressure. A. Sequences from Pool 7 of the mRNA display selection targeting G i1-GDP. This selection used only binding to G i1-GDP to select for several high affinity peptides . B. Sequences from Pool 10, which employed both chymotrypsin digestion and binding to G i1- GDP. 79 Figure S2: CD spectra of cycPRP1-3 in 10mM phosphate buffer pH 7.4 at 60°C. A. GIBP. B. PRP-1. C. PRP-2. D. PRP-3. 80 Chapter 4: Cell-permeable and Biologically Stable Peptide Therapeutics Aimed at Disrupting the HDM2-p53 Interaction 4.1 Introduction HDM2 and p53: The tumor suppressor protein p53 is a transcriptional factor responsible for cell cycle arrest 66 and ultimately pro-apoptotic signaling during cell stress (i.e., DNA damage, hypoxia, etc.). 67, 68 Mouse Double Minute 2 (MDM2), or its human homologue, HDM2, is an E3 ligase which regulates p53 levels inside the cell by poly-ubiquitinating p53 and targeting it for protease degradation. 69 Approximately 50% of human cancers retain the wild-type (wt) tumor suppressor p53 70 , but its function is inhibited by several intracellular inhibitors, among which HDM2 (an E3 ubiquitin ligase) and its homolog MDM4 are major players. 71, 72, 73, 74 Restoration of endogenous p53 activity can inhibit in vivo tumor growth by inducing apoptosis and innate inflammatory responses. 75, 76, 77, 78 Mounting evidence shows that HDM2 is a clinically relevant cellular oncoprotein. HDM2 overexpression has been observed in many human carcinomas such as non-small cell lung cancer, breast cancer, esophageal cancer, leukemia, non-Hodgkin’s lymphoma, melanoma, and sarcoma. 79 A recent study estimates that ~2-3 million cancer patients, out of ~9 million new cases a year, have elevated expression of HDM2/MDM4 and thus would benefit from therapeutics targeting these genes. 71 For this reason, there is great interest in finding molecules that can inhibit HDM2 and restore the function of p53 to cancerous cells, leading to their apoptosis. An easy way to achieve this goal is to disrupt the interaction between HDM2 and p53 81 using ligands that bind to these proteins at the interaction site. Researchers have shown that inhibiting the HDM2/p53 interaction can lead to cell apoptosis and reduction in the size of the tumor in vivo. 80 Targeting the interaction: Disrupting protein-protein interactions (PPIs) is challenging. 81 This problem is especially difficult when the interactions occur intracellularly. Historically, small molecules have been used to target intracellular PPIs due to their superior, though incomplete, membrane permeability. 82 Small molecule therapeutics are limited, in that, due to their small size, they rarely interact with the entirety of the PPI interface, leading to lower affinity and selectivity for their targets. 83 Hence, only ~10-15% of human genes are thought to be viable targets for small molecule drugs 84 , and small molecule drugs are associated with a host of side effects due to their non-specificity. Various pharmaceutical companies have tried to create small molecule therapeutics to inhibit the HDM2/p53 interaction for over 10 years. 85 Although some success has been made in utilizing small molecules as HDM2 antagonists 86, 87, 88 , their µM dissociation constants (K d ) could pose challenges for therapeutic applications 89, 90 . These low affinities also underscore the challenge of designing small molecules to block large and flexible PPI surfaces 91 . Currently 5 small molecule drugs targeting HDM2 are in various stages of phase I clinical trials for treatment of Soft Tissue Sarcoma, Neoplasms, Acute Myeloid leukemia, Lymphoma, and other advanced malignancies. Another class of promising therapeutics targeting PPIs is peptides. Due to their larger size and surface area, peptides are better able to fit the interaction site on proteins, leading to higher affinity binders with greater specificity. 92 Many peptides/peptidomimetics have been developed as HDM2 antagonists and some exhibit nanomolar affinity and can bind both HDM2 and 82 MDM4. 78, 91, 93 There are two main drawbacks for using peptides as therapeutic reagents. The first problem is the peptide’s vulnerability to proteases and peptidases, abundantly present in human serum and inside cells. 45 The second problem involves the peptide’s inability to pass through the cell membrane. Current HDM2 antagonist peptides exhibit low cell permeability and poor pharmacokinetic properties, making them unlikely drug candidates. Therefore, there is an unmet need to develop high affinity compounds with drug-like properties to inhibit p53/MDM interactions 90 . Using mRNA Display: The biological stability and membrane permeability of peptide therapeutics have been studied extensively in the Roberts lab. Li et al. showed that in vitro selection experiments could be conducted using unnatural amino acids 4 , that highly stabilized N- methyl phenylalanine peptides could be incorporated in a display format. 5 More recently, Fiacco et al. showed by performing an N-methyl scanning mutagenesis (wherein a natural residue is substituted with an N-methyl analogue; e.g., N-methyl leucine in place of leucine) 94 that even a single N-methyl residue can confer dramatic increases in protease resistance (70-fold to >1,000- fold). The protease resistance was achieved even when the unnatural amino acid was not inserted at the scissile bond location. These results combined with the construction of trillion member covalent macrocycle libraries provides an ideal start for an protease resistant peptide selection containing unnatural amino acids 8, 53 . A new type of mRNA display selection was devised by Fiacco et al. (manuscript in preparation) which uses scanning libraries of relatively small diversities to find unnatural amino acid containing peptides that are protease resistant. This type of selection was dubbed SUPR mRNA display (Scanning Unnatural Protease Resistant). The Roberts lab has also made headways into making membrane permeable peptides (fiacco et. al, 83 manuscript in preparation) by attaching small molecules (i.e. biotin) or fatty acids (i.e. palmitoleic acid). We planned on using SUPR mRNA display to generate therapeutic reagents which are biologically stable, specific, and cell permeable, which target the HDM2/p53 pathway, and characterize these reagents in vitro and in situ. 84 4.2 Materials and Methods HDM2 synthesis and purification: The HDM2(1-125) sequence was ordered as a gBlock from Integrated DNA technologies and PCR amplified by primers 5HDM2Nde (5’ – CCAATGCAGCATATGAATACCAACATGTCtGTACCTACT – 3’) and 3HDM2_Bio_XhoI_1 (5’ – GAAAATATCGTTCAGGCCTCCAGCGCCACCGCCGTTCTCACTCACAGATGTACCTGA GTC – 3’) followed by a second PCR amplification with primers 5HDM2Nde and 3HDM2_Bio_XhoI_2 (5’ – GTGGTGCTCGAGCTCGTGCCATTCGATTTTCTGAGCTTC GAAAATATCGTTCAGGCCTCC – 3’) with PfuUltra II Polymerase (Agilent). These primers add restriction sites for cloning and a C-terminal Avitag used for in vivo biotinylation. The DNA was purified with a spin PCR cleanup column (Macherey-Nagel), digested with Nde I and Xho I (New England Biolabs), and ligated into pET24a (Novagen). The resulting plasmid was sequenced to confirm the correct plasmid sequence. The HDM2-pET24a plasmid was transformed into BL21(DE3) cells bearing a Kan-resistant plasmid encoding the BirA enzyme (Avidity). Protein was expressed overnight using auto- induction media. 33 The cells were lysed with B-per (Pierce) and run on a NiNTA column using an FPLC. The column was washed with Buffer A (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM imidazole), Buffer A + salt (20 mM Tris-HCl, pH 7.5, 1 M NaCl, 10 mM imidazole), and eluted with a linear gradient with Buffer B (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 100 mM imidazole). Fractions were screened using SDS-PAGE followed by concentration using Amicon spin concentrators (Millipore). The proteins were aliquoted and frozen at -80 ºC. 85 Synthesis of THG73-NMV: THG73 tRNA template (5’- AATTCGTAATACGACTCACTATAGGTTCTATAGTATAGCGGTTAGTACTGGGGACTC TAAATCCCTTGACCTGGGTTCGAATCCCAGT AGGACCGC-3’, 96nt) was ordered from Integrated DNA Technologies (IDT), along with the complementary strand. The tRNA was in vitro transcribed using T7 RNA polymerase, 95 phenol extracted using Phase Lock Gel (Brinkmann Instruments, Inc., Westbury, NY), and desalted by ethanol precipitation. The resulting THG73 tRNA was purified using an 8% denaturing urea page gel and the RNA product was excised from the gel and extracted using an elutrap (Whatman). The synthesis of N-methyl, N-nitroveratrylcarbonyl norvaline cyanomethyl ester was carried out according to the published protocol. 96 The final product was purified by silica gel chromatography in 3:1 ethyl acetate to hexanes. The N-methyl, N-nitroveratrylcarbonyl norvaline cyanomethyl was dissolved in dimethylformamide (DMF) and coupled to phosphodeoxycytidine-phosphoadenosine (pdCpA) tetrabutylammonium salt under nitrogen at room temperature overnight. The reaction was stopped with the addition of water and the final product was purified using HPLC on a C18 column. Following purification, the pdCpA-N- methyl, N-nitroveratrylcarbonyl norvaline was ligated to THG-73 tRNA using T4 RNA ligase in ligation buffer (100mM HEPES pH 7.5, 10 mM DTT, 50 mM MgCl2, 400 µM ATP, 120 mg BSA). The product was phenol-extracted and desalted by ethanol precipitation. Deprotection of the nitroveratryloxycarbonyl group was accomplished by photolysis with a xenon lamp equipped with a 315-nm cutoff filter, and the N-methyl norvaline tRNA was immediately added to the translation reaction. 86 SUPR mRNA Display: Using MIP(R9A) and PMI(N8A), we designed three biased libraries such that each position a mixture of the wild type residue and N-methyl norvaline coded by the UAG amber codon would be present. The library based on MIP(R9A) 97 peptide has a theoretical diversity of 5.6 x 10 8 sequences. The MX 12 K and the MX 10 K libraries based on PMI(N8A) 78, 98 peptide have theoretical diversities of 3.7 x 10 8 and 2.9 x 10 8 sequences respectively (Figure 2). Additional nucleotides were used at specific positions in order to increase the diversity of amino acids at said positions. The DNA for the libraries was ordered from IDT, and was amplified and extended with the 5’ and 3’ primers shown in the Appendix. After PCR, the DNA was phenol-extracted, desalted by ethanol precipitation, and resuspended in 10 mM Tris-HCl pH 8.0. The Round 0 mRNA pool was generated by in vitro transcription (as described above for tRNA transcription) and purified by Urea-PAGE. The purified mRNA was ligated to F30P (5’-dA21[C9]3dAdCdC-P; C9=triethylene glycol phosphate (Glen Research), P= puromycin (Glen Research)), a flexible DNA linker containing puromycin via an oligonucleotide splint (Appendix). Following PAGE purification of the ligation reaction, the template was dissolved in water and quantitated by absorbance at 260 nm. 26 pmol of each library was combined for a single translation reaction. The ligated mRNA was translated in rabbit reticulocyte lysate under standard conditions. 99 The translation reaction was supplemented with 20 µg of NMF-tRNACUA in 1 mM NaOAc (pH=4.5). After 1 hr. of translation at 30 ºC, KOAc and MgOAc were added to a final concentration of 600 mM and 50 mM respectively and the reactions were placed at -20 ºC for 1 hr. 87 Translation mixtures were diluted 1:10 in dT Binding Buffer (10 mg/mL dT cellulose, 1M NaCl, 20 mM Tris, 1 mM EDTA, 0.2% Triton X-100, pH=8) and agitated for 1 hr. at 4 ºC. The dT cellulose was filtered and washed with dT Wash Buffer (1 M NaCl, 20 mM Tris, , 1 mM EDTA, 0.2% Triton X-100, pH 8.0). The fusions were eluted using 2x 100 µL washes with 65 ºC DI water. The elutions were added to 200 µg of chymotrypsin on agarose beads (Sigma-Aldrich) and 200 µg of proteinase K on agarose beads (Sigma-Aldrich), and digested for 30 seconds. After the digest, the proteases were separated from fusions by Spin-X filters (Corning). The supernate was then incubated with 300 pmols of HDM2 immobilized on NeutrAvidin Agarose resin (Life technologies), and resuspended in 1 mL selection buffer (1X PBS, 0.1% BSA, 0.1% Tween20, 100 µg/mL yeast tRNA). After 1 hour, the beads were washed 7x with selection buffer, resuspended in 50 µL DI water. The solution was RT-PCRed using the “SuperScript One-Step RT-PCR with Platinum Taq” kit from life technologies to recover a double stranded DNA library. Further rounds of selections were performed as above using varying amount of proteases, digestion time, and immobilized HDM2 (Table 1). Round 3-4 were also performed using HDM2 immobilized on streptavidin magnetic beads (dynabeads). Pool 4 was analyzed by high throughput DNA sequencing via an Ion Torrent device at EvoRX. Peptide synthesis and purification: Peptides HUS 4.01 (biotin-MSL*LL*WLEQREGK- NH 2 ), HUS 4.02 (biotin- MSL*LL*WLEQRRGREV-NH 2 ), HUS 4.13 (biotin- MSLL*EYWLE*REGK-NH 2 ), HUS 4.26 (biotin-M*Q*LLQSHE*WVGK-NH 2 ), and MIP(R9A) (biotin-MPRFWEYWLALME-NH 2 ), were synthesized by solid phase Fmoc 88 synthesis, using a Biotage Alstra Microwave Synthesizer. 30 The peptides were synthesized on Rink amide MBHA resin using five-fold molar excess of each amino acid and HATU. After the coupling of the last amino acid, methionine, the biotin was coupled to the N-terminal amine, resulting in biotin-labeled peptides. Peptides were cleaved from the resin and deprotection with a solution of 95% (v/v) TFA, 2.5% 1,2-ethanedithiol (EDT), 1.5% (v/v) deionized water (DI), and 1% (v/v) triisopropylsilane (TIS) for 2 hours at room temperature. 100 The resin was filtered out, and the peptide was precipitated using 4-fold (v/v) excess ether. The peptides were dried, resuspended in DMSO, and HPLC purified using a C 18 reverse phase column and a gradient of 10-90% acetonitrile/0.1% TFA in water. Fractions were collected and tested for the correct molecular weight using MALDI-TOF mass spectrometry. The correct fractions were lyophilized, dissolved in DMSO, and flash frozen at -80 °C. Radiolabeled Competition assay: The DNA for the parental peptide, MIP(R9A), was ordered from IDT and PCR amplified using the pool 5’ and its own 3’ primer (Appendix). The DNA was phenol-extracted, ethanol precipitated, and resuspended in 10 mM Tris-HCl pH 8.0. The DNA was then in vitro transcribed, ligated to F30P, and gel purified as described above. The ligated mRNA was translated with 35 S-methionine supplemented translation reaction. The radiolabeled fusions were dT purified and eluted with DI water as described above. Radiolabeled fusions were incubated with 30 pmols of HDM2 in 1mL of selection buffer and washed. The flow through, washes and the beads were counted via a scintillation counter. In separate tubes, the radiolabeled fusions were first mixed with 100 pmols of synthetic MIP(R9A), or 100 nmols of synthetic HUS 4.01, 4.13 or 4.26 peptides. After an hour of incubation with 30 89 pmols of HDM2 immobilized on magnetic beads, the beads were washed and the flow through, the washes and the beads were counted. Serum Half-Life Measurement: Delipidated/lyophilized human serum was purchased from Thermo Scientific and reconstituted as per manufacturer’s instructions. 250 nmoles of peptide in 50 μLs of sodium phosphate buffer (pH 8.0) with 10% DMSO was added to 1 mL of reconstituted serum and incubated at 37˚C. 100 μL aliquots were taken at various time points and quenched in 300 μLs of acetonitrile. Samples were spun down and decanted to remove precipitate followed by dilution in water to 1.5 mLs. Samples were injected onto a C18 reverse phase column and separated by gradient elution (15–90% B in 25 min. Solvent A: 0.1% TFA in water. Solvent B: acetonitrile (0.05% TFA). The area under the curve was corrected for the background peak and analyzed by Microsoft Excel. The data points were normalized to the 0 minute time point, and fit to an exponential decay curve. XTT Cell Viability Assay: HCT116 p53+/+ and HCT116 p53-/- cells (National Cancer Institute, Developmental Therapeutics Program, Bethesda, MD) were maintained in RPMI-1640 supplemented with 10% heat-inactivated FBS. Cells were seeded in 96-well plates (2,000 cells/well) and allowed to attach overnight before the indicated treatments. Cells were treated with peptides for 72 hours. At the end of treatment, cells were incubated with 50 µL of the activated-XTT solution (1:50 volume ratio) for 3 hours at 37 °C. The plates were shaken gently and the absorbance at 450 nm was measured on a microplate reader. All experiments were performed in triplicate. 90 4.3 Results and Discussion SUPR mRNA display: In order to find peptide ligands capable of disrupting the HDM2-p53 interaction, we used Scanning Unnatural Protease Resistant (SUPR) peptide selection via mRNA display (Figure 1.) This method allows for the incorporation of unnatural amino acids into peptides and selects for both protease resistance and function (affinity for desired target) simultaneously. This method starts with a peptide that is known to bind to the desired target, and allows for the wild type amino acid codon or the translational termination signal “TAG” at each amino acid position (scanning at each position). The way that the SUPR library is synthesized also allows for the possibility of several natural amino acids at each codon position alongside the wild type amino acid and the stop codon. During the translation step, the nonsense suppressor tRNA charged with the unnatural amino N-methyl norvaline (NMV) is added to the reaction. At the sites of the “TAG” stop codon, the ribosome will incorporate the unnatural amino acid to the nascent polypeptide chain. The peptide-mRNA fusion library is then sieved for both protease resistance and binding function and taken through the mRNA display cycle ~4-6 times. When convergence is reached, the enriched pool is sequences exhaustively by Next Generation Sequencing. Library Construction: There are several natural peptides in the literature which bind to HDM2 and inhibit its interaction with p53. 78, 97, 98, 101 We chose two such peptides which bind HDM2 and disrupt its interaction with p53: MIP(R9A) 97 and PMI(N8A) 78, 98 . Three libraries were made from these two peptides (a 13 amino acid long library based on MIP(R9A), and a 12 and a 10 amino acid long libraries based on PMI(N8A) peptide). The libraries were constructed 91 so at each amino acid position, the codon TAG or the wild type amino acid codon was used (Figure 2, DNA sequences for the libraries are shown in the Appendix). This means that at most 8 codons (3 positions, and 2 choices per position) and 8 amino acids could appear at each position (i.e. M at position 12 of the MIP(R9A) library). Generally the choices were fewer due to shared nucleotide between the stop codon and the wild type amino acid (i.e. F at position 4 of the MIP(R9A) library), or the degeneracy of the codons(i.e. P at position 2 of the MIP(R9A) library). In order to increase the diversity of amino acids available at certain position, extra nucleotide choices were added to the mix (i.e. L at position 9 of the PMI(N8A) MX 10 K library). The diversities of the libraries were purposefully kept within the same order or magnitude to ensure similar frequency of each unique sequence when the libraries would be combined. Selection: Each library was PCR amplified, transcribed, and ligated to the DNA linker attached to puromycin separately. Before the translation of pool 0, equal moles of each library was combined and taken through the rest of the mRNA display cycles as a single pool. Having similar diversities ensured that the library with the much smaller diversity would not start with a much higher copy number per sequence. The translation reaction was supplemented with THG73 amber suppressor tRNA charged with NMV. This allowed the incorporation of NMV at the sites of TAG stop codons. The fusions were dT purified, and digested with a protease cocktail. The digestion time and the amount of each protease in the cocktail for each round of selection are shown in Table 1. After the digest, the pool was sieved for function (binding to HDM2) and the surviving sequences were reverse transcribed and PCR amplified. After 4 rounds of selection, the pool was exhaustively sequenced by high throughput DNA sequencing using an Ion Torrent device at EvoRX location. The highest frequency clone, HUS 92 4.01, accounted a staggering 14% of the total pool 4 library (Table 2), and ~80% of the pool was derived from the MIP(R9A) MX 13 K library. This was somewhat surprising to us since according to published data, PMI(N8A) has a higher affinity for HDM2 than MIP(R9A) (490 picomolar 98 vs. 36 nanomolar 97, 102 ). Aside from a tryptophan residue at position 8, a leucine at position 9 and a glutamic acid at position 13, the rest of the peptide showed very little resemblance to the parental sequence. Within the pool, however, there was a significant amount of sequence homology (shown in bold in Table 2). The three sequences shown in table 2 were chosen and synthesized via solid phase peptide synthesis, HPLC purified, and checked for the correct molecular weight by MALDI-TOF mass spectrometry. The peptides were made with an N- terminal biotin in order to increase their membrane permeability. The parental peptide, MIP(R9A), was also synthesized and purified with an N-terminal biotin. Relative Affinity: In order to find the highest affinity clone among the three tested peptides, we performed a radiolabeled competition assay. Radiolabeled parental peptide fusions, MIP(R9A), were made by an in vitro translation reaction supplemented with 35 S-labeled methionine. Labeled MIP(R9A) fusions were then incubated with HDM2 immobilized on magnetic beads in the presence of absence of competing peptides (Figure 3a). In the absence of any competing peptide, radiolabeled MIP(R9A) fusions was able to bind to the beads at ~6%. Mixing 100 nM non-labeled MIP(R9A) peptide was sufficient to reduce the binding by more than 10-fold. The addition of each of the three unnatural amino acid containing peptides at 1000- fold higher concentration (100 µM) also reduced the signal from radiolabeled MIP(R9A). However, none of the unnatural containing peptides showed as high of an affinity for HDM2 as 93 the parental peptide. HUS 4.01 showed the lowest IC 50 value of ~100 µM in a dose dependent response (Figure 3b). Protease Stability: The ability of the peptides to resist proteolysis was measured by incubating peptides with human serum at 37 ºC. Human serum contains a variety of peptidases and proteases. Even though during this selection, the peptides were not specifically selected for digest resistance against the collection of proteases present in the human serum, previous experiments have shown us that resistance to a protease such as chymotrypsin can infer a general protease resistance against proteases present in the human serum. 9 Several time points were taken during the digest, and the serum proteins were precipitated and removed by the addition of an equal volume of acetonitrile. The supernatant was then analyzed on a C18 HPLC column as described in Millward et al. 32 The retention profile of the sample with or without human serum addition is shown in Figure 4a. The intact peptide peak (between 29.5 and 30.5 minutes) was isolated for each time point, and the area under the curve was integrated (Figure 4b). By normalizing the integrated areas to the first time point (0 minutes) and fitting the data to a simple exponential decay curve, we obtained a 23 hour half-life for HUS 4.01 (Figure 4c). This result is very impressive since natural peptides generally have a half-life of mere minutes in human serum. 9 Inhibition of Cancer Cell Growth: So far we had shown that peptide HUS 4.01 is resistant to proteolysis and binds to HDM2 at the same site as p53. In order to test the ability of HUS 4.01 to disrupt the HDM2/p53 interaction, we used HCT116 +/+ colorectal carcinoma cells. These cells overexpress HDM2 and are unable to proliferate when the HDM2/p53 interaction is blocked. 85 94 As a negative control, to ensure the specificity of our peptide, we used HCT116 -/- cells lacking a functional copy of the p53 gene. HUS 4.01 or MIP(R9A), each with an N-terminal biotin, were incubated with cells for 72 hours. The result of the cell viability assay (XTT) is shown in Figure 5. Neither MIP(R9A) nor HUS 4.01 peptide affect the cell viability of the HCT116 -/- cells line (Figure 5a). This result was expected, since even if either peptide is able to bind to HDM2 and stop its function inside the cell, lack of p53 gene results in the non-initiation of the apoptosis signal. HUS 4.01 peptide, however, significantly reduces the viability of the HCT116 +/+ cell line when administered at <100 µM, whereas the parental peptide is significantly less potent (Figure 4b-c). The results of the XTT assay demonstrate the fact that it is crucial for peptide therapeutics aimed at disrupting intracellular protein-protein interactions to be resistant to proteolysis. HUS 4.01 peptide’s affinity for HDM2 is over 1,000-fold less than the parental peptide’s, however it is a more potent therapeutic under in situ conditions. The ability of the HUS 4.01 peptide to only reduce the viability of the HCT116 cell line when the functional copy of p53 gene was present demonstrates the specificity of the peptide for its target, and its specific mode of action. 95 4.4 Conclusions Using SUPR mRNA display we developed HUS 4.01, a peptide therapeutic candidate to disrupt the intracellular interaction of HDM2 and p53. HUS 4.01 is a cell-penetrable and highly biologically stable peptide which is able to kill colorectal cancer cells overexpressing HDM2 when a functional copy of the p53 gene is available, and has no significant effect on the same cells without a copy of the p53 gene. HUS 4.01 is also much more potent than its parental peptide, MIP(R9A), even though its affinity for HDM2 is over 1,000-fold lower. The higher therapeutic efficacy of HUS 4.01 considering its much lower affinity argues for the importance of protease resistance in peptide therapeutics aimed at intracellular targets. The lack of sequence homology between HUS 4.01 and the parental MIP(R9A) peptide shows the importance of the mRNA display’s selection process. We have previously demonstrated that peptide analogs after the replacements of their amino acids with N-methyl versions of the amino acids usually lose their binding and selectivity against intended targets. 94 The fact that HUS 4.01retains its binding function while gaining protease resistance is a significant advance in the field, not only due to its potential use as a therapeutic, but because it shows the validity of using SUPR mRNA display as a method for developing effective peptide therapeutics against other intracellular targets. 96 4.5 Figures Figure 1: SUPR mRNA Display. (a) Schematic for SUPR mRNA Display. A double stranded DNA library is in vitro transcribed and ligated to a DNA linker attached to puromycin. The ligated mRNA is then added to a translation reaction supplemented with THG73 amber suppressor tRNA charged with N-methyl norvaline. The unnatural amino acid containing peptide fusions are then incubated with a protease cocktail and affinity sieved for the target of interest (here HDM2). The pool, now enriched in both HDM2 affinity and protease resistance, is reverse transcribed and PCR amplified to generate the DNA pool for the next round of selection. (b) The THG73 amber suppressor tRNA incorporates unnatural amino acids into the nascent polypeptide chain. At the site of the stop codon UAG, THG73 can hybridize with the mRNA template in the A site of the ribosome. The ribosome will then add the unnatural amino acid to the polypeptide chain being formed. 97 Figure 2: The three libraries constructed for the SUPR mRNA display selection against HDM2. The DNA sequence at each amino acid position (yellow row) is a combination of the wild type amino acid codon and the TAG stop codon used to code for the unnatural amino acid. At certain positions, more nucleotides were chosen in order to increase the amino acid diversity. (a) The MX 13 K library based on the MIP(R9A) 97 peptide with the theoretical diversity of 5.6 x 10 8 sequences. The MX 12 K (b) and the MX 10 K (c) libraries based on PMI(N8A) 78, 98 peptide with theoretical diversities of 3.7 x 10 8 and 2.9 x 10 8 sequences respectively. 98 SUPR mRNA Display Pool 0 Pool 1 Pool 2 Pool 3 Pool4 Digestion Enzyme Chymotrypsin (µg) 200 200 400 400 Trypsin (µL) - 10 20 20 Proteinase K (µg) 200 200 400 400 Digestion Time (sec) 30 60 60 120 Amount of HDM2 (pmols) 300 300 30 30 PCR Cycles 23 23 29 27 Table 1: Selection conditions for the SUPR mRNA display selection. The amount of proteases used in the digestion step, the digestion time, the amount of HDM2, and the PCR cycles used to recover the library for each round of selection is shown above. 99 Name Sequence Copy number in Pool 4 Pool 4 Composition HUS 4.01 MSL*LL*WLEQREG 10,033 14.3% HUS 4.13 MSLL*EYWLE*REG 458 0.7% HUS 4.26 M*Q*LLQSHE*WVG 208 0.3% MIP(R9A) Library MPRFWEYWLALMEX - - Table 2: Sequences from pool 4 of the SUPR mRNA display selection against HDM2. The chosen sequences for further characterization are shown above with their copy number and sequence composition in the pool. Conserved amino acids from the library are shown in red, while the consensus sequence in the pool is shown in bold. 100 Figure 3: The unnatural amino acid containing peptides compete with MIP. (a) Radiolabeled MIP(R9A) binds to magnetic beads with immobilized HDM2. When radiolabeled MIP(R9A) and 100 nM synthetic MIP(R9A) (unlabeled) are mixed before incubation with beads, the signal is lowered significantly. While HUS 4.01, 4.13, and 4.26 all reduce the signal from radiolabeled MIP(R9A), HUS 4.01 shows the highest level of affinity for HDM2. (b) HUS 4.01 reduces the signal for radiolabeled MIP(R9A) binding to immobilized HDM2 on magnetic beads in a dose dependent fashion. 101 Figure 4: Protease stability of the HUS4.01 sequence. (a) HPLC retention time for the HUS 4.01 peptide. The elution profile from the C18 column for the peptide (red line), peptide in human serum (blue line), and two negative controls (buffer, and human serum) are shown. (b) The area under the curve from 29.5 minute mark to the 30.5 minute mark was selected to measure the amount of intact peptide. (c) HUS 4.01 peptide has a half-life of 23 hours in human serum at 37 ºC. The amount of peptide was calculated by integrating the area under the curve from the 29.5 to 30.5 time marks, and subtracting the background signal. The area under the curve was then normalized to the 0 time point value. 102 Figure 5: HUS 4.01 peptide is able to inhibit the growth of colorectal carcinoma cells more efficiently than the parental MIP(R9A) peptide. For these experiments, HUS 4.01 peptide and the parental MIP(R9A) peptides were made with an N-terminal biotin conjugate in order to increase cell permeability. Cells were incubated for 72 hours with the peptides in an XTT assay. (a) Neither the MIP(R9A) peptide nor the HUS4.01 peptide significantly changes the cell viability of HCT116 cells which lack the p53 gene. (b) HUS 4.01 peptide is able to reduce cell viability significantly (40% reduction at 125 µM peptide concentration) while the parental peptide did not reduce cell viability significantly. (c) HUS 4.01 peptides results on colorectal carcinoma cells are repeatable. 103 4.6 Appendix The DNA Sequences for ordered and used for this project DNA Libraries: MIP(R9A) MX 13 K: TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATGYMSYDGTWST NGKHGYAKTNGYWBKMSYWBWWSKHGNNSAAGGGAGGTAGCTCAGGCGGA PMI(N8A) MX 12 K: TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATGKMKKMKYWSK MSKMGYAKTNGKMSYWSYWSKMKYMSAAGGGAGGTAGCTCAGGCGGA PMI(N8A) MX 10 K: TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATGKVKYWSKHSB MGYWKYNGKHSYHSYWSKVKAAGGGAGGTAGCTCAGGCGGA Mixed nucleotide guide: R A/G Y C/T M A/C K G/T S C/G W A/T H A/C/T B C/G/T V A/C/G D A/G/T N A/C/G/T 5’ PCR Primer: TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATG 3’ Primer: TCCGCCTGAGCTACCTCC Splint: TTTTTTTTTTTNTCCGCCTGAGCT 104 MIP(R9A) Template: GGGACAATTACTATTTACAATTACAATGCCCAGATTCTGGGAGTACTGGCTGGCCC TGAT MIP(R9A) 3’ Primer: TCCGCTGCCAGATCCGCT CTTCTCCATCAGGGCCAGCC MIP(R9A) Splint: TTTTTTTTTTTNTCCGCTGCCAGA 105 Chapter 5: Generating Ultrahigh Affinity Peptide Ligands Using a Fragment-based Approach This work has been adapted from the following publication: Takahashi TT, Jalali-Yazdi F, Roberts RW. Generating Ultrahigh Affinity Peptide Ligands. Manuscript submitted to Nature Chem Bio (2015). 5.1 Introduction Peptides are potentially attractive reagents for protein capture and detection, due to their simplicity, size, stability, and ease of production and chemical modification 103 . A major concern regarding peptide ligands is that they tend to have relatively modest target affinity compared to antibody-based reagents, giving decreased sensitivity. A second concern regarding peptides is their specificity. Peptides that have been isolated from screening protocols can possess unacceptably high nonspecific binding. This issue is particularly acute for hydrophobic peptides or where the target protein has many related homologs in the genome. Small molecule-based discovery screens have benefitted from a fragment approach, wherein a lead compound (or set of compounds) is coupled with a library of fragments, in an effort to walk across the surface of the protein target, thereby gaining improved affinity and selectivity 104 . To date, a fragment strategy has not been explored as a means to systematically improve peptides. Using this strategy, significant improvements could be achieved, particularly given observations that affinity can scale with contact surface, arguing that constructing ultrahigh affinity peptide- protein complexes should be feasible 105, 106 . Here, we implemented a step-wise fragment-based discovery strategy in an effort to create high affinity peptides. To do this, we coupled mRNA display with a three-step approach. First, a 106 random library (X 9 ) is sieved to isolate a family of peptides that bind a model target analyte, here, Bcl-x L . After the initial selection, we pursued a convergent strategy wherein all the binders were extended from their N-terminus with a second naïve library, followed by further selection. Finally, the best member of this extension library was subjected to mutagenesis and reselection to reveal both the sequence conservation at each position and to give highly optimized binders. The results indicate that low picomolar affinity can be easily achieved with peptides only 21 residues in length. 107 5.2 Materials and Methods First Round of mRNA Display Selection: DNA coding for the Primary (Extlib) library was ordered from the Keck Oligonucleotide Facility at Yale (5’ – GCTGGAGCCACTGCCAGATCCCA5655655655655655655655655655CATTGTAATTGTA AATAGTAATTGTCCC – 3’; where 5 = 30% of A and C, and 20% of G and T; 6 = 60% C and 40% G. 31 These molar ratios of were selected to adjust for the different coupling rates of the phosphoramidites; the target percentages of the randomized positions were 5 = 25% each of A, C, G, and T; 6 = 50% G and C). The DNA was gel purified on 8% denaturing urea-PAGE, electroeluted, and ethanol precipitated. The Extlib ssDNA library was PCR amplified for six cycles using the 5’ primer 47T7 (5’ – GGATTCTAATACGACTCACTATAGGGACAATTACTATTTACAATTAC – 3’), the 3’ primer 3Extlib (5’ – GC TGGAGCCACTGCCAGATCCCA – 3’), and 10 nM Extlib library in 1X PCR buffer (10 mM Tris-HCl, pH 9.0, 50 mM KCl, 2 mM MgCl 2 , 0.1% (v/v) Triton X-100, 0.2 mM each dNTP) using Taq polymerase in a total volume of three mL. The DNA was phenol extracted and ethanol precipitated. Half of the resulting DNA was transcribed into RNA using T7 RNA polymerase (80 mM Hepes-KOH, pH 7.5, 2 mM spermidine, 40 mM DTT, 25 mM MgCl 2 , 20 mM each NTP, pH 8, 10 µg/mL of T7 RNA Polymerase) in a total volume of three milliliters. EDTA (pH 8.0) was added to stop the reaction, and the mRNA was purified by phenol extraction, ethanol precipitation, 8% urea-PAGE, and electroelution. Five thousand pmol of mRNA was ligated to 5,000 pmol of pF30P (5’ – phosphate-(dA) 21 - [S9] 3- dAdCdC-P; S9 = spacer 9, P = puromycin (Glen Research)) using 7,500 pmol of ExtlibSplint (5’ – TTTTTTTTTTTGCTGGAGCCAC – 3’) and T4 DNA Ligase in 1X ligase 108 buffer (50 mM Tris-HCl pH 7.5, 10 mM MgCl 2 , 1 mM ATP, 10 mM DTT). The ligated RNA was gel purified via 8% urea-PAGE, electroeluted, and ethanol precipitated. The ligated mRNA-F30P was quantified on a nanodrop and used to program an in vitro translation reaction. Four hundred pmol of ligated mRNA-F30P was used in a volume of 1 mL in 1X translation buffer (100 KOAc, 0.5 mM Mg(OAc) 2 , 20 mM Hepes-KOH pH 7.6, 100 mM creatine phosphate, 2 mM DTT, 312.5 μM of each amino acid, and 40% (v/v) rabbit reticulocyte lysate (Green Hectares; prepared according to the method of Jackson and Hunt 99 ). After incubation at 30 °C for one hour, 2.5 M KCl and 1 M MgCl 2 were added to final concentrations of 0.5 M and 50 mM, respectively, to facilitate peptide-mRNA fusion 95 . After 30 minutes at room temperature, the fusions were purified by dT chromatography. Seventy-five mg of dT cellulose (GE Healthcare) was washed with 1X Isolation Buffer (100 mM Tris-HCl pH 8.0, 1 M NaCl, 0.2% (v/v) Triton X-100) three times, and added to the fusion in a total volume of 10 mL of isolation buffer. The samples were incubated at 4 °C for one hour, transferred to three 0.45 µm Spin-X filters (Corning) and washed five times with 700 µL of isolation buffer. The fusions were eluted with two, 250 µL elutions of 65 °C water and ethanol precipitated in the presence of linear acrylamide (30 µg/500 mL elution). The desalted fusions were adjusted to 1X RT buffer (50 mM Tris-HCl pH 8.3, 75 mM KCl, 3 mM MgCl 2 , 2.4 mM 3Extlib primer, 200 mM each dNTP,) and the sample was heated to 65 °C for 5 minutes and cooled on ice to anneal the 3’ primer. After cooling, 33.3 µL of Superscript II enzyme was added and the reaction incubated at 42 °C for one hour. Superscript II was inactivated by heating to 65 °C for 5 minutes, after which the samples were cooled on ice. Bcl-x L agarose beads were prepared by adding 1,000 pmol of biotinylated Bcl-x L to 50 µL of prewashed neutravidin agarose beads (bed volume) in 1X PBST++ (10 mM Na 2 HPO 4 , 1.8 mM 109 KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl, pH 7.4, 0.2% Tween-20 (v/v), 1 mg/mL (w/v) Bsa, 50 µg/mL yeast tRNA) and incubating at 4 °C for one hour. The beads were washed with 700 µL of 1X PBST++, then transferred to the RT reaction containing cDNA/mRNA peptide fusions. Ten µL of 2 mM D-biotin in 1X PBS buffer (10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl, pH 7.4) was added to block any unbound sites on neutravidin. The cDNA/mRNA peptide fusions and immobilized Bcl-x L were incubated at 4 °C for 1 hour, then washed five times with 700 µL of 1X PBST++. The bound fusions were PCR amplified as above, using 13 cycles of PCR and a final volume of 1 mL. Subsequent Rounds of mRNA Display Selection: In subsequent rounds of mRNA display, the mRNA display selection cycle was performed as described above with the following exceptions: The transcription volume was reduced to 1 mL, and after phenol extraction and ethanol precipitation, the mRNA was desalted using a NAP-25 column (GE Healthcare) and 7,500 pmol of mRNA was used in a 250 µL ligation reaction. The translation volume was reduced to 100 µL, after which 100 µL of a 25% (v/v) slurry of dT cellulose in isolation buffer was added. After dT purification, the fusions were eluted with two, 50 µL elutions of 65 °C ddH 2 O, and desalted through a Centrisep column (Princeton Separations). The RT volume was reduced to 125 µL, and selections incubated in 1 mL of 1X PBST++. After binding to immobilized Bcl-x L and washing with 1X PBST++, the bound fusions were eluted with two, 100 µL elutions of 0.15 % (w/v) SDS, and the SDS removed by incubation with SDS-out (Pierce), following the manufacturer’s recommendations. The number of PCR cycles was determined empirically by PCR until the samples were approximately 0.1 µM (final concentration). Binding 110 for all selection rounds was performed at 4 °C, except for the last round, where binding was performed at room temperature. Radioactive Binding Assays: Radiolabeled mRNA peptide fusions were in vitro translated as above, except total translation volume was 50 µL and 35 S-labeled methionine (Perkin Elmer; 40 µCi) was substituted for the cold methionine. After dT purification, 100,000 cpm were added to immobilized Bcl-x L and incubated at 4 °C for 1 hour. The samples were transferred to a Spin-X tube, washed three times with 700 µL of 1X PBST++, and the flowthrough, washes and beads counted via scintillation. Percent bound was calculated by dividing cpm on beads by total cpm (flowthrough + washes + beads). Construction of the Extension Library: DNA from the first Bcl-x L selection or naïve library were PCR amplified as above, except the 5’ primer 5AcuExtLib (5’ – TAATACGACTCACTATAGGGACACTGAAGATTTACAATTACAATG – 3’) was used to amplify the DNA from the 5 th Round of the Bcl-x L selection and the 3’ primer 3BpmExtLib (5’ – GTTCTGCTTGCTGGAGCCACTGCCAGATCCCA – 3’) was used to amplify the naïve Extlib library. The one mL PCR reactions were phenol extracted, ethanol precipitated, and the resulting DNA resuspended in 500 µL of 1X TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0). For the Bcl-x L selection DNA, the PCR products were digested with AcuI (1X NEB Buffer #2 (10 mM Tris-HCl pH 7.9, 50 mM NaCl, 10 mM MgCl 2 , 1 mM DTT), 0.8 mM S-adenosylmethionine, and 50 U of AcuI (NEB)) at 37 °C for four hours. The naïve DNA generated from PCR with the 3BpmExtLib primer was digested with BpmI (1X NEB Buffer #3 (50 mM Tris-HCl, pH 7.9, 100 mM NaCl, 10 mM MgCl 2 , 1 mM DTT), 100 mg/mL Bsa (NEB), and 20 U of BpmI (NEB)) at 37 111 °C for four hours. The DNA was gel purified via a PCR cleanup column (Macherey-Nagel). Equal volumes of the AcuI-cut Bcl-x L selected library and the BpmI-cut naïve library were mixed in 1X ligase buffer (50 mM Tris-HCl pH 7.5, 10 mM MgCl 2 , 1 mM ATP, 10 mM DTT) with 50 µL of T4 DNA ligase in a total volume of 1 mL. The DNA was purified on an 8% native polyacrylamide gel, and electroeluted. The yield of the NExt library was 3.3 µg (~40 pmol). The purified DNA was PCR amplified as above with the 47T7 and 3Extlib primers, with 1.6 pmol of template in a total volume of 3 mL of PCR (for the N-terminal library) using 9 cycles. The DNA was purified by phenol extraction and ethanol precipitation, and transcribed in a total volume of 2 mL. mRNA was purified via 6% urea-PAGE and electroeluted. For ligation of the pF30P, 2,500 pmol of mRNA was ligated in a total volume of 250 µL. mRNA was gel purified as above, and 200 pmol of ligated mRNA-F30P used to program a 0.5 mL translation. Fusions were dT purified, reverse transcribed, and selected as described above. Doped Selection: A doped library (5’ – GGGACAATTACTATTTACAATTACAATG1442121351441351444171134131121122352352 13313443433142433142GGTAGTGGAAGCGGCTCCAGC – 3’; where 1 = 7.1% T, 10.7% C, 75.0% A, 7.1% G; 2 = 9.1% T, 13.6% C, 13.6% A, 63.6% G; 3 = 7.1% T, 75.0% C, 10.7% A, 7.1% G; 4 = 63.6% T, 13.6% C, 13.6% A, 9.1% G; 5 = 60% C, 40% G; 7 = 93.1% C, 6.9% G) based on the E1 peptide sequence was ordered from the Keck Facility at Yale. The ssDNA was purified and PCR amplified in a two mL reaction as above with the 5’ 47T7 primer and 3’ 3BclDoped primer (5’ – GCTGGAGCCGCTTCCACTACC – 3’). The 3’ primer was purposefully changed from the original library to avoid cross contamination. The DNA was transcribed into mRNA in a one mL reaction and ligated in a 750 µL reaction with 112 BclDopedSplint (5’ – TTTTTTTTTTTNGCTGGAGCCGC – 3’) as described above. The purified ligated mRNA-pF30P construct was translated in a 500 µL reaction, purified, and reverse transcribed as above, to yield a total of ~1 x 10 12 unique sequences with ~3 copies each. Radiolabeled Off-Rate Assay: To determine the off-rate of mRNA-peptide fusions, dT purified fusions were allowed to bind to immobilized Bcl-x L for 1 hour in 1X PBST++. The beads were transferred to a Spin-X column, and washed 3 times with 700 µL of 1X PBST++. The beads were resuspended in 500 µL of 1X PBST++ and transferred to 581 µL of 17.2 µM non-biotinylated Bcl-x L (10,000 pmol; ~100X molar excess relative to biotinylated Bcl-x L on beads) in 1X PBS buffer. At various time points, 100 µL of slurry was removed from the binding reaction, and spun through a Spin-X column (~20 s). The percent remaining at each time point was determined by dividing the cpm on beads by total cpm (beads + flowthrough). Binding reactions were performed at room temperature, unless otherwise stated. To determine the off-rate of radiolabeled peptides, 1 µM desalted mRNA coding for the appropriate peptide followed by an HA-tag (YPYDVPDYA) was translated with 35 S-methionine as above. The HA-tagged peptides were purified by immunoprecipitation with 10 µL of anti-HA magnetic beads (Pierce) in 250 µL of 1X TBST (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% (v/v) Tween-20), incubated at 4 C for two hours, and washed with 800 µL of 1X TBST eight times. The peptides were eluted with 100 µL of 50 mM NaOH for two minutes at room temperature, then neutralized with 30 µL of 1 M Tris-HCl, pH 8.0. The radiolabeled HA- tagged peptides were then used in binding and off-rate assays as described above, except 20 pmol of Bcl-x L was immobilized on streptavidin magnetic beads (Dynal). 113 Off-Rate Selections: The optimal time for competition was determined using the equations of Boder and Wittrup 107, 108 . The approximate off-rate of each pool was determined using the radiolabeled off-rate assay above. After binding the reverse transcribed fusions for 1 h, 10,000 pmol of non-biotinylated Bcl-xL was added, similar to the off-rate assay. The competition times ranged from 2.3 h in Round 4 to 137 h in Round 6. Bcl-x L(1-209) and Biotin-labeled Bcl-x L expression and purification: The gene for the first 209 amino acids of Bcl-x L (Clone HsCD00004711; Dana Farber/Harvard Cancer Center DNA Resource Core) was PCR amplified using Pfusion polymerase with primers 5NdeBclxl (5’ – GAAGGTCGTCATATGTCTCAGAGCAACCGGGAG – 3’) and 3XhoBclxl (5’ – CCTATCGT ACTCGAGGCGTTCCTGGCCCTTTC – 3’) and cloned into pET24a using NdeI and XhoI. For Biotin-Bcl-x L , an N-terminal avitag (AGGLNDIFEAQKIEWHEGG) 34 was added via PCR using the 5’ primer 5BiotagBclxl (5’ – AAGGTCGTCATATGGCTGGAGGCCTGAACGATATTTTCGAAGCTCAGAAAATCGAA TGGCACGAGGGTGGGATGTCTCAGAGCAACCGGGAG – 3’) and 3XhoBclxl and similarly cloned into pET24a. Plasmid DNA was miniprep purified and sequenced (Laragen). The plasmids were transformed into BL21(DE3) cells and expressed overnight at 37 C using auto-induction media. 33 Cells were lysed using Bper (Pierce), and loaded onto a Ni-NTA superflow column using an FPLC (Bio-Rad), washed with >10 column volumes of Buffer A (Buffer A: 25 mM Hepes pH 7.5, 1 M NaCl, 10 mM imidazole), then eluted using a gradient from 100% Buffer A to 100% Buffer B (Buffer B: 25 mM Hepes pH 7.5, 1 M NaCl, 400 mM imidazole). The purity of both biotinylated and non-biotinylated Bcl-x L was analyzed by SDS- PAGE. 114 Pure fractions of non-biotinylated Bcl-x L were concentrated in a stir cell protein concentrator and buffer exchanged into 1X PBS (10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl, pH 7.4). For the radiolabeled binding assays, the protein was further purified using a Superdex size exclusion column. The protein was frozen on liquid nitrogen and stored at -80 C. Pure fractions of the avi-tagged Bcl-x L were combined, concentrated, and desalted into 50 mM Tris-HCl, pH 8.0. Bcl-x L was biotinylated in vitro using BirA biotin ligase (0.1 mg/mL in 50 mM Tris-HCl, pH 8.3, 10 mM ATP, 10 mM Mg(OAc) 2 , 50 µM biotin) at 30 C for two hours. The protein was buffer exchanged into 1X PBS, frozen in liquid nitrogen, and stored at -80 C. Testing the Activity of Bcl-x L protein and In Vitro Biotinylation: The antisense DNA coding for the Bak peptide (5’ – GGGACAATTACTATTTACAATTACAATGGGCGGCCAGGTGGG CCGCCAGCTGGCCATCATCGGCGACGACATCAACCGCTCGGGGTCTGG – 3’) was purified and PCR amplified as described above using either the 3’ primer 3GlySerRP (5’ – ACC GCTGCCAGACCCCGA – 3’) to make the BakShort peptide (Amino acid sequence: NH 2 – MGGQVGRQLAIIGDDINRSGSGSG–COOH) or the 3’ primer 3BakLongExt (5’ – ACCGCTGCCAGACCCCGACTGGAACTCGGAGTCGTACCGGCGGTTGATGTCGTCGCC GAT – 3’) to make the BakLong peptide (Amino acid sequence: NH 2 – MGGQVGRQLAIIGDDINRRYDS EFQSGSGSG–COOH). Radiolabeled mRNA-peptide fusions corresponding to the BakShort or BakLong were synthesized, purified, and tested for binding to immobilized Bcl-x L as described above. No binding was observed without in vitro biotinylation (data not shown). 115 Peptide Synthesis: Peptides were synthesized on a Biotage Alstra Peptide Synthesizer using Rink Amide MBHA resin (loading 0.45 mmol/g) at the 0.2 mmol scale (444 mg of resin). Fmoc amino acids were dissolved in N-methyl-2-pyrrolidine (NMP) to a concentration of 0.5 M. The resin was swollen in for 20 minutes at 70 ºC in 5 mL of NMP then drained. Five milliliters of 25% (v/v) 4-methylpiperidine was added, the solution heated to 70 ºC for 1 minute, then the vial was drained. An additional 5 mL of 25% 4-methylpiperidine was added and the solution heated to 70 ºC for 3 minutes. The resin was washed five times with 4.5 mL of NMP. Fmoc amino acid (0.5 M in NMP; 5 equivalents), HATU (1-[Bis(dimethylamino)methylene]-1H-1,2,3- triazolo[4,5-b]pyridinium 3-oxid hexa-fluorophosphate; 0.5 M in NMP; 4.9 equivalents), and DIEA (N,N-diisopropylethylamine; 2 M in NMP; 10 equivalents) were added, and the solution heated for 8 minutes at 70 ºC, followed by 6 washes of 4.5 mL of NMP. The Fmoc was then deprotected using 25% 4-methylpiperidine (as above) 109 and the procedure repeated for synthesis of the appropriate peptide. After addition of the final amino acid, the final Fmoc was deprotected as above. Synthesis of Biotinylated Peptides: Biotinylated peptides were synthesized by incorporating Fmoc-Lys(Mtt)-OH at the C-terminus of the peptides. Rink amide MBHA resin was swollen, and Fmoc-Lys(Mtt)-OH added as above. After addition of Fmoc-Lys(Mtt)-OH, the resin was removed from the synthesizer and washed 5 times with dichloromethane (DCM). The Mtt group was removed using 50-150 mL of 1% (v/v) trifluoroacetic acid (TFA) in DCM, which was poured over the resin with the synthesis vial attached to vacuum until no more yellow color appeared in the eluate. Care was taken not to let the resin dry and avoid potential TFA concentration caused by DCM evaporation. The resin was then washed with DCM (5 times), 116 NMP (5 times), 10% DIEA in NMP (2 times), and NMP (5 times). A Kaiser test was used before (negative) and after deprotection (positive) to confirm successful mtt protecting group removal. The synthesis vial was then placed back in the synthesizer, washed with NMP and swollen in 5 mL NMP for 5 minutes at 70 ºC. D-Biotin (0.25 M in 50% (v/v) DMF/50% (v/v) DMSO; 2.5 equivalents), HATU (2.4 equivalents), and DIEA (5 equivalents) were added and the solution heated to 70 ºC for 8 minutes. The vial was drained and the biotin coupling was repeated. The resin was washed five times with 4.5 mL NMP, and then the rest of the peptide was synthesized as above. Binding Specificity Analysis: The binding specificity of the selected peptides was determined using previously published protocols. 110 The genes for Bcl-x L (Amino acids 1-209; 5’ primer: 5BclxlFLAGIVT, 5’ – GGGACAATTACTATTTACAATTACAATGTCTCAGAGCAACCGGGAGCTGGTG – 3’; 3’ primer: 3BclxlFLAGIVT, 5’ – TCACTTGTCATCGTCGTCTTTGTA- GTCGCCGCGTTCCTGGCCCTTTCGGCT – 3’), Bcl-w (Amino Acids 1-172; 5’ primer: 5BclwFLAGIVT, 5’ – GGGACAATTACTATTTACAATTACAATGGCGACCCCAGCCTCGTCCCCAGA – 3’; 3’ primer: 3BclwFLAGIVT, 5’ – TCACTTGTCATCGTCGTCTTTGT- AGTCGCCTGTCCTCACTGATGCCCAGTTC – 3’; DF/HCC Clone# HsCD00000436), A1 (Amino Acids 1-152; 5’ primer: 5A1FLAGIVT, 5’ – GGGACAATTACTATTTACAATTACAATGACAGACTGTGAATTTGGATATATTTAC – 3’; 3’ primer: 3A1FLAGIVT, 5’ – TCACTTGTCATCGTCGTCTTTGTAGTCGCCAGATTTAGGTTCAAACTTCTTTAC – 3’; 117 DF/HCC Clone# HsCD00002275), Bcl-2 (Amino Acids 1-217; 5’ primer: 5Bcl2FLAG-IVT, 5’ – GGGACAATTACTATTTACAATTACAATGGCGCACGCTGGGAGAACGG – 3’; 3’ primer: 3Bcl2FLAGIVT 5’ – TCACTTGTCATCGTCGTCTTTGTAGTCGCCCAGAGACAGCCAGGAGAAATCAAACAG A – 3’; DF/HCC Clone# HsCD00005947), and Mcl-1 (Amino Acids 172-327; 5’ primer: 5MclFLAGIVT 5’ – GGGACAATTACTATTTACAATTACAATGGACGAGTTGTACCGGCAGTC – 3’; 3’ primer: 3MclFLAGIVT, 5’ – TCACTTGTCATCGTCGTCTTTGTAGTCGCCGCCACCTTCTAGGTCCTCTACAT – 3’; DF/HCC Clone# HsCD00004566) were PCR amplified using PfuUltra II polymerase as directed by the manufacturer, and transcribed into mRNA as described above. After transcription, 1/10 (v/v) 0.5 M EDTA, pH 8.0 was added, and the samples phenol extracted. The mRNA was desalted by spinning through a Microcon YM-30 at 14,000  g, and washed with ddH 2 O three times. The mRNA was quantified by UV absorbance and 6 µg of mRNA was used to program a 35 S-methionine-labeled in vitro transcription reaction as described above. Three hundred microliters of M2 FLAG beads were washed with FLAG buffer (50 mM Tris-HCl, pH 7.5; 150 mM NaCl; 0.05% (v/v) Tween-20) and resuspended in FLAG buffer as a 50/50 (v/v) slurry. To the translation reactions, 60 µL of washed FLAG beads (slurry), and 0.5 mL of FLAG buffer were added, and the samples incubated at 4 ºC for 1 hour. The binding reactions were transferred to a 0.45 µm Spin-X tube, washed 13 times with 700 µL of FLAG buffer, and eluted with 100 µL of 3X FLAG peptide in FLAG buffer. The specific activity of the eluted protein was determined by scintillation counting. 118 To prepare the peptide resin, 100 µL neutravidin agarose was washed three times with 1 mL of 1X PBST++. To the washed resin, 1,000 pmol of biotinylated peptide was added, the samples incubated for 20 minutes at room temperature, then were washed three times with 1 mL of 1X PBST++. Fifty µL of 1X PBST++ was added to the beads to make a 50/50 (v/v) slurry. As a positive control, we used a synthetic Bim peptide (NH 2 - DMRPEIWIAQELRRIGDEFNAYYARRGK(Biotin)-NH 2 that binds approximately equally to all proteins that we tested. For each binding reaction, 1 mL of 1X PBST++, 10 µL of peptide beads, and 150,000 to 250,000 cpm of each protein were incubated at 4 ºC for 1 hour. The samples were transferred to a Spin-X column, washed 3X with 700 µL of 1X PBST++, and counted in a scintillation counter. The percent bound was calculated from the cpm on beads divided by total cpm (beads + flowthrough + washes). All proteins, except for A1, were functional as judged by radioactive pull down with immobilized Bim peptide. All binding data for A1 samples were omitted from further analysis. Data represents the average of three trials, and error represents standard error. Calculation of K i for Peptide A10: From Matsumura et al. 111 , the IC 50 of peptide A10 interacting with Bcl-x L was determined by competitive inhibition using fluorescence polarization assays. In these experiments, 190 nM fluorescent-Bak peptide (K d 337  32 nM) and 400 nM GST- Bcl-x L were used to determine an IC 50 of 900 nM for peptide A10. Using the Cheng- Prusoff equation: 𝐾 i = IC 50 1 + [L] 𝐾 d we determined K i for the peptide A10/Bcl-x L interaction to be 575 nM. 119 Illumina Sequencing and Analysis: The final DNA from the Doped selection was PCR amplified with primer 5IluSeqHS1_GGG (5’ – GCATTCCTGCTGAACCGCTCTTCCGATCTNNNGAGGGGACAATTACTATTTAC   AATTACA – 3’) and primer 3BD_HS1_GCT (5’ – CCCTACACGACGCTCTTCCGATCTNNNGCTGGAGCCGCTTCCACTACC – 3’) followed by a second PCR with primer 5IluSeqHS2 (5’ – CAAGCAGAAGACGGCATACGAGATCGGTCTCGGCATTCCTGCTGAACCGCT – 3’) and 3IlusxUniv (5’ – AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATC T – 3’) to add adapters for Illumina sequencing. The final PCR product was gel purified and gel extracted, quantitated by Nanodrop, and sequenced at the USC Genome and Cytometry Core. The Illumina data were processed by Python scripts developed in house that extract the Doped sequences with the appropriate bar code, translate the DNA sequences to peptide sequences, and count the number of times each peptide sequences occurs in the Pool. The composition of the sequences was analyzed using Excel. Bcl-x L Quantitation ELISA: Fifty µL of 30 nM streptavidin in 1X PBS was incubated with an ELISA plate overnight at 4 °C. The plate was washed 3X with wash buffer (1X PBS + 0.1% (v/v) Tween-20) and blocked with 350 μL of 1X PBS + 5% (w/v) Bsa for two hours. One hundred µL of 30 nM D1 peptide in assay buffer (1X PBS, 0.1% (v/v) Tween-20, 1 mg/mL Bsa) was added to wells and incubated for 1 hour. The plate was washed, and 100 µL of serially diluted Bcl-x L in assay buffer was added to each well and incubated for 2 hours. The plate was washed and incubated with 100 µL of probe antibody (54H6 mAb, Cell Signaling Technologies) 120 in assay buffer for an hour. The plate was then washed and incubated with a 1:1,000 dilution of anti-rabbit antibody conjugated to HRP (54H6 mAb, Cell Signaling Technologies). After the wash, the plate was incubated with 100 µL of TMB substrate, and the reaction was stopped after approximately 10 minutes by the addition of 100 µL of 2 M sulfuric acid. The absorbance at 450 nm was measured via a plate reader (Molecular Devices). The OD450 values and the concentration of each sample were fit to a four parameter logistics curve using the least absolute deviation method and Excel Solver. Enzymatic K d Calculation Assay: The K d values of the peptides were determined using a protocol modified from Friguet et al. 112 A set of serially diluted Bcl-x L standards, at 2X the desired concentration, were made in sample buffer. For each peptide ligand, a set of dilutions at 2X the desired concentrations were also prepared. The Bcl-x L samples were either mixed 1:1 with sample buffer (standards) or ligands (samples), and allowed to incubate at room temperature for 6 days. To analyze the samples, the ELISA plates were incubated overnight at 4 C with 1.5 nmol of streptavidin in 1X PBS. Plates were washed 3X with wash buffer (1X PBS + 0.1% (v/v) Tween- 20) and blocked with 1X PBS + 5% (w/v) Bsa for two hours. 100 µL of a 30 nM solution of the D1 peptide (capture ligand) was added to wells and incubated for 1 hour. After the capture ligand incubation, 100 µL of sample or standards were added in each well, and incubated for 1 hour at room temperature. Plates were washed, incubated with HRP-conjugated anti-HIS tag antibody in sample buffer for 1 hour, washed, and incubated with TMB substrate (Thermo Scientific). Reactions were stopped after approximately 10 minutes with 2 M sulfuric acid, and the absorbance at 450 nm was measured via a plate reader (Molecular Devices). 121 The OD450 for the standards and their concentration values were fit to a four parameter logistic curve (standard curve). The concentration of the free Bcl-x L in solution (responsible for the signal) for each sample was calculated using the standard curve, and converted into percent of Bcl-x L bound by peptide in solution. For each peptide, the values for all the tested concentration of Bcl-x L and peptide in solution were fit simultaneously to the monovalent equilibrium model to obtain the K d . 122 5.3 Results and Discussion Bcl-x L Binding Peptides from the Primary X 9 Pool: Bcl-x L was selected as a model target because 1) it is an important cancer drug target (overexpression suppresses apoptosis and plays an important role in adenocarcinoma of the pancreas, prostate, and colon) 113 , 2) because it is a member of a larger family (the Bcl-2 family) with ~ 20 naturally occurring homologs, and 3) because this protein has been the target of numerous development efforts including drugs (e.g., ABT-737) 114 , linear peptides 111 , and chemically cyclized (e.g., stapled) peptides 52 . Ligands that target Bcl-x L could thus serve as potential therapeutics or diagnostics. There are two concepts we wished to explore with this work: 1) to test the method of extending functional peptides to achieve high affinity binders and 2) to gauge the maximal affinity vs. chain length that can be achieved by a peptide ligand. The overall strategy is shown in Figure 1a- c. Subjecting a target to a naïve library often generates many different independent peptide ligands (Fig. 1a, b). Each of these could be improved by extension (Fig. 1c) 115 , but it is likely that some peptides could be improved more than others, and it is impossible to know a priori which sequence would benefit the most from this strategy. Thus, extending the entire family of binders at once provides a route to create more optimized, higher affinity ligands from a large number of potential leads in an unbiased and systematic fashion. We began with a naïve primary library of the form MX 9 3, (X = all 20 natural amino acids; 3 = is M, L, or V) and performed an mRNA display selection vs. Bcl-x L (Fig. 2a) (see Online Methods). The Primary library’s theoretical complexity was 1.5  10 12 and we began the selection with an experimental library of at least 7 x 10 12 peptides, giving an average of 3 copies of each sequence in the starting library. After five rounds of selection, we observed significant 123 binding above background for the pool (Fig. 2a) and for individual clones (e.g., clone P1 (Fig. 2b) and clones P2 and P3 (Supplementary Fig. 1)). All eight sequences obtained by Sanger sequencing were unique (Fig. 2c), and the best binding peptide (P1) shows homology with known Bcl-x L ligands Bim and Bad at a universally conserved Asp residue and the hydrophobic positions H2, H3 and H4 (Fig. 2d). Sequences P2-8 show no homology with natural Bcl-2 family ligands 116, 117 . The selection thus produced ligands (clone P1) that likely mimic the natural ligands in their sequence and binding site (i.e., give a convergent solution) and others (clones P2- 8) that bind Bcl-x L through a different mechanism of recognition. A key goal of this work is addressing how target affinity scales with the chain length of the peptide ligand. Binding analysis of P1 indicated its off-rate is 6.3  10 -3 s -1 and its K d = 65 nM 118 (Fig. 2d). This is consistent with previous work with the highly optimized peptide ligands R6A1 (8 residues, K d = 200 nM) 119 and linear GiBP (10 random positions, K d = 31 nM) 32 . Left unanswered though is the affinity limit that could be achieved by extension of the library. Extension Selection Yields Picomolar Peptide Ligands: Our MX 9 3 library was designed so that it could be extended at either the N-terminus or the C-terminus. Here, we have explored extensions to the N-terminus of the binders, fusing a second X 9 3 segment to the entire Round 5 pool (Fig. 1c), creating an N-terminally extended library (NExt). After four rounds of selection, a high proportion of the pool bound to immobilized Bcl-x L (~70% of the library, Fig. 3a) as compared with the shorter Primary peptides (~15% of the library, Fig. 2a), indicating that the NExt library possessed sequences with higher affinity than the Primary pool. To further enrich the highest affinity peptides, we used the competition approach of Boder and Wittrup to isolate sequences with the slowest dissociation rates 108 in rounds 5-7 of the selection. As the selection 124 progressed, the off-rate of the NExt pool consistently decreased, from 5  10 -4 s -1 in Round 3, to 7  10 -6 s -1 in Round 7 (Supplementary Fig. 2). The coupling of mRNA display with high-throughput sequencing methods has greatly enhanced both the speed and quality of ligand generation 120 . High-throughput sequencing of the Round 7 library showed that one family of sequences dominated the pool and one sequence, clone E1, a 21-residue peptide, was most abundant (Fig. 3b). The E1 family has some similarity to the P1 sequence, shows homology to natural Bcl-2 family ligands at the Asp, H3 and H4 positions, and bears positively charged residues between H2 and H3 (Fig. 3c). Taken together the sequence information argues E1 binds the same site and with a similar orientation to the natural peptides. E1 differs from the majority of Bcl-2 family ligands at H2, where most natural peptides have leucine, but E1 utilizes a hydrophobic tyrosine instead. Binding analysis of two peptides derived from the NExt library, E1 and E2 (an N9K mutant of E1), indicates they bind with very high affinity, with K d = 39  6 pM and 300  14 pM, respectively (Fig. 3b, Supplementary Fig. 3). For comparison, the four highest affinity approved therapeutic monoclonal antibodies (adalimumab, basiliximab, infliximab, cetuximab) bind with K d values between 100 to 200 pM 121 . E1 thus has better target affinity than this group despite being over 60-fold smaller by molecular weight. E1 has ~1,000-fold higher affinity than the Bcl- x L -directed stapled peptide SAHB A , a 21 residue peptide with K d = 38 nM 52 and 17,000-fold higher affinity than a previous peptide selected for binding to Bcl-x L (K i = 575 nM) 111 . E1 also has 1,500- to 6,000-fold higher affinity than designed Bcl-2 family binding miniproteins (K d = 52 nM) 122 or helical scaffolds (K d = 220 pM) 123 . The picomolar affinity of the relatively short 21- residue E1 peptide demonstrates that a structured protein scaffold is not required for ultrahigh binding affinities. Finally, based on the abundance of these sequences, high throughput 125 sequencing indicates that there are potentially >2,000 independent peptides with K d ’s less than 300 pM in the round 7 library. Doped Selection: Review of the literature indicated that the E1 peptide is the highest affinity protein-directed peptide ligand known. On the other hand, it was unclear if this represented the maximal affinity possible, since there are more than 20 20 = 10 26 possible 20 residue peptide sequences, a number that cannot be searched exhaustively by any experimental technique. To address the question of the maximal affinity, we constructed a Doped library based on the E1 peptide and performed additional rounds of selection with this library against Bcl-x L . The Doped library was constructed with each residue as ~40% wild type and ~60% the other 19 amino acids. After four rounds of binding and off-rate selection for Bcl-x L , we sequenced the library and identified Peptides D1 and D2 (Fig. 3d). Analysis of the Doped family of sequences indicates they have even higher affinity than the NExt family and are thus the highest affinity protein-directed peptides known. The D1 and D2 sequences have off-rates ~2  10 -6 s -1 (Fig. 3d) and D1 gives a K d = 8.5  2 pM (Supplementary Fig. 3). For comparison, the off-rate of the biotin-streptavidin complex under similar conditions is 2.4 x 10 -6 s -1 (See Supplementary Note 1), a higher dissociation rate constant than the D1 peptide. Interestingly, high throughput sequencing indicates that the D1 sequence is far from unique and that >10,000 independent sequences have equal or greater abundance in the final pool. We next worked to determine the extent that the Doped family of sequences was optimal (Fig. 4). Examining the high throughput sequence data, we noticed that many of the Doped family clones were very similar to the original E1 sequence, even though the naïve doped library was 126 moderately randomized (1 in 1 x 10 8 sequences were expected to be the E1 sequence). If a residue was optimal, that position in the chain would be expected to show little or no variation. Among the top 10,000 Doped family clones, eight positions show no variation whatsoever and three more show only one highly homologous substitution (Fig. 4a). This result is striking because some positions are directly adjacent to each other and define a core motif of nine consecutive residues that is essentially intolerant of any mutation. Many of the observed changes are conservative (e.g., E3D, F17Y, A18S) with an average of 2.4 conservative mutations in the top 100 sequences. It is difficult to imagine that sidechains at each of these positions contact the protein in a sequence selective fashion. Rather, our result implies that the NExt and Doped family peptides are optimized for both structure-specific contacts and sequence context along the length of the chain. This dual optimization is not possible via one bead-one compound or single-position mutagenesis methods and may be one source contributing to the ultrahigh affinities we observe here. A second metric of this sequence optimization is the Hamming distance for every Doped family clone vs. the original E1 starting sequence. The Hamming distance is defined as the number of individual residue changes needed to convert one sequence into another 124 . Plotting the Hamming distance for the top 1,000 sequences before and after selection (Fig. 4b) also indicates the Doped sequences are highly optimized. Prior to selection, the Doped pool clones have a Hamming distance of 11-12 changes vs. E1, consistent with every position having ~40% chance of being the wt residue. The selected Doped sequences show an average of only 3.5 mutations vs. E1 in the 100 most abundant clones. This result argues that the Doped family of sequences are highly optimized along the entire length of the chain. 127 Maximal Affinity vs. Chain Length: The off-rates for P1, E1, and D1 clones give a general idea of how affinity scales with chain length (Fig. 4c). For the 10 residue P1 sequences, clones give ∆G° 298 of -9.9 kcal/mol. If the entropic penalty for bringing the partners together is +3-5 kcal/mol 125 , then the total binding energy is 1.3 to 1.5 kcal/mol per residue. The initial extension screen results in an improvement of ~750 fold in K d (-4 kcal/mol) and doping optimization results in yet another four-fold improvement. Thus moving from 10 residues to 20 residues gives ∆∆G° of -4.8 kcal/mol, giving a crude scaling law of -0.5 kcal/mol per residue. This data is consistent with the initial peptide library localizing to a protein hot spots 126 while the extensions explore regions where the ability to gain binding energy is somewhat more limited. We note that our observations are significantly less than the "maximal" affinity per heavy atom previously predicted by Kuntz et al. 105 (1.5 kcal per heavy atom). On the other hand, we have achieved K d = 8.5 ± 2 pM for a 21 residue peptide and our approach argues that even higher affinity ligands could be constructed by further extension, opening the door to femtomolar reagents and perhaps beyond. Specificity of Binding: Peptide D1 competes with the natural Bcl-x L ligand, Bim, for binding (Supplementary Fig. 4), showing that D1 recognizes the same site and represents a convergent solution, similar to the function of the BH3 peptides. Bim shows poor specificity in discriminating between Bcl-2 family members Bcl-2, Bcl-W, and Mcl-1, which have 37% to 63% sequence similarity to Bcl-x L (Fig. 5a). One motivation for designing protein-directed reagents is to generate highly target-selective ligands that can discriminate even highly sequence related proteins. At the outset of this work it was unclear if improved affinity and specificity 128 could be achieved simultaneously. Arguments exist that selections for improved affinity result in improved specificity 127 , that improved affinity does not result in improved specificity 128 , or that the relationship between affinity and specificity is idiosyncratic and difficult to predict 129 . The P1 peptide derived from the primary selection showed somewhat improved selectivity for Bcl-x L compared with Bim (Fig. 5b), particularly vs. Mcl-1, perhaps reflecting that it was selected vs. a single target. Peptide P1 is specific for Bcl-x L , but also binds moderately well to Bcl-W (Fig. 5b). The selectivity for Bcl-x L increases dramatically with the E1 peptide (Fig. 5c), which shows a small amount of binding to Bcl-2 and essentially no binding to Bcl-W and Mcl-1 (Fig. 5c). The D1 peptide shows a truly remarkable level of specificity, particularly given its high affinity, binding only Bcl-x L and showing no affinity for the other three homologs (Fig. 5d). This result indicates that it is possible to achieve very high specificity and affinity simultaneously with a protein-directed peptide ligand. Peptide-Based Diagnostics: Finally, we tested to see if the ultrahigh affinity D1 peptide could replace a monoclonal antibody in a diagnostic setting (Fig. 6). To do this, we designed a sandwich detection scheme involving analyte capture by biotinylated D1 peptide and detection via a commercially available anti-Bcl-x L antibody (Fig. 6a). We then used this scheme to perform analysis of serial dilutions and generate a standard curve (Fig. 6b). The resulting assay shows excellent sensitivity with a lower limit of quantitation of 400 fM (Fig. 6b). The resulting assay is not only equivalent, but more than 15-fold more sensitive than the stated detection limit for Bcl-x L commercial antibody ELISA assays (See Supplementary Note 2). 129 5.4 Conclusions We have developed a general strategy to engineer ultrahigh affinity protein-directed peptide ligands through selection and extension. Our approach allows optimizing segments of peptide sequence (here 9 residues at a time) rather than single positions and this appears to have optimized most of the residues along the entire chain. A key and somewhat unexpected outcome of this work is that protein-directed peptide ligands can have truly remarkable affinities, and that many examples with single digit picomolar K d values can be found. The D1 sequence we have characterized is both the highest affinity Bcl-x L directed ligand known (antibody, peptide, or small molecule) and also the highest affinity protein-directed peptide reagent that has been described to date. Utility: What use are ultrahigh affinity peptides? One obvious answer is that they provide a simple route to create ultrasensitive diagnostic reagents and tests. The extraordinarily high selectivity we observe argues that peptide diagnostics are both possible and can be designed to outperform antibody-based tests. A second possible use is in therapeutic targeting, particularly of drug conjugates. The recent advent of antibody drug conjugates (ADCs), such as TDM-1, argue that readily synthesized peptide-drug conjugates could have use in therapeutic applications. The combination of our extension selection method described here with the ability to generate protease resistant peptides 9 argues that ultrahigh affinity peptides that can resist proteolytic degradation are within reach. 130 5.5 Figures Figure 1: Extension selection strategy. (a) An initial library of more than 1 trillion unique sequences contains a small fraction of functional sequences (red). (b) This library is sieved against a target (grey) to generate an enriched library of sequences with nanomolar affinities against the target (c) The enriched library is extended with random sequence and reselected to generate extended peptides (red/yellow) that bind the target with picomolar affinity. 131 Figure 2: Primary selection against Bcl-x L . (a) Binding of radiolabeled 9-mer library fusions to immobilized Bcl-x L (black bars) or to beads without target (red bars). (b) Binding of radiolabeled P1 fusions to immobilized Bcl-x L (black) or to beads without target (red) (c) Round 5 Primary selection sequences that bind Bcl-x L (See Supplemental Figure 1), but that do not resemble known Bcl-x L -binding sequences.(d) Sequence alignment of P1 with Bcl-x L -binding peptides Bim and Bad. H1-4 denote hydrophobic positions characteristic of BH3 domains that interact with hydrophobic pockets on Bcl-x L . A conserved Asp is shown in bold. 132 Figure 3: Sequences from the Extension and Doped selections. (a) Binding of radiolabeled Extension library fusions to immobilized Bcl-x L (black bars) or to beads without target (red bars). (b) Two peptides, E1 and 2, from the Extension selection show picomolar dissociation constants and slow dissociation rate constants. (c) Sequence alignment of Peptide E1 with Bcl- x L -binding peptides Puma and Bim. A highly conserved Asp is shown in bold while H1-4 denote the characteristic hydrophobic positions of BH3 domains. (d) Peptide D1 from the doped selection shows a 8.5 pM dissociation constant and dissociate rate constants of ~2  10 -6 /s. Red amino acids denote positions that differ from wild type. 133 Figure 4: (a) Mutational analysis of the E1 sequence. Each column denotes the frequency of the amino acid composition of the top 10,000 clones from the Doped selection at that position, with the exception of the N-terminal Met, which is fixed (b) Hamming distance analysis of the top 1,000 sequences from the Doped selection. The percent of sequences with a particular number of mutations from wild type (the Hamming distance) is shown for the original starting library (black bars, 11-12 mutations on average) and the final Doped library (red bars, 3-4 mutations on average). (c) The improvement of off-rate from peptide P1 (blue curve) to E1 (black curve) to D1 (red curve). The improvement due to extension is ~750-fold while doping increases affinity by ~4-fold. 134 Figure 5: Binding specificity of selected Bcl-x L -binding peptides. Biotinylated synthetic peptides (a) Bim, (b) P1, (c) E1, and (d) D1 were immobilized on neutravidin agarose and incubated with each respective 35 S-labeled protein (Bcl-x L , Bcl-2, Bcl-w, or Mcl-1). Binding data is normalized to Bcl-x L binding. Error bars represent standard deviation. Inset: magnified view of normalized binding. 135 Figure 6: ELISA quantitation assay for Bcl-x L using peptide D1 (a) Schematic for the Bcl-xL quantitation assay. Bcl-x L is captured by the Doped peptide that is immobilized on the ELISA plate. Bound Bcl-x L is bound with a non-competing anti-Bcl-x L antibody, which is then detected by an anti-rabbit/HRP conjugate. (b) Bcl-x L quantitation assay. Two-fold dilutions of Bcl-x L were tested via the quantitation assay shown in (a) and fit to a simple four-parameter logistics curve (4PL Fit). The assay shows a limit of detection (LOD) of 200 fM and a lower limit of quantitation (LLOQ) of 400 fM. 136 5.6 Supplementary Figures Supplementary Figure 1: Binding of Round 5 clones from the Primary selection. Two clones, P2 and P3, that do not resemble known BH3 peptides were radiolabeled with 35 S-methionine and tested for binding against neutravidin beads (red bars) or neutravidin beads with immobilized Bcl-x L (black bars). 137 Supplementary Figure 2: Off-rates for Rounds 3-7 from the N-terminal Extension selection. 35 S-radiolabeled mRNA-peptide fusions were first bound to neutravidin beads with immobilized Bcl-x L , washed, and 100X molar excess of free, non-biotinylated Bcl-x L was added. Time points were taken and spun through a 0.45 m filter to separate bound (cpm on beads) and unbound fusions (cpm in the flowthrough). The percent bound remaining was then calculated by dividing the bound fraction by the total counts at that point. Theoretical curves are shown for off-rates ranging from 1 10 -3 to 1 10 -6 s -1 (dashed lines). 138 Supplementary Figure 3: Determination of K D by ELISA competition. A simple sandwich ELISA assay to detect Bcl-x L was created, using synthetic D1 peptide as the capture reagent. Three concentrations of Bcl-x L were equilibrated with varying amounts of either synthetic D1 or E1 peptide prior to incubation with the ELISA plate. When the peptide ligands and Bcl-x L form a complex, the complex is unable to bind to D1 peptide immobilized on the ELISA plate, thus reducing the signal. The concentration of free Bcl-x L was determined by quantitating the reduction in signal (panel a-b, data points). The data points for all three Bcl-x L concentrations were fit to the equilibrium equation simultaneously to obtain the K d of the peptides for Bcl-x L . Supplementary Figure 4: Binding competition with the Bim peptide. The D1 peptide binds to the same epitope of Bcl-x L as Bim. A simple sandwich ELISA assay to detect Bcl-x L was created, using synthetic D1 peptide as the capture reagent. Pre-incubation of 1 nM Bcl-x L with [10 - 0.4] nM Bim Peptide reduces the Bcl-x L signal, since the Bcl-x L -Bim complex is unable to bind to the D1 peptide immobilized on the ELISA plate. 139 Off-Rate Temperature Reference 2.4  10 -6 Room Temp (22-25 C). Prian and Riordan 130 2.8  10 -6 25 C Green 131 5.4  10 -6 25 C Chilkoti and Stayton 132 5.2  10 -5 37 C Howarth et al. 133 6.8  10 -5 37 C Chivers et al. 134 Supplementary Table 1: In the literature, there are several measurements of the biotin- streptavidin off-rate. Here, we have reported the slowest off-rate of 2.4  10 -6 s -1 , as reported by Prian and Riordan 130 . Company Protein Product Number Sensitivity (ng/mL) Sensitivity (fM) MyBioSource.com Human Bcl-x L MBS046545 2.0 77,000 Antibodies-Online Human Bcl-x L ABIN832352 0.156 5,990 R&D Systems Rat/Human Bcl-x L DYC894 0.625 24,000 Supplementary Table 2: There are three companies that we found that sell human Bcl-x L detection kits. The sensitivity for these tests is reported in ng/mL and was converted to concentration in fM using the Bcl-x L molecular weight of 26,049 g/mol. 140 Chapter 6: Analysis of Proteins using Peptide Immunoreagents by an Acoustic Resonant Sensor This work has been adapted from the following publication: Jalali-Yazdi F, Corbin JM, Takahashi TT, Roberts RW. Robust, quantitative analysis of proteins using peptide immunoreagents, in vitro translation, and an ultrasensitive acoustic resonant sensor. Anal Chem 86, 4715-4722 (2014). 6.1 Introduction: Advances in mass spectrometry 135, 136 and microarrays 137, 138 have provided a better perspective of biological systems, but a long-term goal in proteomics is the development of affinity reagents against all members of the proteome. Monoclonal antibody methods 139 , phage display 140, 141 , ribosome display 142 , and mRNA display 143 can all generate tens to hundreds of potential polypeptide ligands against targets of interest. Recent advances combining in vitro selection with high-throughput sequencing has greatly accelerated the process of generating a large list of potential ligands. 120 Many of these techniques do not use antibodies as affinity reagents, but rather use either small protein scaffolds or peptides for protein recognition, offering the potential of antibody-free diagnostics. Peptides are especially attractive as antibody replacements because they can be chemically synthesized (avoiding issues with expression and purification), are renewable, and are generally stable without refrigeration. 103 However, in order to generate a set of proteome-wide affinity reagents, high-throughput methods will not only be required for initial ligand discovery, but will also be needed for screening and characterization to determine the best ligand for an application. Current methods (e.g., ELISA) are slow, laborious, and do not correct for differences in ligand expression levels. 144, 145 While new technologies such as optical 141 resonators 146 and nanowire sensors 147 provide the advantage of direct ligand-target affinity measurements, these methods are challenging to implement in complex media, at physiological salt, and in a high-throughput fashion. There is thus a pressing need for high-throughput, robust methods that are sensitive, utilize little reagent, and function in complex media. In vitro translation provides an appealing route to screen ligands in a high-throughput fashion, since no cloning is required, allowing ligands to be generated quickly. Typically, very little material is synthesized and translation levels are highly variable, requiring both a highly sensitive assay and the ability to normalize the signal for expression. Additionally, detection of proteins in crude translation reactions (a complex media) is essential because purification steps after translation can be costly, inconsistent, and significantly hinder the throughput of the method. Here, we developed a method to accurately assess the relative affinity of multiple ligands for a specific protein. To do this, we quantify the specific activity of a clone (i.e., binding to the protein of interest) and normalize the binding signal for protein expression. This approach was implemented in a general way on a commercial acoustic resonant biosensor platform, the ViBE BioAnalyzer (BioScale, Lexington, MA.) The ViBE BioAnalyzer uses Acoustic Membrane Micro-Particle technology (AMMP) to detect the presence of the analyte using a sandwich format assay. The analyte is linked to the surface of the sensor using an anti-hapten antibody at one end and to a magnetic bead on the other end. 148 The sensor uses piezoelectric properties of a vibrating membrane, where the membrane’s in-liquid resonance frequency shifts when an external mass influences the sensor acoustic loading. 149 The beads act to amplify the loading, resulting in much higher levels of sensitivity. 142 An advantage of our approach, using the acoustic resonant devices, is that the assays are sensitive enough to be done using in vitro synthesized proteins, enabling rapid analysis in less than one day starting with synthetic DNA. 143 6.2 Materials and Methods: Peptide Synthesis: Peptide 1 (Pep1), containing a C-terminal flexible linker and HA tag, (NH 2 –MIETITIYKYKKAADHFSMSMGSGSGSYPYDVPDYA–NH 2 ) was chemically synthesized using Fmoc solid phase peptide synthesis via a PS3 peptide synthesizer (Protein Technologies Inc.) 30 A 200 µmol synthesis was performed with five molar equivalents of each amino acid and HATU. N-methyl-2-pyrrolidone (NMP) was used as the solvent and a solution of 25% (v/v) 4-methyl piperidine in NMP was used as the deprotection solution. 109 The peptide was synthesized on a Rink amide MBHA resin resulting in a C-terminal amide. Cleavage and deprotection of the peptide was performed with 95% (v/v) trifluoroacetic acid (TFA), 2.5% 1,2- ethanedithiol (EDT), 1.5% deionized water (DI), and 1% triisopropylsilane (TIS). 100, 150 The peptide was purified via HPLC with a C 18 reverse phase column (Vydac) using a gradient of 10- 90% acetonitrile/0.1% TFA in water. Fractions were collected and tested for the correct molecular weight using MALDI-TOF Mass Spectrometry (ABI). The correct fractions (mass expected: 4,024.57 g/mol, mass found: 4,025.0 g/mol) were lyophilized, dissolved in DMSO to a final concentration of 4 mM, and stored at -80 C. Fluorescein Labeling of the Anti-HA Antibody and the Synthetic Peptide: Anti-HA antibody (Thermo Scientific) was buffer exchanged to 1X PBS (10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl, pH 7.4) using a NAP-25 column (GE Healthcare) to remove sodium azide present in the storage solution. A twenty-fold molar excess of NHS- fluorescein (Pierce) in DMF was then added to the buffer-exchanged antibody or synthetic Pep1 and incubated for one hour at room temperature in the dark. The reactions were quenched with 1 144 M Tris-HCl, pH 8.0, and buffer exchanged into 1X PBS using NAP-25 columns to remove the un-reacted NHS-fluorescein. Since the HA tag does not contain any lysine residues, the HA tag on synthetic Pep1 was not modified with a label. The concentration of the peptide and anti-HA antibody, and the average number of fluorescein labels per molecule were calculated as per manufacturer’s instructions. The purified solutions contained an average of three fluorescein labels per antibody and one fluorescein label per synthetic peptide. Biotin-labeled Bcl-x L Expression and Purification: The gene for the first 209 amino acids of Bcl-x L (Clone HsCD00004711; Dana Farber/Harvard Cancer Center DNA Resource Core) was PCR amplified using Pfusion polymerase, and cloned into pET24a using NdeI and XhoI. An N- terminal avitag (AGGLNDIFEAQKIEWHEGG) was added for in vivo biotinylation using the BirA enzyme. 34 Bcl-x L was expressed overnight at 37 C in BL21(DE3) cells using auto- induction media. 33 Cells were lysed using Bper (Pierce), and purified using Ni-NTA superflow resin on an FPLC (Bio-Rad), using a gradient from 10 mM to 400 mM imidazole (Buffer A: 25 mM Hepes pH 7.5, 1 M NaCl, 10 mM imidazole; Buffer B: 25 mM Hepes pH 7.5, 1 M NaCl, 400 mM imidazole). Fractions with pure Bcl-x L were combined, concentrated, and desalted into 50 mM Tris-HCl, pH 8.0. Bcl-x L was biotinylated in vitro using BirA biotin ligase (0.1 mg/mL in 50 mM Tris-HCl, pH 8.3, 10 mM ATP, 10 mM Mg(OAc) 2 , 50 µM biotin) at 30 C for two hours. The protein was buffer exchanged into 1X PBS, frozen in liquid nitrogen, and stored at - 80 C. Bead Loading: Anti-HA tag antibody was immobilized on magnetic beads by incubating 72 pmol of unlabeled anti-HA antibody with 0.6 mg of epoxy magnetic beads (Life Technologies) 145 in 1X PBS buffer at 4 C. After 48 hours, the reaction was quenched with 100 µL of 1 M Tris- HCl, pH 8.0. The beads were then washed and re-suspended in 1 mL of 1X PBS + 1% (w/v) BSA + 0.1% (v/v) Tween-20. Bcl-x L was immobilized on magnetic beads by incubating 60 pmol of biotinylated Bcl-x L with 0.5 mg of streptavidin magnetic beads (Life Technologies) at 4 C overnight. To block any unbound sites on the streptavidin, 100 nmols of biotin was added and incubated with the Bcl-x L loaded beads for 30 minutes at room temperature. The beads were then washed with assay buffer (1X PBS + 1% (w/v) BSA + 0.1% (v/v) Tween-20), and resuspended in 600 µL of the same buffer. In Vitro Translation of Peptides, Proteins, and mRNA-Peptide fusions: The DNA sequences coding for the peptides and fibronectin protein were ordered from Integrated DNA Technologies (IDT). Each DNA construct contained a T7 RNA Polymerase promoter, and a 5’ UTR that was a deletion mutant of the Tobacco Mosaic Virus (ΔTMV). 95 The C-terminal portion of the peptides were elongated with a flexible serine-glycine linker (six amino acids long) and an HA tag (Appendix). After gel purification using urea-PAGE, the DNA sequences were PCR amplified using Taq polymerase and in vitro transcribed into mRNA using T7 RNA polymerase. 95 After transcription, the mRNA was urea-PAGE purified and resuspended in deionized water to a final concentration of 30 µM. The mRNA was then ligated to fluorescein- F30P (phosphate–dA 21 –[dT-fluor]–[C9] 3 –dAdCdCP; where [dT-fluor] is fluorescein dT (Glen Research), [C9] is spacer 9 (Glen Research), and P is puromycin (Glen Research); synthesized at the Keck Oligo Facility at Yale) using T4 DNA ligase. 151 The ligation was performed using a splint complementary to the 3’ end of the RNA and the 5’ end of the DNA-linker (Appendix). The ligated mRNA was urea-PAGE purified and resuspended in deionized water to final 146 concentration of 30 µM. The samples were in vitro translated in the translation solution—150 mM KOAc, 750 µM MgCl 2 , 2 µM mRNA, in 1X translation mix (20 mM Hepes-KOH pH 7.6, 100 mM creatine phosphate, 2 mM DTT, and 312.5 µM of each amino acid) and 60% (v/v) rabbit reticulocyte lysate (Green Hectares; prepared according to the method of Jackson and Hunt 99 ). To prepare radiolabeled peptides or proteins, non-labeled methionine was substituted with 35 S-labeled methionine (Perkin Elmer; 20 µCi for each 25 µL of translation). The translation reactions were incubated at 30 C for one hour. To form mRNA-protein fusions, KCl and MgCl 2 were added to the reaction to final concentrations of 250 mM and 30 mM respectively after translation. Samples were frozen at -20 C and used within two days. Radiolabeled Binding Assays: 35 S-labeled mRNA-protein fusions translated as described above were purified using oligo-dT cellulose beads (GE Healthcare) as described elsewhere. 151 Briefly, samples were incubated at 4 C for 1 hour, washed with 1X dT buffer (50 mM Hepes- KOH pH 7.5, 1 M NaCl, 1 mM EDTA, 0.05% (v/v) Tween-20), and eluted using 65 C water. RNase (Roche) was added to the samples and incubated at 37 C for 15 minutes to digest the attached mRNA. To the RNase treated samples, 100 pmol of biotinylated Bcl-x L immobilized on streptavidin magnetic beads were added, and incubated for one hour at room temperature in assay buffer. The beads were then washed four times using a magnetic separator. The percent bound of each peptide was calculated from the number of counts remaining on the beads, divided by the total number of counts added to the binding reaction. AMMP Assays: in vitro translated peptides and proteins were diluted 1:10 initially using assay buffer (1X PBS + 1% (w/v) BSA + 0.1% (v/v) Tween-20) and the subsequent dilutions 147 were performed in 10% translation solution in assay buffer. A sample of synthetic Pep1 at 30 nM in 10% translation solution was prepared and diluted serially to generate the standard curve. All samples and standards were run in duplicate. Samples and standards were incubated with magnetic beads and fluorescein-labeled analyte for four hours. Because each AMMP assay run takes 10 minutes, constant incubation time for all samples was achieved by separating the start of incubation for each column on the 96 well plates by 10 minutes. Run buffer was 1X PBS + 1% (v/v) Tween-20 + 1% (v/v) heat-treated FBS (Invitrogen; FBS was heat treated for 15 minutes at 65 C and filtered.) BioScale Universal Detection Cartridges were used in performing all of the assays. The device was used per the manufacturer’s instructions. 148 For the AMMP HA tag competition assay, 60 µL of sample was incubated with 30 µL of anti- HA antibody immobilized on epoxy magnetic beads (6 µg of beads/mL) and 30 µL of 240 pM fluorescein-labeled synthetic Pep1 for four hours, and run on the ViBE BioAnalyzer. Both anti- HA beads and Pep1 were diluted to the appropriate concentration using assay buffer. For the AMMP target-binding assay using in vitro translated peptides, 60 µL of sample was incubated with 30 µL of biotinylated Bcl-x L immobilized on streptavidin magnetic beads (8.3 µg of beads/mL) and 30 µL of 10 nM fluorescein-labeled anti-HA antibody for four hours. For the AMMP target-binding assay using mRNA-peptide fusions, 60 µL of sample (diluted in 1X assay buffer) was added to 30 µL of biotinylated Bcl-x L immobilized on streptavidin magnetic beads and 30 µL of assay buffer and incubated for four hours. The concentration of the HA tag competition assay samples were determined using synthetic Pep1 standards and fit to a simple 4-parameter logistics curve (Figure 3a). 148 ELISA Target-Binding Assay: ELISA plates were incubated overnight at 4 C with 1.5 nmol of anti-HA antibody per well. Plates were washed with wash buffer (1X PBS + 0.1% (v/v) Tween-20) and blocked with 1X PBS + 5% (w/v) BSA for two hours. In each well, 100 µL of sample and 100 µL of 33 nM biotinylated Bcl-x L were added and incubated for four hours. Plates were washed, incubated with streptavidin horseradish peroxidase conjugate (Strep-HRP, Thermo Scientific) for 1 hour, washed, and incubated with TMB substrate (Thermo Scientific). Reactions were stopped after approximately 10 minutes with 2 M sulfuric acid, and the absorbance at 450 nm was measured via a plate reader (Molecular Devices.) 149 6.3 Results and Discussion Quantitation of in Vitro Translated Proteins: Overall, our goal was to develop a general quantitative method to rank-order protein-binding ligands using in vitro translated polypeptides. This simple aim is complicated by several factors: (1) the small amount of protein produced by in vitro translation systems, (2) detection of binding in complex media, and (3) the fact that most existing methods (e.g., sandwich ELISA) require two orthogonal affinity reagents to function. Our solution was to develop a competition assay that required only a single affinity tag added to each sequence (described below). Unfortunately, only modest levels of polypeptides are synthesized via in vitro translation, and the crude sample matrix (the solution containing the sample, which includes proteins, surfactants, nucleic acids, salts, etc…) can interfere with the assay, further reducing sensitivity. 152 Competition assays are also inherently insensitive at concentrations where analyte and affinity reagents are present in well below the dissociation constant (K d .) Under these circumstances, competition assays require the analyte to be present in a large molar excess in order to reduce the signal of a known complex. 153 In line with these views, we were unsuccessful in our attempts to create a competition ELISA assay sensitive enough to detect translated peptides and proteins (data not shown). We decided to use AMMP technology, which was ideal for our analysis due to its high sensitivity, low reagent consumption, ability to detect analytes in complex matrices, and the potential for automation. 148 The AMMP competition assay we developed is shown in Figure 1. In this assay, a synthetic peptide is labeled with fluorescein, allowing it to bind to an anti-fluorescein antibody on the sensor surface. The synthetic peptide also contains an HA tag (NH 2 -YPYDVPDYA-COOH) that binds to an anti-HA antibody on a magnetic bead, completing a sandwich for detection of labeled 150 fluorescein-HA peptide (Figure 1a). As HA tagged (non-fluoresceinated) peptide is added, it competes with the fluorescein-HA peptide for binding to the anti-HA antibody. This decreases the fraction of antibody available to bind the fluorescein-HA peptide, thus reducing the signal in a dose-dependent fashion (Figure 1b). We chose the HA tag in the competition assay because it is small, widely used in biotechnology and several inexpensive anti-HA antibodies are commercially available. Since each in vitro translated peptide or protein in our assay could be designed to include a single HA tag, each peptide or protein should interact with the anti-HA antibody with equal affinity. Thus, appending an HA tag to our peptides or proteins would enable us to quantify in vitro translated polypeptides independent of the sequence N-terminal to the HA tag using this competition assay. In order to test and validate our HA competition assay, we needed to show that the assay could be used to quantify proteins or peptides with different sequences N-terminal to the HA tag. We also needed sequences with different affinities for their target so that we could test our ability to rank-order ligands by binding affinity. We chose two peptides from an mRNA display selection against the B-Cell lymphoma – extra large protein (Bcl-x L ) (T.T. Takahashi, R.W. Roberts, manuscript in preparation). The first peptide, Peptide 1 (Pep1), binds to Bcl-x L with high affinity while the second peptide, Peptide 2 (Pep2), has a lower affinity for Bcl-x L . We also used a scrambled version of Peptide 1 (ScPep1) as a negative control in the specific binding assays to demonstrate the binding selectivity to Bcl-x L . We synthesized the DNA that coded for these peptides followed by an HA tag on the C-terminus of each peptide. The HA Competition Assay: We first generated a standard curve using the synthetic fluorescein-HA peptide. Using the synthetic peptide, we observed significant reduction of signal 151 for dilutions with concentrations higher than 1 nM (Figure 2a). We then in vitro translated Pep1, Pep2, and ScPep1 as samples, as well as Pep1 lacking the HA tag (Pep1ΔHA) as a negative control. We also translated the 10 th fibronectin type III domain of human fibronectin 154 with a C- terminal HA tag (wt-Fn) to determine if we could quantify in vitro translated proteins using our HA competition assay as well. All samples were run in duplicate, starting with a 1:10 dilution followed by four subsequent 1:2 dilutions. We normalized matrix interference effects by diluting all samples and standards into the same solution, which contained 10% (v/v) translation solution. Pep1, Pep2, and ScPep1 all showed a reduction of signal up to a dilution factor of 40. As expected, the Pep1ΔHA control, which lacks the HA tag, did not reduce the signal at any tested dilution. Lastly, wt-Fn containing a C-terminal HA tag also showed a significant reduction in signal, demonstrating we could also quantify in vitro translated proteins in complex matrices. Using the synthetic fluorescein-HA peptide, we generated a calibration curve to calculate the concentration of our translated samples. To do this, we fit the synthetic fluorescein-HA peptide data to a simple four-parameter logistic curve (Figure 2b). We then determined the concentration of the translated samples by interpolation of all four dilutions and taking the mean. The concentration (in nM) and coefficient of variation (CV) for the different in vitro translated samples are shown in Figure 3b. Wt-Fn had the highest translation efficiency, with its concentration measured at 327 nM, followed by ScPep1 (219 nM), Pep1 (136 nM), and Pep2 (62 nM). Our data demonstrate that the amount of peptide or protein synthesized using in vitro translation is highly sequence dependent. The difference between the lowest (Pep2) and the highest (wt-Fn) concentrations of in vitro translated polypeptides was a factor of five. This variability could drastically skew rank-ordering of potential ligands. Moreover, the difference in 152 expression level between Pep1 and ScPep1 is approximately two-fold. This result is somewhat surprising as the sequences use ~70% identical codons (Appendix). In summary, we demonstrate that different sequences translate with significantly different efficiencies. Thus, only determining a clone’s specific binding is not enough for ranking-ordering ligands. Instead, it is necessary to normalize each clone’s expression level in order to determine the highest affinity clones. We also needed to validate that the signal generated in the assay was due to both the HA/antibody interaction and the fluorescein/antibody interactions (Figure 1a). Theoretically, generated signal could also arise from antibody modified magnetic beads, with or without non- fluoresceinated peptide, adhering nonspecifically to the sensor surface. To show this was not the case, we first eliminated the peptide analyte from the complex and observed background levels of signal (~10%, Figure 4). We then added an excess (30 nM) of non-fluoresceinated peptide to the antibody modified magnetic beads, and still observed background levels of signal (Figure 4). Design of a Target-Binding Assay: Once we developed a method to quantify the amount of sample in an in vitro translation reaction, we then designed an assay to measure each sample’s relative binding affinity. We used a simple sandwich assay where we could test different peptides for binding to their target (here, Bcl-x L ). In the AMMP target-binding assay (Figure 6a), we combined Bcl-x L on streptavidin magnetic beads with HA tagged peptide and fluoresceinated anti-HA antibody. We then analyzed the mixture using the ViBE BioAnalyzer, where the fluorescein on the anti-HA antibody binds to the anti-fluorescein antibody on the sensor surface. To demonstrate that our peptides bound to Bcl-x L with different affinities, we used a radiolabeled binding assay where we immobilized biotinylated Bcl-x L on streptavidin magnetic beads and tested the binding of 35 S-labeled, HA tagged Pep1, Pep2, and ScPep1 (Figure 5). Pep1 153 shows the highest level of binding to immobilized Bcl-x L (~80%) while Pep2 shows modest binding (~1%.) Scrambled Peptide 1 (ScPep1) shows negligible binding to immobilized Bcl-x L (<0.1%). No peptide showed any appreciable binding to streptavidin magnetic beads without Bcl-x L . These results corroborate previous experiments that show Pep1 possesses a higher affinity for Bcl-x L than Pep2 (T.T. Takahashi, R.W. Roberts, unpublished observations). The data also show that the addition of the HA tag (with a six-amino acid spacer) does not interfere with the binding of the peptides to Bcl-x L . Measuring Relative Binding Affinity Using the AMMP Target-Binding Assay: We translated the Bcl-x L binding peptide (Pep1 and Pep2) as well as the negative control peptides (ScPep1 and Pep1ΔHA), and ran dilutions of each sample on the AMMP target-binding assay (Figure 6b-c). Both Pep1 (the high affinity binder) and Pep2 (the moderate affinity binder) show saturated signal levels for the 1:10 dilution samples. Dilutions of 1:80 for Pep2 and over 1:1000 for Pep1 give robust signal over background. In vitro translated ScPep1 and Pep1ΔHA (two negative controls) showed background levels of signal at every dilution. Omitting various components of the sandwich (Synthetic Pep1 or fluoresceinated anti-HA antibody) results in no signal over background levels (Figure 7). In order to accurately rank the relative affinities of Pep1 and Pep 2, we used the concentrations calculated for each from the HA tag competition assay (Figure 6c). Pep2 provides robust signal over background at 1 nM, whereas Pep1 gives robust signal even at 100 pM. ScPep1 has undetectable levels of binding for Bcl-x L at the tested concentrations. These results are in agreement with the radiolabeled binding assay, and previous observations. The performance 154 characteristics of Pep1 are comparable to antibody based ligands, previously analyzed using this device, in terms of assay range and sensitivity. 148 Comparison of AMMP Technology and ELISA: In order to directly compare AMMP technology to widely used ELISA methods, we used the same reagents (in a different orientation) to analyze the same standards and samples from the AMMP assay on a similarly formatted ELISA assay (Figure 6d). The results show that in vitro translated Pep1 is functional in ELISA, and gives a dilution profile similar to what would be expected of an antibody. In vitro translated Pep2 also gives significant ELISA signal over background for the lowest two dilutions (Figure 6e). The highest dilutions for Pep1 and Pep2 are at background levels of signal for ELISA, while they are significantly above the background level of signal on the AMMP assay. In this experiment the difference in sensitivity between the ELISA and the AMMP assays for the in vitro translated Pep1 is ~20-fold. Pep1ΔHA and ScPep1 showed no appreciable binding to Bcl-x L at any dilution. We also performed several independent experiments with known concentrations of synthetic Pep1 (Table 1) in order to confirm the sensitivity difference we observe between AMMP assays and ELISA. In buffer, the AMMP assay is on average an order of magnitude more sensitive than ELISA. In 10% translation solution, the AMMP assay remains more sensitive than ELISA by a factor of four. Perhaps more notable, with regards to high-throughput ligand analysis, is that the AMMP assay offers this higher sensitivity with five-fold less antibody and over 100-fold less target per sample. One of the advantages of higher sensitivity assays is lower sample consumption and less sample matrix interference, since higher dilutions in a simple buffer reduces the fraction of interfering matrix. The decreased amount of antibody required in AMMP 155 assays has the additional benefit of enabling highly sensitive competition assays. Under concentrations where the target protein and the ligand are <K d , any competition assay will require a vast molar excess of sample protein to deplete available ligand sites and reduce the binding of the reference molecule. The AMMP assay, by using significantly lower amounts of ligand, can attain a level of sensitivity in competition assays that is difficult to reach, or unreachable, using ELISA. Comparison of Different Assays to Determine Translated Peptide Concentration: We were interested in using the three assay formats described above (competition assay, target- binding assay, and ELISA-based assays) to determine the concentration of a translated sample for comparison. Ideally, each assay would give the same result for a single sample, but previous comparisons of methods 155, 156, 157, 158 have shown biases using different assay formats. To compare these assays, we calculated the concentration of in vitro translated Pep1 using synthetic Pep1 standards and a four-parameter logistic calibration curve for each assay. The concentration of translated Pep1 was measured at 136 nM using the HA tag competition assay, 300 nM using AMMP target-binding assay, and 1,100 nM using the ELISA target-binding assay (Figure 8a). A value between 100-300 nM is more consistent with previous analyses on this system. 95 The bias in concentration measurement was consistent (R 2 = 0.92) over three translated samples tested in three independent trials when comparing the target-binding assays using AMMP or ELISA (Figure 8b). Though the reason for the bias is unknown, consistent biases of different methods of concentration measurement are common. 155, 156, 157, 158 Lastly, as long as the relative concentrations that are determined for a set of samples using a single assay format are consistent, the bias from different assays will not affect our ability to rank order different ligands. Since we 156 are determining the relative activity of a set of ligands, normalizing each ligand’s activity by its relative concentration should not change the overall rank-ordering. Sample Matrix Effects: For the AMMP and ELISA assays, we diluted all samples and standards into assay buffer containing 10% translation solution to negate matrix effects on the generated signal. In the competition assay, translation solution present in samples gives artifactual signal in the positive direction, and in the target-binding assay, translation solution gives artifactual decreased sensitivity (Figure 9). Thus, assays that do not adjust for matrix effects can over- or under-estimate the level of analyte present in solution, depending on the assay used. Binding of Peptide-mRNA Fusion Molecules: One application of the AMMP sandwich assay we developed is for analysis of mRNA-peptide fusions used in mRNA display. 143 In mRNA display, an mRNA template is covalently linked to the polypeptide that it encodes using a puromycin linker. This step is essential in mRNA display, which is widely used for the generation of high affinity ligands by in vitro selection of high diversity libraries. 151 Currently, radioactive binding assays are used to evaluate the function of mRNA-peptide fusions. 95 We were interested to test if our AMMP assays were sensitive enough to evaluate mRNA-peptide function, which would increase throughput and avoid the use of radiation. To do this, we designed a target-binding assay for the mRNA-peptide fusion molecule (Figure 10a). We ligated the mRNA of our peptides to the puromycin-containing DNA linker, where the DNA linker possessed a fluorescein label. By using the fluorescein-labeled linker, we could avoid the use of a C-terminal HA tag/antibody to form the sandwich. In this assay, the fluorescein tag on 157 the mRNA-puromycin molecule binds to the anti-fluorescein antibody on the sensor surface. The peptide of the fusion molecules binds to the biotinylated Bcl-x L immobilized on streptavidin magnetic beads, thus connecting the magnetic bead to the sensor surface (Figure 10a). To test this assay, we in vitro translated mRNA or mRNA-puromycin linked molecules. 99 Samples of in vitro translated peptide were incubated with fluorescently-labeled anti-HA antibody and Bcl-x L immobilized on the magnetic beads, while samples of in vitro translated mRNA-peptide fusions were incubated simply with Bcl-x L magnetic beads. The scrambled ScPep1 negative control showed low levels of signal as either peptide or fusion (Figure 10b). Interestingly, we observed the Pep1-mRNA fusion gives higher signal levels than Pep1 peptide. This is surprising given that generally less than half of the translated peptides are fused to their encoding mRNA. 95 This result, which increases the sensitivity of the experiment by five-fold, is likely due to eliminating the HA tag/anti-HA antibody interaction necessary in the peptide format. In the fusion format, every peptide is covalently joined through puromycin to the fluorescein that immobilizes the construct on the sensor chip. In addition to the advantage of not needing a protein tag, this assay can be used in cases where higher sensitivities are needed either due to poor target-ligand interaction or poor translation efficiency, to distinguish between ligands’ relative affinities. However, because we do not use the HA tag/anti-HA antibody in the sandwich, we cannot calculate the concentration of fusions by competition with a known standard. Peptides as Diagnostic Reagents: Peptides as affinity reagents for immunoassays offer several advantages over antibodies in terms of stability, storage, cost of production, and purification. 103 They have largely been ignored as immunoreagents due to their instability in the 158 presence of protease and peptidases, as well as their generally lower affinities for their targets. Recent advances in display technologies have enabled generation of high affinity peptides with the potential of being attractive immunoreagents. 78 Previous studies have shown that protecting the C-terminus of the peptides can greatly increase resistance to proteolysis. 159, 160, 161 We sought to demonstrate that our assays could be performed in serum, a matrix which would be encountered frequently in diagnostic settings. To do this, we tested Pep1 in a solution containing 1% non-heat treated FBS in assay buffer for 4 hours. FBS is known to contain active proteases that could degrade linear peptides. 162 Both the mRNA-peptide fusion (C-terminus covalently bound puromycin) and synthetic peptide (C-terminal amide) show very similar signal levels in FBS and in assay buffer (Figure 11a-b). This observation is consistent with the molecules being stable in FBS over the four hour incubation time. On the other hand, in vitro translated Pep1 (natural carboxy C-terminus and therefore sensitive to carboxypeptidase degradation) 163 shows distinctly lower signal in FBS versus assay buffer, consistent with peptide degradation. Overall, our results show that C- terminally blocked peptides (fusions and amidated peptides) show excellent performance in 1% FBS and argue that these reagents can be used in complex media containing proteases and peptidases. 159 6.4 Conclusions We have demonstrated a robust, sensitive, specific, and scalable method to assess the relative affinity of ligands for the protein of interest in a complex matrix, while normalizing for expression levels. We measured the difference in expression levels of two similar sequences to be a factor of two, while different sequences showed a five-fold difference in expression levels, which can introduce significant error in relative affinity assessments. Our method is approximately an order of magnitude more sensitive than ELISA in a similarly formatted target- binding assay, while using 100-fold less target, and five-fold less antibody per sample. Lastly, we used AMMP technology to test mRNA-peptide fusion function without the need for the HA tag/anti-HA antibody, and showed a more general implementation of the method. The ranking method described above is not limited to in vitro translated proteins and peptides. We have shown that detection can occur in complex matrices; therefore our methods can likely be modified to work in different cell lysates. Our methods can also likely be adapted to screen antibodies, for which the HA tag and the anti-HA antibody can be switched with the Fc region of antibodies, and immunoreagents against the Fc region. As diagnostic reagent candidates, peptides offer several advantages over antibodies: they are smaller, have a longer shelf-life, can be stored at room temperature while lyophilized, and are much easier to synthesize and purify. The main disadvantages over antibodies in diagnostics are peptide instability in complex matrices containing peptidase and proteases as well as lower affinities. We have demonstrated that with a simple modification to the C-terminus, our peptides were able to detect target in a serum solution containing active proteases. We have also shown 160 that peptides are capable of detecting target with very high affinities in complex media, and are suitable as high affinity diagnostic reagents. 161 6.5 Figures Figure 1: Schematic of HA tag competition assay. (a) Fluorescein-labeled HA tagged peptide binds to both anti-HA antibody on magnetic beads and anti-fluorescein antibody on sensor surface to generate signal. (b) Non-fluorescein-labeled peptides or proteins cannot interact with sensor surface. Signal levels are reduced if a large excess of un-labeled HA tagged proteins reduces the fraction of bound fluorescein-labeled peptide. 162 Figure 2: Results of the HA tag competition assay. (a): AMMP signal vs. the dilution factor of tested samples. Translated Peptide1 without HA tag (Pep1ΔHA) does not compete with the fluorescein-labeled HA tagged peptide in the wells, therefore signal is not reduced. All other samples with C-terminal HA tags compete with the fluorescein-labeled peptide resulting in a reduction of signal. (b) Concentrations of the various samples were calculated based on a simple 4-parameter logistics curve and the results are plotted to show dilutional linearity. The concentration of Pep1ΔHA cannot be estimated using this method. 163 (a) Concentration = min + max − min 1 + ( Signal EC 50 ) slope (b) Figure 3: The 4 parameter logistics curve used to measure the concentration of the in vitro translated peptides in the HA tag competition assay. (a) The equation for the standard curve, as well as the parameters used to construct the standard curve. (b) The calculated concentration and coefficient of variance for the 4 tested samples. All 4 dilutions were used in calculating these numbers. Min 0.117 Max 0.570 EC 50 11219 Slope 0.896 Sample Concentration (nM) CV Pep1 136 37% Pep2 62 29% ScPep1 219 16% wt-Fn 327 11% 164 Figure 4: The controls for the AMMP HA tag competition assay. HA tag competition assay: Epoxy magnetic beads coated with anti-HA antibody are incubated with fluorescein-labeled and non-labeled HA tagged peptides. Background level of signal is observed when there is no fluorescein-labeled peptide. Presence of 30nM non-labeled peptide does not increase signal levels in the absence of fluorescein labeled peptide. Signal is generated when fluorescein-labeled peptide is present at 60pM. The signal is then reduced when 30nM of the non-labeled peptide is presents. Figure 5: Ability of the three selected peptides to bind to immobilized Bcl-x L . 35 S-Methionine labeled peptides, bound to a DNA-puromycin linker, were purified and incubated with immobilized Bcl-x L protein on magnetic beads (black filled bars) or magnetic beads without Bcl- x L (white bars). Peptide1 (Pep1) shows the highest level of binding to Bcl-x L , followed by Peptide2 (Pep2.) Scrambled Peptide1 (ScPep1) shows virtually no binding to Bcl-x L . None of the sequences show a significant level of binding to the magnetic beads without Bcl-x L . 165 Figure 6: AMMP and ELISA target-binding assays. (a) AMMP target-binding assay schematic: Biotin-labeled Bcl-x L immobilized on the streptavidin magnetic beads is incubated with translated or synthetic HA tagged peptides and the fluorescein-labeled anti-HA antibody. The anti-fluorescein antibody on the sensor surface binds to the fluorescein-labeled anti-HA antibody to link the magnetic beads to the sensor surface and generate signal. (b) Signal generated from AMMP target-binding assay vs. dilution factor of translated peptides. Signal generated by Pep1 is much higher than Pep2. ScPep1 and Pep1ΔHA show background levels of signal. (c) Signal generated from AMMP target-binding assay vs. concentration of translated peptides, measured using HA tag competition assay. (d) ELISA target-binding assay schematic: Biotin-labeled Bcl-x L and HA tagged peptides are incubated in HA antibody-coated wells. The plate is then incubated with streptavidin conjugated to HRP. Signal is generated by HRP cleavage of TMB substrate. (e) Signal from ELISA target-binding assay vs. dilution factor of the translated peptides. For the same dilution, Pep1 generates a higher signal level than Pep2. ScPep1 and Pep1ΔHA generate no signal over the background level. (f) Signal generated from ELISA target-binding assay vs. concentration of translated peptides, measured using HA tag competition assay. 166 Figure 7: The controls for the AMMP target binding assay. Target binding assay: biotin- labeled Bcl-x L immobilized on streptavidin magnetic beads is incubated with fluorescein-labeled anti-HA antibody and synthetic Pep1. The signal is at background level when either analyte is absent. High level of signal is achieved when both analyte are added. ELISA AMMP Mean LOD in Assay Buffer (pM) 51 5 Mean LOD in 10% Translation Solution (pM) 288 67 Range (logs of Concentration) 2 2 Sample Volume (µL) 100 60 Anti-HA antibody/sample (fmols) 1500 300 Biotin-labeled Bcl-x L /sample (fmols) 3300 30 Table 1: Comparison between ELISA and AMMP target-binding assays. On average, AMMP assay is approximately an order of magnitude more sensitive than ELISA in assay buffer and four-fold more sensitive in 10% translation solution. Both assays had a similar range of quantitation. AMMP assay used five-fold less anti-HA antibody and 110-fold less biotin-labeled Bcl-x L . 167 (a) (b) Figure 8. The difference between sample concentrations calculated from the AMMP the ELISA assays. Synthetic Pep1in 10% translation solution was used as standards in each of the assays. (a) The concentration of the Pep1 translated sample is determined using three different methods. The HA tag competition assay reports the concentration of the sample as 136nM, while the AMMP target binding assay reports the sample concentration as 304nM. The ELISA target binding assay reports the sample concentration as 1 µM. (b) The concentration of translated Pep1 sample was measured using the synthetic Pep1 standard curve for the AMMP and ELISA target binding assays. The results were generated from three separate runs on three different preparations of samples. Translated Pep1 Quantitation Method Concentration (nM) CV AMMP HA tag competition assay 136 37% AMMP target binding assay 300 12% ELISA target binding assay 1,100 20% 168 Figure 9. The effects of 10% translation solution on the assays. Synthetic Pep1 was diluted either in the assay buffer without translation solution or the assay buffer with the addition of 10% translation solution. (a) Result of the 10% translation solution on the HA tag competition assay. The signal is reduced by approximately 25% when translation solution is added. The limit of detection (LOD) of the assay is also increased from ~1nM to ~3nM after the addition of the translation solution. (b) The effects of the translation solution on the target binding assay. For both AMMP and ELISA assays, the sensitivity of the assays in translation solution is lower than the sensitivity of the assays in buffer. 169 Figure 10. Use of peptide-mRNA fusion molecules in target-binding assay. (a) mRNA of peptides, ligated to fluorescently-labeled DNA linker bound to puromycin, were in vitro translated to generate peptide-mRNA fusion molecules. The fluorescein tag on the fusion molecules binds anti-fluorescein antibody on the sensor surface, and peptide binds to immobilized Bcl-x L on a magnetic bead, linking it to the sensor surface. (b) Low levels of signal are measured for translated ScPep1 peptide or ScPep1-mRNA fusion molecule. High levels of signal are measured for translated Pep1 peptide and Pep1-mRNA fusion molecule. 170 Figure 11. Serum stability of the translated peptides. Synthetic Pep1, translated Pep1, and Pep1-mRNA fusion were diluted in the assay buffer or assay buffer containing 1% FBS. Synthetic peptide contains a C-terminal amide, and fusion molecules have a C-terminal puromycin. Translated peptide contains a natural carboxy C-terminus. (a) Pep1-mRNA fusion signal levels in 1% FBS buffer are slightly lower than assay buffer. (b) Signal levels for the Synthetic Pep1 in 1% FBS or assay buffer are nearly identical. (c) Signal level for the translated Pep1 is much lower in the buffer containing 1% FBS. 171 6.6 Appendix Name DNA Sequence Pep1 TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATGATCGAGACAATTACTATTT ACAAATACAAGAAGGCGGCGGACCACTTCTCCATGTCCATGGGCAGTGGTTCCGGCAGCTATCC ATACGACGTACCGGATTATGCT ScPep1 TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATGGCGAAAATCTACACATTCG CGTACGAGACTATGATGGACCACAAGTCCGGGAGCGGAAGCGGGAGTTACCCTTATGATGTTCC CGACTACGCA Pep2 TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATGTGGCGGTGGAAGATGATCG CGGACCAGCTGGGATCTGGCAGTGGCTCCTACCCCTACGACGTGCCCGACTACGCC P ep 1 Δ HA TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATGATCGAGACAATTACTATTT ACAAATACAAGAAGGCGGCGGACCACTTCTCCATGTCCATGGGGAGCGGAAGCGGGGGT wt-Fn TAATACGACTCACTATAGGGACAATTACTATTTACAATTACAATGCTCGAGGTCAAGGCTGCGA CTCCGACCAGCCTCCTGATCAGCTGGGATGCTCCTGCTGTCACAGTGCGCTACTACCGCATCAC CTACGGTGAAACAGGTGGCAATAGCCCTGTCCAGGAATTCACCGTGCCTGGGAGCAAGTCTACA GCTACCATCAGCGGCCTGAAACCTGGTGTCGACTATACCATCACGGTGTACGCCGTCACGGGCA TTAATGACAGCCCCGCAATCAGCAAGCCGATCTCCATCAACTACCGCACCACCGGATCCGGTTC CGGCAGCTATCCATACGACGTACCGGATTATGCT Pep1 Splint TTTTTTTTTTTNAGCATAATCCGGTAC ScPep1 Splint TTTTTTTTTTTNTGCGTAGTCGGGAAC Pep2 Splint TTTTTTTTTTTNGGCGTAGTCGGGCAC The DNA sequence of polypeptides used for in vitro translation. The first 20 nucleotides is recognized by T7 RNA polymerase for transcription. The next 23 nucleotides is a deletion of the Tobacco Mosaic Virus (ΔTMV) which has shown to be helpful for translation via reticulocyte lysate. The Splint sequences were used to hold together the mRNA to a synthetic DNA linker (F30P) during ligation via T4 DNA ligase. 172 Chapter 7: A General, Label-Free Method for Determining K d and Ligand Concentration Simultaneously 7.1 Introduction Immune assays remain the most widely used method for protein detection, tracking, and characterization. The generation of proteome-wide immune reagents provides an important route to address cancer biology, immunology, and basic research. However, a problem with most antibody-based assays is that neither the antibody concentration ([L] 0 ) nor the K d for the target are generally known. This is suboptimal in a variety of important situations ranging from antibody screening to quantitative immunoassays, and in the development of therapeutic antibodies where efficacy directly relates to affinity and specificity. 164 Generally, it is assumed that in order to determine K d for a ligand-protein interaction, one must know the concentration of the ligand. In this view, titration of ligand with target over a concentration range, combined with a method for detecting the ligand-target complex provides a direct means to determine K d values. 165, 166 For this reason, quantitative analysis using antibody- based assays is difficult since the concentration of the antibody is often unknown. A second issue with antibody-based diagnostics is that the prevailing model for analyzing equilibrium data treats antibodies as monovalent reagents. 112, 167, 168 This approach can work for the forward assay (target in solution) but produces erroneous results for the reverse assay (target immobilized). A third major issue is that measuring K d for high affinity ligands can be challenging because long off-rates can bias results, while some indirect methods require chemical labeling of ligands, which can alter K d . 169, 170, 171 173 Here, we developed a robust approach to determine both K d and [L] 0 values simultaneously by fitting the data to the equilibrium model. To do this, we use a modified version of the equilibrium assay first developed by Friguet et. al. 112 At the core of our work is combining simultaneous fitting of both parameters with data from two separate target concentrations. Modeling and error analysis of experimental and simulated data enable us to demonstrate that the method is robust and define conditions where our results are accurate. Using our approach, we have analyzed and overcome a significant problem in the field—proper modeling and analysis of reverse assays. Overall, our approach is platform independent and can be used in any system where protein concentrations can be accurately determined. We have demonstrated our approach using two established methods of protein analysis—classical ELISA and Acoustic Membrane MicroParticle technology (AMMP) with three different classes of reagents—peptides, antibodies, and small molecule ligands. 174 7.2 Materials and Methods Protein Expression and Purification: The gene for the first 209 amino acids of Bcl-x L (Clone HsCD00004711; Dana Farber/Harvard Cancer Center DNA Resource Core) was PCR amplified with Pfusion polymerase. An N-terminal avitag (AGGLNDIFEAQKIEWHEGG) was added via the PCR reaction for in vivo biotinylation using the BirA enzyme. 34 The product was purified via PCR purification column and cloned into the pET24a vector using NdeI and XhoI. Bcl-x L was expressed overnight at 37 °C in BL21(DE3) cells using auto-induction media. 33 Cells were lysed using Bper (Pierce), and purified using Ni-NTA superflow resin on an FPLC (Bio-Rad), using a gradient from 10 mM to 400 mM imidazole (Buffer A: 25 mM Hepes pH 7.5, 1 M NaCl, 10 mM imidazole; Buffer B: 25 mM Hepes pH 7.5, 1 M NaCl, 400 mM imidazole). Fractions with pure Bcl-x L were combined, concentrated, and desalted into 50 mM Tris-HCl, pH 8.0. Bcl-x L was biotinylated in vitro using BirA biotin ligase (0.1 mg/mL in 50 mM Tris-HCl, pH 8.3, 10 mM ATP, 10 mM Mg(OAc) 2 , 50 μM biotin) at 30 °C for two hours. The protein was buffer exchanged into 1X PBS, frozen in liquid nitrogen, and stored at -80 °C. Peptide Synthesis: Peptides E1 (NH 2 -MIETITIYNYKKAADHFSMSMGSK-NH 2 ), E2 (NH 2 - MIETITIYKYKKAADHFSMSMGSK-NH 2 ), D1 (NH 2 -MIAISTIYNYKKAADHYAMTKGSK- NH 2 ) and Bim (NH 2 -MDMRPEIWIAQELRRIGDEFNAYYARRGK-NH 2 ) were synthesized by solid phase Fmoc synthesis, using a Biotage Alstra Microwave Synthesizer. 30 The peptides were synthesized on Rink amide MBHA resin using five-fold molar excess of each amino acid and HATU. After the coupling of the first amino acid, (Fmoc-Lys(Mtt)-OH), the primary amine in the side-chain of the lysine for each peptide was deprotected using a solution of 1% (v/v) 175 trifluoroacetic acid (TFA) in Dichloromethane (DCM). Biotin was then coupled to the side-chain primary amine before the synthesis was resumed, resulting in biotin-labeled peptides. Peptides were cleaved from the resin and deprotection with a solution of 95% (v/v) TFA, 2.5% 1,2- ethanedithiol (EDT), 1.5% (v/v) deionized water (DI), and 1% (v/v) triisopropylsilane (TIS) for 2 hours at room temperature. 100 The resin was filtered out, and the peptide was precipitated using 4-fold (v/v) excess ether. The peptides were dried, resuspended in DMSO, and HPLC purified using a C 18 reverse phase column and a gradient of 10-90% acetonitrile/0.1% TFA in water. Fractions were collected and tested for the correct molecular weight using MALDI-TOF mass spectrometry. The correct fractions were lyophilized, dissolved in DMSO, and flash frozen at -80 °C. Radiolabeled Off-Rate Assay: The DNA sequences coding for the peptides were ordered from Integrated DNA Technologies (IDT). Each DNA construct contained a T7 RNA Polymerase promoter, and a 5’ deletion mutant of the Tobacco Mosaic Virus (ΔTMV). 95 The C- terminal portion of the peptides were elongated with a flexible serine-glycine linker (six amino acids long) and an HA tag. After gel purification using urea-PAGE, the DNA sequences were PCR amplified using Taq polymerase and in vitro transcribed into mRNA using T7 RNA polymerase. 95 After transcription, the mRNA was urea-PAGE purified and resuspended in deionized water to a final concentration of 30 µM. The samples were in vitro translated at 30 C for 1 hour in the translation solution—150 mM KOAc, 750 µM MgCl 2 , 2 µM mRNA, 1X translation mix (20 mM Hepes-KOH pH 7.6, 100 mM creatine phosphate, 2 mM DTT, and 312.5 µM of each amino acid excluding methionine), 35 S- labeled methionine (Perkin Elmer; 20 µCi for each 25 µL of translation), and 60% (v/v) rabbit 176 reticulocyte lysate (Green Hectares; prepared according to the method of Jackson and Hunt) 99 . Radiolabeled peptides were purified using magnetic HA beads (Life Technologies) and eluted with 100 µL, 50 mM NaOH, then immediately neutralized with 20 µL of 1 M Tris-HCl, pH 8.0. The radiolabeled peptides were allowed to bind to 30 pmol immobilized Bcl-x L for 1 hour in sample buffer (1X PBS, 1% (w/v) BSA, 0.1% (v/v) Tween 20, 10 µM biotin). The beads were magnetically separated, and washed 5X with sample buffer. The beads were resuspended in 1 mL of sample buffer containing 3 μM non-biotinylated Bcl-x L (~100X molar excess relative to immobilized biotinylated Bcl-x L ). At various time points, 100 µL of slurry was removed and the beads were magnetically separated and washed. The percent remaining at each time point was determined by dividing the counts per minute (cpm) on beads by total cpm (beads + washes). The peptide off-rate was determined by an exponential fit of the Percent counts on beads vs. Time (s). Bead Loading: 54H6 mAb was immobilized on magnetic beads by incubating 400 pmol of the antibody with 1.5 mg of tosyl magnetic beads (Life Technologies) in 1X PBS buffer at 4 C. After 48 hours, the reaction was quenched with 100 µL of 1 M Tris-HCl, pH 8.0. The beads were then washed and re-suspended in 1 mL of 1X PBS + 1% (w/v) BSA + 0.1% (v/v) Tween-20. Bcl-x L and D1 peptide were immobilized on magnetic beads by incubating 60 pmol of each biotinylated compound with 0.5 mg of streptavidin magnetic beads (Life Technologies) at 4 C overnight. To block any unbound sites on the streptavidin, 100 nmol of biotin was added and incubated with the beads for 30 minutes at room temperature. The beads were then washed with sample buffer, and resuspended in 600 µL of the same buffer without biotin. 177 Fluorescein Labeling of the Anti-HIS and Anti-Rabbit Antibodies: Anti-HIS (Thermo Scientific) or Anti-Rabbit (Thermo Scientific) antibodies were buffer exchanged to 1X PBS using a NAP-25 column (GE Healthcare) to remove sodium azide or other preservatives in the storage solution. A twenty-fold molar excess of NHS-fluorescein (Pierce) in DMF was then added to each buffer-exchanged antibody and incubated for one hour at room temperature in the dark. The reactions were quenched with 1 M Tris-HCl, pH 8.0, and buffer exchanged into 1X PBS using NAP-25 columns to remove the unreacted NHS-fluorescein. The concentration of the peptide and anti-HA antibody were calculated as per manufacturer’s instructions. Sample Preparation: A set of serially diluted Bcl-x L standards, at 2X the desired concentration, were made in sample buffer. For each ligand to be tested, a set of dilutions at 2X the desired concentration was also prepared. The Bcl-x L samples were either mixed 1:1 with sample buffer (standards) or ligands (samples), and allowed to incubate at room temperature for 6 days. After the incubation, the standards and samples were analyzed using ELISA or the ViBE BioAnalyzer (Fig. 1). ELISA Assays: ELISA plates were incubated overnight at 4 C with 1.5 nmol of streptavidin (for D1 or Bcl-x L capture ligands) or 54H6 mAb in 1X PBS. Plates were washed 3X with wash buffer (1X PBS + 0.1% (v/v) Tween-20) and blocked with 1X PBS + 5% (w/v) BSA for two hours. For the D1 or Bcl-x L capture ligands, 100 μL of a 30 nM solution of the reagents was added to wells and incubated for 1 hour. This step was skipped for the 54H6 mAb capture ligand (already immobilized on the plate). After the capture ligand incubation, 100 µL of sample or standards were added in each well, and incubated for 1 hour at room temperature. Plates were 178 washed, incubated with HRP-conjugated probe antibody in sample buffer for 1 hour, washed, and incubated with TMB substrate (Thermo Scientific). Reactions were stopped after approximately 10 minutes with 2 M sulfuric acid, and the absorbance at 450 nm was measured via a plate reader (Molecular Devices). Ligand of interest ABT 737 Peptide Ligands mAb (Forward Assay) mAb (Reverse Assay) Capture Ligand D1 Peptide D1 Peptide 54H6 mAb Bcl-x L Target Bcl-x L Bcl-x L Bcl-x L 54H6 mAb Probe Ligand Anti-HIS-HRP Anti-HIS-HRP Anti-HIS-HRP Anti-Rabbit-HRP AMMP Assays: For the AMMP assays, 90 μL of each sample or standards was incubated with 30 μL of magnetic beads (12 µg of beads/mL) and fluorescein-labeled antibody (8 nM) in sample buffer for 1 hour. The experiment’s run buffer was 1X PBS + 1% (v/v) Tween-20 + 1% (v/v) heat-treated FBS (Invitrogen; FBS was heat treated for 15 minutes at 65 C and filtered). BioScale Universal Detection Cartridges were used in performing all of the assays. The device was used per the manufacturer’s instructions. 148 Assay ABT 737 Peptide Ligands Forward mAb Assay Reverse mAb Assay Ligand on Beads D1 Peptide D1 Peptide 54H6 mAb Bcl-x L Target Bcl-x L Bcl-x L Bcl-x L 54H6 mAb Probe Ligand Anti-HIS-Fl Anti-HIS-Fl Anti-HIS-Fl Anti-Rabbit-Fl Monovalent and Divalent Analysis: The data for both sets of target concentrations were simultaneously fit for K d (in the K d only fit) or K d and [L] 0 . The data is fitted to the equilibrium model using the lowest absolute deviation method, by varying either only K d or both K d and [L] 0 179 simultaneously. The monovalent assay fitting was done by Excel Solver (GRG Non-Lin method) using a set of five initial values. The set of values which provided the lowest error after the fitting were chosen as the final values. For the Divalent assays, the fitting was performed by MATLAB’s fminsearch function and a set of 10 initial values for K d1 , K d2 , and [L] 0 . In order to calculate the %C EQ value, first the concentration of monovalently bound antibody was found by finding the real, positive root of the cubic function in Supplementary Figure 2. For the divalent reverse assay, an extra parameter, C f , was also determined by fitting (Supplementary Fig. 2 Reverse Assay). Simulated error analysis: To prepare the 3D-error plot in Figure 2d, we used 8 simulated data points where two [T] 0 values (T H is high [T] 0 and T L is low [T] 0 , and T H =10 x T L ) and 4 [L] 0 values were chosen (starting from 10 x T H diluted serially with a dilution factor of 1:10). We constructed a 2D matrix in MATLAB where the x-coordinate represents the deviation in K d over a 2 order of magnitude window, and the y-coordinate represents the deviation in [L] 0 . We then evaluated the total difference between %C EQ when calculated using the deviated K d and [L] 0 values, vs the True K d and [L] 0 values for all 8 data points, and dubbed this difference the error. The error matrix also depends on the relationship between the true K d value and T H . Six values for K d /T H ratios were tested [100 – 0.01, going by factors of 10], and the result of one of these (where true K d = T H ) is shown in figure 2d. These 2D error matrices were also used in the step-wise analysis for Supplementary Figure 3. To perform this type of analysis, we chose a specific column (deviation in K d ) in the matrix. The row with the lowest error for the chosen column represents the optimum [L] 0 value for the specific deviation in K d . If the initial chosen column also represents the lowest error in the 180 optimum [L] 0 row, then the pair of K d and [L] 0 are a stable pair. If not, then the lowest error in the row should be used to find the new optimum deviation in K d , and this iterative method should be continued until a stable pair of values are reached. 181 7.3 Results and Discussion Target and Ligands Used: In our experiments we used ligands directed against B-cell Lymphoma extra-large (Bcl-x L ), an oncogenic protein which is up-regulated in several types of human carcinomas 172 and a target for therapeutic development. This protein has three distinct classes of known ligands—antibodies, peptides, and small molecules. We purchased two of these classes from commercial vendors—1) monoclonal antibody 54H6 and 2) the high affinity small molecule ABT-737. 114 For peptides, we synthesized a 26-residue fragment of Bim (a pro- apoptotic natural ligand of Bcl-x L 173 ) and three ultrahigh affinity peptides (K d ≤ 1 nM) that bind to Bcl-x L (Takahashi and Roberts, manuscript in preparation). Importantly for these assays, the peptides and small molecule bind one site in Bcl-x L and the antibody binds a second, noncompeting site on the protein. Forward (Target in Solution) Equilibrium Assay: In order to determine the K d for our ligands, we modified the method described by Friguet et. al. 112 (Fig. 1a). In this assay, a capture ligand pulls down the free target in solution. A competing ligand (of unknown K d ) is incubated with target and allowed to equilibrate, reducing the amount of free target in solution. The K d of interaction can then be determined by quantifying the amount of free target. The response curve for target quantitation is shown in Fig. 1b. Using the response curve, we chose two target concentrations that gave signal that was above background yet not saturated (111 pM and 1 nM, indicated with arrows) for our analysis. At each of these concentrations, competing ligand (Bim in Fig. 1) was equilibrated with the sample to reduce the signal (Fig. 1c). These data are fit to yield a single K d and result in two curves that correspond to the different target concentrations 182 (Fig. 1d, equilibrium models for monovalent and divalent ligands are shown in Supplementary Fig. 1-2). This equilibrium assay is transferable to any method capable of sensitive measurement of analyte concentration. To show this, we used a commercially available quantitation platform, the ViBE BioAnalyzer, capable of high throughput automatic sample analysis (Fig. 1e). Comparing the AMMP (ViBE Platform) and ELISA methods demonstrates that antibody, small molecule, and peptide ligands give the same K d values independent of the measurement method (Fig. 1f and Supplementary Table 1). This validates the AMMP approach for K d measurements as the accuracy of the equilibrium ELISA method has been shown extensively. 112 Additionally, our measured K d value for the Bim peptide (130 ± 40 pM) matches the reported value in the literature (140 pM) 174 , and our calculated k on values for all tested peptides fall within 10 4 -10 6 (M - 1 s -1 ) typically observed for most protein-protein interactions 175 (Supplementary Table 2). The forward assay is especially useful for screening multiple ligands to find the best binding sequences that can block a specific interaction (e.g., generating therapeutic monoclonal antibodies), as it can rapidly determine the dissociation constants of multiple competing ligands for a single target. If all ligands bind to the same epitope, only a single capture ligand is needed to create a target response curve, greatly reducing the number of samples needed to accurately measure K d for all ligands. We used this feature to measure the K d of multiple ligands with a single capture ligand and corresponding standard curve. Measuring K d in the Case of Unknown Ligand Concentration: Measuring the K d by equilibrium assays or directly measuring the formation rate constant (e.g., by surface plasmon resonance) is very sensitive to the concentrations of target and ligand. It is generally assumed 183 that knowing the ligand concentration is required to determine accurate K d values. This prerequisite is one issue that makes high-throughput K d screens for ligands difficult, time consuming, and infeasible on a proteomic scale. However, measuring the ligand concentration is not always simple. Many factors can complicate accurate measurement of ligand concentration such as unknown expression levels, or unknown fraction of functional/correctly-folded ligand, or the desire to use crude, unpurified samples for highest throughput. 118 The key insight of this work is in exploring whether K d and [L] 0 can be determined simultaneously, and defining under what conditions this analysis is accurate and robust. In our experiments to determine K d values for Bcl-x L ligands, we followed a traditional approach and determined the value of [L] 0 for each of our binders (Fig. 1). We also re-analyzed this same data without inserting the value of [L] 0, and attempted to determine both K d and [L] 0 simultaneously. Remarkably, this analysis revealed the same values of K d (Fig. 2a) and [L] 0 (Fig. 2b) obtained using standard approaches for all three classes of ligands. The correspondence between the two approaches is excellent, giving the same values of K d over the entire range studied. This might be possible in our work as compared to prior work in the field due to the fact that we had used two sets of target concentrations to generate our equilibrium response curves, vs. the more typical approach of using a single target concentration. Fidelity of the Fit and Parameter Sensitivity: While the results from the simultaneous fit for K d and [L] 0 look very promising, it is important to learn about the potential weaknesses of this type of analysis. We explored the sensitivity of our fitting to each of the input values of K d and [L] 0 . Figure 2c shows a rudimentary measure of the fidelity of each parameter. After obtaining K d and [L] 0 values through simultaneous fitting, we kept one parameter constant and changed the 184 other parameter by an order of magnitude in each direction to show the accuracy of the obtained values (light and dark gray dashed lines). Plots of data similar to Figure 2c are often seen in the literature as proof that the fit values for K d and [L] 0 are correct. 176, 177, 178, 179, 180 When a pair of K d and [L] 0 values are fit, the error between the data and the equilibrium model is plotted as one parameter is fixed, and the other is scanned over a range. Values are accepted when each parameter produces the minimum level of error when the other parameter is fixed (Supplementary Fig. 3a-b). The shortcoming of this iterative fitting analysis is that it cannot show how changing one parameter can compensate for changing the other. This approach can result in self-consistent pairs of K d and [L] 0 that are incorrect and far from true K d and [L] 0 values (Supplementary Fig. 3c). Fitting for two variables simultaneously can result in a situation where varying one parameter can compensate for the error generated when the other parameter is moved. To address this problem, we carried out a more rigorous analysis of parameter sensitivity. To do this, we needed a way to visualize how the overall error changes for all combinations of fit K d and [L] 0 values. Given true K d and [L] 0 , and two target concentrations each with 4 dilutions of ligand, we simulated 8 data points. We then varied K d and [L] 0 within a four orders of magnitude window, and calculated binding percentages at equilibrium. Error was defined as the total distance between the two sets of data points (Fig. 2d). This type of analysis produces an error surface where the z-axis corresponds to the error and the x- and y-axis values show the changes in K d and [L] 0 using the true values of each as a reference point. Hence, at the center of the plot (where K d and [L] 0 = their true values) the error (z-axis) is defined as zero. Looking at the plot in Figure 2d, it is clear that many different combinations of [L] 0 and K d result in relatively large error values. Put another way, the error surface approaches the x-y plane 185 (where error is lowest) for a very restricted set of values of both parameters—the ravine running down the middle of the surface. This approach to viewing the data obscures whether there is a unique solution where error is minimized, or whether there are a family of solutions of K d and [L] 0 that give error values very near the x-y plane. To address this, we projected the error surface (Fig. 2d) onto the [L] 0 -error plane (Fig. 2e) or the K d -error plane (Fig. 2f), and only retained the lowest error values for each projection (details shown in Supplementary Fig. 4). A point on each line in Figure 2e thus represents the minimum error for a given variation in [L] 0 , over all tested K d values. The lines corresponding to the error surface in Figure 2d can be seen as purple dashed lines in Figures 2e and 2f. Viewed in this way, it becomes apparent that the accuracy of this analysis depends on the K d value in relation to the concentrations of the target (low target concentration—T L and high target concentration—T H ) used in the experiments. The purple dashed lines in Figure 2e and 2f produce a unique, unambiguous solution approaching the x-axis at a single point, the true value of [L] 0 and K d respectively. Some choices of target concentrations vs. K d can give even more clear solutions (blue, green, and orange curves in Fig. 2e and the orange curve in Fig. 2f). Importantly, there are choices of target concentrations that give ambiguous results (e.g., the red curve in Fig. 2e-f). In these cases, it is clear that these target concentrations cannot be used to determine accurate values of K d and [L] 0 . Indeed, this type of analysis can be formulated as a set of rules that direct where K d and [L] 0 can be determined. When the high and low target concentrations are 10-fold apart and the ligand concentration ranges from 10 x T H to 0.1 x T L , accurate K d values can be obtained for T H > K d > 0.1 x T L . The accuracy of fit [L] 0 follows a significantly different rule: the fit for [L] 0 is accurate when T H > K d , and is improved continuously as the K d is lowered with respect to initial target 186 concentration. These ranges are guidelines for assessing the accuracy of the obtained K d and [L] 0 values. If the obtained K d value is within the T H > K d > 0.1 x T L range, the fits can be trusted. However, if the obtained K d is outside the window, the experiment must be repeated with new initial target concentrations. This same type of analysis can be used to demonstrate that accurate K d and [L] 0 values cannot be determined using a single target concentration (Supplementary Fig. 5), showing that at least two concentrations of target are needed. The validity of the above ranges is shown in Figure 3. When the true K d is within the optimum range, a 5-fold deviation in fit K d cannot be compensated for by adjusting the [L] 0 value (Fig. 3a). Here, the erroneous K d and [L] 0 values do not fit the data. However if a single target concentration is used (Fig. 3b), or K d is outside the specified range (Fig. 3c-d), the data points and the erroneous K d and [L] 0 values match and would be falsely interpreted as “correct” K d and [L] 0 values. There are examples in the literature where others have worked to determine K d and activity or K d and [L] 0 iteratively. Our work indicates that these approaches are flawed as they either use a single target concentration 178, 180 or target concentrations outside of the window of accuracy 176 . Importantly, those analyses do not specify the ranges where the calculations are valid. Our results demonstrate that these approaches are prone to generate erroneous data while giving no indication the fits are incorrect. One outcome of our approach is that any experimental method that extends the quantitative range of the response curve (e.g., vs. standard ELISA) provides a means to determine high affinity binding constants with high accuracy. Here, the commercial AMMP device used for some of our analysis provides this extended range. The AMMP assay is more sensitive than the ELISA (Supplementary Fig. 3b) and on average yielded a ~5-fold increase in sensitivity, in line 187 with our previous observations. 118 The higher sensitivity of the AMMP assay makes K d measurements possible even with sub-picomolar interactions. Treating Antibodies as Divalent Ligands: Antibodies (e.g., IgGs) are divalent with two chemically equivalent target-binding sites. Most analyses model antibody binding by considering each antibody molecule as two separate ligands, each with a single identical binding site. 112 This approach poorly represents the true situation, because binding of the second target often has a weaker K d value than the first, likely due to a combination of effects such as excluded volume from binding the first target molecule. 181, 182, 183 Using the reverse assay to derive quantitative data has also been problematic due to inadequate consideration of avidity issues, where many times the measured parameters can be more indicative of the experimental conditions than the ligand kinetics. 184 This is unfortunate because the reverse assay is much more efficient at determining the K d of one ligand vs. many targets than the forward assay (e.g., to determine the specificity of a ligand). To address these issues, we systematically attempted to fit data in the forward and reverse assays with monovalent and explicit divalent models, toward the goal of quantitating valency effects and developing a useful version of the reverse assay. Divalent Ligands: Forward Assay: The forward assay (Fig. 4a-b) is the same for both monovalent and divalent ligands. When only fitting for K d , the divalent model provides better fits for the data than the monovalent model (Fig. 4c) and gives markedly different results for K d (38 pM for the monovalent model vs. 14 pM for the divalent model). When fitting for both K d and [L] 0 simultaneously (Fig. 4d), both models give curves that fit the data well and produce K d values identical to the divalent K d -only fit (K d = 11 pM). However, the monovalent model 188 produces a fit [L] 0 that is equivalent to the antibody concentration and thus half of the total concentration of sites. Our data indicate that for the forward assay to give accurate K d values, one must use the antibody concentration (rather than the number of sites) with the monovalent equilibrium model, a marked change from current practice. This is due to the negligible contribution of the divalently bound ligand at equilibrium for the forward assay (Fig. 5e), essentially turning antibodies into monovalent ligands under these conditions. We showed in the first segment of this paper how a pair of erroneous K d and [L] 0 values can match the data points when a single target concentration is used. Since most equilibrium immunoassays to determine antibody K d values use a single target concentration, previous studies have failed to uncover this discrepancy. This issue is only observed when multiple concentrations of target are used, however it is often simply attributed to ligand activity. An activity coefficient of 0.5 is often obtained, arguing that half of antibody sites are non-functional (i.e., mean activity coefficient for various antibodies reported as 0.47 ± 0.07 178 and 0.53 ± 0.05 179 ). Divalent Ligands: Reverse Assay: The Schematic for the reverse assay is shown in Figure 5a-b. Unlike the forward assay, in the reverse assay the target is immobilized and used to capture the free ligand in solution (Fig. 5a). The main difference between the forward and the reverse assay is that for multivalent ligands, monovalently bound ligands are still able to interact with the immobilized target (Fig. 5b). The strength of this interaction depends on the cooperativity of the binding sites as well as the immobilized target density. Due to this effect, the use of the reverse assay has been discouraged in the past. 185 For the divalent equilibrium model, we added a cooperativity term to account for the strength of interaction between the target and a free ligand 189 vs. a monovalently bound ligand. The cooperativity factor (C f ) measures the percent of the monovalently bound ligand which does not interact with the immobilized target. This means that for the divalent model, the effective complex concentration at equilibrium is the concentration of the divalently bound ligand (unable to interact with the immobilized target) plus the concentration of the monovalently bound ligand multiplied by the cooperativity factor (concentration of the monovalently ligand which is unable to interact with immobilized target). Much like the forward assay, two concentrations of the species in solution (here, the monoclonal antibody) were used to obtain accurate K d and [L] 0 values. Data from a sample reverse assay is shown in Figure 5c. When the high and low ligand concentrations are fit to equilibrium models, only the divalent model simulates the behavior of the obtained data points. Interestingly, simultaneously fitting for both K d and [L] 0 does not help the monovalent model match the data better than fitting for K d only (Fig. 5d). For the reverse assay, both the monovalently bound and divalently bound species are present at significant quantities and contribute to the effective complex composition at equilibrium. While at low target concentrations the monovalently bound ligand dominates the signal, at high target concentration the divalently bound ligand has the most significant contribution (Fig. 5e). The cooperativity constant depends on several factors such K d1 , K d2 , and immobilized target density. The value of the cooperativity factor was obtained by fitting and remained consistent for all experiments: 74% ± 4% for the K d fit only and 73% ± 3% for the simultaneous K d -[L] 0 fit. While the data from the forward assay is convincing that divalent modeling of the antibody is more accurate than monovalent modeling with twice the concentration, it is still possible that our antibody was simply ~50% inactive. Obtaining accurate K d and [L] 0 values from the reverse 190 assay that match the forward equilibrium assay solves a persisting problem in the field and removes any doubt that our antibody was not inactive, rather, all antibody sites are functional. 191 7.4 Conclusions Antibodies are often used non-quantitatively because neither [L] 0 nor K d values for the target are known. Here, we determine both [L] 0 and K d values simultaneously using a direct, label-free, and general approach. In addition our approach gives accurate values for both peptide and small molecule ligands. Our approach is platform independent, reproducible and robust. The K d and [L] 0 values are obtained by performing quantitative equilibrium immunoassays with two different concentrations of target and fitting the data simultaneously to the equilibrium model. We tested the validity of our approach vigorously by performing detailed error analysis, and we demonstrate that our fitting gives unique and reproducible solutions. Further, we defined where K d and [L] 0 measures are reliable and where they are underdetermined. By using a divalent equilibrium model for antibody binding, we have shown that obtaining reliable K d and [L] 0 values is only possible when the cooperativity factor between the two antibody binding sites has been taken into account. This approach solves a long-term problem of obtaining quantitative data from reverse assays 192 7.5 Figures 193 Figure 1. Measuring K d via forward equilibrium immunoassays (target in solution). (a) Schematic for generating ELISA signal: Target protein (Bcl-x L ) binds capture-ligand immobilized on the ELISA plate. Bound target is quantified with a detection-antibody/HRP conjugate. Target equilibrated with increasing concentrations of a competing ligand (here Bim) reduces the signal, since the pre-formed ligand/target complex cannot interact with the ELISA plate. (b) ELISA signal for known concentrations of Bcl-x L fit to a 4 parameter logistic model. Two target concentrations (1 nM and 111 pM) were chosen for pre-incubation with Bim. (c) Loss of ELISA signal resulting from equilibrating 1 nM or 111 pM Bcl-x L (red diamonds and squares, respectively) with Bim. The signal represents the unbound target (concentration calculated using panel b). (d) Determining the K d value for the ligand. The fraction of Bcl-x L bound (%C EQ , diamonds and squares) and ligand concentrations are fit to the equilibrium model (Supplementary Fig. 1-2). Data from both high and low target concentrations are fit simultaneously to obtain a K d value. (e) Schematic for K d measurements generated via AMMP. The capture-ligand is immobilized on magnetic beads and incubated with target and fluoresceinated detection-antibody. Binding of the detection-antibody to the anti-fluorescein antibody on the sensor surface connects the magnetic bead to the sensor, generating signal. As with the ELISA assay, signal is reduced when a competing ligand is equilibrated with the target. (f) The K d values obtained using the AMMP assay are equivalent to the results obtained by ELISA for peptides, small molecule and antibody ligands. 194 195 Figure 2. Simultaneous fitting of K d and [L] 0 produces accurate results. (a) Fitting for K d and [L] 0 simultaneously yields K d values that are equivalent to the values obtained when [L] 0 is known. (b) Ligand concentrations determined by simultaneous fitting of K d and [L] 0 match the known [L] 0 . (c) Simultaneous fitting of K d and [L] 0 for peptide E1 using the monovalent equilibrium model yields a unique solution (red line). Light grey and dark gray dashed lines demonstrate the fidelity of the fit to the high (T H , red diamonds) and low (T L , red squares) target concentration samples when K d and [L] 0 are each varied ± 10-fold while the other variable is held constant. Here, the x-axis is given as relative concentration (DF -1 ) since [L] 0 is unknown. (d) 3-D surface plot showing the error (absolute deviation, z-axis) between a simulated data set calculated from true [L] 0 and K d values, and data sets where [L] 0 and K d are allowed to vary ±100-fold from their true values (see Online Methods). A unique and accurate solution for [L] 0 and K d can be determined if the error surface only approaches the x-y plane at the true values of [L] 0 and K d . (e, f) The lowest values of the projected error surface as viewed on the error vs [L] 0 or error vs K d planes, respectively (details in Supplementary Fig. 2). A higher error projection (e.g., the blue projection in panel c) corresponds to higher sensitivity of the measured parameter resulting in better accuracy and precision. 196 Figure 3. Fitting using two or more target concentrations that bracket K d is required to derive accurate values for K d and [L] 0 . The above data points were simulated to illustrate the range where simultaneously fitting for K d and [L] 0 produce accurate results. For each plot, the fit K d value was set to 5-fold the true K d value, and the fit [L] 0 value was chosen to minimize the error. The data points and the black lines represent the true K d and [L] 0 values for each plot. (a) Within the optimal range for accurate K d and [L] 0 measurement by simultaneous fitting (T H > K d > 0.1 x T L , obtained from Figure 2e-f), the erroneously fit K d and [L] 0 (red dashed lines) do not match the data. However, when using a single target concentration (b) or working outside the appropriate target concentration ranges (c-d), plots using the erroneous values (red dashed lines) can show good overlap with the data, despite a five-fold deviation in K d . 197 Figure 4. In the forward assay, accurate K d and [L] 0 values can be determined by modeling antibodies as monovalently bound ligands. (a) Schematic to generate the standard curve for the forward assay. The target can bind to immobilized antibody on solid support (here, ELISA plate) in a monovalent or divalent format. (b) Schematic for the forward assay at equilibrium. Equilibration of target and antibody generates both monovalently-bound and divalently-bound target-ligand complexes. Neither complex can interact with the immobilized antibody on the solid support, lowering the signal similarly to Figure 1c. (c) The traditional approach to determine binding constants (a monovalent model using the number of antibody sites as the ligand concentration) results in both large errors and erroneous K d values (dashed black lines) when fit for both target concentrations. A model treating the ligand as divalent results in better fits at both target concentrations (red lines). (d) Simultaneous fitting of K d and [L] 0 results in excellent fits for both monovalent and divalent models and gives identical values for K d , but results in a two-fold difference in the fit ligand concentration (R L = the ratio of the fit [L] 0 to known [L] 0 ). The K d values from the simultaneous fits also match well with the divalent K d only fits in panel (c). (e) Fraction of signal due to monovalent (red dashes) and divalent (red dots) antibody-target complexes. In the forward assay, >99% of the signal arises from the monovalent complex. 198 Figure 5. In the reverse assay (target immobilized), determining K d and [L] 0 can only be done accurately when a divalent model is used. (a) Schematic to generate the standard curve for the reverse assay. The antibody can bind to a single immobilized target on solid support or it can bridge two nearby target proteins. (b) Schematic for the reverse assay at equilibrium. The monovalently bound ligand can bind to the immobilized target and give rise to signal whereas the divalently bound ligand cannot. (c) Calculating the K d values for the reverse immunoassays. The best-fit curve of the monovalent equilibrium model does not match the experimental data for either high (blue diamonds) or low (blue squares) ligand concentration sets. In contrast, the divalent model (solid line) matches the data very closely. (d) Simultaneous fitting of K d and [L] 0 for the reverse assay. The monovalent model does not match the data when K d and [L] 0 are fit simultaneously. Both the divalent and the monovalent K d values are similar to the calculated values in panel c (e) The divalent complex has a very significant contribution in the reverse assay. At low target concentrations, the monovalent complex dominates the signal, whereas at high target concentrations, the divalent complex has a greater contribution. This effect can be treated using a negative cooperativity term (C f ) corresponding to the percent of monovalently bound ligand that does not interact with the immobilized target. 199 Ligand Equilibrium Model K d Determined Using Known [L] 0 (pM) K d Determined by Fitting for [L] 0 (pM) Ratio of Fit [L] 0 to Known [L] 0 D1 Pep Monovalent 8.5 ± 2 14 ± 5 109% ± 7% E1 Pep Monovalent 39 ± 6 27 ± 12 88% ± 10% Bim Pep Monovalent 130 ± 40 150 ± 80 110% ± 12% E2 Pep Monovalent 300 ± 14 240 ± 94 96% ± 40% ABT-737 Monovalent 3,100 ± 360 1,900 ± 790 83% ± 24% 54H6 mAb Divalent K d1 = 21 ± 6 K d2 = 3,300 ± 1,300 K d1 = 19 ± 4 K d2 = 4,000 ± 1,800 90% ± 11% Table 1. Measured K d values and [L] 0 ratios for the tested ligands. Mean K d values and [L] 0 ratios with associated standard errors are reported. The data are from both the ELISA and the AMMP assays. For the mAb, K d1 refers to the dissociation constant for the free mAb for Bcl-x L . The K d1 values for the 54H6 mAb are obtained by combining the data from both forward (target in solution) and reverse (target immobilized) assays. The mAb K d2 values were obtained using only the reverse assay, as the divalently bound species was a significant contributor to the overall results in this format. 200 7.6 Supplementary Figures Supplementary Figure 1: Formulas governing the equilibrium and transient behavior of a simple binary binding system. The ligand binds to the target to form the target-ligand complex with the rate constant k on . The complex dissociates back into the target and ligand in solution with the rate constant k off . The total concentration of ligand or target at any point in the reaction is restricted such that the amount in complex ([C]) and the amount free in solution ([L] or [T]) must add up to the initial amount added to the reaction ([L] 0 or [T] 0 ). The transient solution can be used to ensure enough time has been allocated for the samples to reach equilibrium. 201 Supplementary Figure 2: Formulas governing the equilibrium behavior of divalent ligand. The ligand binds to the target to form the monovalently bound target-ligand complex ([TL]) with the rate constant k on1 . The complex dissociates back into the target and ligand in solution with the rate constant k off1 . The monovalently bound target-ligand complex ([TL]) binds to the target to form the divalently bound target-ligand complex ([T 2 L]) with the rate constant k on2 . The complex dissociates back into the target and monovalently bound target-ligand complex ([TL]) with the rate constant k off2 . The concentration of the monovalently bound target-ligand complex at equilibrium ([TL] EQ ) is the real positive root to the cubic function shown above. The concentration of the divalently bound target-ligand complex at equilibrium ([T 2 L] EQ ) can be calculated once the [TL] EQ has been found. 202 Ligand Equilibrium Model K d Determined by ELISA (pM) K d Determined by AMMP (pM) D1 Monovalent 7 ± 2 12 ± 2 E1 Monovalent 34 ± 9 45 ± 8 Bim Monovalent 170 ± 55 77 ± 11 E2 Monovalent 290 ± 29 315 ± 9 ABT-737 Monovalent 2,700 ± 260 3,500 ± 660 54H6 Divalent 20 ± 5 12 ± 7 Supplementary Table 1: The K d values for the ligands as determined by the ELISA or the AMMP assays. Mean values and standard errors are reported. 203 Ligand Type K d (M) k off (s -1 ) k on (M s -1 ) D1 Peptide Ligand 8.5E-12 2.0E-6 2.4E5 E1 Peptide Ligand 3.9E-11 1.2E-5 3.1E5 Bim Peptide Ligand 1.3E-10 1.2E-4 9.2E5 E2 Peptide Ligand 3.0E-10 1.6E-5 5.3E4 Supplementary Table 2: Calculating the kinetic on-rate for Bcl-x L binding peptides. The K d for the peptides is measured by the equilibrium ELISA/AAMP assays (Table 1). The off-rate for these peptides was obtained by measuring the dissociation rate for radiolabeled peptide-mRNA fusions bound to immobilized Bcl-x L . The on-rate was calculated based on the equilibrium K d measurements and the radiolabeled off-rate. 204 Supplementary Figure 3: Iterative fitting methods can produce stable but erroneous pairs of K d and [L] 0 values. Panels a and b show the calculated error using true K d and [L] 0 vs the iterative optimization method developed by Darling and Brault (red). Sequential optimization can result in stable pairs for the fit K d and fit [L] 0 that minimize the calculated error, but do not match the true K d and [L] 0 . Plotting the target bound vs. dilution factor for the example in panel c, demonstrates that the true K d and [L] 0 values accurately fit all the data (black lines), whereas the sequential method (red dashed lines) does not. 205 Supplementary Figure 4: Obtaining lowest error values as a tool for assessing parameter sensitivity. (a) Deviation between the true %C EQ and %C EQ obtained by varying K d and [L] 0 each by 2 orders of magnitude (error) where T H = true K d . (b) Projection of panel (a) on the [L] 0 vs. error plane. (c) Projection of panel (a) on the K d vs. error plane. (d) The lowest error obtained from panels (b) and (c). The lowest error for a given [L] 0 deviation on the graph to the left provides the minimum error generated by testing all K d values. 206 Supplementary Figure 5: Using a single target concentration leads to underdetermined K d and [L] 0 values. (a, b) Minimum values for the 3D-error surface as viewed on the [L] 0 vs. error plane or K d vs. error plane, respectively (details of this process are shown in Supplementary Fig. 2). The error projections are much broader than when two concentration of target are used (Fig. 2c and 2d) making it difficult to uniquely determine accurate values for K d and [L] 0 , since there are multiple values of K d or [L] 0 , that result in small minimum errors. A single target concentration thus results in lower precision and accuracy of the fit K d and [L] 0 , values. 207 Supplementary Figure 6: Advantages of the AMMP assay over ELISA. (a) AMMP assay signal for Bcl-x L Standards is fit to a 4 parameter logistic model. The magnetic beads collected on the AMMP sensor surface are washed at three flow rates: low (blue circles, highest sensitivity), medium (red diamonds) and high (black squares). The use of the three flow rates extends the dynamic range of the assay to ~3 log units. (b) The AMMP assay is more sensitive than ELISA for identical samples and affinity reagents. The Lower Limit of Quantification (LLOQ) for the assays are marked with a green arrow (ELISA, 37 pM) and a blue arrow (AMMP 4 pM). 208 Chapter 8: High-throughput Binding Kinetics Measurement by mRNA Display and Next-Generation Sequencing 8.1 Introduction Various in vitro selection techniques (e.g., phage display, 140 ribosome display, 142 and mRNA display 143 ) have facilitated the generation of polypeptide ligands against targets of interest. Recent advances combining in vitro selection with high-throughput sequencing has greatly accelerated the process of generating large lists of potential ligands. 120 The challenge, increasingly, is ranking the molecules based on their desirable properties, chiefly, their affinity for their targets. 118, 186 Our initial hypothesis was that the affinity of a ligand would directly correlate to its frequency rank-order in this list of potential ligands. Although we have shown that higher ranked sequences do exhibit functionality, 120 we observe that a sequence’s rank shows poor correlation with its binding affinity, i.e, a higher ranked sequence is not always better than a lower ranked one. For example, a ligand with the frequency rank of over 15,000 (D1) exhibited a higher affinity than a sequence with a much higher frequency in the pool (E1, ranked 3rd) (Takahashi and Roberts, manuscript in preparation). Due to this discrepancy, there is a great need for methods that evaluate the affinity of a ligand for its target in a high-throughput manner. Advances in the field have been able to increase the throughput of Kd measurements using radioactivity, 186 SPR or fluorescent microarrays, 187, 188 and ELISA assays (Jalali-Yazdi and Roberts, manuscript in preparation) 118 . All of these methods require individually expressed and purified ligands, greatly 209 reducing their throughput. Measuring the Kd for thousands of potential ligands simultaneously has not yet been realized. In this work, we combined high throughput DNA sequencing with mRNA display to obtain kinetic on- and off-rates, and thus K d values, for tens of thousands of ligands simultaneously. To do this, we chose two enriched pools from our selection against B-cell lymphoma extra-large protein (Bcl-x L ) (Takahashi and Roberts, manuscript in preparation). These pools are the final enriched pools from an extension selection (selection that resulted in 21 amino acid long peptide ligands against Bcl-x L ) and a doped selection (the top sequence from extension was used to create a biased library, to further optimize binding). The mRNA of both pools were ligated to a 3’ DNA linker attached to puromycin, in vitro translated, purified and reverse transcribed. 143 A small fraction of each pool was also translated using radiolabeled methionine to track pool binding. 210 8.2 Materials and Methods Protein Expression and Purification: The gene for the first 209 amino acids of Bcl-x L (Clone HsCD00004711; Dana Farber/Harvard Cancer Center DNA Resource Core) was PCR amplified with Pfusion polymerase. An N-terminal avitag (AGGLNDIFEAQKIEWHEGG) was added via the PCR reaction for in vivo biotinylation using the BirA enzyme. 34 The product was purified via PCR purification column and cloned into the pET24a vector using NdeI and XhoI. Bcl-x L was expressed overnight at 37 °C in BL21(DE3) cells using auto-induction media. 33 Cells were lysed using Bper (Pierce), and purified using Ni-NTA superflow resin on an FPLC (Bio-Rad), using a gradient from 10 mM to 400 mM imidazole (Buffer A: 25 mM Hepes pH 7.5, 1 M NaCl, 10 mM imidazole; Buffer B: 25 mM Hepes pH 7.5, 1 M NaCl, 400 mM imidazole). Fractions with pure Bcl-x L were combined, concentrated, and desalted into 50 mM Tris-HCl, pH 8.0. Bcl-x L was biotinylated in vitro using BirA biotin ligase (0.1 mg/mL in 50 mM Tris-HCl, pH 8.3, 10 mM ATP, 10 mM Mg(OAc) 2 , 50 μM biotin) at 30 °C for two hours. The protein was buffer exchanged into 1X PBS, frozen in liquid nitrogen, and stored at -80 °C. Peptide Synthesis. Peptides E1 (NH 2 -MIETITIYNYKKAADHFSMSMGSK-NH 2 ), E2 (NH 2 - MIETITIYKYKKAADHFSMSMGSK-NH 2 ), D1 (NH 2 -MIAISTIYNYKKAADHYAMTKGSK- NH 2 ), and D79 (NH 2 - MIDTNVILNYKKAADHFSITMGSK-NH 2 ) were synthesized by solid phase Fmoc synthesis, using a Biotage Alstra Microwave Synthesizer. 30 The peptides were synthesized on Rink amide MBHA resin using five-fold molar excess of each amino acid and HATU. After the coupling of the first amino acid, (Fmoc-Lys(Mtt)-OH), the primary amine in the side-chain of the lysine for each peptide was deprotected using a solution of 1% (v/v) 211 trifluoroacetic acid (TFA) in Dichloromethane (DCM). Biotin was then coupled to the side-chain primary amine before the synthesis was resumed, resulting in biotin-labeled peptides. Peptides were cleaved from the resin and deprotection with a solution of 95% (v/v) TFA, 2.5% 1,2- ethanedithiol (EDT), 1.5% (v/v) deionized water (DI), and 1% (v/v) triisopropylsilane (TIS) for 2 hours at room temperature. 100 The resin was filtered out, and the peptide was precipitated using 4-fold (v/v) excess ether. The peptides were dried, resuspended in DMSO, and HPLC purified using a C 18 reverse phase column and a gradient of 10-90% acetonitrile/0.1% TFA in water. Fractions were collected and tested for the correct molecular weight using MALDI-TOF mass spectrometry. The correct fractions were lyophilized, dissolved in DMSO, and flash frozen at -80 °C. Radiolabeled Off-Rate Assay: The DNA sequences coding for the peptides were ordered from Integrated DNA Technologies (IDT). Each DNA construct contained a T7 RNA Polymerase promoter, and a 5’ deletion mutant of the Tobacco Mosaic Virus (ΔTMV). 95 The C- terminal portion of the peptides were elongated with a flexible serine-glycine linker (six amino acids long) and an HA tag. After gel purification using urea-PAGE, the DNA sequences were PCR amplified using Taq polymerase and in vitro transcribed into mRNA using T7 RNA polymerase. 95 After transcription, the mRNA was urea-PAGE purified and resuspended in deionized water to a final concentration of 30 µM. The samples were in vitro translated at 30 C for 1 hour in the translation solution—150 mM KOAc, 750 µM MgCl 2 , 2 µM mRNA, 1X translation mix (20 mM Hepes-KOH pH 7.6, 100 mM creatine phosphate, 2 mM DTT, and 312.5 µM of each amino acid excluding methionine), 35 S- labeled methionine (Perkin Elmer; 20 µCi for each 25 µL of translation), and 60% (v/v) rabbit 212 reticulocyte lysate (Green Hectares; prepared according to the method of Jackson and Hunt) 99 . Radiolabeled peptides were purified using magnetic HA beads (Life Technologies) and eluted with 100 µL, 50 mM NaOH, then immediately neutralized with 20 µL of 1 M Tris-HCl, pH 8.0. The radiolabeled peptides were allowed to bind to 30 pmol immobilized Bcl-x L for 1 hour in sample buffer (1X PBS, 1% (w/v) BSA, 0.1% (v/v) Tween 20, 10 µM biotin). The beads were magnetically separated, and washed 5X with sample buffer. The beads were resuspended in 1 mL of sample buffer containing 3 μM non-biotinylated Bcl-x L (~100X molar excess relative to immobilized biotinylated Bcl-x L ). At various time points, 100 µL of slurry was removed and the beads were magnetically separated and washed. The percent remaining at each time point was determined by dividing the counts per minute (cpm) on beads by total cpm (beads + washes). The peptide off-rate was determined by an exponential fit of the Percent counts on beads vs. Time (s). Enzymatic K d Calculation Assay: The K d values of the peptides were determined using a protocol modified from Friguet et al. 112 A set of serially diluted Bcl-x L standards, at 2X the desired concentration, were made in sample buffer. For each peptide ligand, a set of dilutions at 2X the desired concentrations were also prepared. The Bcl-x L samples were either mixed 1:1 with sample buffer (standards) or ligands (samples), and allowed to incubate at room temperature for 6 days. To analyze the samples, the ELISA plates were incubated overnight at 4 C with 1.5 nmol of streptavidin in 1X PBS. Plates were washed 3X with wash buffer (1X PBS + 0.1% (v/v) Tween- 20) and blocked with 1X PBS + 5% (w/v) Bsa for two hours. 100 µL of a 30 nM solution of the D1 peptide (capture ligand) was added to wells and incubated for 1 hour. After the capture ligand 213 incubation, 100 µL of sample or standards were added in each well, and incubated for 1 hour at room temperature. Plates were washed, incubated with HRP-conjugated anti-HIS tag antibody in sample buffer for 1 hour, washed, and incubated with TMB substrate (Thermo Scientific). Reactions were stopped after approximately 10 minutes with 2 M sulfuric acid, and the absorbance at 450 nm was measured via a plate reader (Molecular Devices). The OD450 for the standards and their concentration values were fit to a four parameter logistic curve (standard curve). The concentration of the free Bcl-x L in solution (responsible for the signal) for each sample was calculated using the standard curve, and converted into percent of Bcl-x L bound by peptide in solution. For each peptide, the values for all the tested concentration of Bcl-x L and peptide in solution were fit simultaneously to the monovalent equilibrium model to obtain the K d . [𝐶 ] 𝐸𝑄 = [𝑇 ] 0 + [𝐿 ] 0 + 𝐾 𝐷 − √([𝑇 ] 0 + [𝐿 ] 0 + 𝐾 𝐷 ) 2 − 4[𝑇 ] 0 [𝐿 ] 0 2 Preparing the Pools: The DNA for the final enriched pools from the extension and the doped selection against Bcl-x L was obtained from Dr. Takahashi (Takahashi and Roberts, manuscript in preparation). The DNA pools were PCR amplified using Taq polymerase and in vitro transcribed into mRNA using T7 RNA polymerase. 95 After transcription, the mRNA was urea-PAGE purified and resuspended in deionized water to a final concentration of 30 µM. The mRNA was then ligated to fluorescein-F30P (phosphate–dA 21 –[dT-fluor]–[C9] 3 –dAdCdCP; where [dT-fluor] is fluorescein dT (Glen Research), [C9] is spacer 9 (Glen Research), and P is puromycin (Glen Research); synthesized at the Keck Oligo Facility at Yale) using T4 DNA ligase. 151 The ligation was performed using a splint complementary to the 3’ end of the RNA and the 5’ end of the DNA-linker. The ligated mRNA was urea-PAGE purified and resuspended in deionized water to 214 final concentration of 30 µM. The samples were in vitro translated in the translation solution— 150 mM KOAc, 750 µM MgCl 2 , 2 µM mRNA, in 1X translation mix (20 mM Hepes-KOH pH 7.6, 100 mM creatine phosphate, 2 mM DTT, and 312.5 µM of each amino acid) and 60% (v/v) rabbit reticulocyte lysate (Green Hectares; prepared according to the method of Jackson and Hunt 99 ). To prepare radiolabeled peptides or proteins, non-labeled methionine was substituted with 35 S-labeled methionine (Perkin Elmer; 20 µCi for each 25 µL of translation). The translation reactions were incubated at 30 C for one hour. To form mRNA-protein fusions, KCl and MgCl 2 were added to the reaction to final concentrations of 250 mM and 30 mM respectively after translation, and the samples were frozen at -20 C. To purify the fusion molecules, 100 µL of dT cellulose (25% (v/v) slurry, GE Healthcare) in isolation buffer (100 mM Tris-HCl pH 8.0, 1 M NaCl, 0.2% (v/v) Triton X-100) was added and incubated for 1 hour. The beads were washed five times with 700 µL of isolation buffer, and the fusions were eluted with 3X 80 µL of 65 ºC water and desalted through Centrisep columns (Princeton Separations). The desalted fusions were adjusted to 1X RT buffer (50 mM Tris-HCl pH 8.3, 75 mM KCl, 3 mM MgCl 2 , 2.4 mM 3’ primer, 200 mM each dNTP,) and the sample was heated to 65 °C for 5 minutes and cooled on ice to anneal the 3’ primer. After cooling, 33.3 µL of Superscript II enzyme was added and the reaction incubated at 42 °C for one hour. Superscript II was inactivated by heating to 65 °C for 5 minutes, after which the samples were cooled on ice, and used within the same day. On- and off-rate experiments: To obtain high-throughput sequencing kinetic (HTSK) on- rates, mRNA-peptide fusions of each pool from a 50 µL translation reaction (radiolabeled and non-labeled fusions separately) were first mixed with 7.5 pmols of Bcl-x L immobilized on 215 magnetic beads, and adjusted to 1 mL in 1X Selection buffer (1X PBS, 0.1% (w/v) BSA, 0.1% (v/v) Tween20, 100 µg/mL yeast tRNA, 0.05% (w/v) sodium azide, 10 µM biotin). At each time point, 100 µL of the solution was removed. The non-radiolabeled samples were magnetically separated and washed, PCR amplified with the appropriate primers, and sent for next-generation sequencing. The radiolabeled samples were washed 3X, and the the beads were counted via a scintillation counter. To obtain the HTSK off-rates, after the kinetic on-rate experiment, the remaining beads were washed 5X with selection buffer. The beads were then resuspended in 800 µL of selection buffer without biotin and supplemented with 2 µM Bcl-x L in solution. The excess Bcl-x L in solution prevents binding of dissociated ligands back to the beads. At specific time points, 100 µL of the solution was removed. The non-radiolabeled samples were washed, PCR amplified, and sent for next-generation sequencing. The radiolabeled samples were washed and counted via a scintillation counter. Next Generation DNA Sequencing Analysis: The mRNA-peptide fusions from all of the time points and pools were PCR amplified using unique identifying barcodes, combined into a single sample and sent for high throughput DNA sequencing using a HiSeq 2500 machine at the USC genome core. The file containing the results from the DNA sequencing run (FASTQ format) was first stripped of all content except for the DNA sequences using python code developed in house. Then the file was split into separate files for each on- and off-rate time point based on the DNA bar code. Each DNA sequence in each file was then translated (only the region after the start codon until the 3’ primer, using biopython and in house developed code) and the frequency of each translated sequence in the pool was calculated. Then, the fractional composition (frequency 216 of the sequence divided by the total sequences in the pool) for each sequence was calculated. A separate file was created per selection to track the frequency composition for each sequence throughout the various time points. An example of this data can be seen in Figure 1a-b in the left panels. Obtaining the on- and off-rates by HTSK: To obtain the on-rate for each sequence, the fractional composition for each sequence was multiplied by the radiolabeled counts for that pool’s time point. This results in the radiolabeled counts per sequence as a function of time. These values (representing [C]), the concentration of immobilized Bcl-x L on magnetic beads, and time in seconds were fit to the on-rate equation in supplementary figure 1 to obtain [L] 0 (asymptotic maximum) and k on for each sequence. The fitting was done using the fminsearch function in MATLAB to minimize the error (Least Absolute Deviation method) between the real data and the model by changing [L] 0 and k on . To obtain the off-rate, the same procedure was performed with the off-rate portion of the fraction composition data for each sequence. MATLAB was used to fit the product of the fractional composition and the radiolabeled pool counts at each time point, to the off-rate formula in Supplementary Figure 1. To obtain the on- and off-rates for each sequence without using the radiolabeled data, it is possible to use another method to quantitating the amount of pool bound to the beads at each time point. We quantitated the amount of DNA bound to the beads by measuring the intensity of the DNA bands in the agarose gels using ImageJ’s intensity measurement tool, and using the DNA ladder (NEB 100bp ladder) as our standards. The amount of DNA bound to the beads can be exchanged for the radiolabeled counts on beads, in order to obtain the on- and off-rate values for sequences by HTSK. 217 Number of sequences for which HTSK analysis can give produce reliable results, depends on the initial diversity and the convergence of the library. We could only do the analysis for ligands with a statistically significant representation in a pool. For a pool that had converged to a large degree (extension pool), where the top 50 sequences accounted for ~78% of the pool, we were able to obtain HTSK results for approximately 2,000 sequences. However for a less converged pool (Doped) where the top 50 sequences accounted for ~3% of the pool, we were able to obtain HTSK results for ~20,000 sequences. The HTSK analysis cannot however provide kinetics constants for any sequences if the diversity of the pool was too high (where the highest represented sequence in the library accounted for less than 1 PPM of the library). 218 8.3 Results and Discussion In this work, we combined high throughput DNA sequencing with mRNA display to obtain kinetic on- and off-rates, and thus K d values, for tens of thousands of ligands simultaneously. To do this, we chose two enriched pools from our selection against B-cell lymphoma extra-large protein (Bcl-x L ) (Takahashi and Roberts, manuscript in preparation). These pools are the final enriched pools from an extension selection (selection that resulted in 21 amino acid long peptide ligands against Bcl-x L ) and a doped selection (the top sequence from extension was used to create a biased library, to further optimize binding). The mRNA of both pools were ligated to a 3’ DNA linker attached to puromycin, in vitro translated, purified and reverse transcribed. 143 A small fraction of each pool was also translated using radiolabeled methionine to track pool binding. To obtain high-throughput sequencing kinetic (HTSK) on-rates, a library of mRNA-peptide fusions were first mixed with Bcl-x L immobilized on magnetic beads. A portion of the beads were removed at various time points, washed, PCR amplified, and sent for next-generation sequencing (Fig. 1a, left panel). Sequencing results yield all the ligands bound to the beads at that point, allowing the calculation of each ligand’s frequency and thus fractional composition. Separately, using the radiolabeled samples, we measured the amount of peptide bound to the beads at each time point (Figure 1a, middle panel). The amount of radiolabeled binding at each time point represents the sum of all the peptides bound to the beads at that point. To obtain the kinetic on-rates for each ligand, we simply multiplied each ligand's fractional composition by the total radiolabeled binding. This results in a measure of binding for each sequence as a function of time (Fig. 1a right panel). Using this analysis, and knowing the concentration of immobilized 219 Bcl-x L , we obtained the kinetic on-rate for each sequence by fitting the binding data to a simple kinetic on-rate equation (Supplementary Fig. 1). The contribution of the dissociation-rate to the binding equation has been removed because in the small time scale of this experiment (~45 minutes) and given the slow off-rate of the sequences tested (2 x 10-6 s-1 on average), the contribution of the dissociation rate is minimal. This allows for independent calculation of on- and off-rates. To obtain the HTSK off-rates, we followed a similar approach. After the kinetic on-rate experiment, the remaining beads were washed and excess Bcl-x L was added in solution to prevent rebinding of dissociated ligands to the beads. Periodically, a fraction of beads were removed, washed, PCR amplified, and sent for next-generation sequencing (Fig. 1b left panel). By multiplying each sequence’s fractional composition by the total radiolabeled peptides still bound at each time point (Fig. 1b, middle panel), we were able to obtain the amount of each peptide still bound as a function of time. A simple exponential fit was then used to calculate the kinetic off-rate (Fig. 1b, right panel). Calculating the amount of peptide bound to the beads Figure 2a shows the Kd obtained for the 50 highest frequency ligands in each tested pool. As expected, the ligands in the doped pool show a higher affinity on average than the ligands in the extension pool. It is also clear that frequency rank poorly correlates to sequence affinity. To show the reproducibility of the obtained kinetic constants, we compared the obtained values for the 40 ligands that appeared in both the extension and doped pools (Fig. 2b). The results show that the HTSK values are remarkably reproducible and highly precise. In order to check the validity of the obtained results, we tested the off-rate of several ligands using in vitro translated radiolabeled peptides. The peptide ligands were made using a C-terminal HA tag, and affinity purified. The off-rate of the radiolabeled peptides was then determined using similar methods as 220 the radiolabeled pool off-rate. Figure 2c shows the HTSK vs. radiolabeled peptide off-rates. The HTSK off-rates correlate very well to the radiolabeled peptide off-rates, however, there is a consistent bias between the two methods. The measured bias is ~7-fold for the fastest off-rate clone, and less than 2-fold for the slowest off-rate clones. This bias is relatively small in comparison to biases measured between other established methods for affinity measurement, which are frequently as high as 60-fold different. 188, 189, 190 One contributing factor to this difference could be the context of binding. The HTSK results are obtained for mRNA-DNA- peptide fusion molecules whereas the radiolabeled koff values are for the peptide with a short C- terminal HA tag. Using HTSK, we identified peptide D79 (frequency rank of 79 in the doped selection pool) with a koff value of 5.9 x 10 -7 , over three times slower than the previously identified slowest off- rate peptide ligand (D1) or the biotin-streptavidin interaction (Fig. 2d). Another interesting result was the identification of peptide E1452 (frequency rank of 1452 from the extension selection pool) with the koff value of 8.5 x 10 -7 , over two fold slower than D1 (Supplementary Fig. 2). These results point to the ability of the extension selection to generate ultra-high affinity ligands without the need for a biased (doped) selection to improve affinity further. Our shortcoming was not the generation of ultrahigh affinity ligands, but the inability to identify the highest affinity ligands from the pool of thousands of peptides. Indeed, in this modest chain length (21 amino acids long), using HTSK, we have identified thousands of sequences with 10 pM Kd or better (Supplementary Dataset). 221 8.4 Conclusions By combining mRNA display with next-generation sequencing, we were able to find the kinetic on- and off-rate for thousands of ligands without the need to individually synthesize or purify them. Our approach is not limited to mRNA display. In fact, it is directly transferable to aptamer selection techniques or any monomeric genotype-phenotype linked display system such as ribosome display. The off-rate portion of the analysis can even be used for non-monomeric displays such as yeast or phage display (in these methods the on-rate calculations would be difficult due to the multivalancy of the binding entity and the complications of avidity). This technique is most suitable for high affinity reagents (K d ≤ 10 nM) since the slower off-rates allow for more precise measurements. We have shown our HTSK method to be repeatable, precise and accurate as compared to other methods of affinity measurement, and have identified the highest affinity peptide-protein interaction yet discovered. 222 8.5 Figures Figure 1: Obtaining kinetic rates for ligands using HTSK. a) Obtaining the kinetic on-rate. The pool of mRNA-peptide fusion molecules was incubated with Bcl-x L (immobilized on beads). At specific time points, a fraction of beads were collected and washed. The molecules bound to the beads were sequenced via next-generation sequencing. The fraction of each ligand at each time point was calculated from the sequencing data and normalized with respect to the final data point (left). Separately, the pool was in vitro translated using radiolabeled methionine, and its binding was determined at each time point (middle). By multiplying each ligand’s composition fraction by the radiolabeled binding at each data point, we obtained the ligand’s contribution to the radiolabeled binding, and subsequently, the on-rate (right). b) Obtaining the kinetic off-rate. At the end of the on-rate experiment, the remaining beads were washed and placed in a solution containing 100X excess Bcl-x L in solution, preventing ligands from re-binding to the beads. At specific time points, a fraction of beads were collected and washed. The molecules still bound to the beads were analyzed by next-generation sequencing. The fraction of each ligand at each time point was calculated from the sequencing data and normalized with respect to the first data point (left). The counts remaining on the beads at each time point were measured using the radiolabeled sample (middle). By multiplying each ligand’s composition fraction by the radiolabeled binding at each data point, we obtained the ligand’s contribution to the radiolabeled binding and the off-rate (right). 223 Figure 2: The HTSK results are reproducible and accurate. a) The obtained K d for the top 50 clones in the extension and doped pools. While the extension pool on average (dashed red line) is comprised of lower affinity binder than the doped pool (dashed blue lines), some sequences in the extension pool show higher affinity than the doped pool average. b) The obtained HTSK values are reproducible. 40 sequences appeared in both the extension and the doped pools. Comparing the kinetic constants for these sequences shows that the results are reproducible. c) The k off value obtained by HTSK correlate well to the values obtained using radiolabeled peptides. There is a consistent bias in the measured off-rate values for the two methods of measurements. d) The radiolabeled peptide off-rate for the previously identified sequences E1 and D1, and the HTSK identified sequence D79. The off-rate for sequence D79 is over 3 times slower than the off-rate of D1, the previously identified highest affinity binder. The slowest reported value for the off-rate of biotin and streptavidin in the literature (2.4 x 10 -6 ) 130 is shown as a reference. 224 Sequence Peptide k o f f (s -1 ) HTSK k o f f (s -1 ) Enzymatic K d (pM) HTSK K d (pM) E1 MIETITIYNYKKAADHFSMSM 7.4 x 10 -6 2.5 x 10 -6 39 ± 6 23 ± 2 D1 --AIS-----------YA-TK 2.0 x 10 -6 1.0 x 10 -6 9 ± 2 15 D79 --D-NV-L----------IT- 5.9 x 10 -7 3.3 x 10 -7 2.4 Table 1: Validity of the HTSK results. The kinetic off-rates and the dissociation constant for three selected clones obtained by HTSK vs. radiolabeled peptides (k off ) and ELISA (K d ). 225 8.6 Supplementary Figures Supplementary Figure 1: On- and off-rate equations for the HTSK experiments. Due to the relatively short time period for the on-rate segment of the experiment (~45 minutes) and the very slow off-rate for the clones (~2 x 10 -6 on average) the contribution from the off-rate can be ignored during the binding phase. This allowed the transient complex concentration equation under excess target concentration conditions to reduce to the kinetic on-rate expression above. Supplementary Figure 2: Ligand E1452 (green circles, frequency rank of 1452 in the extension selection pool) was identified by HTSK and tested as a radiolabeled peptide. Its off- rate is slower than D1, the previously identified highest affinity peptide from the doped selection. 226 Supplementary Figure 3: Histogram of the obtained K d values for the extension and doped pools. 227 Chapter 9: Conclusions mRNA display is an immensely powerful approach to ligand selection and characterization. Its complete in vitro design allows a wide array of post-translational modifications (such as cyclization and protease digest) and co-translational incorporation of unnatural amino acids (discussed in detail in Chapters 2-4). Using mRNA display we have developed ligands against the hepatitis C virus nucleocapsid protein (Chapter 2), Gαi1 (Chapter 3), Human Double Minute 2 (HDM2, Chapter 4), and B-Cell Lymphoma extra Large (Bcl-x L , Chapter 5). Throughout this work we have dispelled the belief that peptides are poor therapeutic and diagnostic reagents by showing biologically stable peptides (Chapters 3-4) which are cell-permeable (Chapter 4) and capable of extremely high affinity and selectivity (Chapter 5). Our hepatitis C nucleocapsid protein ligands are capable of inhibiting virus particle production in human liver cells without detectable off-target effect and our HDM2 ligand kills colorectal carcinoma cells with overexpressed HDM2. We have even demonstrated sub-picomolar sensitivity for a diagnostic Bcl-x L assay which is 15-fold more sensitive than the current ELISA kits on the market, as well as peptides with off-rates slower than streptavidin-biotin interaction. Our goal has always been to streamline the mRNA display process. There are three main steps in the work flow: first, the synthesis, preparation, and validation of the desired target. This step can be very variable in length, but takes approximately 1-3 months. Second is the selection itself. An advanced mRNA display user can finish a cycle in less than two days, which makes the 5-9 rounds of selection needed to obtain ligands take from 1-3 months as well. The final step is clone selection and characterization. Traditionally, this step was done by TOPO cloning ~40-200 sequences from the final mRNA display pool, obtaining the sequences via Sanger sequencing, 228 and synthesizing and purifying several ligands (one such method described in chapter 6). Once the optimal peptide was selected, we would obtain its affinity constant by SPR or equilibrium exclusion assays (described in chapter 7). This process was consistently protracted and labor intensive. Depending on the ease of expression and purification of the clones and the number of clones to be tested, it could take up to a year to characterize the selection. More recently, with the incorporation of next generation sequencing, we have been able to use high-throughput DNA sequencing and obtain thousands of potential ligands for our target. This has made ligand characterization more challenging, since we have realized that sequence affinity and frequency in the pool are poorly correlated, and now have thousands of potential ligands to test. Developing the high throughput sequencing kinetics methods described in chapter 8 is a serious breakthrough in addressing the characterization challenges of screening and selection methods. By combining mRNA display with next-generation sequencing, we were able to find the kinetic on- and off-rates for thousands of ligands simultaneously without individually synthesizing or purifying them. This data was used to identify the highest affinity peptide interaction ever recorded in literature. mRNA display technique has advanced greatly in recent years with the inclusion of unnatural amino acids, 4, 5, 6 covalent modifications, 7 post translational cyclization, 8 selection for protease resistance, and as an analytical tool to obtain protein structural information. 19, 35 The potential of combining mRNA display and high throughput DNA sequencing to develop and characterize ligands with desired properties has not been fully realized; future tools and techniques will improve upon this method, facilitating faster identification of desired properties and streamlining the selection process. 229 References: 1. Roberts RW, Szostak JW. RNA-peptide fusions for the in vitro selection of peptides and proteins. Proc Natl Acad Sci USA 94, 12297-12302 (1997). 2. Liu R, Barrick J, Szostak JW, Roberts RW. Optimized Synthesis of RNA-Protein Fusions for In Vitro Protein Selection. Methods in Enzymology 317, 268-293 (2000). 3. Takahashi TT, Austin RJ, Roberts RW. mRNA display: ligand discovery, interaction analysis and beyond. Trends in Biochemical Sciences 28, 159-165 (2003). 4. Li S, Millward S, Roberts RW. In vitro selection of mRNA display libraries containing an unnatural amino acid. J Am Chem Soc 124, 9972-9973 (2002). 5. Frankel A, Millward SW, Roberts RW. Encodamers: Unnatural Peptide Oligomers Encoded in RNA. Chemistry & Biology 10, 1043-1050 (2003). 6. Frankel A, Roberts RW. In Vitro Selection for Sense Codon Suppression. RNA 9, 780- 786 (2003). 7. Li S, Roberts RW. A Novel Strategy for In Vitro Selection of Peptide-Drug Conjugates. Chemistry & Biology 10, 233-239 (2003). 8. Millward SW, Fiacco S, Austin RJ, Roberts RW. Design of Cyclic Peptides that Bind Protein Surfaces with Antibody-Like Affinity. ACS Chem Biol 2, 625-634 (2007). 9. Howell SM, et al. Serum stable natural peptides designed by mRNA display. Sci Rep 4, 6008 (2014). 10. Barrick JE, Takahashi TT, Balakin A, Roberts RW. Selection of RNA-Binding Peptides Using mRNA-Peptide Fusions. Methods 23, 287-293 (2001). 11. Ja WW, Roberts RW. In vitro Selection of State-Specific Modulators of G Protein Signaling Using mRNA Display. Biochemistry 43, 9265-9275 (2004). 12. Ja WW, Olsen BN, Roberts RW. Epitope Mapping Using mRNA Display and a Unidirectional Nested Deletion Library. Protein Engineering, Design and Selection 18, 309-319 (2005). 13. Ja WW, Wiser O, Austin RJ, Jan LY, Roberts RW. Turning G proteins on and off using peptide ligands. ACS Chemical Biology 1, 570-574 (2006). 14. Austin RJ, Ja WW, Roberts RW. Evolution of class-specific peptides targeting a hot spot of the Gαs subunit. J Mol Biol 377, 1406-1418 (2008). 230 15. Tsang J, Joyce GF. In Vitro Evolution of Randomized Ribozymes. Methods in Enzymology 267, 410-426 (1996). 16. Olson CA, Liao H-I, Sun R, Roberts RW. mRNA display selection of a high-affinity, modification-specific phospho-IkappaBalpha-binding fibronectin. ACS Chem Biol 3, 480- 485 (2008). 17. Olson CA, Liao HI, Sun R, Roberts RW. mRNA display selection of a high-affinity, modification-specific phospho-IkappaBalpha-binding fibronectin. ACS Chem Biol 3, 480- 485 (2008). 18. Gross GG, et al. Recombinant probes for visualizing endogenous synaptic proteins in living neurons. Neuron 78, 971-985 (2013). 19. Olson CA, Wu NC, Sun R. A comprehensive biophysical description of pairwise epistasis throughout an entire protein domain. Current biology : CB 24, 2643-2651 (2014). 20. From World Health Report 1996 (World Health Organisation, Geneva). J Commun Dis 28, 215-219 (1996). 21. Dubuisson J. Hepatitis C virus proteins. World J Gastroenterol 13, 2406-2415 (2007). 22. Poynard T, Yuen MF, Ratziu V, Lai CL. Viral hepatitis C. Lancet 362, 2095-2100 (2003). 23. Tanaka Y, et al. Selective binding of hepatitis C virus core protein to synthetic oligonucleotides corresponding to the 5' untranslated region of the viral genome. Virology 270, 229-236 (2000). 24. Kuiken C, Yusim K, Boykin L, Richardson R. The Los Alamos hepatitis C sequence database. Bioinformatics 21, 379-384 (2005). 25. Kuiken C, Hraber P, Thurmond J, Yusim K. The hepatitis C sequence database in Los Alamos. Nucleic Acids Res 36, D512-516 (2008). 26. Arumugaswami V, et al. High-resolution functional profiling of hepatitis C virus genome. PLoS Pathog 4, e1000182 (2008). 27. Jolivet-Reynaud C, et al. HCV core immunodominant region analysis using mouse monoclonal antibodies and human sera: characterization of major epitopes useful for antigen detection. J Med Virol 56, 300-309 (1998). 28. Menez R, et al. Crystal structure of a hydrophobic immunodominant antigenic site on hepatitis C virus core protein complexed to monoclonal antibody 19D9D6. J Immunol 170, 1917-1924 (2003). 231 29. Ladaviere L, Deleage G, Montserret R, Dalbon P, Jolivet M, Penin F. Solution Structure of The Hepatitis C Virus N-Terminal Capsid Protein 2-45 [C-HCV(2-45)]. PDB ID: 1CWX, (1999). 30. Fields GB, Noble RL. Solid phase peptide synthesis utilizing 9- fluorenylmethoxycarbonyl amino acids. Int J Pept Protein Res 35, 161-214 (1990). 31. Bartel DP, Szostak JW. Isolation of new ribozymes from a large pool of random sequences [see comment]. Science 261, 1411-1418 (1993). 32. Millward SW, Fiacco S, Austin RJ, Roberts RW. Design of cyclic peptides that bind protein surfaces with antibody-like affinity. ACS Chem Biol 2, 625-634 (2007). 33. Studier FW. Protein production by auto-induction in high density shaking cultures. Protein Expr Purif 41, 207-234 (2005). 34. Beckett D, Kovaleva E, Schatz PJ. A minimal peptide substrate in biotin holoenzyme synthetase-catalyzed biotinylation. Protein Sci 8, 921-929 (1999). 35. Millward SW, Takahashi TT, Roberts RW. A general route for post-translational cyclization of mRNA display libraries. Journal of the American Chemical Society 127, 14142-14143 (2005). 36. Jones S, Thornton JM. Analysis of protein-protein interaction sites using surface patches. Journal of molecular biology 272, 121-132 (1997). 37. Kyte J, Doolittle RF. A simple method for displaying the hydropathic character of a protein. Journal of molecular biology 157, 105-132 (1982). 38. Pace CN, Scholtz JM. A helix propensity scale based on experimental studies of peptides and proteins. Biophys J 75, 422-427 (1998). 39. Young L, Jernigan RL, Covell DG. A role for surface hydrophobicity in protein-protein recognition. Protein Sci 3, 717-729 (1994). 40. Azzarito V, Long K, Murphy NS, Wilson AJ. Inhibition of alpha-helix-mediated protein- protein interactions using designed molecules. Nat Chem 5, 161-173 (2013). 41. Wada A. The alpha-helix as an electric macro-dipole. Adv Biophys, 1-63 (1976). 42. Hol WG, van Duijnen PT, Berendsen HJ. The alpha-helix dipole and the properties of proteins. Nature 273, 443-446 (1978). 232 43. Marqusee S, Robbins VH, Baldwin RL. Unusually stable helix formation in short alanine-based peptides. Proceedings of the National Academy of Sciences of the United States of America 86, 5286-5290 (1989). 44. Rizzo MA, Springer GH, Granada B, Piston DW. An improved cyan fluorescent protein variant useful for FRET. Nat Biotechnol 22, 445-449 (2004). 45. McGregor DP. Discovering and improving novel peptide therapeutics. Curr Opin Pharmacol 8, 616-619 (2008). 46. Donzel B, Goodman M, Rivier J, Ling N, Vale W. Synthesis and conformations of hypothalamic hormone releasing factors: two QRF-analogues containing backbone N- methyl groups. Nature 256, 750-751 (1975). 47. Dado GP, Gellman SH. Intramolecular Hydrogen-Bonding in Derivatives of Beta- Alanine and Gamma-Amino Butyric-Acid - Model Studies for the Folding of Unnatural Polypeptide Backbones. Journal of the American Chemical Society 116, 1054-1062 (1994). 48. Seebach D, Matthews JL. beta-peptides: a surprise at every turn. Chemical Communications, 2015-2022 (1997). 49. Owens RA, Gesellchen PD, Houchins BJ, DiMarchi RD. The rapid identification of HIV protease inhibitors through the synthesis and screening of defined peptide mixtures. Biochem Biophys Res Commun 181, 402-408 (1991). 50. Simon RJ, et al. Peptoids: A modular approach to drug discovery. Proc Natl Acad Sci USA 89, 9367-9371 (1992). 51. Blackwell HE, et al. Ring-closing metathesis of olefinic peptides: Design, synthesis, and structural characterization of macrocyclic helical peptides. J Org Chem 66, 5291-5301 (2001). 52. Walensky LD, et al. Activation of apoptosis in vivo by a hydrocarbon-stapled BH3 helix. Science 305, 1466-1470 (2004). 53. Millward SW, Takahashi TT, Roberts RW. A General Route for Post-Translational Cyclization of mRNA Display Libraries. J Am Chem Soc 127, 14142-14143 (2005). 54. Hedstrom L, Szilagyi L, Rutter WJ. Converting trypsin to chymotrypsin: the role of surface loops. Science 255, 1249-1253 (1992). 55. Ja WW, Roberts RW. In vitro selection of state-specific peptide modulators of G protein signaling using mRNA display. Biochemistry 43, 9265-9275 (2004). 233 56. Johnston CA, et al. Structure of Galpha(i1) bound to a GDP-selective peptide provides insight into guanine nucleotide exchange. Structure (Camb) 13, 1069-1080 (2005). 57. Folk JE, Cole PW. Chymotrypsin C. Ii. Enzymatic Specificity toward Several Polypeptides. The Journal of biological chemistry 240, 193-197 (1965). 58. Schellenberger V, Braune K, Hofmann HJ, Jakubke HD. The specificity of chymotrypsin. A statistical analysis of hydrolysis data. European journal of biochemistry / FEBS 199, 623-636 (1991). 59. Hedstrom L. Serine protease mechanism and specificity. Chemical reviews 102, 4501- 4524 (2002). 60. Fung BK, Nash CR. Characterization of transducin from bovine retinal rod outer segments. II. Evidence for distinct binding sites and conformational changes revealed by limited proteolysis with trypsin. The Journal of biological chemistry 258, 10503-10510 (1983). 61. Hurley JB, Simon MI, Teplow DB, Robishaw JD, Gilman AG. Homologies between signal transducing G proteins and ras gene products. Science 226, 860-862 (1984). 62. Marqusee S, Baldwin RL. Helix stabilization by Glu-...Lys+ salt bridges in short peptides of de novo design. Proceedings of the National Academy of Sciences of the United States of America 84, 8898-8902 (1987). 63. Compton LA, Johnson WC, Jr. Analysis of protein circular dichroism spectra for secondary structure using a simple matrix multiplication. Anal Biochem 155, 155-167 (1986). 64. Powell MF, et al. Peptide stability in drug development. II. Effect of single amino acid substitution and glycosylation on peptide reactivity in human serum. Pharmaceutical research 10, 1268-1273 (1993). 65. Wildes D, Wells JA. Sampling the N-terminal proteome of human blood. Proceedings of the National Academy of Sciences of the United States of America 107, 4561-4566 (2010). 66. el-Deiry WS, et al. WAF1, a potential mediator of p53 tumor suppression. Cell 75, 817- 825 (1993). 67. Xiong Y, Hannon GJ, Zhang H, Casso D, Kobayashi R, Beach D. p21 is a universal inhibitor of cyclin kinases. Nature 366, 701-704 (1993). 68. Dittmer D, et al. Gain of function mutations in p53. Nat Genet 4, 42-46 (1993). 234 69. Brown C, Lain S, Verma C, Fersht A, Lane D. Awakening guardian angels: drugging the p53 pathway. Nature reviews Cancer 9, 862-873 (2009). 70. Vousden KH, Lane DP. p53 in health and disease. Nat Rev Mol Cell Biol 8, 275-283 (2007). 71. Toledo F, Wahl GM. Regulating the p53 pathway: in vitro hypotheses, in vivo veritas. Nat Rev Cancer 6, 909-923 (2006). 72. Marine JC, Dyer MA, Jochemsen AG. MDMX: from bench to bedside. J Cell Sci 120, 371-378 (2007). 73. Vogelstein B, Lane D, Levine AJ. Surfing the p53 network. Nature 408, 307-310 (2000). 74. Vousden KH, Lu X. Live or let die: the cell's response to p53. Nat Rev Cancer 2, 594-604 (2002). 75. Xue W, et al. Senescence and tumour clearance is triggered by p53 restoration in murine liver carcinomas. Nature 445, 656-660 (2007). 76. Ventura A, et al. Restoration of p53 function leads to tumour regression in vivo. Nature 445, 661-665 (2007). 77. Martins CP, Brown-Swigart L, Evan GI. Modeling the therapeutic efficacy of p53 restoration in tumors. Cell 127, 1323-1334 (2006). 78. Pazgier M, et al. Structural basis for high-affinity peptide inhibition of p53 interactions with MDM2 and MDMX. Proceedings of the National Academy of Sciences of the United States of America 106, 4665-4670 (2009). 79. Momand J, Jung D, Wilczynski S, Niland J. The MDM2 gene amplification database. Nucleic Acids Res 26, 3453-3459 (1998). 80. Chang YS, et al. Stapled alpha-helical peptide drug development: a potent dual inhibitor of MDM2 and MDMX for p53-dependent cancer therapy. Proceedings of the National Academy of Sciences of the United States of America 110, E3445-3454 (2013). 81. Yin H, Hamilton AD. Strategies for targeting protein-protein interactions with synthetic agents. Angew Chem Int Ed Engl 44, 4130-4163 (2005). 82. Goodnow RA, Jr., Guba W, Haap W. Library design practices for success in lead generation with small molecule libraries. Comb Chem High Throughput Screen 6, 649- 660 (2003). 83. Hopkins AL, Groom CR. The druggable genome. Nat Rev Drug Discov 1, 727-730 (2002). 235 84. Russ AP, Lampel S. The druggable genome: an update. Drug Discov Today 10, 1607- 1610 (2005). 85. Vassilev L, et al. In vivo activation of the p53 pathway by small-molecule antagonists of MDM2. Science (New York, NY) 303, 844-848 (2004). 86. Vassilev LT, et al. In vivo activation of the p53 pathway by small-molecule antagonists of MDM2. Science 303, 844-848 (2004). 87. Shangary S, et al. Temporal activation of p53 by a specific MDM2 inhibitor is selectively toxic to tumors and leads to complete tumor growth inhibition. Proc Natl Acad Sci U S A 105, 3933-3938 (2008). 88. Ding K, et al. Structure-based design of potent non-peptide MDM2 inhibitors. J Am Chem Soc 127, 10130-10131 (2005). 89. Klein C, Vassilev LT. Targeting the p53-MDM2 interaction to treat cancer. Br J Cancer 91, 1415-1419 (2004). 90. Cheok CF, Verma CS, Baselga J, Lane DP. Translating p53 into the clinic. Nat Rev Clin Oncol 8, 25-37 (2011). 91. Murray JK, Gellman SH. Targeting protein-protein interactions: lessons from p53/MDM2. Biopolymers 88, 657-686 (2007). 92. Lien S, Lowman HB. Therapeutic peptides. Trends Biotechnol 21, 556-562 (2003). 93. Liu M, et al. D-peptide inhibitors of the p53-MDM2 interaction for targeted molecular therapy of malignant neoplasms. Proc Natl Acad Sci U S A 107, 14321-14326 (2010). 94. Fiacco SV, Roberts RW. N-Methyl scanning mutagenesis generates protease-resistant G protein ligands with improved affinity and selectivity. Chembiochem : a European journal of chemical biology 9, 2200-2203 (2008). 95. Liu R, Barrick JE, Szostak JW, Roberts RW. Optimized synthesis of RNA-protein fusions for in vitro protein selection. Methods Enzymol 318, 268-293 (2000). 96. Saks ME, et al. An engineered Tetrahymena tRNAGln for in vivo incorporation of unnatural amino acids into proteins by nonsense suppression. The Journal of biological chemistry 271, 23169-23175 (1996). 97. Shiheido H, Takashima H, Doi N, Yanagawa H. mRNA display selection of an optimized MDM2-binding peptide that potently inhibits MDM2-p53 interaction. PloS one 6, e17898 (2011). 236 98. Li C, et al. Systematic mutational analysis of peptide inhibition of the p53- MDM2/MDMX interactions. Journal of molecular biology 398, 200-213 (2010). 99. Jackson RJ, Hunt T. Preparation and use of nuclease-treated rabbit reticulocyte lysates for the translation of eukaryotic messenger RNA. Methods Enzymol 96, 50-74 (1983). 100. Huang H, Rabenstein DL. A cleavage cocktail for methionine-containing peptides. J Pept Res 53, 548-553 (1999). 101. Pazgier M, et al. Structural basis for high-affinity peptide inhibition of p53 interactions with MDM2 and MDMX. Proceedings of the National Academy of Sciences of the United States of America 106, 4665-4670 (2009). 102. Nagata T, et al. Structural basis for inhibition of the MDM2:p53 interaction by an optimized MDM2-binding peptide selected with mRNA display. PloS one 9, e109163 (2014). 103. Meloen RH, Puijk WC, Langeveld JP, Langedijk JP, Timmerman P. Design of synthetic peptides for diagnostics. Curr Protein Pept Sci 4, 253-260 (2003). 104. Erlanson DA, et al. In situ assembly of enzyme inhibitors using extended tethering. Nature biotechnology 21, 308-314 (2003). 105. Kuntz ID, Chen K, Sharp KA, Kollman PA. The maximal affinity of ligands. Proceedings of the National Academy of Sciences of the United States of America 96, 9997-10002 (1999). 106. Lancet D, Sadovsky E, Seidemann E. Probability model for molecular recognition in biological receptor repertoires: significance to the olfactory system. Proceedings of the National Academy of Sciences of the United States of America 90, 3715-3719 (1993). 107. Boder ET, Midelfort KS, Wittrup KD. Directed evolution of antibody fragments with monovalent femtomolar antigen-binding affinity. Proceedings of the National Academy of Sciences of the United States of America 97, 10701-10705 (2000). 108. Boder ET, Wittrup KD. Optimal screening of surface-displayed polypeptide libraries. Biotechnology progress 14, 55-62 (1998). 109. Hachmann J, Lebl M. Alternative to piperidine in Fmoc solid-phase synthesis. J Comb Chem 8, 149 (2006). 110. Austin RJ, Ja WW, Roberts RW. Evolution of class-specific peptides targeting a hot spot of the Galphas subunit. Journal of molecular biology 377, 1406-1418 (2008). 237 111. Matsumura N, et al. mRNA display selection of a high-affinity, Bcl-X(L)-specific binding peptide. FASEB journal : official publication of the Federation of American Societies for Experimental Biology 24, 2201-2210 (2010). 112. Friguet B, Chaffotte AF, Djavadi-Ohaniance L, Goldberg ME. Measurements of the true affinity constant in solution of antigen-antibody complexes by enzyme-linked immunosorbent assay. Journal of immunological methods 77, 305-319 (1985). 113. Bai J, Sui J, Demirjian A, Vollmer CM, Jr., Marasco W, Callery MP. Predominant Bcl- XL knockdown disables antiapoptotic mechanisms: tumor necrosis factor-related apoptosis-inducing ligand-based triple chemotherapy overcomes chemoresistance in pancreatic cancer cells in vitro. Cancer research 65, 2344-2352 (2005). 114. Oltersdorf T, et al. An inhibitor of Bcl-2 family proteins induces regression of solid tumours. Nature 435, 677-681 (2005). 115. Ja WW, Wiser O, Austin RJ, Jan LY, Roberts RW. Turning G proteins on and off using peptide ligands. ACS chemical biology 1, 570-574 (2006). 116. Boersma MD, Sadowsky JD, Tomita YA, Gellman SH. Hydrophile scanning as a complement to alanine scanning for exploring and manipulating protein-protein recognition: application to the Bim BH3 domain. Protein science : a publication of the Protein Society 17, 1232-1240 (2008). 117. Chen L, et al. Differential targeting of prosurvival Bcl-2 proteins by their BH3-only ligands allows complementary apoptotic function. Molecular cell 17, 393-403 (2005). 118. Jalali-Yazdi F, Corbin JM, Takahashi TT, Roberts RW. Robust, quantitative analysis of proteins using peptide immunoreagents, in vitro translation, and an ultrasensitive acoustic resonant sensor. Anal Chem 86, 4715-4722 (2014). 119. Ja WW, Adhikari A, Austin RJ, Sprang SR, Roberts RW. A peptide core motif for binding to heterotrimeric G protein alpha subunits. The Journal of biological chemistry 280, 32057-32060 (2005). 120. Olson CA, et al. Single-round, multiplexed antibody mimetic design through mRNA display. Angew Chem Int Ed Engl 51, 12449-12453 (2012). 121. Carter PJ. Potent antibody therapeutics by design. Nature reviews Immunology 6, 343- 357 (2006). 122. Chin JW, Schepartz A. Design and Evolution of a Miniature Bcl-2 Binding Protein. Angewandte Chemie 40, 3806-3809 (2001). 123. Procko E, et al. A computationally designed inhibitor of an Epstein-Barr viral Bcl-2 protein induces apoptosis in infected cells. Cell 157, 1644-1656 (2014). 238 124. Bertman MO, Jungck JR. Group graph of the genetic code. The Journal of heredity 70, 379-384 (1979). 125. Cantor CR, Schimmel PR. The behavior of biological macromolecules. W. H. Freeman (1980). 126. Clackson T, Wells JA. A hot spot of binding energy in a hormone-receptor interface. Science 267, 383-386 (1995). 127. Eaton BE, Gold L, Zichi DA. Let's get specific: the relationship between specificity and affinity. Chemistry & biology 2, 633-638 (1995). 128. Carothers JM, Oestreich SC, Szostak JW. Aptamers selected for higher-affinity binding are not more specific for the target ligand. Journal of the American Chemical Society 128, 7929-7937 (2006). 129. Greenspan NS. Cohen's Conjecture, Howard's Hypothesis, and Ptashne's Ptruth: an exploration of the relationship between affinity and specificity. Trends in immunology 31, 138-143 (2010). 130. Piran U, Riordan WJ. Dissociation rate constant of the biotin-streptavidin complex. Journal of immunological methods 133, 141-143 (1990). 131. Green NM. Avidin and streptavidin. Methods in enzymology 184, 51-67 (1990). 132. Chilkoti A, Stayton PS. Molecular-Origins of the Slow Streptavidin-Biotin Dissociation Kinetics. Journal of the American Chemical Society 117, 10622-10628 (1995). 133. Howarth M, et al. A monovalent streptavidin with a single femtomolar biotin binding site. Nature methods 3, 267-273 (2006). 134. Chivers CE, Crozat E, Chu C, Moy VT, Sherratt DJ, Howarth M. A streptavidin variant with slower biotin dissociation and increased mechanostability. Nature methods 7, 391- 393 (2010). 135. Gavin AC, et al. Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature 415, 141-147 (2002). 136. Ho Y, et al. Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature 415, 180-183 (2002). 137. Schena M, Shalon D, Davis RW, Brown PO. Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 270, 467-470 (1995). 239 138. MacBeath G, Schreiber SL. Printing proteins as microarrays for high-throughput function determination. Science 289, 1760-1763 (2000). 139. Kohler G, Milstein C. Continuous cultures of fused cells secreting antibody of predefined specificity. Nature 256, 495-497 (1975). 140. McCafferty J, Griffiths AD, Winter G, Chiswell DJ. Phage antibodies: filamentous phage displaying antibody variable domains. Nature 348, 552-554 (1990). 141. Scott JK, Smith GP. Searching for peptide ligands with an epitope library. Science 249, 386-390 (1990). 142. Mattheakis LC, Bhatt RR, Dower WJ. An in vitro polysome display system for identifying ligands from very large peptide libraries. Proceedings of the National Academy of Sciences of the United States of America 91, 9022-9026 (1994). 143. Roberts RW, Szostak JW. RNA-peptide fusions for the in vitro selection of peptides and proteins. Proceedings of the National Academy of Sciences of the United States of America 94, 12297-12302 (1997). 144. Colwill K, Graslund S. A roadmap to generate renewable protein binders to the human proteome. Nat Methods 8, 551-558 (2011). 145. Pershad K, et al. Generating a panel of highly specific antibodies to 20 human SH2 domains by phage display. Protein Eng Des Sel 23, 279-288 (2010). 146. Washburn AL, Gunn LC, Bailey RC. Label-free quantitation of a cancer biomarker in complex media using silicon photonic microring resonators. Anal Chem 81, 9499-9506 (2009). 147. Ishikawa FN, et al. Importance of controlling nanotube density for highly sensitive and reliable biosensors functional in physiological conditions. ACS nano 4, 6914-6922 (2010). 148. Dickerson M, Leong K, Sheldon K, Madison L. Developing custom chinese hamster ovary-host cell protein assays using acoustic membrane microparticle technology. J Vis Exp, (2011). 149. Muramatsu H, Kajiwara K, Tamiya E, Karube I. Piezoelectric Immuno Sensor for the Detection of Candida-Albicans Microbes. Analytica Chimica Acta 188, 257-261 (1986). 150. Lee JH. Reduction of a 4-pyrrole phenylacyl-containing peptide with trifluoroacetic acid- triisopropylsilane-phenol-H2O during solid-phase peptide synthesis and its protein kinase C alpha inhibitory activity. Bioorg Med Chem Lett 15, 2271-2274 (2005). 240 151. Takahashi TT, Roberts RW. In vitro selection of protein and peptide libraries using mRNA display. Methods Mol Biol 535, 293-314 (2009). 152. Mutayoba BM, Meyer HH, Schams D, Schallenberger E. Development of a sensitive enzyme immunoassay for LH determination in bovine plasma using the streptavidin- biotin technique. Acta Endocrinol (Copenh) 122, 227-232 (1990). 153. Porstmann T, Kiessig ST. Enzyme immunoassay techniques. An overview. Journal of immunological methods 150, 5-21 (1992). 154. Olson CA, Roberts RW. Design, expression, and stability of a diverse protein library based on the human fibronectin type III domain. Protein Sci 16, 476-484 (2007). 155. Cho JH, Han SM, Paek EH, Cho IH, Paek SH. Plastic ELISA-on-a-chip based on sequential cross-flow chromatography. Anal Chem 78, 793-800 (2006). 156. Lofgren JA, et al. Comparing ELISA and surface plasmon resonance for assessing clinical immunogenicity of panitumumab. J Immunol 178, 7467-7472 (2007). 157. Einav S, Itshayek E, Kark JD, Ovadia H, Weiniger CF, Shoshan Y. Serum S100B levels after meningioma surgery: A comparison of two laboratory assays. BMC Clin Pathol 8, 9 (2008). 158. Pedersen L, Pedersen SM, Brasen CL, Rasmussen LM. Soluble serum Klotho levels in healthy subjects. Comparison of two different immunoassays. Clin Biochem, (2013). 159. Wettergren A, Pridal L, Wojdemann M, Holst JJ. Amidated and non-amidated glucagon- like peptide-1 (GLP-1): non-pancreatic effects (cephalic phase acid secretion) and stability in plasma in humans. Regul Pept 77, 83-87 (1998). 160. Tamamura H, et al. Development of specific CXCR4 inhibitors possessing high selectivity indexes as well as complete stability in serum based on an anti-HIV peptide T140. Bioorg Med Chem Lett 11, 1897-1902 (2001). 161. Stromstedt AA, Pasupuleti M, Schmidtchen A, Malmsten M. Evaluation of strategies for improving proteolytic resistance of antimicrobial peptides by using variants of EFK17, an internal segment of LL-37. Antimicrob Agents Chemother 53, 593-602 (2009). 162. Chen Y, Chen C, Liu-Chen LY. Dynorphin peptides differentially regulate the human kappa opioid receptor. Life Sci 80, 1439-1448 (2007). 163. Starck SR, Roberts RW. Puromycin oligonucleotides reveal steric restrictions for ribosome entry and multiple modes of translation inhibition. RNA 8, 890-903 (2002). 241 164. Thakkar S, Nanaware-Kharade N, Celikel R, Peterson EC, Varughese KI. Affinity improvement of a therapeutic antibody to methamphetamine and amphetamine through structure-based antibody engineering. Sci Rep 4, 3673 (2014). 165. McGhee JD, von Hippel PH. Theoretical aspects of DNA-protein interactions: co- operative and non-co-operative binding of large ligands to a one-dimensional homogeneous lattice. Journal of molecular biology 86, 469-489 (1974). 166. Fried M, Crothers DM. Equilibria and kinetics of lac repressor-operator interactions by polyacrylamide gel electrophoresis. Nucleic Acids Res 9, 6505-6525 (1981). 167. Luo L, et al. Structure-based affinity maturation of a chimeric anti-ricin antibody C4C13. J Biomol Struct Dyn 32, 416-423 (2014). 168. Hakimi H, et al. Development of Monoclonal Antibodies That Target 1-Cys Peroxiredoxin and Differentiate Plasmodium falciparum from P. vivax and P. knowlesi. Trop Med Health 41, 55-59 (2013). 169. Fei Y, et al. Fluorescent labeling agents change binding profiles of glycan-binding proteins. Mol Biosyst 7, 3343-3352 (2011). 170. Sun YS, et al. Effect of fluorescently labeling protein probes on kinetics of protein-ligand reactions. Langmuir 24, 13399-13405 (2008). 171. Blake RC, 2nd, Li X, Yu H, Blake DA. Covalent and noncovalent modifications induce allosteric binding behavior in a monoclonal antibody. Biochemistry 46, 1573-1586 (2007). 172. Boise LH, et al. bcl-x, a bcl-2-related gene that functions as a dominant regulator of apoptotic cell death. Cell 74, 597-608 (1993). 173. O'Connor L, et al. Bim: a novel member of the Bcl-2 family that promotes apoptosis. EMBO J 17, 384-395 (1998). 174. Sleebs BE, et al. Discovery of potent and selective benzothiazole hydrazone inhibitors of Bcl-XL. J Med Chem 56, 5514-5540 (2013). 175. Schlosshauer M, Baker D. Realistic protein-protein association rates from a simple diffusional model neglecting long-range interactions, free energy barriers, and landscape ruggedness. Protein Sci 13, 1660-1669 (2004). 176. Xie L, Mark Jones R, Glass TR, Navoa R, Wang Y, Grace MJ. Measurement of the functional affinity constant of a monoclonal antibody for cell surface receptors using kinetic exclusion fluorescence immunoassay. Journal of immunological methods 304, 1- 14 (2005). 242 177. Rathanaswami P, Richmond K, Manchulenko K, Foltz IN. Kinetic analysis of unpurified native antigens available in very low quantities and concentrations. Analytical biochemistry 414, 7-13 (2011). 178. Bee C, et al. Exploring the dynamic range of the kinetic exclusion assay in characterizing antigen-antibody interactions. PloS one 7, e36261 (2012). 179. Bee C, Abdiche YN, Pons J, Rajpal A. Determining the binding affinity of therapeutic monoclonal antibodies towards their native unpurified antigens in human serum. PloS one 8, e80501 (2013). 180. Darling RJ, Brault PA. Kinetic exclusion assay technology: characterization of molecular interactions. Assay and drug development technologies 2, 647-657 (2004). 181. Harms BD, Kearns JD, Iadevaia S, Lugovskoy AA. Understanding the role of cross-arm binding efficiency in the activity of monoclonal and multispecific therapeutic antibodies. Methods 65, 95-104 (2014). 182. Krishnamurthy VM, Estroff LA, Whitesides GM. Multivalency in Ligand Design. In: Fragment-based Approaches in Drug Discovery (ed^(eds). Wiley-VCH Verlag GmbH & Co. KGaA (2006). 183. Mack ET, Snyder PW, Perez-Castillejos R, Whitesides GM. Using covalent dimers of human carbonic anhydrase II to model bivalency in immunoglobulins. Journal of the American Chemical Society 133, 11701-11715 (2011). 184. Kaufman EN, Jain RK. Effect of bivalent interaction upon apparent antibody affinity: experimental confirmation of theory using fluorescence photobleaching and implications for antibody binding assays. Cancer Res 52, 4157-4167 (1992). 185. Rich RL, Myszka DG. Survey of the year 2006 commercial optical biosensor literature. J Mol Recognit 20, 300-366 (2007). 186. Larsen AC, Gillig A, Shah P, Sau SP, Fenton KE, Chaput JC. General approach for characterizing in vitro selected peptides with protein binding affinity. Anal Chem 86, 7219-7223 (2014). 187. Landry JP, Fei Y, Zhu X. Simultaneous measurement of 10,000 protein-ligand affinity constants using microarray-based kinetic constant assays. Assay and drug development technologies 10, 250-259 (2012). 188. Wassaf D, et al. High-throughput affinity ranking of antibodies using surface plasmon resonance microarrays. Analytical biochemistry 351, 241-253 (2006). 189. Estep P, et al. High throughput solution-based measurement of antibody-antigen affinity and epitope binning. mAbs 5, 270-278 (2013). 243 190. Landry JP, Ke Y, Yu GL, Zhu XD. Measuring affinity constants of 1450 monoclonal antibodies to peptide targets with a microarray-based label-free assay platform. Journal of immunological methods 417, 86-96 (2015).

Abstract (if available)

Abstract We explore the effectiveness of peptides as therapeutic reagents in chapters 2-4 of this work. In Chapter 2 we show successful inhibition of hepatitis C virus particle production in infected human hepatocellular carcinoma cells. In chapter 3 we construct peptides composed entirely of natural amino acids which resist proteases and peptidases. Subsequently, in chapter 4, we develop cell permeable, biologically stable peptides that inhibit the growth of colorectal cancer cells in situ. In chapter 5, we use a fragment-based design approach to select ultrahigh affinity peptide reagents. Using this method, we select picomolar affinity peptides that outperform most clinical antibodies in terms of affinity and specificity, capable of target recognition in complex matrices. Chapters 6 and 7 explore methods to best identify and characterize the highest affinity ligands through either enzyme linked immunosorbent assays (ELISAs) or using an ultrasensitive acoustic resonant sensor. Chapter 8 describes the combination of mRNA display and next generation DNA sequencing to create a powerful tool capable of high-throughput analysis of ligand affinities. Using this technique, we obtain the dissociation constant for tens of thousands of ligands against Bcl-xL and identify the highest affinity peptide-protein interaction ever known.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Directed evolution of high affinity, cyclic, protease resistant peptides using mRNA display

PDF

Customization of the mRNA display cycle – from protease resistance to high throughput analysis of peptide ligands

PDF

The selective targeting of Ras: a tale of kinetics, mRNA display, and vaccines

PDF

Intricate microfluidic devices for biopharmaceutical processes: forging ahead with additive manufacturing

PDF

mRNA display as a method to generate stable fibronectin ligands for the alpha 1 subunit of the nicotinic acetylcholine receptor

PDF

Rapid microfluidic prototyping for biological applications via stereolithographic 3-D printing: a journey

PDF

Developing peptide and antibody-mimetic ligands for the cell surface receptors β2AR and DC-SIGN

PDF

Engineering immunotoxin and viral vectors for cancer therapy

PDF

Engineering therapeutics for the improved antitumor efficacy of chimeric antigen receptor T cell therapy

PDF

Engineering lentiviral vectors for gene delivery and immunization

PDF

Engineering chimeric antigen receptor-directed immune cells for enhanced antitumor efficacy in solid tumors

PDF

Simultaneous monomer deposition and polymerization at low substrate temperatures for the formation of porous polymer membranes

PDF

Engineering lentiviral vectors for gene therapy and DC-vaccine

PDF

Dendritic cell-specific vaccine utilizing antibody-mimetic ligand and lentivector system

PDF

Cell penetrating peptide-based polyplexes for sirRNA delivery

PDF

Directed evolution provides stable, cell permeable, efficacious, and orally bioavailable peptidomimetics

PDF

Study of dendritic cell targeting by engineered lentivectors

PDF

High-throughput nanoparticle fabrication and nano-biomembrane interactions

PDF

Engineering nanoparticles for gene therapy and cancer therapy

PDF

Imaging molecular transport across and nanomaterial interaction with lipid membranes

Asset Metadata

Creator Jalali-Yazdi, Farzad (author)

Core Title Generation and characterization of peptide theranostics by mRNA display

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Chemical Engineering

Publication Date 07/22/2017

Defense Date 06/23/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag affinity,hepatitis c virus,mRNA display,OAI-PMH Harvest,peptide,protein engineering

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Roberts, Richard W. (committee chair), Fraser, Scott E. (committee member), Malmstadt, Noah (committee member), Wang, Pin (committee member)

Creator Email jalaliya@usc.edu,thyzad@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-605269

Unique identifier UC11300275

Identifier etd-JalaliYazd-3678.pdf (filename),usctheses-c3-605269 (legacy record id)

Legacy Identifier etd-JalaliYazd-3678.pdf

Dmrecord 605269

Document Type Dissertation

Format application/pdf (imt)

Rights Jalali-Yazdi, Farzad

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA