De novo mapping of α-helix recognition sites on protein surfaces using unbiased libraries

Significance To target disease-causing proteins that are inaccessible to small-molecule or antibody therapies, we are exploring the use of a nature-inspired drug modality that exploits the intrinsically favorable properties of synthetically constrained α-helical polypeptides (Helicons). Here, we report a screening method that enables the de novo discovery of Helicons for protein targets without prior information on their α-helix binding properties, which has significantly limited the proteins and diseases for which Helicons could be discovered. We applied this method to six structurally diverse protein domains and found Helicons that block relevant protein–protein interactions, inhibit enzymatic activity, induce conformational rearrangements, and dimerize their targets, highlighting the strength of the approach in interrogating targets that have been considered “undruggable”.

between 5 x 10 12 and 1 x 10 13 phage particles in each amplified library. Phage-displayed Helicon libraries were covalently crosslinked (stapled) by diluting the phage particle solution in TBS to an OD600 of 1.0 and adding dithiothreitol to a concentration of 1 mM, followed by dialysis against 100 volumes of 20 mM NH4CO3, 2 mM EDTA, pH ~8 for 30-60 min, followed by addition of the dialyzed phage to a solution of crosslinker prepared in 20 mM NH4CO3, 2 mM EDTA, pH ~8 (final crosslinker concentration is 200 μM. As the crosslinker does not completely dissolve in buffer, we briefly sonicated the solution immediately prior to mixing with phage to disperse the solid into a fine suspension) and incubation with rotation for 2 hours at 32˚C. Excess crosslinker was removed first by pelleting at 5000 x g and decanting, followed by addition of dithiothreitol to a concentration of 0.25 mM with incubation for 10 minutes, and then addition of iodoacetamide to a concentration of 0.75 mM with incubation for a further 10 minutes. Ellman's reagent (5,5′-dithiobis-(2-nitrobenzoic acid) was used to track the quenching of DTT until all thiols were capped. Phage particles were further purified by repeating the precipitation, pelleting, and resuspension steps described above for purification from E. coli culture, then are stored as solutions in 50% v/v glycerol in TBS at -80˚C at >10 12 pfu/mL. Nextgeneration sequencing was performed to assess the library quality (details can be found in the Phage NGS section). We typically sequenced between 10 6 -10 7 phage particles and found that on average, between 95%-98% of all reads have the correct library structure. Mass spectrometry analysis of crosslinked phage was performed by adding 15 μL of phage samples to 2.5 μL of a solution of Trypsin at 1.0 mg/ml freshly prepared in 5 mM acetic acid and then pH adjusted by mixing 1:1 with 100 mM Tris pH 8. After 1 hour of digestion at room temperature, each cleavage reaction was quenched with 15 μL of 20% ACN + 1% formic acid, and analyzed a Q-Exactive Plus mass spectrometer equipped with an Ultimate 3000 LC system (Thermo Electron) and a Aeris™ C18 column (Phenomenex). Finally, individual phage library members were characterized by DNA sequencing. Well-separated blue plaques were picked from the LB/IPTG/Xgal Agar plates in 50 μL of water. 2 μL of resuspended template was mixed with 23 μL of the amplification master mix containing OneTaq DNA polymers (NEB, Ipswitch, MA) and two 10 μM M13KE sequence-specific amplification primers (NEB, Ipswitch, MA). Routine PCR was performed, and samples were submitted for standard Sanger sequencing (GENEWIZ, Cambridge, MA). Prior to library screening, we performed and recommend deep sequencing (as described below) of the library to ensure that it is high in sequence diversity and is not dominated by a small number of individual sequences.

5
Phage Library Screening Phage screening was performed using biotinylated proteins bound to streptavidin magnetic beads (Dynabeads MyOne Streptavidin T1, Thermo Fisher Scientific, Waltham, MA). 10 10 phage particles were added to each phage screening sample, to ensure approximately 100 copies of each of the 10 8 library members. Phage display libraries were incubated with streptavidin magnetic beads for 1 hour at room temperature in a buffer made of 1X TBS, 1 mM MgCl2, 1% w/v BSA, 0.1% Tween-20, 0.02% w/v sodium azide, 5% w/v nonfat milk to remove bead-binding library members. Briefly, Dynabeads were prepared in a 15ml Falcon tube according to a manufacturing protocol from Thermo Fisher Scientific, diluted phage display libraries were added to the magnetic beads. After an incubation period, the tube was placed on a magnet for 1 min to separate bead-bound and non-bead-bound phage library members. Supernatant containing bead depleting phage library was collected and beads were discarded. For each screening condition, 100 μL of 2 μM biotinylated protein was captured with 0.5 mg of streptavidin-coated magnetic beads that have been previously blocked with 1% BSA, 0.1% Tween, 2% glycerol in 1x TBS pH 7.4 at room temperature for 15 minutes in 96-well plates, followed by removal of the supernatant using a plate magnet and prompt but gentle resuspension of the beads in 50 μL of the same buffer. Next, 150 μL of the depleted phage library was added to each well for 200 μL final volume, the plates were sealed, and the screening reactions incubated at room temperature for 45 minutes, with rotation to maintain beads in solution. We inspected these solutions to confirm that beads had not aggregated or crashed out of solution, which can be indicative of protein aggregation. Following binding, beads were washed 5x with ice-cold 1x TBS, 1 mM MgCl2, 1% w/v BSA, 0.1% Tween-20, 0.02% w/v sodium azide, 2% w/v glycerol. Washing steps can be performed, as in our case, with an automated bead handler such as a KingFisher (Thermo Fisher) or manually. If washing beads manually for a screen of 48 wells or fewer, we recommend working quickly to ensure that washing occurs consistently between samples and occurs in 20 minutes or less. Given the speed required to complete all steps, manual washing of greater than 48 samples is not recommended. Target-bound phage library members are directly processed for NGS.

6
Phage Next-Generation Sequencing (NGS) This protocol is used to perform NGS for newly built phage display libraries and to identify targetbound phage library members after a phage screen. To sequence the phage-displayed peptide library members, phage particles are removed from the beads by a denaturation step at 95˚C for 15 min in 25 mM Tris pH 8, 50 mM NaCl, 0.5% Tween-20. Prior to boiling, 10,000 copies of a phage clone of known sequence (not a library member) are spiked in to each well to enable crosswell normalization of sequence reads. The sequence of the spike-in clone is TCTCACTCTGCGCCGGAATGCATTCTGGATTGCCATGTGGCGCGCGTGTGGGGTGGTTCT. A two-step low-cycled PCR is performed to introduce Illumina adaptors and 10bp TruSeq DNA UD Indexes (Illumina, San Diego, CA) to the 3' and 5' ends of amplicons with M13KE Forward and M13KE Reverse primers (M13KE Forward: 5'-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGTTCGCAATTCCTTTAGTGG-3' and M13KE Reverse: 5'-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGATTTTCTGTATGGGATTTTGCTAA-3') similar to Illumina's 16S Metagenomic Sequencing Library Preparation protocol. The NGS libraries are sequenced by an Illumina NovaSeq platform using a 2x150-bp high-output kit (Illumina, San Diego, CA).

7
Hit ID and Clustering NGS reads were trimmed for quality (Phred score ≥ 18) and filtered for sequences that matched the design of the phage library (Fig 2A). Counts for each unique sequence were tallied, and then normalized by the counts of the spike-in sequence added to each sample. A metric called Hit Strength was computed for each sequence as the fold change between the normalized counts in the highest target concentration sample and the normalized counts in the blank bead samples (averaged across experimental replicates). When 0 counts were observed for a sequence in blank bead samples, a count of 0.5 was used to prevent dividing by zero. Sequences with a hit strength greater than 5 were then subjected to hierarchical clustering to identify sequence families. Pairwise distances between sequences i and j were computed using where scoreij is the alignment scores based on a modified BLOSUM62 substitution matrix. We decreased the tryptophan-tryptophan match score from 11 to 7 in the BLOSUM62 substitution matrix to prevent overly biasing clustering towards tryptophans. Hierarchical clustering using average linkage was used to group the sequences into families. To avoid clustering using a large number of sequences, which is computationally intensive and can make it difficult to identify small clusters, we performed multiple rounds of clustering. First, we sorted sequences by descending hit strength. We then took the top 1000 (first round of clustering), top 2000 (2 nd round of clustering), or top 3000 (≥3 rd round of clustering) sequences, and subjected them to clustering as described above. Clusters of sequences with high sequence similarity (sequence "families") were identified at each round and removed from the pool of sequence for subsequent rounds. Sequences subjected to three rounds of clustering without falling into a sequence family were similarly dropped from subsequent rounds of clustering under the assumption that they did not belong to a sequence family. The process was halted after 10 rounds of clustering, or when no sequences remained in the list.

β-Catenin Surface Plasmon Resonance (SPR)
SPR experiments were performed on a Biacore™ 8K (Cytiva) instrument at 25°C. Test peptides were diluted into running buffer (50 mM Tris pH 8.0, 300 mM NaCl, 2% glycerol, 0.5 mM TCEP, 0.5 mM EDTA, 0.005% Tween-20, 1% DMSO). Compounds were diluted to 10 μM or 1 μM and serially diluted 1:3 for seven concentrations and two blanks (7-point three-fold peptide dilution series with top concentration = 10 M). Biotinylated β-catenin residues 134-665 (Uniprot ID P35222) was immobilized to the active surface of the sensor chip for 25 seconds at 10 μL/min using the Biotin CAPture Kit, Series S (Cytiva) and compounds were injected over the reference and active surfaces for 180 seconds at 65 μL/min then allowed to dissociate for 400 seconds. Results were analyzed using the Biacore™ Insight Evaluation software, with double-referencing and fitted to a 1:1 binding affinity model.

9
β-catenin-TCF Competition by Fluorescence Polarization Compounds at 10 mM in DMSO were serially diluted 1:3 in DMSO for a total of 11 concentrations using a Mosquito LV (SPT Labtech), then diluted 1000-fold in buffer (50 mM HEPES, pH 7.5, 125 mM NaCl, 2% glycerol, 0.5 mM EDTA, 0.05% v/v pluronic acid) in duplicate by the Mosquito LV (SPT Labtech) into a black polystyrene 384-well plate (Corning) (11-point three-fold peptide dilution series with top concentration = 10 M) Probe solution (10 nM fulllength β-catenin (Uniprot ID P35222), mixed with 10 nM 5FAM-labeled TCF4 residues 10-53 (Uniprot ID Q9NQB0) peptide (FP04872) in buffer) was prepared and plated using the MultiDrop Combi (Thermo Fisher) for a total reaction pool of 40 L. The plate was incubated and protected from light for 1 hour at room temperature prior to read. Reads were performed on a CLARIOstar plate reader (BMG Labtech) with excitation at 485 nm, emission at 525 nm, and cutoff at 504 nm. Data were fitted to a 1:1 binding model with Hill slope using an in-house script.

β-catenin-Axin Competition by Fluorescence Polarization
Compounds at 10 mM in DMSO were serially diluted 1:3 in DMSO for a total of 11 concentrations using the Mosquito LV (SPT Labtech), then diluted 1000-fold in buffer (50 mM HEPES, pH 7.5, 125 mM NaCl, 2% glycerol, 0.5 mM EDTA, 0.05% v/v pluronic acid) in duplicate by the Mosquito LV (SPT Labtech) into a black polystyrene 384-well plate (Corning). Probe solution (15 nM full-length β-catenin (Uniprot ID P35222), mixed with 20nM FITC labeled fStAx-33 (5) peptide (FP00013) in buffer) was prepared and plated using the MultiDrop Combi (Thermo Fisher) for a total reaction pool of 40 L. The plate was incubated protected from light for 1 hour at room temperature prior to read. Reads were performed on a CLARIOstar plate reader (BMG Labtech) with excitation at 485 nm, emission at 525 nm, and cutoff at 504 nm. Data were fitted to a 1:1 binding model with Hill slope using an in-house script.

11
Measurement of the Cell Association of Helicons A "Source" plate of 47 test compounds was prepared at a concentration of 1mM in 90% DMSO in a 500 μL 96-well plate (2 replicates for each compound with 2 DMSO blanks). The 96-well format with one compound per well was maintained for all transfers throughout the protocol. A 2000 μL 96-well v-bottom plate ("Cells" plate) was used to dilute 1.25 μL of compounds from the Source plate into 500 μL of Expi293™ Expression Medium. Each well also received 500 μL of Expi293™ cells (Thermo Fisher) for a total concentration of 1x10 6 cells/mL in 1 mL of Expi293™ Expression Medium. Baseline cell health at the time of compound addition was measured using CellTiter-Glo™ 2.0 Reagent (CTG) and a GloMax Discover reader (Promega). The Cells plate was incubated in an Infors HT Multitron Pro shaking incubator at 1000rpm, 37°C, 8.0% CO2, ~55% humidity, for 22 hours, along with four plates of DI water to maintain humidity. After 22 hours, the cells were sampled again for post-incubation CTG analysis of cell health. CellTiter-Glo foldchange is calculated as the luminescence readout at Tfinal divided by the luminescence readout at Tinital (time at which peptide was added). The cells were washed twice with 200 μL of Dulbecco's Phosphate Buffered Saline (DPBS) and transferred to a 500 μL 96-well "Final Assay" plate. After the second wash, the cells were resuspended in 80 μL of buffer (90% DPBS, 10% dimethyl sulfoxide (DMSO), 10 μM of a mixture of nonstapled 14-mer peptides with randomized sequences) in the Final Assay plate. Cell lysis was induced by the addition of 240 μL of ammonium hydroxide and 2 hours of shaking at 37°C. All solvents were removed via 23 hours in a SpeedVac vacuum concentrator (Thermo Fisher). The dried compounds and cell debris were resuspended using 180 μL of resuspension buffer (47.5% Acetonitrile (ACN) with 0.1% formic acid (FA), 47.5% H2O with 0.1% FA, 5% DMSO) and shaking for 3 hours at 600 rpm. Once resuspended, the cell debris was separated from the resuspended compounds via centrifugation at 3220 rcf for 20 min. A portion of compound-containing supernatant from each well was transferred to a corresponding well in a 384-well plate for mass spectrometry analysis. Matching "cell-free" wells for all "cells" wells were plated in the same 384-well plate. The "cellfree" wells were prepared by adding 1 μL of 0.1 mM compounds from the Source plate to 19 μL of input buffer (47.5% ACN with 0.1% FA, 47.5% H2O with 0.1% FA, 5% DMSO, 10 μM of a mixture of nonstapled 14-mer peptides with randomized sequences), then adding 2 μL of this 5 μM-dilution to 58 μL of input buffer in the 384-well mass spectrometry plate. All samples were analyzed using mass spectrometry. The percentage of compound in cells after treatment and wash was computed as the percentage of the total compound added to the Cells samples that was present in the cell fraction after removal of the extracellular media. Compounds were quantitated by mass spectrometry, and the signal in the cell-free samples were used as a singlepoint calibration curve to convert from signal intensity in the Cells sample to the percentage of the total amount added.
Proteins were eluted isocratically and fractions containing pure protein were collected and pooled.
PUB domain RNF31 PUB domain protein (residues 1-179) with N-terminal MBP-TEV-6xHis-YBBR-3C tags was recombinantly expressed in E. coli BL21 (DE3) pLysS cells (Thermo Fisher) from pET-derived expression vectors (Novagen). The cells were induced at OD600=0.6 with 0.15 mM IPTG for 16 hours at 16°C, then harvested and resuspended in 50 mM HEPES pH 7.5, 500 mM NaCl, 1 mM TCEP, 2 mM ATP, 5 mM MgCl2, 5% Glycerol, and 1 mM PMSF. For purification, the pellet was lysed with a tip sonicator, pelleted at 22,000 x g for 30 minutes at 4°C, then the supernatant was purified with Ni-NTA resin (Qiagen), eluting with 250 mM imidazole. For crystallography efforts, protein was cleaved by adding 3C protease at a ratio of 1:40 protease to protein and incubating overnight at 4°C. For biochemical experiments, protein was cleaved by adding TEV protease at a ratio of 1:10 protease to protein and incubating overnight at 4°C overnight. TEV cleaved proteins were then biotinylated via the yBBr reaction. Proteins were concentrated and injected over a Superdex® HiLoad 16/600 75pg SEC column pre-equilibrated with 25 mM HEPES pH 7.5, 150 mM NaCl, 1 mM TCEP. Proteins were eluted isocratically and fractions containing pure protein were collected and pooled.
14 RNF31 UBA and PUB SPR All SPR experiments were performed on a Biacore 8K (Cytiva) instrument at 25°C. For kinetics experiments, the instrument was primed with 10 mM HEPES, pH7.5, 150 mM NaCl, 0.05% Tween 20, 1% DMSO. A CAP Series S sensor chip was docked and pre-conditioned with 3 injections of 1X CAP regeneration solution to remove unbound capture reagent from the surface. Biotinylated RNF31 UBA and PUB domain proteins were diluted to 1 µM in running buffer. FP06655 was diluted to 1 µM in running buffer and serially diluted 1:3 for a total of 8 concentrations and a blank (8-point four-fold peptide dilution series with top concentration = 1 M.). Otulin and test peptides were diluted to 10 µM in running buffer and serially diluted 1:3 for a total of 8 concentrations and a blank. Proteins were captured to the active surface of the sensor chip for 60 seconds at 5 µL/min and the peptides were injected over the reference and active surfaces for 180 seconds at 50 µL/min then allowed to dissociate for 360 seconds. Surface was regenerated with a 120-second injection of CAP regeneration solution each cycle. Data was analyzed using Biacore Insight Evaluation software (Cytiva). Sensorgrams were double-referenced and fit to 1:1 binding affinity model.

15
RNF31-Otulin competition Fluorescence Polarization RNF31 PUB domain was diluted to 1.6 µM in assay buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 0.05% Tween 20) and pipetted into a 384-well black microplate (Corning) in a final volume of 20 μL. Test peptides were added to the plate (40 nL each) serially diluted 3-fold from 10 mM and the plate was incubated at room temperature for 20 minutes (11-point three-fold peptide dilution series with top concentration = 10 M). FITC-labeled Otulin peptide (residues 49-67) (FP16923) was diluted to 40 nM in assay buffer, then 20 μL of the stock was added to the plate for a final volume of 40 μL. The plate was incubated for 60 minutes at room temperature, then fluorescence anisotropy was recorded on a CLARIOstar (BMG LabTech) with excitation at 485 nm, emission at 525 nm, and cutoff at 515 nm. Data were plotted using Prism (Graphpad) and fit to a one-site specific binding model with Hill coefficient.

SPR ABA Competition
All SPR experiments were performed on a Biacore 8K (Cytiva) instrument at 25°C. For kinetics experiments, the instrument was primed with 10 mM HEPES, pH 7.5, 150 mM NaCl, 0.05% Tween 20, 1% DMSO. A SA Series S sensor chip was docked and pre-conditioned with three injections of 50mM NaOH/1M NaCl to remove unbound streptavidin from the surface. Biotinylated RNF31 PUB domains were diluted to 2 μM in running buffer. FP06649 and FP06652 were diluted to 10 μM in running buffer. Otulin peptide (residues 49-67) was diluted to 10 μM in running buffer. Proteins were captured on the active surface of the sensor chip for 300 seconds at 1 μL/min. For each injection, compounds were injected over the surface for 120 seconds at 30 μL/min to achieve equilibrium binding. Otulin was then injected for 60 seconds at 30 μL/min in the absence or presence of competing compound over the surface. Surface was regenerated with an injection of 1M sodium chloride after each cycle. Data was analyzed using Biacore Insight Evaluation software (Cytiva). Sensorgrams were double-referenced and evaluated for competition.

16
RNF31-Sharpin competition by fluorescence polarization RNF31 UBA domain, and FAM-labeled RNF31-binding peptide, (FP12122), were diluted to 400 nM and 40 nM, respectively in assay buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 0.05% Tween 20) and pipetted into a 384-well black microplate (Corning) in a final volume of 20 μL. Recombinant Sharpin/SIPL1 UBL protein (residues 153-256), as well as selected control peptides, were added to the plate (20 L each), serially diluted 3-fold from 10 M (10-point three-fold peptide or Sharpin dilution series with top concentration = 3.3 M). The plate was incubated for 60 minutes at room temperature, then fluorescence anisotropy was recorded on a CLARIOstar (BMG LabTech) with excitation at 485 nm, emission at 525 nm, and cutoff at 515 nm. Data were plotted using Prism (Graphpad) and fit to a one-site specific binding model with Hill coefficient.

17
CDK2 and PPIA protein production CDK2 Full length CDK2 (residues 1-298) with an N-terminal GST-3C-6xhis-TEV tag was recombinantly expressed in Sf9 cells according to the Bac-to-Bac protocol (Thermo Fisher). Briefly, Sf9 cells were plated at 1x10 6 cells in 2mL Sf-900™ II media (Thermo Fisher) into a 6-well cell culture plate. Cells were transfected with purified bacmid diluted in OptiMem™ media using Cellfectin™ reagent. Cells were incubated at 27°C for 5 days. Cells and supernatant were removed from plate and centrifuged. P1 virus was collected and cell pellet was evaluated for protein expression by Western blot. P2 virus was generated by infecting 2x10 6 cells/mL Sf9 in 50 mL Sf-900™ II media with 500 mL P1 virus. Cells were incubated with shaking at 27°C for 5 days. Cells were centrifuged at 1500 rpm at room temperature for 5 minutes. Supernatant was stored at 4°C as P2 virus stock and pellet was evaluated by Western blot for protein expression. Protein was expressed by seeding Sf9 cells at 2x10 6 cells/mL in Sf-900™ II media and infecting at an MOI of 1:200. Cultures were incubated with shaking for 72 hours at 27°C. Cells were harvested and supernatant was discarded. Pellets were resuspended in 25mM HEPES, pH 7.5, 300 NaCl, 10% glycerol, 0.5 mM PMSF and then sonicated with a tip sonicator. Lysates were centrifuged at 22,000 x g for 30 minutes at 4°C. Clarified lysate was purified with a GSTrap™ (Cytiva) pre-equilibrated in 25 mM HEPES, pH 7.5, 300 mM NaCl, 10% glycerol. Protein was eluted with 25 mM HEPES, pH 7.5, 300 mM NaCl, 10% glycerol, 10 mM GSH. Eluted protein was cleaved by combining protein with TEV protease at a ratio of 1:10 protease to protein and incubating at room temperature for 40 hours. Cleaved protein was then dialyzed into 25 mM HEPES, pH 7.5, 300 mM NaCl, 10% glycerol and re-injected over a GSTrap™ to remove cleaved tags. Purified protein was concentrated and centrifuged at 22,000 x g for 10 minutes at 4°C to remove soluble aggregates. Protein was then loaded onto a Superdex™ HiLoad 16/600 75pg SEC column pre-equilibrated with 20 mM HEPES, pH 7.5, 150 mM NaCl, 2% glycerol, 2 mM DTT. Protein was eluted isocratically at 0.5 mL/min. Finally, protein was centrifuged at 22,000 x g for 10 minutes at 4°C to remove soluble aggregates, then aliquoted and frozen.

PPIA
Full length PPIA (residues 1-165) with an N-terminal 6xhis-yBBR-TEV tag was recombinantly expressed in E. coli BL21 (DE3) CodonPlus RIPL cells (Agilent) from pET-derived expression vectors (Novagen). The cells were induced at OD600=0.6 with 0.15 mM isopropyl β-D-1thiogalactopyranoside (IPTG) for 16 hours at 16°C, then harvested and resuspended in PBS pH 7.4 with 1mM PMSF. For purification, the pellet was lysed with a tip sonicator, pelleted at 22,000 x g for 30 minutes at 4°C. Supernatant was collected then centrifuged again at 22,000 x g for 30 minutes at 4°C. The supernatant was purified with Ni-NTA resin (Qiagen), eluting with 250 mM imidazole. Protein was TEV cleaved by adding TEV protease at a ratio of 1:10 protease to protein and incubated for 4 hours at 4°C. Protein was then concentrated and diluted into 20 mM HEPES pH 7.0, 5% glycerol and centrifuged at 22,000 x g for 10 min at 4°C. The supernatant was loaded onto a SP HP (Cytiva) column pre-equilibrated with 20 mM HEPES pH 7.0, 5% glycerol. Purified protein was eluted with a gradient from 0 mM to 1 mM NaCl. Protein fractions were pooled, concentrated then centrifuged at 22,000 x g for 10 min at 4°C. Supernatant was collected and loaded onto a Superdex™ 16/600 75pg (Cytiva) SEC column pre-equilibrated with PBS pH 7.4. Purified proteins were eluted isocratically in PBS pH 7.4. Protein fractions were collected, concentrated, aliquoted and frozen.

18
SPR for CDK2 and the CDK2/CCNE1 complex SPR experiments were performed on a Biacore S200 or 8K (Cytiva) instrument at 25°C in 1x HBS-P+ buffer (Cytiva) with 1% DMSO. A SA Series S sensor chip was docked and preconditioned with three injections of 50 mM NaOH/1 M NaCl to remove unbound streptavidin from the surface. CDK2 protein, or CDK2: CCNE1 complex, was diluted to 5 μg/mL in running buffer and immobilized to channels 1 through 8 at 5 μL/min for 50-80 seconds for a final immobilization level of ~1800 RU. Peptides were diluted to 5 μM in running buffer then serially diluted 2-fold for a total of seven concentrations with one blank (7-point two-fold peptide dilution series with top concentration = 5 μM.). Compounds were injected over the immobilized and reference surfaces at 30 μL/min for 60 seconds and then allowed to dissociate for 180 seconds without surface regeneration. Sensorgrams were double-referenced and fit to a 1:1 steady state affinity model.

19
ATP competition of CDK2 For ATP competition experiments, 50 M of selected CDK2-binding compounds were added into 20 nM Bodipy-ATP-γS (Thermo Fisher) and 2 M CDK2 in an assay buffer contained 20 mM Tris pH 8, 300 mM NaCl, 10% (v/v) glycerol, 2 mM TCEP, and 10 mM MgCl2, and pipetted into a 384-well black microplate (Corning) in a final volume of 40 μL. The plate was incubated for 60 minutes at room temperature, then fluorescence anisotropy was recorded on a CLARIOstar (BMG LabTech) with excitation at 485 nm, emission at 525 nm, and cutoff at 515 nm. Final data were normalized against DMSO control and Bodipy-ATP-γS-free control.

20
SPR for PPIA, including cyclosporine competition All SPR experiments were performed on a Biacore 8K (Cytiva) instrument at 25°C. For kinetics experiments, the instrument was primed with 10 mM HEPES, pH 7.5, 150 mM NaCl, 0.05% Tween 20, 1 mM DTT, 1% DMSO. A SA Series S sensor chip was docked and pre-conditioned with 3 injections of 50 mM NaOH/1 M NaCl to remove unbound streptavidin from the surface. PPIA protein was diluted to 5 μg/mL in running buffer and immobilized to channels 1 through 8 at 5 μL/min for 50 seconds for a final immobilization level of ~1900 RU. Peptides were diluted to 5 μM in running buffer then serially diluted 2-fold for a total of seven concentrations with one blank (7-point two-fold peptide dilution series with top concentration = 5 μM.). Compounds were injected over the immobilized and reference surfaces at 30 μL/min for 60 seconds then allowed to dissociate for 180 seconds. The surface was regenerated after each cycle with an injection of 1 M sodium chloride. Sensorgrams were double-referenced and fit to a 1:1 steady state affinity model. For ABA competition experiments, PPIA was immobilized similarly to a level of ~850 RU. Compounds were diluted to 10 μM in running buffer and cyclosporine A (CsA) was diluted to 100 nM in running buffer. For each injection, peptides were injected over the surface for 120 seconds at 30 μL/min to achieve equilibrium binding. CsA was then injected for 60 seconds at 30 μL/min in the absence or presence of competing compound over the surface. Surface was regenerated after each cycle with an injection of 1 M sodium chloride. Data was analyzed using Biacore Insight Evaluation software (Cytiva). Sensorgrams were double-referenced and evaluated for competition.

21
PPIA inhibition assay PPIA inhibition assays were performed at Eurofins Discovery (Ongar, Essex, United Kingdom). Briefly, a 1.5 mL of assay buffer (35 mM HEPES pH 7.8, 50 mM DTT) is pipetted into a 3-mL glass cuvette and cooled to 10°C with stirring. Test compounds are diluted in 100% DMSO then added to the buffer to establish a blank. PPIA is then added at a final concentration of 2 nM and substrate is added to a final concentration of 60 μM. The absorbance at 330 nm is measured for 300 seconds. The resulting data were fit to a first order rate equation and the catalytic rate was calculated. An exponential curve was generated using the catalytic rate versus the inhibitor concentration to obtain a Ki value. CsA is included at a single concentration as a positive control.

22
Production of PDL1 E.coli Protein For crystallography, human PD-L1 protein (residues 18-134) with a C-terminal 6xHis tag was recombinantly expressed in E. coli BL21 (DE3) CodonPlus RIPL cells (Agilent) from pET-derived expression vectors (Novagen). The cells were induced at OD600=0.6 with 1.0 mM IPTG for 4 hours at 37°C, then harvested and resuspended in 20 mM Tris-HCl pH 8.0, 300 mM NaCl, 10% glycerol, 1 mM PMSF, DNase I. For purification, the pellet was lysed with a tip sonicator, pelleted at 22,000 x g for 50 minutes at 4°C. Protein was located in the inclusion bodies, so the cell pellet was collected and resuspended in 50 mM Tris-HCl pH 8.0, 200 mM NaCl, 10 mM EDTA, 10 mM β-mercaptoethanol (β-ME), 0.5% Triton X-100 and stirred by magnetic stir bar at room temperature for 30 minutes. The suspension was centrifuged at 22,000 x g for 30 minutes at 4°C. The process was repeated three times. The pellet was resuspended in 50 mM Tris-HCl pH 8.0, 200 mM NaCl, 10 mM EDTA, 10 mM β-ME and stirred by magnetic stir bar at room temperature overnight. The suspension was centrifuged at 22,000 x g for 30 minutes at 4°C and the supernatant was collected. Supernatant was loaded onto a HisTrap (Cytiva) pre-equilibrated with 50 mM Tris-HCl pH 8.0, 200 mM NaCl, 8 M urea, 10 mM β-ME and eluted with 250 mM imidazole. Protein was refolded by diluting into 100 mM Tris-HCl pH 8.0, 1 M L-Arginine, 0.235 mM GSH, 0.25 mM GSSG with incubation overnight at 4°C. Protein was then dialyzed against PBS pH 7.4 for 4 hours at 4°C. Dialysis was repeated three times. Protein was concentrated and injected onto a Superdex™ HiLoad 16/600 75pg SEC column pre-equilibrated with 10 mM Tris-HCl pH 8.0, 20 mM NaCl. Fractions containing pure protein were collected and pooled.

Mammalian Protein
For biochemical assays, human PD-L1 (residues 18-239) with a C-terminal human IgG Fc-Avi™tag or a C-terminal TEV-10xhis-Avi™ tag were recombinantly co-expressed with BirA in Expi293™ cells from pcDNA-derived plasmid (Thermo Fisher) using the Expifectamine™293 expression system. For Fc-tagged protein, cells were harvested after 5 days of expression and supernatant was collected and passed over a MabSelect sure affinity column (Cytiva) preequilibrated with PBS pH 7.4. Protein was eluted with 100 mM sodium citrate, pH 3.0 and neutralized in 1M Tris-base, pH 9.0. Elution was concentrated and injected over a Superdex™ HiLoad 16/600 200pg pre-equilibrated with PBS pH 7.4. Fractions containing pure protein were collected and pooled. For 10xhis-tagged protein, cells were harvested after 5 days of expression and supernatant was collected and passed over a Ni-NTA affinity column (Qiagen) preequilibrated with PBS pH 7.4. Protein was eluted with 300 mM imidazole. Elution was concentrated and injected over a Superdex™ HiLoad 16/600 75pg pre-equilibrated with PBS pH 7.4. Fractions containing pure protein were collected and pooled.

23
SPR for PDL1, including PD1 competition All SPR analysis was performed on a Biacore S200 (Cytiva) in at 25°C. BMS PD-L1 interacting small molecules (BMSpep-57, BMS-1, BMS-1001) were obtained from MedChem Express. A Protein G Series S Sensor Chip (Cytiva) was docked into the instrument primed with PBS pH 7.4 with 0.05% Tween 20 and 1% DMSO. PD-L1-Fc was diluted in running buffer to 50 nM and captured on the surface for 60 seconds at 5 μL/min. Compounds were diluted to 2 μM then serially diluted 3-fold in running buffer (7-point two-fold dilution series with top concentration = 5 M (FP30790), or 6-point two-fold dilution series with top concentration = 2 M (others)). Diluted compounds were injected over the surface at 30 μL/min for 180 seconds and allowed to dissociate for 360 seconds. The surface was regenerated every cycle with a 60-second injection of 10 mM glycine-HCl, pH 2.5. The resulting sensorgrams were double-referenced and fit to a 1:1 binding model using Biacore Insight Evaluation software (Cytiva). For ABA competition experiments, PD-L1 was immobilized on a Streptavidin Series S Sensor Chip (Cytiva) to ~250 RU. Compounds were diluted to 10 μM in running buffer and PD-1 was diluted to 400 nM in running buffer. For each injection, compounds were injected over the surface for 120 seconds at 30 μL/min to achieve equilibrium binding. PD-1 was then injected for 60 seconds at 30 μL/min in the absence or presence of competing compound over the surface. The surface was regenerated every cycle with a 60-second injection of 10 mM glycine-HCl, pH 2.5. Data was analyzed using Biacore Insight Evaluation software (Cytiva). Sensorgrams were doublereferenced and evaluated for competition.

24
ELISA for PDL1 PD-1/PD-L1 ELISA competition assays were performed according to manufacturer's instruction (Acro Biosystems). Human PD-L1 was diluted to 2 μg/mL in PBS + 0.05% Tween 20. High-binding 96-well plates (Corning) were coated with PD-L1 at 2 μg/ 100 μL per well. Human PD-1-Avi was diluted to 0.6 μg/mL in ELISA wash buffer (PBS + 0.05% Tween 20 + 0.5% BSA + 0.09% DMSO) and added to the coated wells to form complexes. Test peptides were diluted in ELISA wash buffer to 20 µM then serially diluted 4-fold and added to the PD-L1/PD-1-avi complexes. After incubation with the ligand, the plate was washed, and the bound ligand was detected with the addition of streptavidin-HRP, followed by development with 1-Step™ Ultra TMB-ELISA substrate solution (Thermo Fisher). The HRP reaction was stopped by adding ELISA stop solution (Thermo Fisher). Absorbance at 450 nm was determined on a CLARIOstar plate reader. Samples were blank-subtracted and normalized using in-plate controls. Data were plotted using Prism (Graphpad) and fit to a one-site specific binding model with Hill coefficient.

25
PD-L1 Dimerization Assays, analytical SEC and TR-FRET Analytical SEC All analytical SEC methods were performed at Viva Biotech (Shanghai, China) on an Agilent Bio-1260 Infinity II HPLC system. Complexes were prepared by mixing PD-L1 and peptide at a 1:2 protein to peptide ratio. Complexes were injected in separate analyses onto a Superdex® Increase 5/150 200pg column pre-equilibrated with 10mM Tris-HCl pH 8.0, 20mM NaCl containing 1 μM of corresponding peptide (for complexes) or no peptide (for apo-protein). Data was processed using Agilent ChemStation software to determine retention time shifts.

TR-FRET
For TR-FRET dimerization experiments, biotinylated mammalian PD-L1 was diluted to 50 nM, Alexa Fluor™ 488 labeled PD-L1 was diluted to 120 nM and Terbium-labeled streptavidin (Cis-Bio) was diluted to 20 nM in assay buffer (10 mM HEPES pH 7.5, 150 mM NaCl, 0.05% Tween20) in a final volume of 40 μ L per well of a black 384-well plate (Costar). Compounds were serially diluted in 90% DMSO and 80 nL of compound (11-point three-fold peptide dilution series with top concentration = 20 M) was added to the plate and the samples were incubated for 60 minutes at room temperature. FRET signal was determined using a LanthaScreen™ filter on a PheraStar (BMG Biotech) plate reader (Ex: 337 nm; Em1: 490 nM; Em2: 520 nM). The ratio of Em520 to Em490 was calculated and plotted against compound concentration. Resulting data was fit to a four-parameter dose-response curve with variable slope.   and an alanine-scanning series across the non-crosslinked residues of the peptide demonstrates the importance of the conserved logo residues for binding (n=2 or 3; data are presented as mean ± SD). n.d., no data.   (20), and FP19711 induces a rotation of the N-terminal lobe. (B) The FP19711-CDK2 co-structure indicates that the N-and C-terminal-most residues of the peptide do not interact with CDK2 (seen also in Fig. S7). Truncation of FP19711 to FP33215 improves its CDK2binding affinity to ~300 nM. (C) By SPR (Biacore), FP19711 and FP24322 bind CDK2. FP19711 and its truncated version, FP33215, also bind the active CDK2 (with T160 phosphorylated) and active CDK2: Cyclin E1 (CCNE1) complex. In addition, FP19711 retains its CDK2-binding affinity in the presence of ATP-analog (AMPPNP) or ATP-competitive inhibitors (Dinaciclib and Zotiraciclib), suggesting it is not ATP-competitive. (D) Compared with ATP-competitive CDK2 inhibitors, FP19711 does not compete with fluorescently labeled ATP analog (BODIPY-ATP-S) for binding CDK2 as assessed by fluorescence polarization. This is consistent with its binding to an allosteric site on CDK2. (E) By SPR (Biacore), Cluster C53 peptide FP29092 and Cluster C54 peptide FP29103 bind to PPIA. (F) Co-crystal structure of Cluster C54 peptide FP29102 Helicon with PPIA, like that of FP29103 (Fig. 5D), shows that it binds a site similar to that of Cyclosporine A and peptide substrates. (G) Helicons FP29092 and FP29103 compete with Cyclosporine A for binding to PPIA as monitored by Surface Plasmon Resonance (SPR, Biacore).

Figure S6. Characterization of PD-L1-binding Helicons (A) SPR binding and competition assays
show that Cluster C62 peptide FP30790 binds to PD-L1 (left) and while it competes for binding to PD-L1 with PD-L1 receptor PD-1, Cluster C61 Helicon FP28312 and the small-molecule inhibitor of the PD-1/PD-L1 interaction, BMS-1001 compete more potently (right). FP28141 is a point mutant of FP28132 with an alanine residue replacing a conserved tryptophan from the phage cluster logo, that does not bind PD-L1 (see S6B). (B) Surface plasmon resonance (SPR, Biacore) shows binding of FP28132 and additional Cluster C61 Helicons FP28135 and FP28136 to PD-L1, while FP28141 does not bind. (C) As with FP28132 (here and in Fig 6B), the co-crystal structure of FP28135 and FP28136 with PD-L1 shows a symmetric dimer of two Helicon/PD-L1 complexes. (D) Close examination of the co-structure reveals an extensive series of contacts between both the two FP28132 protomers and the two PD-L1 protomers. (E) Analytical SEC suggests that the apo-PD-L1 and FP30790-PD-L1 are monomeric, while FP28132-PD-L1 complex is a dimer in solution.  n=2; data are presented as mean ± SD. b. In vitro assay for competition with β-catenin-binding TCF peptide, as in Figure 3. c. Measure of ATP levels, reflective of cell health (see Materials and Methods). d. Full sequences of control peptides can be found in Table S3.