Cyclic peptides can engage a single binding pocket through highly divergent modes

Significance Large DNA-encoded libraries of cyclic peptides are emerging as powerful sources of molecules to tackle challenging drug targets. The structural and functional diversity contained within these libraries is, however, little explored. Here we demonstrate that one such library contains members that use unexpectedly diverse mechanisms to recognize the same surface on the same target proteins with high affinity and specificity. This range of binding modes is much larger than observed in natural ligands of the same proteins, demonstrating the power and versatility of the technology. Our data also reveal opportunities for the development of more sophisticated approaches to achieving specificity when trying to selectively target one member of a family of closely related proteins.

Pellets from cells expressing proteins for SPR were lysed via sonication in a buffer composed of 50 mM Tris pH 8.0, 500 mM NaCl, 20 mM imidazole, 5 mM β-mercaptoethanol (β-ME), 0.1% Triton X-100, 1× cOmplete EDTA-free protease inhibitor, 10 µg/mL DNase I, 10 µg/mL RNase, and 100 µg/mL lysozyme. The lysate was clarified via centrifugation at 18,000 ×g for 30-60 min. The soluble fraction of the clarified lysate was subjected to immobilised nickel ion affinity chromatography using a 1-mL HisTrap column. The bound protein was eluted using a 20-250 mM imidazole gradient elution. The protein containing eluates from the affinity chromatography step were pooled and concentrated to a small volume. The concentrated sample was subjected to size exclusion chromatography (SEC) using a HiLoad 16/600 Superdex 75 column. Protein was eluted from the column using 50 mM Tris pH 8.0, 150 mM NaCl, and 1 mM DTT. Protein containing eluates were pooled and either aliquoted directly or concentrated before aliquoting. Aliquots were snap frozen in liquid nitrogen and stored at -80 ⁰C. Protein purification was analysed by SDS-PAGE and monitoring UV absorbance at 280 nm.
Cell pellets for protein for all other purposes were lysed in a buffer composed of 50 mM Tris pH 7.2, 500 mM NaCl, 5 mM β-mercaptoethanol (β-ME), 0.1% Triton X-100, 1× cOmplete EDTA-free protease inhibitor, 10 µg/mL DNase I, 10 µg/mL RNase, and 100 µg/mL lysozyme via sonication. Lysate was clarified via centrifugation at 18,000 ×g for 30-60 min. The soluble fraction of the clarified lysate was subjected to GSH-affinity chromatography using a 5-mL GSTrap column. Bound protein was step eluted with 50 mM Tris pH 7.2, 150 mM NaCl, 10 mM reduced glutathione, and 5 mM β-ME. Protein containing eluates were pooled and incubated with HRV-3C protease overnight at 4 ⁰C to enable cleavage of the GST-tag. The pooled eluate was concentrated and subjected to size exclusion chromatography using a HiLoad 16/600 Superdex 75 column. The protein was eluted from the column in a buffer comprising 10 mM Tris pH 7.2, 100 mM NaCl, and 1 mM dithiothreitol (DTT). Fractions containing the desired protein were pooled and concentrated before aliquoting. 15 N-labelled protein was concentrated to 5-10 mg/mL and unlabelled protein was concentrated to 10-15 mg/mL. Protein aliquots were snap frozen in liquid nitrogen and stored at -80 ⁰C.

RaPID screening
DNA sequences containing a T7 polymerase binding site, ribosome binding site, variable length randomised peptide coding region containing a fixed central 'ATG' codon for incorporation of the fixed AcK, (Gly-Ser)3, linker, 'TAG' stop codon and sequence for puromycin ligation were constructed by extensions and PCR from oligos purchased from Eurofins Genomics K.K. (Japan) to produce AcK-focused libraries. Library DNA - DNA libraries were transcribed to mRNA using T4 RNA polymerase and mRNA libraries were ligated to a puromycin-PEG-DNA splint using T4 RNA ligase following standard reaction conditions. For the first selection round libraries were mixed in the following proportions: For codon reprogramming, the 3,5-dinitrobenzyl ester of N  -acetyl-lysine was synthesised as previously described and aminoacylated onto tRNA Asn CAU using dFx (2hr, RT, standard aminoacylation conditions) for incorporation into library peptides 1 . Peptides were initiated with N-(chloroacetoxy)-L-tryptophan (ClAc-L-Trp), which was aminoacylated via the cyanomethyl ester onto tRNA fMet CAU using eFx (2hr, RT, standard aminoacylation conditions). Flexizymes were prepared as described previously 2 .
T7g10M_F46 -TAATACGACTCACTATAGGGTTGAACTTTAAGTAGGAGATATATCC CGS3-CH.R22 -TTTCCGCCTACCTACCTAAGAC Following RaPID selections, double indexed libraries (Nextera XT indices) were prepared from recovered library DNA from rounds 3-5 and sequenced on a MiSeq platform (Illumina) using a v3 chip as single 151 cycle reads 3 . Each DNA sequence was converted to a peptide sequence and ranked by total read number (Dataset S1).

Peptide synthesis
Peptides were synthesised as C-terminal amides using standard fluorenylmethyloxycarbonyl (Fmoc)-strategy solid-phase chemistry using a Syro I automated synthesiser (Biotage) and NovaPEG Rink Amide resin (Novabiochem). Couplings were performed with HBTU/HOBt (1:1) and 6 equivalents of each amino acid. Double couplings were performed for arginine residues. Following the final amino acid coupling reaction, the Fmoc group was removed and resin incubated with N-(chloroacetoxy)succinimide (0.2 M in DMF, 1 h, RT). Resin was washed with DMF (5 times) and DCM (5 times) and dried in vacuo.
Crude peptides were purified by reverse-phase high-performance liquid chromatography using a Chromolith Prep column (Merck) on a Prominence LC-20AP system (Shimadzu) (Solvent A: 0.1% TFA in H2O, Solvent B: 0.1% TFA in acetonitrile) to >95% purity as determined by mass spectrometry (Fig. S17). Purified peptides were reconstituted in DMSO and concentrations determined from their absorbance at 280 nm in 5% DMSO using predicted extinction coefficients.

Surface plasmon resonance (SPR)
Measurements were conducted on a T200 or S200 (GE Healthcare) and data analysed using the Biacore Insight Evaluation Software. Experiments were performed at 4 C in single cycle kinetics mode. Biotinylated BET bromodomains were immobilised on a CAP chip (GE Healthcare) with a target density of ~1000-1500 RU. 50 mM HEPES, 150 mM NaCl, 0.05% Tween-20, 0.1% DMSO, pH 7.5 was used as the running buffer. Between cycles the chip was regenerated following the manufacturer's protocol.

X-ray crystallography
Crystallisation of bromodomain-peptide complexes was performed using a sitting-drop vapourdiffusion technique. Purified bromodomains (10-15 mg/mL) were combined with ~1.5 molar equivalents of peptide and incubated on ice for at least 0.5 h. In situations where the bromodomain concentration was diluted below 5 mg/mL after addition of peptide, the bromodomain-peptide mixture was concentrated to bring the bromodomain concentration back to ~10 mg/mL. Initial crystallisation trials were performed using commercial 96-well crystallisation screens. Bromodomain-peptide mixtures were dispensed into MRC two-drop chamber, 96-well crystallisation plates using a Mosquito crystallisation robot and each condition was screened at a 1:1 or 2:1 protein to precipitant ratio (maintaining a final drop volume of 300 nL). In certain cases, crystallisation was performed without pre-mixing of the bromodomain and peptides by separately dispensing 100 nL of protein and peptide directly into the crystallisation plate prior to mixing with the precipitant. Where required, initials hits were optimised by gradient refinement of the original condition, scaling up drop sizes, and microcrystal seeding. All experiments were performed at 18 C. Protein crystals generally took days to weeks to appear. Crystals were frozen by plungefreezing in liquid nitrogen following cryoprotection with 10% glycerol in the mother liquid from which the crystals were grown. X-ray diffraction data were collected from frozen crystals at the Australian Synchrotron using the Macromolecular Crystallography MX1 (bending magnet) and MX2 beamlines (microfocus) at 100 K and a wavelength of 0.9537 Å 4,5 . Data were integrated using XDS and were processed further using the CCP4i suite 6,7 . AIMLESS was used for indexing, scaling, and merging of the data and the initial phases were calculated by the molecular replacement program PhaserMR using existing xray structures of bromodomains as the molecular replacement models (PDB IDs 4UYF for BRD2-BD1, 3ONI for BRD2-BD2, 3S91 for BRD3-BD1, 3S92 for BRD3-BD2, 4LYI for BRD4-BD1, and 5UVV for BRD4-BD2) [8][9][10][11][12] . Manual model building was performed using COOT and refinement was performed by iterative rounds of manual building in COOT followed by refinement using Phenix 8,13 . The quality of the final model was validated with the wwPDB server and submitted to the PDB. Structure diagrams were generated using PyMol. Protein:peptide interfaces were analysed using PDBePISA. The data collection and refinement statistics for all structures described in this study are outlined in Supplementary Tables 2-16.

NMR spectroscopy
NMR samples of BET BDs were prepared at ~50-100 µM concentrations and were titrated with different molar equivalents of peptides. NMR spectra were acquired at 298 K using Bruker Avance III 600-or 800-MHz NMR spectrometers fitted with TCI probe heads and using standard pulse sequences from the Bruker library. TOPSPIN3 (Bruker) and NMRFAM-SPARKY were used for analysis of spectra 14 . Spectra were internally referenced to 10 M 4,4-dimethyl-4-silapentane-1sulfonic acid. Chemical shift perturbation experiments were performed by collecting 15 N-HSQC spectra of 15 N-labelled bromodomains before and after titration of unlabelled peptide into the samples. Interactions were assessed by monitoring the chemical shift perturbations induced by addition of the peptides to the bromodomains.
Chemical shift assignments were made for BRD3-BD1 using the standard triple resonance approach. Chemical Shift Perturbation (CSP) plots were generated using the following approach. A reference HSQC was recorded without the ligand present (ref HSQC), and then the ligand was added and a second HSQC (shift HSQC) recorded. The two HSQCs were peak picked using the APES algorithm in NMRFAM-Sparky to generate peak lists. The distance between each peak on the ref HSQC and all peaks on the shift HSQC was computed as follows: The minimum of the distances computed for each reference peak was taken as the CSP. Although this approach does not ensure every peak is correctly matched, it has the desirable property that the reported CSPs are guaranteed to be real and not an artefact of the algorithm, as the reported CSPs are a minimum of all possible CSP allocations.
The structure of 3.1B was determined using NOESY, TOCSY and DQF-COSY spectra recorded at 5 °C. Spectra were analysed using CCPNMR Analysis and structures were calculated in CYANA3.98.5 15,16 . Cyclization was introduced according to the protocol provided at the CYANA wiki (http://www.cyana.org/wiki/). The topology for the acetyllysine residues was created manually. For structures run with hydrogen bond restraints, the hydrogen bonds introduced were (a) Trp9-HN to Lys2 O and (b) Ile4 HN to AcK7 O. The restraints were added following the CYANA wiki recommendations. The final ensemble was the lowest-energy set of 20 models from 1000 calculated models.