Structure Based Design of Bicyclic Peptide Inhibitors of RbAp48

Abstract The scaffolding protein RbAp48 is part of several epigenetic regulation complexes and is overexpressed in a variety of cancers. In order to develop tool compounds for the study of RbAp48 function, we have developed peptide inhibitors targeting the protein–protein interaction interface between RbAp48 and the scaffold protein MTA1. Based on a MTA1‐derived linear peptide with low micromolar affinity and informed by crystallographic analysis, a bicyclic peptide was developed that inhibits the RbAp48/MTA1 interaction with a very low nanomolar K D value of 8.56 nM, and which showed appreciable stability against cellular proteases. Design included exchange of a polar amide cyclization strategy to hydrophobic aromatic linkers enabling mono‐ and bicyclization by means of cysteine alkylation, which improved affinity by direct interaction of the linkers with a hydrophobic residue on RbAp48. Our results demonstrate that stepwise evolution of a structure‐based design is a suitable strategy for inhibitor development targeting PPIs.


Synthetic methods
Peptides were synthesized according to one of the following methods:

Method A (linear peptide synthesis):
Synthesis was performed on Rink Amide MBHA resin (0.4 mmol/g) using an automated CEM Librety microwave peptide synthesizer. Amino acid coupling reactions were done at 75°C (40 W) for 5 minutes using 5 eq amino acid, 5 eq PyBOP and 10 eq DIPEA in DMF. Fmoc-deprotection was performed by addition of 20% piperidine in DMF and reaction at 75°C (40 W) for 30 seconds followed by addition of fresh reagents and reaction at 75°C (40 W) for 3 minutes. Peptides were cleaved using TFA/ODT/TIPS/H2O (90/2.5/2.5/5) for 2 x 1 h or for 1 h per arginine in the sequence. Peptides were precipitated in cold Et2O followed by centrifugation. The supernatant was removed and the pellet resuspended in fresh cold Et2O, the procedure was repeated once more. The crude product was dissolved in H2O/ACN (1:1) and lyophilized.

Method B (linear peptide synthesis):
Synthesis was performed on Rink Amide MBHA resin (0.61 mmol/g) using an automated Syro II parallel peptide synthesizer. Amino acid coupling reactions were done at rt for 40 minutes using 4 eq amino acid, 4 eq HATU and 8 eq DIPEA in DMF. All amino acids were double coupled. Fmoc-deprotection was performed by addition of 40% piperidine in DMF for 3 min followed by addition of fresh reagents and reaction for 10 minutes. Peptides were cleaved using TFA/TIPS/H2O (90/2.5/2.5) for 2 h. Peptides were precipitated in cold Et2O followed by centrifugation. The crude product was dissolved in H2O/ACN (1:1) and lyophilized.

Method C (cyclization via amide bond formation):
The amino acids to be cyclized were introduced as Fmoc-Glu(All)-OH and Fmoc-Lys(Alloc)-OH. After linear synthesis, using the same method as described for linear peptides, the allyl protecting groups were removed by treating the resin with Pd(PPh3)4 (0.25 eq) and PhSiH3 (25 eq) in dry DCM for 1 h. The liquid was removed and fresh reagents were added and reacted for another hour. The resin was washed with DCM (4x), DMF (2x), 0.5% diethyldithiocarbamic acid trihydrate sodium salt in DMF (4x) and DMF (4x). Next the resin was treated with PyBOP (4 eq) and DIPEA (8 eq) in DMF for 2 h followed by peptide cleavage and purified following the same protocol as for linear peptides.

Method D (Cyclization via cysteine alkylation):
Lyophilized linear peptides were dissolved in 20 mM NH4HCO3/ACN (3:1) and 1.1 eq of the appropriate bromide reagent was added as a solution in a small amount of ACN. After 1 h the reaction mixture was lyophilized and peptides were purified as described below.

Method E (Cyclization via cysteine alkylation):
To a 0.33 mM solution of crude peptide dissolved in a 1:1 solution of H2O/MeCN, the core (1.2 eq) predissolved in a minimum amount of MeCN was added, followed by a solution of 60 mM NH4HCO3 (to afford a final peptide concentration of 0.5 mM). The resulting clear reaction mixture was stirred at rt for 1h, and freeze dried to afford the crude peptide as a white fluffy solid.

Fluorescently labelled peptides
After linear synthesis, as described above, the peptides were manually modified with Fmoc-O2Oc-OH (4 eq), PyBOP (4 eq) and DIPEA (8 eq) in DMF. Next Fmoc-deprotection was done by treating the resin with 20% piperidine in DMF for 5 min and once more for 10 min. Next, the peptides were treated with FITC (4 eq) and DIPEA (8 eq) twice for 60 min. Cleavage and purification was similar as described above.

Peptide purification Method A:
Peptides were purified using a preparative scale C18 column (Macherey-Nagel, 5µM, 125 x 21 mm) at a flow rate of 20 ml/min. The peptides were eluted with a binary mixture of H2O and ACN, both containing 0.1% TFA, using a linear gradient of 5-50% ACN over 50 min.

Method B:
Similar to method A but MeOH was used instead of ACN.

Method C:
Peptides were purified using a preparative scale reverse-phase column (Atlantis T3 OBD column, 5µM, 150 x 19 mm) at a flow rate of 30 ml/min. The peptides were eluted with a binary mixture of buffer A (H2O + 0.15% TFA) and buffer B (CAN). Gradients used were determined by elution profiles obtained from analytical RP-UPLC:

Protein expression and purification
RbAp48 was expressed and purified by the Dortmund Protein Facility. Full length RbAp48 was cloned into a pOPIN vector ligating it to an N-terminal His tag followed by a 3C cleavage site. The protein was expressed from Sf9 cells for 48 h. Cells were harvested and lysed and the protein was purified by loading onto a HisTrap FF crude 5 ml column followed by on-column cleavage using Precission protease. The cleaved protein was then further purified using a HiLoad 26/60 Superdex 200 column.

Fluorescence polarization assay
Fluorescence polarization assays were performed in 384 wells plates in 80 µl volumes. The assay buffer contains 20 mM Tris-HCl, 150 mM NaCl, 0.01% Tween-20 at pH 7.5. The concentration of the fluorescently labelled peptides was 1 nM plates were incubated at room temperature for 30 min before they were read on a Tecan Spark plate reader (λex = 485 nm, λem = 535 nm).
For competitive assays the same conditions were used with a tracer concentration of 1 nM was used and a protein concentration of 15 nM.

Isothermal Titration Calorimetry
Isothermal titration calorimetry measurements were performed using a Microcal ITC-200 device. The buffer for all measurements consisted of 20 mM Tris, 150 mM NaCl, pH 7.5 and all measurements were performed at 25°C in duplicate except for 8. RbAp48 was dialyzed overnight using a 10 kDa MWCO Slide-A-lizer dialysis membrane and peptide samples were dissolved in the dialysis buffer. Direct measurements of 40 were perfomed by titrating the peptide (800 µM) into the cell containing RbAp48 (40 µM) using 2 µl injection after an initial injection of 0.5 µl.
Direct titration of 2 and 33 gave curves which were hard to fit and therefore their affinity was measured by competitive ITC where the protein is premixed with a weak ligand (40). [1] Competitive measurements were performed by titrating 2, 8 or 33 (300 µM) into the cell containing RbAp48 (30 µM) and 40 (416 µM). Other conditions were the same as for the direct measurement described above.
Ki was calculated according [1] using the following formula:

Circular dichroism
Peptides were dissolved at a concentration of 50-100 µM in a buffer containing 10 mM NaH2PO4 at pH 7.5. CD spectra were recorded using a Jasco J-815 CD spectrometer using a wavelength range from 190 -260 nm and spectra were averaged over 5 measurements. Peptides were measured at 25°C at a scanning speed of 50 nm/min and a bandwidth of 1 nm.

Stability assay
Peptide stability was tested in whole cell lysate prepared from MDA-MB-231 cells using the freezethaw method. Peptides were dissolved in lysate (normalized to 5mg/ml protein using PBS) at 600 µM and incubated at 37°C. Samples were taken at different time points and mixed with an equal amount of MeOH containing 0.05 mg/ml ethylparaben as an internal standard. The samples were mixed and kept on ice for 15 minutes before centrifugation at 14000 rpm at 4°C for 10 min. The supernatant was carefully removed and analyzed by HPLC. Peptide peaks were integrated and surface areas compared to blank samples which did not contain lysate. Measurements were performed in duplicate.
The generated images were processed with the CellProfiler package (https://cellprofiler.org/, version 3.0.0) on a computing cluster of the Max Planck Society to extract 1716 cell features (parameters) per microscope site. The data was then further aggregated as medians per well (9 sites -> 1 well), then over the three replicates.
From the total set of 1716 parameters a subset of highly reproducible and robust parameters was determined using the procedure described by Woehrmann et al. [3] in the following way: Two biological repeats of one plate containing reference compounds were analysed. For every parameter, its full profile over each whole plate was calculated. If the profiles from the two repeats showed a similarity >= 0.8 (see below), the parameter was added to the set. This procedure was only performed once and resulted in a set of 579 robust parameters out of the total of 1716 that was used for all further analyses.
To determine the phenotypic fingerprint for each test compound Z-scores were then calculated for each parameter as how many times the Median Absolute Deviation (MAD) of the controls the measured parameter value of a test compound deviates from the Median of the controls: The morphological compound fingerprint is then the list of z-scores of all parameters for one compound.
In addition to the morphological fingerprint, an induction value was determined for each compound as the fraction of significantly changed parameters, in percent: Similarities of morphological fingerprints were calculated from the correlation distances between two fingerprints (https://docs.scipy.org/doc/scipy/reference/generated/scipy.spatial.distance.correlation.html; Similarity = 1 -Correlation Distance) and the compounds with the most similar fingerprints were determined from a set of 3000 reference compounds that was also measured in the assay.
p53 expression level assay U2OS cells were incubated with 10 µM compound 49, 50 or DMSO (DMSO 0.5%) for 24 h followed by lysis for western blotting analysis. p53 expression was monitored using fluorescent antibody. Signal ration between p53 and loading control (alpha tubulin) was measured. Data are mean value of two biological replicates.

Peptide stability in cell lysate
Supplemental    Crystallization and structure determination 5 mg/ml of full length human RBBP4 (residues 1-425, with a GP overhang on the N-terminus) was cocrystallized with a 2-fold molar excess of the macrocyclic peptides by mixing 100nl of the protein/peptide solution with 100nl reservoir at 20°C. Crystals with peptide 8 appeared in the "Classics II" screen condition H7 (Qiagen, Hilden) containing 20%w/v PEG3350 and 0.15M DL-malic acid, pH 7.1. The peptide 33 yielded thin, plate-like crystals in condition G6 of the "Classics II" screen containing 25%w/v PEG3350, 0.2M ammonium acetate and 0.1M Bis-TRIS pH 5.5. Crystals were flash-frozen in liquid nitrogen and data was collected at 100K using a Pilatus 6M detector at the X10SA beamline of SLS in Villigen, Switzerland. All data sets were integrated and scaled using XDS and XSCALE (Kabsch, 2010).
The structures were solved with PHASER (CCP4 suite) in space groups C2 (peptide 8) and P21 (peptide 33) by molecular replacement using chain A of RbAp48 in complex with MTA1 (PDB ID 4PBZ) as template. Both spacegroups have two molecules in the asymmetric unit and were refined using PHENIX with torsional NCS restraints, in case of peptide 8 also with TLS. The structure containing peptide 8 indexes in space group C2221, but refinement in this space group did not converge and showed huge unexplained blobs in the electron density. In contrast, refinement in C2 worked very well and revealed a peculiar swap of one beta strand (residues 160 to 181) between each of the molecules in the asymmetric unit and its symmetry related molecule that is also observed in a structure of RbAp48 in complex with MTA1 (PDB ID 6G16 In both structures, the peptide is involved in crystal contacts, but both xtal structures show the same binding mode of the helical part and most of the backbone as explained in the following, indicating that the binding is not significantly disturbed by crystal packing. In the peptide 33 structure, the largest contact by far (approx. 867Å 2 in monA and 833Å 2 in monB) of the peptide 33 involves the canonical binding site where many related structures (e.g. 4PXY) have an alpha helix bound. The peptide in monomer A has a well defined electron density that shows all three cysteine links, whereas the peptide in monomer B shows a superposition of probably at least two conformations. The side chain of Phe30 in monomer B is not as parallel to the mesitylen moiety as in monomer A and is moved slightly closer to the peptide. The reason for this difference is most likely that the peptide in momomer B has extensive contacts to the N-terminal alpha helix of one and to the 5G loop (that is artificially ordered by crystal contacts) of another symmetry related molecule (440Å 2 in total), whereas the peptide in monomer A has only very few crystal contacts (81Å 2 ).
Compared to the linear peptide of 4PBZ, the triple link introduces some strain that slightly shifts the Cterminus of the peptide, but the main contacts to the helix and the tryptophan are undisturbed. The slightly higher affinity of the linear peptide could arise from additional contacts of the N-terminal residues of the peptide in the 4PBZ structure that are not present in the peptides 8 and 33. Due to the additional link, the C-terminus of the cyclic peptide is pushed slightly farther away from its N-terminus so that it would clash with a symmetry related molecule in the C2 space group. This probably causes the change in space groups for the mesitylene linker peptide compared to the xylene linker.
In the peptide 8 structure, one peptide sits in the middle between the two monomers in the asymmetric unit with 887 Å 2 buried contact area to monomer A and 433 Å 2 to monomer B. The second peptide bound to monomer B buries 852 Å 2 within the main binding site, and 431 Å 2 with a symmetryrelated monomer A molecule. Thus, the secondary binding site that buries the much smaller surface area is most likely a crystal packing artefact and is not seen as canonical binding site in any other structure. The main binding site has a very similar size of the contact surface area compared to the peptide 33 structure, corroborating that the two peptides bind in a similar mode. The swapped beta strand is located on the opposite side of the beta propeller compared to the main peptide binding site, so it does not influence the peptide binding mode.
The structures have been deposited with the PDB database under accession numbers 6ZRC (peptide 8) and 6ZRD (peptide 33).