Structure‐Enabled Discovery of a Stapled Peptide Inhibitor to Target the Oncogenic Transcriptional Repressor TLE1

Abstract TLE1 is an oncogenic transcriptional co‐repressor that exerts its repressive effects through binding of transcription factors. Inhibition of this protein–protein interaction represents a putative cancer target, but no small‐molecule inhibitors have been published for this challenging interface. Herein, the structure‐enabled design and synthesis of a constrained peptide inhibitor of TLE1 is reported. The design features the introduction of a four‐carbon‐atom linker into the peptide epitope found in many TLE1 binding partners. A concise synthetic route to a proof‐of‐concept peptide, cycFWRPW, has been developed. Biophysical testing by isothermal titration calorimetry and thermal shift assays showed that, although the constrained peptide bound potently, it had an approximately five‐fold higher K d than that of the unconstrained peptide. The co‐crystal structure suggested that the reduced affinity was likely to be due to a small shift of one side chain, relative to the otherwise well‐conserved conformation of the acyclic peptide. This work describes a constrained peptide inhibitor that may serve as the basis for improved inhibitors.


Introduction
Transducin-like enhancer (TLE) proteins are transcriptional corepressors that modulate key pathways for developmental and oncogenic signalling, such as the Notcha nd Wnt pathways. The TLE proteins do not bind directly to DNA to exert their repressivee ffecto ng ene transcription;i nstead, they utilise their WDR domains to bind to DNA-bound transcription factors. [1] Given their role in pathways known to be deregulated in many cancers,i ti sn ot surprising that members of the TLE family, particularly TLE1,h ave been implicated in the development and maintenance of malignancies. Elevated levels of TLE1 have been observed in ag rowingl ist of tumours, including cervical, lung and colon carcinomas,a nd TLE1 has been recognised as ap utative oncogene. [2] Given that TLE1 does not bind to DNA directly,a nd that its repressive and potentially oncogenic role relies on the ability of the WDR domain to bind to transcription factors, blockingo ft his interaction has been suggested as ap ossible treatment for cancers with elevatedT LE1 activity. [3] However, to date, no TLE inhibitors have been described in the literature.
The crystal structures of the WDR domainof TLE1 in complexesw ith peptides derived from two different transcription factorb inding partnersh ave been solved;t hus characterising the binding interface in detail. [3] One of these peptides (SMWRPW) shows relatively potent (K d = 1 mm)b inding to TLE1. As discussed in more detail below,t he bioactive conformation of this peptide is characterised by ac ompactly folded core formedb yt he central three amino acids. This compactc ore engages in extensive interactions with the WDR1 domain and positions key amino acid side chainss uch that they can form additional polar and non-polar interactions (see below). [3] Given that this peptide binds with micromolar activitya nd that detailedk nowledge of its binding mode and bioactive conformationa re available, it represents an attractive starting point for the discoveryofT LE inhibitors. Herein, we report apeptidomimetica pproach based on the hypothesis that the compact conformation of this peptidec an be stabilised by ah ydrocarbon linker.
Hydrocarbon-stapled macrocyclic peptidesa re increasingly being explored as drug candidates and chemical probes, particularly for challenging targets, such as protein-protein interactions. [4] Introducing conformationalc onstraint through macrocyclisation hasanumber of benefits. It particularly reduces the entropic penalty upon binding to the target andh as been shown to have the potential to improve cell penetration and metabolic stability. [5] Designing and synthesising constrained macrocyclic peptides still remains af ormidable challenge. [4] Nevertheless, successfulexamples have been reported, particularly for constraining and stabilising a helices, b sheetsand b turns. [4] However,i nt he case of TLE1,t he bound SMWRPW peptide adopts neither at ypical a-helical nor a b-sheet conformation, and constraining the peptide thus required ad ifferent strategy. As described in more detail below,w eh ypothesised that connectingt wo amino acids, the side chain of the first tryptophan and the proline-which wereacriticalp art of the binding epitope-through ah ydrocarbonl inker would stabilise the bioactive conformation.H erein, we report the design and development of ac hemical route to this hydrocarbon-linker-constrained, proof-of-concept peptide. Furthermore, we tested the binding affinity of the constraineda nd correspondinga cyclic peptides and solved the structure of the constrained peptide bound to the WDR domain of TLE1.

Design
The crystal structure of the SMWRPW peptide bound to the WDR domain of TLE1 was obtained by soakinga po TLE1 crystals in as olution of the slightly extended SMWRPWp eptide. [3] The indole moiety of the N-terminal tryptophan (Trp5) andt he central proline (Pro3) of the bound peptide tightly pack against each other to form the core of the binding epitope ( Figure 1). This core engages in extensive hydrophobic interactions with the protein.
The compactc onformation positions side chains and backbone moieties of the peptide such that they are ideally placed to engagei na dditional polar andh ydrophobic interactions. [3] The N-terminal serine residue of the SMWRPW peptide is not resolved, which suggestst hat it is disordered and does not make any specific interactions.
Our strategy to generate ac onstrained macrocyclic inhibitor is illustrated in Figure1:w eh ypothesisedt hat connecting the Ca-atom of the proliner esidue and the N1 nitrogen of the Nterminalt ryptophan with ah ydrocarbon linker would lock the peptidei nt he bioactive conformation. We modelled various linker lengths in MOE by introducing the linker in silico into the bound conformationo ft he peptide( PDB code 2CE9) and minimising the energy of modified peptides in the TLE binding site. The resulting poses were visually inspected for minimal movement of the peptides ide chains and low-energy conformations of the linker.T hese experiments, together with an analysis of synthetic accessibility (see below), suggested an ideal length of four carbon atoms and compound 2 ( Figure 1) as ap romising synthesis target.

Retrosynthesis
Our retrosynthetic analysis is depictedi nS cheme 1. We envisioneds ynthesising the constrained hydrocarbon-stapled peptide 2 from the macrocyclic intermediate 4 through addition of the N-terminal methionine and C-terminal tryptophan through peptide coupling chemistry.F urthermore, we hypothesised that intermediate 4 could be prepared from acyclic tripeptide 5 through ring-closing metathesis (RCM), followedb y concomitant saturation of the double bond and removal of the Cbzp rotecting group under hydrogenation conditions. To preparet he acyclicR CM precursor 5,t wo unnatural amino acids were required:s ubstituted proline 6 and allyl substituted tryptophan 7.
This approacho ffered the advantage of conducting the critical RCM in solution,w hilst all polar groups, particularly the basic arginines ide chain, were fully protected. Furthermore, cyclic intermediate 4 offered the opportunity of late-stage modification of the C-and N-terminal amino acids.

Synthesis and characterisation
The synthesis of proline derivative 6 was described in the literature and we followed the protocols with minor modifications. We next turned our attention to the synthesis of Cbz-protected 1-allyl-l-tryptophan 7 (Scheme 2). We decided to use the Cbz protecting group because it was stable to basic and acidic conditions and because we anticipated that it could easily be removed during hydrogenation of the double bond arising from RCM;t hus making an additional step unnecessary.A tt he start of this work, direct allylation of unprotected l-tryptophan by using either ac opper tetramethylethylenediamine (TMEDA) catalysto rs odium metal had been described. [6] We tested the copper-mediated conditions, but did not observe any conversion. More recently,ateam from Sanofipublished the synthesis of 1-allyl-l-tryptophan protectedwith the tert-butyloxycarbonyl (Boc) group, but this work was not in the public domain when we undertook our work. [7] We hypothesised that selectivea llylation of unprotected tryptophan could be achieved after de- protonating the carboxyl group and the NH indole with two equivalents of as trong base, such as NaH, because under these conditions the deprotonated indole nitrogen represented the strongest nucleophile.P leasingly,r eacting l-tryptophan with 2.5 equivalents of NaH and one equivalent of allyl bromide in DMFg ave the desired mono-allylated product in 40 % yield after HPLC purification. To avoid HPLC purification of the polar,u nprotected amino acid, we decided to attempta llylation ands ubsequent Cbz protection with benzyl chloroformate in ao ne-pot procedure.G ratifyingly,t his procedureg ave the desired, protected amino acid 7 in an acceptable yield of 28 %o ver two steps.
We next prepared the tripeptide RCM precursor 5 by coupling the allylated proline with protected arginine (Scheme3) under known conditions for this prolined erivative. However, we only isolated the cyclised side product 8.T he formation of this side product is likely to be due to steric hindrance of the amine functionality. Gratifyingly,i ncreasing the reaction temperature and concentrations of the reactants to favour biomolecular reaction led tot he desired dipeptide in 48 %y ield.
Next, we attempted removal of the Fmoc protectingg roup from dipeptide 9 (Scheme 3). However,s tandard conditions with piperidinea st he base gave the undesired side product 11 as as ingle diastereomer.
Repeating the reactionw ith one equivalent of piperidine and at al ower temperature( 08C) resulted in am ixture of the unprotected dipeptide ands ide product. Unfortunately,a ll attempts to isolate the unprotected dipeptidea nd to remove piperidine resulted in complete conversion to the diketopiperazine 11 side product. To solve this conundrum, we reasoned that protonation after Fmoc deprotection would lower the nucleophilicity of the free amino group sufficientlyt op revent cyclisation;t hus allowing isolation by evaporation of the solvent. Furthermore, we tested alternative bases,p articularly bases that were not likely to affect the subsequentp eptide coupling step. This approach indeed proved successful and compound HCl-10 was obtaineda sasingle stereoisomer through clean Fmocd eprotection in EtOH by using one equivalent of NaOEt as ab ase. Subsequent protonation of the amine and residual traces of NaOEt through the addition of as olution of HCl in MeOH thwarted formation of the side product upon solvente vaporation (Scheme3). Coupling of the crude product with the HATU derivative of the allyl-substituted tryptophan 7 and DIPEA as ab ase yieldedt he metathesis precursor 5 in 70 %o verall yield.
To our delight, the pivotal RCM proceeded readily by using the Grubbss econd-generation catalyst [8] in the presence of 1,4-benzoquinone [9] to yield the desired product 12 in 83 % yield as a9:1 mixture of trans and cis isomers(Scheme4).
We next investigated concomitant reduction of the double bond andr emoval of the Cbz group by hydrogenation (Scheme 4). Commonly used conditions, such as 10 %p alladium on carbon and hydrogen at atmospheric pressure, left the startingm aterial intact.E levated temperature, addition of acid or increaseo fc atalyst loading did not significantly improve turnover. We next investigated other catalysts and found that the Pearlman catalyst both reduced the double bond andr emoved the Cbz protecting group. Complete conversionr equired one equivalent of Pd(OH) 2 /C and the addition of two equivalents of HCl, butr esulted in ay ield of 76 %o ft he reduced and deprotected intermediate being isolated.
With intermediate 4 in hand, we next performed coupling to Boc-protected methionine. Although this couplingp roceeded readily,w er eproducibly observed a + 16 Da increasei nm olecular weighta fter isolation and purification. Wea ttributed this increaset oo xidation of methionine to the corresponding sulfoxide derivative 14 (Scheme5). This oxidationh as precedent in the literature;h owever,t he degree and rapidness of the reaction is surprising, given that methioninei sf requently incorporatedi nto peptides.
As we discussi nmore detail below,t his methioniner esidue can be replaced in the acyclicp eptide by phenylalanine withoutl oss of activity.W et hus focused our attention on the phenylalanine derivative. Coupling of 4 with Boc-protected phenylalanine proceeded in 68 %y ield after purification (Scheme5).
To complete the synthesis, we hydrolysed the ester by using LiOH in methanol( 86 %y ield) and added the final amino acid by coupling this intermediate onto tryptophan boundt o ac ommerciallya vailable solid support (Scheme 5).
Cleavage of the solid support of 17 and concomitant removal of the remaining two protecting groups provided the desired macrocyclic peptide 18 in 10 %y ield over three steps (Scheme 5).
Despite initial challenges, our synthetic approach enabled us to access 14 mg of the desired, constrainedp eptide. Some of the optimised steps, for example, the one-pot alkylation and protection of tryptophan, as well as the convenient and mild We next investigated the bindingo ft his macrocycle, as well as acyclic MWRPW and FWRPW peptides, to TLE1. We used two orthogonal binding assays, the thermals hifta ssay [10] and isothermal titration calorimetry (ITC), [11] to test binding of 18 and the linear peptides to the TLE1 WD40 domain (TLE1 residues 443-770). The thermal shift data for the three peptides are shown in Figure 2a nd Ta ble 1.
All three peptides showed significant thermal shifts that were indicative of binding to the protein. Interestingly,t he MWRPW peptide, which is derived from the sequence of TLE1 binding partners, shows the smallest thermal increase. The mutant FWRPW peptide causes as ignificantlyl arger thermal shift (9.4 versus 6.3 8C). The cyclic peptide cycFWRPW (18)a t 100 mm shows at hermal shift comparable to that of the corresponding acyclicp eptide (Table 1). However,t he thermal shift decreases when the concentration is furtheri ncreased from 100 to 200 mm.This decreaseislikely to be due to precipitation of the peptidea th igherc oncentrations. Our thermals hift data thus suggested that all three peptides bound to TLE1.
To confirm thesef indings and to explore the enthalpic and entropic contributions to bindingo ft he linear and constrained peptides, we performed ITC experiments.Given conformational restriction, one might expect the constrained peptide to show as maller entropic penalty upon binding. However,a ll three peptides showed potent binding driven by strong enthalpy contributions.   Interestingly,f or each peptidew eo bserved ab iphasic curve. This was initially more pronounced for FWRPW and 18,b ut also recognisable fort he MWRPW peptide( see Figure S1 in the Supporting Information). We repeated the MWRPW titration at slightly higherp rotein and peptidec oncentrations to achieve ah ighere nthalpy signal, and therefore, better resolution of the titration event. Under these conditions, we also observed ac learb iphasic curve (Figure 3). The biphasic curves are indicative of two binding events and we calculated the thermodynamic data for both ( Table 2).
The first phase of the curves corresponded to am olar ratio of approximately 0.2 (that is, 20 %o ft he protein is bound) and the second phase corresponded to an approximatem olarr atio of 0.8;t hus, the overall curve reached saturationa tamolar ratio close to 1. This suggested that only one binding site per molecule of protein waso ccupied by the ligand. Ap ossible explanation for the biphasic curve is that in the binding experiment the protein exists in two conformationsw hich do not rapidly interconvert and show different binding affinities for the peptides. The observation that the molarr atios for the two parts of the biphasic curve correspond to different peptides is in agreement with this hypothesis.
In the following paragraph, we focus the discussion of the ITC resultso nt he second binding event (K d 2, DH2a nd ÀTDS2) for three reasons. The second binding event coversb inding to the large majority of the protein ( % 80 %);t he K d values are in agreement with published values;a nd, finally,d ue to the ex-perimental set up, the relative errors are smaller.H owever, we include data for the first binding event (K d 1, DH1a nd ÀTDS1) and they broadly follow the same trend.
The rank order based on the ITC K d values (K d 2) confirms the rank order from the thermal shift assay described above.T he acyclic FWRPW peptides hows the highest affinity with a K d of 79 nm.I tt hus shows almost 10-foldm ore potent binding than the peptider epresenting the original MWRPW sequence from the TLE1 binding partners. The K d 2v alue for cyclic peptide 18 is 522 nm and thus less potent than the corresponding acyclic peptide, which suggests that introduction of the hydrocarbon linker leads to as mall loss of activity.I nterestingly,b inding of 18 is accompanied by ar educed loss of entropy compared with the acyclic peptide, which is in agreement with the hypothesis that the introductiono faconstraint reduces the entropic penalty (albeit that this reduced entropicl oss is overcompensated for by al arger enthalpic loss, leadingt oahigher K d compared with that of the acyclic FWRPW peptide).
To be able to interprett hese thermodynamic data in light of the binding modes,w es et out to determine the crystal structure of the cyclic peptide 18.B riefly,w eg rew apo crystals of the TLE1 WD40 domain by using slightly modified previously published conditions [3] and succeeded in solving the structure of cyclic peptide 18 bound to TLE1 to 2.18 resolution by soakingw ith a2 .5 mm solution of 18.T he asymmetric unit containedt wo TLE1 monomers and the electron density was evidenti nb oth binding sites. However,t he quality of the elec-  tron density differed in the two independentT LE1 monomers. Chain As howeds trong ligand density and allowed us to model cyclic peptide 18 with full occupancy.T he ligand density in chain Bw as weaker and refineda talower occupancy (0.83). Therefore, we focus the discussion on the peptide bound to chain A. Figure 4d epicts the constrained peptide boundt oT LE1 and an overlay with the published structure of our acyclic design template, SMWRPW( pdb code 2CE9).
Overall, the binding mode of the constrained peptide is almosti dentical to that of the published acyclicp eptidebound structure. The overall root-mean-square deviation (RMSD) between the two structures is 0.55 .T he N-terminal phenylalanine side chain of the constrained peptideo ccupies as imilarp osition to that of the methionine side chain, with the aromatic side chain efficiently packing against the hydrophobic part of Glu 550;t his potentially explains the slightly higher affinity of acyclicF WRPW,c ompared with that of the MWRPW peptide. The most significant difference between the cyclic peptidec onformationa nd the bound SMRWPWc onformation is the linker,w hich appears to causeaslight change in positiono ft he N-terminal tryptophan and could go some way to explain the lower affinity of 18 compared with that of the linear FWRPW peptide. This slight movement mayc reate an unfavourable, modestly repulsive, interaction that outweighs the gain achieved through constraining the peptide. The observation that our cyclic peptides hows ah igher K d value, despite replicating the bioactivec onformation very accurately, underscorest he challenge of designing constrained peptides. Minor differences that are outside the predictivep ower of current structure-based design tools, even if high-resolution crystal structures are available, can have as ignificant effect on bioactivity.

Conclusion
We developed ac oncises ynthetic route to ac onstrained proof-of-concept peptide 18.B iophysical analysisb yI TC and thermals hift assays andX -ray crystallography confirmed that the constrained peptideb ound to the WD40 domain of TLE1. Furthermore, the observation that the constrained peptide shows binding thermodynamicst hat are entropically favoured, relative to the acyclicF WRPW peptide, is in agreement with the hypothesis that rigidifying the peptidelowered the entrop-ic penalty upon binding to the target.H owever, the constrained peptide also showed an approximately six-foldl ower affinity than that of the acyclic peptide. The crystal structure of the constrained peptideb ound to TLE1 suggests that the linker causes some strain in the molecule that may,a tl east partially, explain the lower affinity.T hese observations underscore the known challenge of designing constrained peptides. Our constrained peptide replicated the bioactive conformation very well with an RMSD of 0.55 and yet as light deviation caused as ufficient penalty to compensate for the gain achieved by introducing the constraint.

Experimental Section
Experimental and characterisation details for all new compounds, computational data, ITC data, assays data, crystallographic data, and NMR spectra are provided in the Supporting Information.

Accession codes
Atomic coordinates and structure factors for the crystal structure of TLE1 with constrained peptide 18 can be accessed by using PDB code 5MWJ.