An α‐Helix‐Mimicking 12,13‐Helix: Designed α/β/γ‐Foldamers as Selective Inhibitors of Protein–Protein Interactions

Abstract A major current challenge in bioorganic chemistry is the identification of effective mimics of protein secondary structures that act as inhibitors of protein–protein interactions (PPIs). In this work, trans‐2‐aminocyclobutanecarboxylic acid (tACBC) was used as the key β‐amino acid component in the design of α/β/γ‐peptides to structurally mimic a native α‐helix. Suitably functionalized α/β/γ‐peptides assume an α‐helix‐mimicking 12,13‐helix conformation in solution, exhibit enhanced proteolytic stability in comparison to the wild‐type α‐peptide parent sequence from which they are derived, and act as selective inhibitors of the p53/hDM2 interaction.

Amongst am ultitude of foldamer classes where structural/ conformational determinants have been mapped, [1][2][3][4] bpeptides and hybrid a/b-peptides,inwhich b-amino acids are dispersed along an a-peptide backbone,c an inhibit a-helixmediated protein-protein interactions [22][23][24][25][26][27][28][29] and mimic the structure and the function of protein surfaces. [30,31] Nonetheless foldamers that more accurately mimic the topology and topography of the a-helix might prove advantageous in comparison to b-and a/b-peptides,which may not fully mimic the spatial presentation of a-helix side chains.S everal foldamer scaffolds have been hypothesized to have potential for the inhibition of a-helix-mediated PPIs, [32][33][34][35] but they have not yet been shown to do so experimentally. b/g-Peptide sequences fall into this category:adipeptide of b-a nd gresidues forming a13-membered hydrogen-bonded ring (C = O(i)-NH(i + 3)) is analogous to atripeptide of a-amino acids forming the 13-membered hydrogen-bonded ring (C = O(i)-NH(i + 4)) of the native a-helix. The1 3-helix represents am ore accurate topographical mimic of the natural a(4 13 )helix and represents an attractive template on which to elaborate inhibitors of protein-protein interactions.W hilst both the Gellman and Balaram groups have previously demonstrated that the introduction of b and g residues is tolerated within sequences of a-amino acids,which retain the secondary structure of the a-helix, [36][37][38] the approach described herein is quite distinct in that anovel-fold is designed in abottom-up manner to mimic the topology and side-chain presentation of an a-helix.
We have recently demonstrated that an alternating sequence of the b-amino acid trans-2-aminocyclobutanecarboxylic acid (tACBC) and g-amino acids can adopt a9 /8ribbon [39] or arobust 13-helix [40] in solution, depending on the absence or presence of branching within the g-amino acid monomer.W et herefore examined the ability of b/g-peptide manifolds to behave as a-helix mimetics by designing mimetics of the N-terminal helical domain (residues 19-26) of the transcription factor p53 ( Figure 1). b/g-Peptide mimetics were designed to display three known hot-spot residues of p53 at the correct positions:P he (i), Tr p( i + 4), and Leu (i + 7). Thep rimary sequence of p53 19-26 and a b/g-peptide backbone with alternating tACBC and g-amino acids were aligned in order to map appropriately positioned side chains. Although Ser20 from p53 appears to align well with the gresidue at position 2inthe a/b/g-peptides,itisnot ahot-spot residue and hence for this first generation, g 4 -Ala was used in this position to ensure that ah elical conformation would be promoted. [40] Twos eries of four a/b/g-hexapeptides (1-8; Figure 1) were proposed: N-Boc-protected (1-4)a nd N-acetamide (5-8)p eptides.B oth series featured g 4 -Leu and Leu at the Cterminus,s ince only the amino acid side chain was aprerequisite,aswell as g 4 -Trp and g 4 -Phe,since previous studies have shown that the replacement of indole with phenyl in the Tr p21-mimicking position does not necessarily alter the affinity of peptidomimetics for hDM2. [18] Peptides 1-8 were prepared by using standard solution-state methods (see the Supporting Information), and conformational analysis was performed by using solution-state spectroscopic techniques and molecular modelling.
The 1 HNMR spectra of the N-Boc-protected peptides 1-4 in CDCl 3 were well-defined and the signals were conveniently dispersed, thus allowing complete residue assignment and unambiguous attribution of all signals pertinent for conformational analysis by using standard 1D and 2D NMR sequences. ROESY experiments revealed correlations between nonadjacent residues ( Figure 2a); similar ROEc orrelation patterns were observed for all four N-Boc-protected a/b/gpeptides.I nb/g-segments,R OEs characteristic of 13-membered hydrogen-bonded rings were detected:H b(i)-NH(i + 2), Hb(i)-Ha(i + 2), Hg(i)-NH(i + 2). [40] In N-terminal a/g/bsegments,r elated ROEs were detected:H a(i)-NH(i + 2), Ha(i)-Ha(i + 2). [40] These latter correlations are indicative of a12-membered hydrogen-bonded ring in this segment of the peptide,w hich was corroborated by the down-field chemical shift values and the high [D 6 ]DMSO titration coefficients observed for the amide NH signals from Phe1 and g 4 -Ala2 in the 1 HNMR spectra (see the Supporting Information for more details). These data fully support the proposal that the a/b/g-peptides 1-4 predominantly adopt awell-defined folded conformation containing one C12 and three C13 features (Figure 2b), which we refer to as a1 2,13-helix.
Solution-state IR absorption spectra of the N-Boc-protected peptides 1-4 recorded in CDCl 3 (Figure 2c)f urther support these conclusions.I na ll cases,i na ddition to af ree NH absorption (around 3425 cm À1 ), as trong, low-frequency amide NH absorption band was observed (around 3325 cm À1 ), as would be expected for 12-and 13-membered H-bonded features.Afree indole NH absorption band was also observed (around 3480 cm À1 )f or peptides 1 and 2.
Thel ower solubility of peptides 5-8 in aprotic solvents (< 1mm in CDCl 3 )precluded similar NMR and IR studies for these compounds.However,the far-UV CD spectra of all a/b/ g-peptides 1-8 were recorded in 0.2 mm MeOH solution and each showed am arked Cotton effect, presenting am inimum around 206 nm and am aximum around 224 nm ( Figure 2d). These data compare closely with the methanol-solution signatures of both the 13-helix adopted by b/g-peptides and the 12-helix adopted by b-peptides, [40][41][42] thus suggesting that a/b/g-peptides 1-8 adopt asimilar folded conformation in the same solvent. Collectively,t he NMR, IR, and CD data provide strong evidence that a/b/g-peptides 1-8 are capable of adopting ah ydrogen-bonded helical conformation in hydrogen-bonding and non-hydrogen-bonding solvents.
Ah ybrid Monte Carlo multiple minima (MCMM) molecular mechanics conformational search [43] was carried out on a/b/g-peptides using MacroModel 10.6 and the MMFFs force field without restraints;inchloroform, octanol, or water for peptides 1-4,and in octanol or water for peptides 5-8.T he conformational landscapes were largely dominated by aw ell-defined 12,13-helix (relative abundance > 67 %i n chloroform, 100 %inoctanol) comprised of the C12 and C13  features as deduced from the spectroscopic analyses.A sw e anticipated, asolvent with ahigher dielectric constant (water) does not significantly change the conformational landscape of the a/b/g-peptides;s ome fraying at the N-terminus is observed, which reduces the population of 12,13-and 13helical conformers to the range 47-79 %. However,t he key central residues are essentially locked in a1 3-helical conformation (see the Supporting Information). Peptides 1-8 were subjected to ab initio geometry optimization by DFT using Gaussian09 at the B3LYP/6-311G(d,p) level of theory. Thel owest-energy structures of 1-4 and 5-8 were superimposed (Figure 3a,b) and the backbone of peptide 2 (as ar epresentative example) was overlaid with the crystal structure of p53 [16][17][18][19][20][21][22][23][24][25][26][27][28][29] (Figure 3c)u sing g 4 -amino acid C(a) atoms as the basis of the alignment. Thesuperimposition gave an excellent RMSD value of around 0.9 (Figure 3d,e), thus strongly suggesting that the a/b/g-peptides 1-8 effectively mimic an a-helix. Gratifyingly,s elected side chains can be accommodated in the required positions to mimic those of an ative a-peptide without affecting the ability to adopt ah elix, which suggests that a/b/g-peptides might have wider use as a-helix-mimetic scaffolds.

Angewandte Chemie
Communications ment of fluorescein-labelledp 53 (FITC-p53 [15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31] )f rom the binding groove of hDM2 is monitored upon titration with the competitor peptide,with the objective of determining ahalfmaximal inhibitory concentrating (IC 50 )v alue (Figure 4b,c). Peptides 3, 4, 5,and 7 were found to bind to hDM2 but were not sufficiently potent ligands to generate full competition curves (IC 50 > 100 mm). Peptides 6 and 8 effectively displaced FITC-p53 15-31 at micromolar concentrations,a lthough the lower asymptote of the competition curve was not achieved for higher concentrations.P eptides 1 and 2 generated full competition curves with good potency: IC 50 values were calculated as 57 AE 22 mm for 1 and 15 AE 2 mm for 2.I ti s noteworthy that the inhibitory potency of 1 and 2 for hDM2 is only one order of magnitude lower than that of the native sequence p53 15-31 (IC 50 = 1.2 AE 0.1 mm)a nd that of nutlin, [44] aw ell-known inhibitor of p53/hDM2 (IC 50 = 0.434 AE 0.024 mm), and is superior to that previously reported for first-generation b-peptides. [27] Thed ifference in behavior between the peptide pair 1 and 2 and the peptide pair 3 and 4 indicates that g 4 -Trp plays as ignificant role in binding to hDM2, as is indeed the case for Trp21 in the native p53 protein. Importantly,i nf urther FA assays conducted for the BODIPY-BAK/Bcl-x L and the FITC-NOXA B/Mcl-1 protein-protein interactions,p eptides 1 and 2 displayed no inhibitory activity,thus indicating that their binding to hDM2 is selective (see the Supporting Information).
Peptide 2 was assessed for its ability to bind to 15 Nlabelled hDM2 at the native p53 protein binding cleft. Peptide 8 was also tested as an egative control for comparison purposes. 1 H-15 NHSQC spectra were recorded in the absence and presence of the designated peptides (Figure 4d for 2;see the Supporting Information for 8). Upon addition of 2,crosspeaks in the 1 H-15 NH SQC spectrum shifted throughout the protein, thus indicating ad irect interaction with hDM2, whereas no significant shifts were observed upon addition of 8.T he chemical shifts were mapped onto the structure of hDM2 by using ap ublished NMR assignment. [45] Thes hift changes were comparable to those induced by p53 [15][16][17][18][19][20][21][22][23][24][25][26][27][28][29] peptide in its interaction with hDM2. [17,45,46] Significant diagnostic changes,c haracterized by strong peak shifts,w ere observed for the amide NH of Phe55 and His73, which are located at opposite edges of the hydrophobic cleft. This supports the hypothesis that 2 binds to hDM2 in the canonical p53 binding site.
In conclusion, a/b/g-peptides constructed from tACBC and g 4 -amino acids are able to fold into 12,13-helices to effectively mimic the a-helix. Ad emonstrated advantage of the a/b/g-peptide motif compared to native a-peptides is resistance to proteolytic degradation. Moreover,the results of the FA assays establish for the first time that a/b/g-peptides can act as functional and selective a-helix-mimetic inhibitors of the p53/hDM2 interaction. Thecurrent design rationale has focused on accurately reproducing the spatial presentation of the three key p53 hot-spot residues (Phe-Trp-Leu);w e anticipate that further optimization studies might increase the potencyo ft hese promising a/b/g-peptide scaffolds through the incorporation of additional side chains other than the key triad and by using affinity-improvement design features identified in previous work. [47][48][49] Moreover,s ubse-quent studies will focus on the use of these foldamer manifolds for the programmable bottom-up design of mimetics of the native a-helix to target other PPIs.