Peptide‐Directed Binding for the Discovery of Modulators of α‐Helix‐Mediated Protein–Protein Interactions: Proof‐of‐Concept Studies with the Apoptosis Regulator Mcl‐1

Abstract Targeting PPIs with small molecules can be challenging owing to large, hydrophobic binding surfaces. Herein, we describe a strategy that exploits selective α‐helical PPIs, transferring these characteristics to small molecules. The proof of concept is demonstrated with the apoptosis regulator Mcl‐1, commonly exploited by cancers to avoid cell death. Peptide‐directed binding uses few synthetic transformations, requires the production of a small number of compounds, and generates a high percentage of hits. In this example, about 50 % of the small molecules prepared showed an IC50 value of less than 100 μm, and approximately 25 % had IC50 values below 1 μm to Mcl‐1. Compounds show selectivity for Mcl‐1 over other anti‐apoptotic proteins, possess cytotoxicity to cancer cell lines, and induce hallmarks of apoptosis. This approach represents a novel and economic process for the rapid discovery of new α‐helical PPI modulators.


Protein-proteininteractions(PPIs)regulatemanyprocesses
in life,b oth in healthy and disease states, [1] and almost two thirds of protein-protein interfaces have a-helical binding motifs. [2] However,t argeting PPIs can be difficult owing to their large hydrophobic binding surfaces. [3] There are currently three commonly employed approaches to develop modulators of PPIs: [4] fragment screening, [5] computational screening and drug design, [6] and the exploration of peptides and peptidomimetics. [7] However,c omputational design and fragment screening require large libraries of molecules and extensive synthetic work, often resulting in non-selective compounds. [3] Peptides are challenging drug leads because in vivo their efficacyc an be compromised owing to al oss of secondary structure,p oor cellular uptake,a nd susceptibility to proteolysis. [8] Thew ork described here exploits the advantages of the above approaches while limiting their weaknesses.T his approach, termed peptide-directed binding,u tilizes the tight and selective binding of a-helical peptides that govern PPIs as aframework for the discovery of small molecules.Sections of the natural peptide are employed to identify asmall-molecule fragment that emulates the peptide.
Inspiration for this approach was taken from the REPLACE strategy of McInnes and co-workers [9] and the chimeric inhibitors of the 14-3-3/Tau PPI developed by Ottmann and co-workers. [10] Thew ork also exploits the advantages of altering peptidic binders,demonstrated by the groups of Gellman, [11] Fairlie, [12] and Wilson. [13] Thetechnique demonstrated here improves these strategies by substituting up to ten amino acids with one small-molecule fragment and rapidly supplanting the entire peptide chain with as mallmolecule modulator.
sessing reactive terminals displayed no appreciable binding affinity for Mcl-1 (IC 50 > 100 mm in ap reviously reported competitive fluorescence anisotropy (FA) assay), [16c] in accordance with results reported by Colman and co-workers. [17] These reactive terminals were then utilized to perform copper-catalyzed azide-alkyne cycloaddition (CuAAC) reactions to attach small-molecule fragments (R 1 and R 2 ). [18] If binding affinity is restored for these small-molecule/peptide hybrids (AAQLRRIGD À R 1 and R 2 À KVNLRQKLLN), then the attached small-molecule fragment may represent ag ood emulator for that particular section of NoxaB.
As an initial control for this strategy,t he two peptide fragments were clicked together to generate an extended NoxaB peptide,w hich had an IC 50 value of 7.23 AE 0.88 mm, compared to 650 AE 130 nm for NoxaB,h ighlighting the suitability of the approach (Scheme 1B).
To economize peptide-directed binding,covalent docking studies using the Schrodinger Suite were employed to assist in the identification of the small-molecule fragments most likely to mimic asection of the NoxaB peptide.The crystal structure geometry of the NoxaB peptide with Mcl-1 (PDB No. 2NLA) [17] was modified;v irtually,a mino acids 85-93 of NoxaB were removed, and propargylglycine was attached to the C-terminus of amino acids 75-84. Covalent docking studies were performed on ac atalogue of azides,m odelling aHuisgen cycloaddition (see the Supporting Information for details). Ther esults were scored and ranked, [19] and those highly ranked structures that were synthetically and economically viable were chosen for synthesis.Similarly,amino acids 75-84 of the peptide were virtually removed, and azidoacetic acid was attached to the N-terminus of amino acids 85-93.
In this manner, we selected sixty hybrid compounds, which were synthesized and screened for their ability to disrupt the Mcl-1/Noxa PPI. Thepeptide consisting of amino acids 85-93 (KVNLRQKLLN) of NoxaB was prepared by solid-phase peptide synthesis (SPPS), and azidoacetic acid was used to cap the peptide (Scheme 2, middle). Subsequently,v arious alkynes were exposed to the peptide on the resin in the presence of Cu(MeCN) 4 PF 6 and DIPEA in DMF to achieve CuAACr eactions.T he use of an N-coordinated Cu I source was found to be advantageous,asother Cu I sources gave lower yields,p resumably owing to sequestration of the copper catalyst by peptidic coordination. Cleavage from the resin and reverse-phase HPLC provided al ibrary of 35 R 2 -KVNLRQKLLN hybrids.Ananalogous method was applied to prepare 25 AAQLRRIGD-R 1 hybrids with ap ropargylglycine-terminated SPPS resin (Scheme 2, top).
Theability of these 60 hybrids to inhibit the interaction of Mcl-1 and FITC-Noxa was examined in an FA assay;2 3 compounds (30 %) were identified as hits (defined as having an IC 50 < 100 mm). Eight of these hits contained amino acids 75-84, and 13 hits were derived from amino acids 85-93 (Table 1; see also the Supporting Information, Table S1).
Theo rthogonal nature of the reaction enabled the combination of azide and alkyne small-molecule fragments to generate alibrary of small-molecule triazoles that in theory have an increased likelihood of possessing characteristics that emulate the entire NoxaB peptide.T he identified hybrid molecules suggested 104 triazoles for preparation. We selected 35 for synthesis and evaluated them in an FA assay (Scheme 2, bottom). Nineteen (54 %) of the triazole compounds showed an IC 50 < 100 mm,and ten (27 %) compounds displayed an IC 50 < 1 mm (Table 2a nd Table S2).
Then ature of this approach allows for structure-activity relationships (SARs) to be drawn without targeting specific modifications.T he most potent compounds, 16 and 17,b oth possess ah eptyl chain, which provokes concerns about nonspecific hydrophobic events.H owever,n ot all of the synthesized compounds with the heptyl chain demonstrated activity (Table S2). Additionally,t he Fmoc-propargylglycine moiety proved effective in several of the identified binders (18, 20, 21, 23,and 25), but again was not afeature that was sufficient for binding on its own (Table S2). Concerns over the Fmoc protecting group were considered, but as these compounds are primarily meant as chemical probes or potential leads, further development may discover more effective alternatives.I ndeed, the Fmoc group is somewhat reminiscent of structural features present in Souers A-1210477 [16d] and Fesiks2 -indole-acylsulfonamides, [20] which are highly potent and selective Mcl-1 binders.I nterestingly,t he most potent small molecules did not result from the combination of the most potent small-molecule peptide hybrids 2 and 9,asmight be expected. Thec ombination of 2 and 9 demonstrated no appreciable ability to inhibit the interaction of Mcl-1 and FITC-Noxa. Additionally,t he small-molecule fragment of 9 did not appear in any small molecule that inhibited the Mcl-1/ FITC-Noxa interaction. Ap ossible explanation for this phenomenon is that some small-molecule fragments,w hen bound to the peptide fragment, alter the helicity of the peptide,perhaps increasing the binding affinity of the peptide segment or altering the binding site. [21] Further studies on the power of small-molecule/peptide hybrids as PPI modulators are underway.
It has been shown that NoxaB is as elective Mcl-1b inder. [17] Thed eployment of the NoxaB peptide as as caffold for discovering new compounds was envisaged to also impart this selectivity onto the new small-molecule mimics.T oexamine this,weemployed two in vitro FA assays with Bcl-2 and Bcl-x L ,u sing an FITC-tagged Bid peptide as our fluorescent marker. [22] Navitoclax (ABT-263) was employed as apositive control and as an additional indication that the FA assay was performing adequately. [16b] Excitingly, compounds 16-25 and an additional nine compounds (see the Supporting Information) displayed no appreciable binding to Bcl-x L or Bcl-2 in our FA assays (Figure 1; see also the Supporting Information), demonstrating am inimum 20-fold selectivity for Mcl-1. To confirm that this result was not an artefact, all compounds were re-examined in the Mcl-1, Bcl-2, and Bcl-x L assays twice more,w ith the NoxaB peptide and Navitoclax performing as expected. These results also provide some relief about the potential of non-specific hydrophobic events caused by the heptyl chain or the Fmoc group.
Compounds 16-25 were examined to determine if they displayed activity towards pancreatic cancer cells,w hich are known to overexpress members of the Bcl-2 family,including Mcl-1. An MTS assay was employed to examine the ability of the compounds to inhibit cell growth and affect metabolism, ap otential indicator of cell death. Several compounds, notably 18, 20, 21, 23,a nd 24,d isplayed activity towards the pancreatic cancer cell lines BxPC-3, known to overexpress Mcl-1, and MiaPaCa-2, which overexpresses both Mcl-1 and Bcl-2. Interestingly,c ompound 19 only had activity against BxPC-3 cells.T he pancreatic cancer cell line AsPC-1, which does not overexpress Mcl-1, was also evaluated ( Table 2). Tw oc ompounds, 18 and 21,w ere ineffective against AsPC-1 cells at the concentrations evaluated in our assay, which may suggest that they are acting through the inhibition of Mcl-1. Thed ifference in the magnitude of activity in cells compared to the in vitro FA assay is ac ommonly observed phenomenon, and largely due to cell permeability. [16d, 23] Indeed, some of the more potent compounds in the FA assay showed no activity in the cellular assays,s uch as compound 17 (IC 50 = 102 AE 14 nm), suggesting an inability to < 100 [b] 15 < 100 [b] [a] IC 50 values determined by non-linear regression of at least three experiments. Errors are the transformed greater extreme of the standard error. [b] The hybrid compound displayed an IC 50 value of < 100 mm and > 10 mm,a ccurate value not determined.F moc = 9-fluorenylmethylcarbonyl. cross the cell membrane.Additionally,subnanomolar binding affinities are often required for small molecules to compete with high-affinity endogenous ligands. [24] Compounds 18, 20, 21,a nd 23 were highlighted by this assay,a nd selected for further examination.
To determine if these compounds are impacting on the intrinsic apoptosis pathway,aswould be expected if they are binding to Mcl-1 in cells, [25] assays were performed that demonstrate induction of the apoptosis pathway.Compounds 18, 20, 21,and 23 induced an increase in caspase-3 activation in BxPC-3 cells 4h after treatment, as indicated by the cleavage of DEVD-pNa and asubsequent increase in optical density at 405 nm. [26] (Figure 2A). Additionally,c ompounds 18, 20, 21,a nd 23 were shown to induce the externalization of phosphatidylserine on the cell surface, [26] as demonstrated by the binding of annexin-V-FLUOS to BxPC-3 cells,resulting in green fluorescence ( Figure 2B).
In conclusion, by utilizing the selective NoxaB peptide as aframework, alibrary of novel Mcl-1 binders that demonstrate selectivity for Mcl-1 has been prepared. Peptidedirected binding provided ah igh percentage of compounds with increased potencyw hen compared to traditional methods of highthroughput screening,b yt he application of an atural peptide framework and simple synthetic manipulations.R ecent literature demonstrates that fragment-based methods require the screening of approximately 15 000 fragments by NMR spectroscopy and extensive subsequent synthetic manipulations to generate as elective potent Mcl-1b inder. [23] Additionally,h ighthroughput screening has been shown to have ah it rate of 0.2 % in recent studies targeting the Bcl-2 family and other prominent cancer PPIs. [27] Them ethod exemplified here represents as ignificant economic improvement, in terms of both cost and time,when compared to both high-throughput screening and fragment-based methods and is ap owerful new approach towards discovering modulators of a-helical PPIs.Asubset of the identified in vitro binders was found to possess activity towards cancer cell lines that overexpress Mcl-1a nd induce hallmarks of the apoptosis pathway.I ti s important to note that these compounds are currently unoptimized but still achieved low micromolar cellular activity,e xemplifying the power of peptide-directed binding to swiftly generate potential selective leads for challenging targets.
This proof-of-concept study has demonstrated that peptide-directed binding is atechnique that rapidly identifies new leads for a-helical protein-protein interactions,aseffectively exemplified for the Mcl-1/Noxa PPI. Importantly,t hese compounds are able to mimic the selectivity of the natural scaffold. Ar ecent review on Mcl-1 inhibitors highlights that less than thirty compounds have been reported with comparable selectivity for Mcl-1. [28] Further studies are underway to structurally confirm the binding sites of the hybrid and small-molecule compounds generated from peptide-directed binding.I ti se xpected that peptide-directed binding is applicable to other a-helical PPIs,s uch as the p53/hDM2 interaction, the Bcl-x L /BIM interaction, or the HIV gp41 hexameric coiled-coil fusion complex. [29]