Structural optimization of reversible dibromomaleimide peptide stapling

Abstract Methods to constrain peptides in a bioactive α‐helical conformation for inhibition of protein‐protein interactions represent an ongoing area of investigation in chemical biology. Recently, the first example of a reversible “stapling” methodology was described which exploits native cysteine or homocysteine residues spaced at the i and i + 4 positions in a peptide sequence together with the thiol selective reactivity of dibromomaleimides (a previous study). This manuscript reports on the optimization of the maleimide based constraint, focusing on the kinetics of macrocyclization and the extent to which helicity is promoted with different thiol containing amino acids. The study identified an optimal stapling combination of X 1 = L‐Cys and X 5 = L‐hCys in the context of the model peptide Ac‐X1AAAX5‐NH2, which should prove useful in implementing the dibromomaleimide stapling strategy in peptidomimetic ligand discovery programmes.


| INTRODUCTION
The development of methodology to constrain peptides in a bioactive conformation for the purposes of inhibiting protein-protein interactions represents an area of significant effort. [1][2][3][4][5][6][7][8][9][10][11][12][13] Widely applied to the pre-organization of helical epitopes and popularized as "stapling", the introduction of a constraint between the i and i + 4 residues ( Figure 1)-or to a lesser extent, i + 7 or i + 11 residues-in a peptide sequence has been shown to be effective in biasing peptides towards a helical conformation. [2,[14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] This pre-organization can lead to improved binding potency towards target, increased (proteolytic) stability and increased cell permeability. [1,3,[31][32][33][34][35][36] Despite this progress, there remains a need to develop new synthetic methodology for constraining peptides that is readily implemented by non-specialists, and which can be applied without recourse to sequences bearing specialized amino acids. Key factors that determine the effectiveness with which introduction of a constraint biases the peptide conformation towards an α-helix include: the length and rigidity of tether between the covalently linked amino acids; the amino acid stereochemistry; the presence of stereogenic centres within the tether and, steric, electrostatic or dipolar interactions between functionality in the tether and the peptide sequence. [37][38][39] Our group recently introduced the first reversible stapling methodology, [25] which exploits native cysteine or homocysteine residues spaced at the i and i + 4 positions in a peptide sequence together with the thiol selective reactivity of dibromomaleimides ( Figure 1). [40][41][42][43] Using N-Alkynyl-dibromomaleimides permits further functionalization through "click" chemistry allowing the facile constraint and diversification of readily and/or commercially available thiol containing peptides.
The widely applied hydrocarbon stapling introduced by Verdine that relies on intramolecular ring closing metathesis of judiciously placed Spentenylalanine residues served as inspiration for our original design. To be isoatomic with the hydrocarbon constraint we used homocysteine to generate constrained peptides of BID (a BH3 only effector member of the BCL-2 family) and RNase S peptide to interact with (anti-apoptotic members of) the BCL-2 family and RNase S protein (a widely studied enzyme which cleaves RNA) respectively. Although, the homocysteine variants were generally more effective than the cysteine variants in promoting a helical conformation and conferring enhanced affinity/ function to either sequence, the later were also clearly tolerated. Moreover, cyclization of the cysteine variants proceeded more rapidly. These observations motivated a more detailed sequence-structure analysis which we describe in the current manuscript; we show that whilst the rate of cyclisation is independent of the length of the constraint, the optimal stapling combination in the context of the model peptide Ac-X 1 AAAX 5 -NH 2 employs X 1 = L-Cys and X 5 = L-hCys.

| MATERIALS AND METHODS
All reagents were purchased from either Sigma-Aldrich, Acros Organics or Fluorochem and were used without further purification.
Solvents used for reactions and workups were purchased from Fisher Chemical (HPLC grade). Deionized and milliQ water were obtained from an Elga Water Purification system. All amino acids were N-Fmoc protected and side chains were protected with; Trt (Cys, hCys); Mtt (Lys) and OPip (Asp). The evaporation of solvents was achieved using a Büchi R3 with a Vacubrand CVC3000 vacuum pump and condenser connected to a recirculating cooler system Julabo F1000. LC-MS analyses were conducted on a ThermoScientific Dionex UltiMate 3000 and high-resolution mass spectrometry (HR-MS) data were recorded using electrospray ionization in positive mode (ESI+) with a Bruker MaXis Impact spectrometer. Preparative HPLC experiments were performed using an Agilent 1260 Infinity instrument with a Jupiter Proteo 90 Å 250 × 21.2 mm, 10 μm preparative column. Analytical HPLC experiments were performed using an Agilent 1290 Infinity LC series system equipped with an Ascentis Express Peptide ES-C18 100 × 2.1 mm column, 2.7 μm particle size on a 5%-50% gradient of acetonitrile in water (with 0.1% formic acid) over 10 min.

| Preparation of constrained peptides 1-2
The syntheses of precursor crude acetylated linear peptides (0.25 mmol, 1 eq) was carried out as described above and these were used in the following stapling reaction without further purification.
F I G U R E 1 Schematic structures for an α-helix together with representative constraints for i to i + 4 residues; the widely employed hydrocarbon constraint, the most effective helix inducing lactam constraint, and the dibromomaleimide constraint explored in this study for the absorption coefficient of maleimide [25] and lactam [38,44] stapled peptides were used for the calculation.

| Kinetic analyses
The kinetic mesurements were performed in 96 well-plates using a Percentage helicity (f Helix ) was then calculated for peptides using the observed mean residue ellipticity of peptides at 215 nm and the following equation Finally, the percentage helicity was normalized relative to the percentage helicity of the lactam constrained peptide 3, which was taken to have 100% α-helicity.

| X-Ray analyses
A solution of 2a (0.5 mg/mL) was dissolved in a mixture of water/ methanol (1:1) and following 2 days of slow evaporation at room temperature, the constrained peptide crystallized as yellow needles. X-ray structure determination was carried out at 120 K on an Agilent Super-Nova diffractometer equipped with an Atlas CCD detector and connected to an Oxford Cryostream low temperature device using mirror monochromated Cu Kα radiation (λ = 1.54184 Å) from a Microfocus X-ray source. The structure was solved by direct methods using SHELXS [38] and refined by a full matrix least squares technique based on F 2 using SHELXL2014. [39] The compound crystallized in a triclinic cell and was solved in the P1 space group, with one molecule and two molecules of water in the asymmetric unit. All non-hydrogen atoms were located in the Fourier Map and refined anisotropically. All carbon-bound hydrogen atoms were placed in calculated positions and refined isotropically using a "riding model." All heteroatom bound hydrogen atoms were located in the Fourier Map and refined isotropically (see Supporting Information structure parameters Table S1).

| RESULTS AND DISCUSSION
We selected a pentameric alanine containing sequence as a model for these studies. This sequence has been widely used by Fairlie and coworkers to compare different synthetic constraints. [40] We prepared seven different peptides 1a-g bearing different combinations of D or L-cysteine and L-homocysteine at the i and i + 4 positions. These could readily be reacted with dibromomaleimide to generate the constrained peptides 2a-g (Scheme 1 and Table 1).
To evaluate our prior qualitative observation that BID and RNase S peptide cysteine variants were more amenable to cyclization, [25] we first compared the rate of reaction of suitably protected amino acids cysteine and homocysteine with dibromomaleimide. These analyses were conducted to allow us to dissect out any differences in the reactivity of the amino acid from differences in ring size for the cyclization; cysteine being anticipated to react more quickly due to the proximal α-carbonyl. The reactions were followed by UV-vis with the bromide for thiol substitution leading to a diagnostic spectroscopic change allowing a qualitative analysis of the reactivity. The analyses indicate that the cyclization proceeds significantly faster than that observed in our original study, [25] although the greater length of the peptide is likely to influence the accessible conformations and sidechains of amino acids proximal to the cysteine's will differentially affect the steric environment, and therefore, rate of cyclization, so caution should be exercised in directly comparing this work and the previous study. A mathematical analysis was hampered by competing hydrolysis of the maleimide ring, [47] nonetheless the experiments indicated clearly only minor differences between the two amino acids ( Figure 2A). Next the rates of cyclization of the peptides 1a-d were assessed. These indicated no dependence of the cyclization rate on amino acid stereochemistry, with all sequences generating near identical cyclization rates ( Figure 2B). The rates of cyclization of peptides with one or both cysteine residues replaced with homocysteine 1e-g also showed a very small effect of the length of the created tether on the cyclization kinetics ( Figure 2C).
We then performed circular dichroism spectroscopy in phosphate buffer and phosphate buffer with 50% trifluoroethanol (TFE). To facilitate comparison, we also prepared the previously reported and optimally constrained sequence 3, bearing a lactam bridge between Lys1 and Asp4 as a control; [40] this lactam constraint was previously shown to be the most effective in a comparative study of i, i + 4 constraints leading to the maximum mean residue helicity for a five residue peptide. Our CD analysis indicated that peptide 2e with X 1 = L-Cys and X 5 = L-hCys to be superior to all other combinations in biasing the conformation of the peptides towards an α-helical conformation ( Figure 3 and Table 1). We used a wavelength of 215 nm (to calculate helicities as it has been shown that the minimum for shorter helices shifts to lower wavelength, [48,49] and this was the value used previously for the reference control lactam bridged 3. We also obtained a crystal structure of peptide 2a. The conformation of the peptide in the solid state is of a "flat" extended nature ( Figure 4) with packing mediated by intermacrocycle hydrogenbonding and stacking of the maleimide. The structure is reminiscent of a constrained peptide reported recently by Li and co-workers [50] and of the cyclic peptide nanotubes described originally by Ghadiri and co-workers. [51,52] In the context of the current study, whilst 2a