Dithiocarbamate-inspired side chain stapling chemistry for peptide drug design

A novel peptide stapling strategy based on the dithiocarbamate chemistry linking the side chains of residues Lys(i) and Cys(i + 4) of unprotected peptides is developed.

Despite its success in peptide drug design, hydrocarbon stapling can be technically cumbersome and costly due to the use of conformationally constrained unnatural amino acids and required transition metal carbene complexes as catalysts for olen metathesis. Additionally, owing to an introduction of severely hydrophobic hydrocarbon stapling, another potential issue of this strategy is the problem of poor aqueous solubility, especially in those cases where the native hydrophilic side chains of Ser, Lys or Arg have to be sacriced. To tackle these problems, we developed a novel peptide stapling strategy by crosslinking the side chains of Lys and Cys at (i, i + 4) positions via a thiocarbonyl group to form the dithiocarbamate (DTC) structure -NH-C(]S)-S-.
Although this work focused on PMI and its derivatives, the DTC stapling chemistry is expected to be applicable to other peptide systems as well. The transactivation domain (TAD) of p53, a peptide of 12-15 amino acid residues, has been extensively studied for its interaction with MDM2 and MDMX. 46,53 We mutated Ser20 to Cys of a TAD peptide of p53, i.e., 16À27 p53 (QETFSDLWKLLP), and stapled it through a DTC linkage between Cys20 and Lys24 (Fig. S3 †). Importantly, when Lys24 was replaced by ornithine, diaminobutyric acid or diaminopropionic acid, the DTC staple failed to form under otherwise identical experimental conditions, suggesting that the side chains of Cys and Lys (or Lys and Cys) at (i, i + 4) positions are optimally paired geometrically for the DTC chemistry.
To furthermore demonstrate the regio-selectivity of the DTC chemistry, we showed with the PMI-derived peptide Ac- TSFAEKWCLLSK-NH 2 , where Cys and two Lys residues are present in the same sequence. The question we asked was: can Cys form two competing DTC staples with the two Lys residues in the same sequence, at (i, i + 4) and (i, i + 2) positions? We recovered only one predominant reaction product containing a DTC staple (Fig. 1c), however. Aer HPLC purication, we subjected the product to tryptic digestion and mass spec analysis, and the data unambiguously demonstrated that the DTC staple had formed between Cys and Lys at (i, i + 4) positions, but not at (i, i + 2) positions (Fig. S4 †).
It is worth noting that formation of the DTC crosslink between Lys and Cys side chains appears stereo-selective despite that Michael addition of Lys-NH-C(]S)S À (product of the reaction between the amino group -NH 2 and CS 2 ) to dehydro-alanine could in theory yield two epimeric compounds (L-Cys and D-Cys) in equal quantities. In reality, however, one predominant isomer was identied and puried by HPLC for subsequent characterization ( Fig. 1c and S2 †), while a very minor isomer of an identical molecular mass was chromatographically resolved but discarded. To ascertain the purity of DTC-stapled peptides, we analyzed PMI(4,8)-a and PMI(8,12)a on HPLC at different gradients. Both PMI(4,8)-a and PMI(8,12)-a, along with the wild type control peptide PMI-0, eluted as single and symmetric peaks at 30-60% and 35-45% acetonitrile over 30 min (Fig. S2 †). While the stereo-and regioselectivity of the DTC chemistry appears to be well-maintained in our study, a more rigorous examination of various reaction conditions and careful analysis of desired/undesired products is obviously warranted in the future to better understand the applicability of this stapling technique for peptide drug design.
We next evaluated the inuence of DTC staple on binding affinities of peptides with target proteins. We quantied the interactions of DTC-stapled PMI peptides with the p53-binding domains of MDM2 and MDMX using uorescence polarization (FP) and surface plasmon resonance (SPR) techniques as described, 46,47,[55][56][57] and the K i and K d values are tabulated in Table 1. In the FP-based competitive binding assay, stapled peptide at increasing concentrations competed off a uorescently tagged PMI peptide (10 nM) complexed with synthetic 25À109 MDM2/ 24À108 MDMX (50 nM), resulting in a progressive decrease in FP. The equilibrium inhibition constant, K i , of stapled peptide for MDM2/MDMX was calculated as described. 54 For SPR-based direct binding, different concentrations of stapled peptide were incubated with MDM2 at 50 nM or MDMX at 100 nM, unless indicated otherwise, and free MDM2/ MDMX was quantied on a 15À29 p53-immobilized CM5 sensor chip to obtain the equilibrium dissociation constant, K d , through non-linear regression analysis. Compared with the Nacetylated and C-amidated wild-type peptide PMI-0, PMI(4,8)a and PMI(8,12)-a bound more strongly to MDM2 and MDMX. In fact, the crosslinked Lys-Cys pair at positions (4,8) enhanced peptide binding to both proteins by one order of magnitude as measured (Fig. 2a-d). Not surprisingly, both PMI(4,8)-a and PMI(8,12)-a partially adopted an a-helical structure in aqueous solution according to CD analyses (Table 1 and Fig. 2e), suggesting that crosslinking Lys-Cys side chains stabilized peptide conformation productive for MDM2 and MDMX binding.
Similarly, the stapled p53 peptide bound to MDM2 and MDMX roughly one order of magnitude stronger than 16À27 p53 (Table 1 and Fig. S3 †). Of note, the reversal of Lys-Cys (a) to Cys-Lys (b) in PMI was in general detrimental to peptide binding to MDM2 and MDMX (Table 1), indicating that the DTC crosslink is functionally unidirectional.
To structurally validate the DTC stapling chemistry, we solved the co-crystal structures of MDM2-PMI(8,12)-a and MDMX-PMI(4,8)-a at 1.8 and 2.7Å resolution (Table S2 †), respectively, and compared them with the structures of MDM2 and MDMX in complex with PMI ( Fig. 3a and b). 47 Both complexes crystallized with multiple copies in the asymmetric unit of the crystal -12 for MDM2-PMI(8,12)-a and 8 for MDMX-PMI(4,8)-a (Table S2 and Fig. S6 †). Whereas all 12 residues could be built into each PMI(8,12)-a peptide complexed with MDM2, PMI(4,8)-a was fully dened in only 3 copies of the MDMX complex with no density observed for Ser11 and/or Pro12 ( Fig. 3c and d). Alignment analysis of the PMI(8,12)a conformation also indicated noticeable variability among the 12 copies of peptide, as evidenced by the root-mean-square deviation (RMSD) between the main-chain atoms in the range of 0.48-1.35Å (Table S3 †). In both complexes, however, the crystallographic density for all atoms of the crosslink formed between Lys(i) and Cys(i + 4) unambiguously dened the geometry of the DTC staple.
As shown in Fig. 3a, MDM2-bound PMI(8,12)-a largely overlapped with PMI, differing mainly in positions of the equivalent C a atoms of residues Thr1-Trp7 with little change in the Cterminal region (Trp7-Ser11) (Table S3 †). More pronounced differences were observed between MDMX-bound PMI(4,8)a and PMI (Table S4 †), with the backbone of the former longitudinally shiing $2Å toward one side of the p53-binding pocket of MDMX and closer to its a2-helix in relation to PMI (Fig. 3b). This shi, while increasing PMI(4,8)-a contacts with the edge of the cavity formed by the a2-helix of MDMX, reduced hydrophobic contacts and lengthened some hydrogen bonds seen in the PMI-MDMX complex (Fig. S7 †). The DTC staple rigidied, at positions (8,12), the C-terminus of PMI in a helical conformation and extended, at positions (4,8), the C-terminal helix of PMI from Leu9 to Ser11 ( Fig. 3a and b). The rigidity of PMI(8,12)-a or PMI(4,8)-a increased to such an extent that the local buried surface area (BSA) slightly decreased as compared with the BSA contributed by PMI to its interface with MDM2/ MDMX (Fig. S8 †). This nding suggests that DTC staplingenhanced binding may be energetically attributable to a reduced loss in entropy afforded by a pre-organized stable helix.
We deduced the DTC structure of the predominant epimer from the crystal structures of PMI(4,8)-a and PMI(8,12)-a in respective complex with MDMX and MDM2, where Cys8 or Cys12 remained as an L-amino acid residue as shown in the electron density maps (Fig. 3e and f). Our biochemical and biophysical ndings on the DTC-stapled peptides unambiguously demonstrated their purity and stereo-selectivity for L-Cys, though.
Side chain stapled peptides are structurally rigidied as compared with their linear counterparts and, thus, expected to be more resistant to proteolysis in vivo. We used HPLC and ESI-MS to evaluate the proteolytic stability of PMI(8,12)-a versus PMI-0 at 100 mM in cell culture medium in the presence of 25 mg ml À1 cathepsin Gan intracellular protease with dual speci-cities for both basic and bulky hydrophobic residues. 58 As shown in Fig. S9, † while PMI-0 was fully degraded by the enzyme within 30 min of co-incubation at room temperature, the DTC-stapled peptide was substantially more stable with a half-life of $8 h under identical conditions. Similar results were obtained using human serum (Fig. S9 †). Of note, the DTC structure is also stable in the presence of reduced glutathione (GST). When PMI(8,12)-a was incubated at 25 C in PBS buffer with GST at 10 mMa physiological concentration, 59 no apparent breakdown of the DTC structure was observed over 24 h (Fig. S9 †).  Verdine and colleagues have shown that structurally permissible stapling of a p53 peptide, while enhancing a-helicity and improving MDM2 binding, is not sufficient to endow the peptide with an ability to kill tumor cells. 22 Although cationicity is not a universal molecular signature of cell-penetrating peptides, it plays a critical role in the ability of stapled peptides to traverse the cell membrane to exert biological activity. 10,22,28 Perhaps not surprisingly, our DTC-stapled peptides carrying a net charge of either 0 or À1 showed little cytotoxicity against HCT116 p53 +/+ and HCT116 p53 À/À cells at up to 100 mM ( Fig. S10 †). Using PMI(4,8)-a as a template, we made two cationicity-enhancing mutations, E5Q and P12R, resulting in a DTC-stapled peptide termed DTC PMI with a +1 net charge ( Fig. 4a and b). Confocal microscopic analysis of HCT116 cells treated with 20 mM DTC PMI N-terminally conjugated to uorescein (FITC) revealed a diffused intracellular localization of the peptide (Fig. S11 †), conrming the ability of DTC PMI to permeabilize the cell membrane.
Compared with its unstapled control peptide, Ac-TSFKQYWCLLSR-NH 2 , DTC crosslinking increased peptide binding affinity for MDM2 and MDMX by 50-fold as measured by SPR (Fig. 4c, d and Table 1) or $20-fold by FP (Fig. 4e, f and Table 1), making DTC PMI (K d ¼ 0.87 and 3.9 nM for MDM2 and MDMX, respectively) a strong dual-specicity peptide antagonist against both proteins. 46,47,60 Of note, DTC PMI also displayed a strong tendency to adopt a-helix on its own in aqueous solution (Table 1 and Fig. 4g), likely contributing energetically to its high-affinity binding to both MDM2 and MDMX. As is the case with DTC PMI, PMI(4,8)-a and PMI(8,12)-a, while staplingenhanced a-helicity qualitatively predicts strong peptide binding to MDM2/MDMX, a quantitative correlation appears lacking, due, in part, to the deciency of CD spectroscopy in accurate measurements of a-helicity of small peptides that are generally disordered and conformationally heterogeneous.
To functionally validate DTC PMI, we subjected it and its unstapled control to a cell viability assay using HCT116 p53 +/+ and p53 À/À cells. Lane and colleagues previously reported that serum proteins were inhibitory against the tumor-killing activity of hydrocarbon-stapled peptide antagonists of MDM2. 27 To mitigate the potential effect of serum binding on peptide activity, we treated cells in serum-free media for 8 h, followed by addition of serum supplements and incubation for 64 h. While the control peptide exhibited no anti-proliferative activity against both cell lines at concentrations of up to 50 mM (Fig. S12 †), DTC PMI displayed p53-dependent growth inhibitory activity against HCT116 p53 +/+ , but not HCT116 p53 À/À , with an IC 50 value of $25 mM at 72 h ( Fig. 4h and S13 †). To investigate the mechanisms of killing of HCT116 p53 +/+ by DTC PMI, we analyzed the expression of MDM2, p53 and p21 by western blotting. As shown in Fig. 4i and S14, † 8 h aer treatment with DTC PMI, dose-dependent induction of p53, MDM2 and p21 became evident in HCT116 p53 +/+ cells. Consistent with this result, dose-dependent induction of apoptosis of HCT116 p53 +/+ cells by DTC PMI was veried by uorescence-activated cell This journal is © The Royal Society of Chemistry 2019 sorting (FACS) (Fig. 4j, k and S15 †). By contrast, no obvious apoptosis of HCT116 p53 À/À cells was observed by FACS under identical treatment conditions (Fig. S16 †). Taken together, these ndings support that DTC PMI actively traversed the cell membrane and killed tumor cells by antagonizing MDM2 to reactivate the p53 pathway. It is worth pointing out that as is oen the case with other stapled peptide activators of p53, 22,26 DTC PMI, despite its low nano-molar binding affinity for MDM2 and MDMX, is rather weak in killing HCT116 p53 +/+ cells. The weak in vitro activity implies that stapling alone is insufficient to achieve optimal therapeutic efficacy of helical peptides, dictated by cell internalization, endosomal escape, proteolytic stability, spatio-temporal distribution, etc.
Of note, at the high concentration of 100 mM, DTC PMI signicantly reduced cell viability of HCT116 p53 À/À cells as well (Fig. 4h). This nding is not entirely surprising in light of the fact that the MDM2 antagonist Nutlin-3 also kills HCT116 p53 À/À at high concentrations, in part by disrupting MDM2 interactions with p73, 61 a member of the p53 family that transcriptionally induces cell-cycle arrest and/or apoptosis. 62 In fact, recent data demonstrate that p73 is elevated to compensate for p53 loss when MDM2 is deleted in p53-null tumor cells. 63 It is therefore plausible that the observed killing of HCT116 p53 À/À by DTC PMI at high concentrations arises from its p53independent on-target activity, potentially extending DTC PMI to the treatment of p53-decient cancers as well.
Aside from the simplicity of using natural amino acids, the DTC chemistry may offer an added advantage over the hydrocarbon stapling technique: peptide solubility. If stapling severely decreases peptide solubility, it can potentially limit drug concentration in vivo, thus therapeutic efficacy. For direct comparison, we stapled Ac-TSFXQYWXLLSR-NH 2 with a hydrocarbon linkage between X residues at positions 4 and 8 (X ¼ (S)-2-(4 0 -pentenyl)alanine), yielding a hydrocarbon stapled peptide termed HC PMI that differs only in the crosslink from DTC PMI. DTC PMI and HC PMI were each suspended at 20 mg ml À1 in PBS, followed by a 2-fold serial dilution and OD measurements at 600 nm. As shown in Fig. S17, † while DTC PMI was soluble at a concentration of >10 mg ml À1 , the solubility of HC PMI was signicantly lower, at $0.3 mg ml À1 . Since dithiocarbamate contains multiple hydrogen bond donors/acceptors, the DTC staple is expected to be more soluble than all-hydrocarbon crosslinks.

Conclusions
We have developed a novel stapling strategy for peptide drug design by taking advantage of the DTC chemistry to crosslink the side chains of the two natural amino acid residues Lys and Cys at (i, i + 4) positions. The DTC staple, structurally validated, induced the formation of and stabilized a productive a-helical conformation of PMIa dual-specicity peptide antagonist of MDM2 and MDMX, enabling it to traverse the cell membrane and kill tumor cells by reactivating the p53 pathway. DTC stapling functionally rescued PMI that, on its own, failed to activate p53 in vitro and in vivo due to its poor membrane permeability and susceptibility to proteolytic degradation. It is worth noting that DTC stapling offers a better peptide aqueous solubility over hydrocarbon stapling. Compared with other known stapling techniques, the solution-based DTC chemistry is simple, cost-effective, regio-specic, and environmentally friendly, promising an important new tool for peptide drug discovery and development for a variety of human diseases.

Author contributions
XL, HGH, MP and WL conceived and designed the study. XL, WDT, NG, YZ, FN, WH and WY performed the experiments. HGH and JCS helped with study design, and edited the manuscript. XL, MP and WL wrote the paper. All authors read and approved the manuscript.

Conflicts of interest
There are no conicts to declare.