Artificial disulfide-rich peptide scaffolds with precisely defined disulfide patterns and a minimized number of isomers

We report the design and synthesis of artificial disulfide-rich peptide scaffolds with precisely defined disulfide patterns and a minimized number of isomers.


Introduction
The discovery and creation of novel organic structures or scaffolds has been a propeller driving the development of smallmolecule drugs. 1 The development of peptide-based drugs would also benet from the discovery and synthesis of natural peptide scaffolds. 2 Indeed, the huge diversity of disulde-rich peptides, including antimicrobial defensins, plant-derived cyclotides, and conotoxins from the venom of predatory marine snails, holds great promise of being exploited for the development of novel therapeutics for diverse human diseases. 2a-e,3 While short peptides are usually structurally ill-dened and extremely susceptible to enzymatic hydrolysis, 4 these privileged peptides containing multiple disulde bonds can signicantly reduce conformational exibility and x the structures of peptides into their bioactive states, therefore displaying an improved binding efficiency and specicity, and can be more tolerant to proteolysis compared to linear peptides. 2e,3a,5 However, the synthesis and reengineering of disulderich peptides have been considered to be an overt challenge, mainly owing to the complexity of oxidative folding processes which amplies rapidly as the number of possible isomers with different disulde patterns increases. 6 Regioselective approaches involving orthogonal protecting groups, stepwise deprotections, and oxidations are most oen exploited to ensure the synthesis of the desired isomers, 7 which is usually sophisticated and laborious; moreover, the desired isomer might evolve towards other disulde isomers due to isomerization of the disulde bonds under thermodynamic control in the presence of thiols in biological uids. 8 Alternative strategies exploited the orthogonal or preferential pairing between deprotected thiols (cysteine (Cys), penicillamine (Pen) or synthetic thiols) or the diselenide formation between selenocysteine residues to direct the oxidative folding of the peptides. 8,9 Though these strategies usually afford the desired isomers as the major folding products, the undesired tendency of forming other disulde isomers can still complicate the folding processes and trigger disulde isomerizations, particularly when the disulde-rich peptides are used as structural scaffolds for drug design applications, because in this circumstance, their primary sequences are subject to extensive manipulation. Accordingly, novel disulde-rich peptide scaffolds that are not besieged by their disulde isomers, and thus are more tolerant to sequence manipulation than conventional natural peptide scaffolds, are still greatly desired.
Here, we report the design and synthesis of articial disulde-rich peptide scaffolds with precisely dened disulde connectivity and a minimized number of isomers. The articial scaffolds have three disulde bonds, which can spontaneously pair with a high degree of accuracy in redox buffers. In theory, natural peptides with three disulde bonds can form 15 possible isomers corresponding to different disulde connectivities. We strategically transformed a specic cysteine framework, which is naturally recruited for the production of human a-defensins (i.e., CXC-C-C-CC), 2a,2b,10 into several articial thiol-frameworks containing three Pen residues and a synthetic dithiol amino acid (Dtaa) (Fig. 1). The oxidative folding of the peptides with unique C/Pen/Dtaa-frameworks leads to the formation of structurally pre-dened disulde-rich peptide scaffolds, a process that is exclusively due to the directed pairing of the disulde bonds, but not the sequence-specic prefolding. These articial peptide scaffolds should be much more tolerant to sequence manipulation and, in principle, can be better at avoiding the problematic isomerization of disulde bonds compared to natural ones.

Results and discussion
A model glycine residue-rich peptide (1) patterned with the CXC-C-C-CC framework was rst designed and synthesized, in which non-glycine amino acids were placed for the later analysis of cysteine pairing in the oxidized products using tryptic digestion liquid chromatography-mass spectrometry (LC-MS). To focus our study on scaffold-dependent disulde pairing, the achiral glycine residues were preferentially placed in the peptide, which disfavour the sequence-dependent prefolding of the peptide towards specic isomers. The oxidative folding of 1 results in the formation of scrambled isomers that cannot be isolated efficiently using high-performance liquid chromatography (HPLC) ( Fig. 2a and b). In addition, the total peak area of the produced isomers is as low as 20% of the initially reduced peptide, suggesting that the overall yield of the folding is rather low, and it is very possible that this involves many undetectable and trapped folding intermediates.
We then rationally and subjectively transformed the cysteine framework of 1 into an articial one by simultaneously replacing the two adjacent cysteines with a synthetic dithiol amino acid and three of the others with Pen residues, 8,9h which generates Fig. 1 Illustration of the oxidative folding of peptides containing six cysteine residues (15 total folds, into which only one corresponds to the native fold); the artificial peptide scaffold designed in this work, which contains the unnatural amino acids Pen and Dtaa; only two expected folds formed after the oxidative folding of the peptides, and one of the two corresponds to the disulfide connectivity of human a-defensin 5 (but of less abundance). -x-x-x-x-x-denotes a peptide segment containing any number of natural amino acids of any type, except cysteine residues. a new peptide 2 with a PenXC-Pen-Pen-Dtaa framework ( Fig. 1 and 2; see the ESI † for details on the synthesis of 2). We hypothesized that the homogenization of the two adjacent cysteines using Dtaa substitution and the spatial separation of the two thiol groups in Dtaa could in principle exclude the formation of 9 specic isomers from the total of 15 possible ones (Fig. 2a). The three-Pen substitution could further reduce the complexity of the oxidized products, leading to a reduction in the number of possible isomers from 6 to 3 due to the orthogonal features of the Cys-Pen disulde pairing (Fig. 2a). 9h In addition, the formation of the intra-CRC motif disulde bond is strongly disfavoured due to the intrinsic constrain of the bridged structure. 9g,11 These inferences, taken together, ultimately restrict the maximum number of nal folding products to 2. Indeed, we only observed the formation of two expected isomers (in 100% yield, and in a 5 : 1 ratio of the two peaks) in the HPLC traces aer the oxidative folding of 2 in an oxidized glutathione (GSSG) buffer, whereas the other peaks (or isomers) were negligibly small (Fig. 2b). In both of the isomers (2a and 2b; Fig. 2a and c), two of the three Pen residues are paired with the Dtaa dithiol, and with the odd one paired to the sole cysteine residue within the peptide (see the ESI for the characterization of the disulde pairing; Fig. S1-S3 †). The isomer 2a is more favorably formed compared to 2b, which is likely to arise from superiority in the change in the folding entropy for 2a, as sequence-related folding processes are largely absent, and thus 2b would correspond to a topologically more compact scaffold compared to 2a.
We demonstrated above that Dtaa and Pen substitution in a six-cysteine peptide can dramatically reduce the complexity of    the oxidative folding products, that is, from a total of 15 isomers to the 2 that are expected. The effect of Dtaa substitution on the overall reductive isomerization is quite clear in theory, 8 whereas the parallel effect of Pen might not be as straightforward. To further reveal the contribution of the Cys/Pen substitution on the reductive isomerization, a peptide (3, CXC-C-C-Dtaa framework), that is analogous to 2 but without replacing the three cysteines with Pen residues, was designed and synthesized (Fig. 3). Although it is unlike peptide 1, the oxidation of 3 results in the formation of 5 distinct and highly resolved HPLC peaks which correspond to the 5 expected isomers (except for the 1-2, 3-5, and 4-5 connectivities shown in Fig. 2a, because of the prohibition of forming an intra-CRC disulde; see the ESI for the characterization of disulde pairing; Fig. S4-S9 †). However, the total peak area is still signicantly diminished (50% of the initially reduced peptide), implying a relatively low oxidative folding yield (similar to that for 1). In addition, we found that the most abundant product of 3 (i.e., the bell-shaped isomer 3e) is not of the same folding as that obtained from 2 (Fig. 3). The two isomers with disulde connectivities equivalent to those of 2a and 2b are 3c and 3d, respectively, which are the second to least and the least abundant products of 3, respectively. This nding suggests that the driving force arising from the orthogonal disulde pairing between the thiols of cysteine/Dtaa and the sterically hindered thiols of Pen residues can overcome the topology-preferred folding propensity to the bell-shaped isomer and can direct the oxidative folding of the peptide to the two specic and expected isomers.
2a, 2b and 3e were then isolated using HPLC and dissolved in a glutathione (GSH)/GSSG buffer (0.5 mM, pH 7.4). Interestingly, 2a and 2b are both substantially more stable than 3e (Fig. 4). 3e is subject to very rapid disulde isomerizations, which result in the formation of the other four isomers and an equilibrium was achieved within 10 min. In contrast, the equilibrium of the disulde isomerizations for 2a and 2b was only achieved aer 2 h. The enhanced stability is considered to stem from the steric hindrance of the Pen residues (i.e., the two methyl groups adjacent to the sulfur atom). More importantly, side-products were not obviously observed during the disulde isomerizations (even for the less stable peptide 2b), suggesting that the oxidative folding pathways used by the articial six-thiol framework (but not the natural six-cysteine framework or the non-Pen analogue) should have high efficiency and precision towards the expected isomers.
Next, we examined if the directed oxidative folding of the articial scaffolds is tolerant to the manipulation of the primary sequence, and if the articial disulde-rich scaffold can be used for graing bioactive sequences. An integrin binding motif RGD (-Arg-Gly-Asp-) was selected for the test. 12 In the rst example, three RGD motifs were inserted into the variable segments (i.e., "-") of the PenXC-Pen-Pen-Dtaa framework to design peptide 4 (Fig. 5a). Secondly, a yeast-selected sequence (PRPRGDNPPLT; 13 containing a RGD motif) was graed into the second variable segment of the framework, and the length of the other two variable segments was shortened by the depletion of two amino acid residues (for the ease of synthesis; 5, Fig. 5a). Then, the peptides (4 and 5) were oxidized under the same conditions as those used for the folding of 2. As can be seen in Fig. 5a, the oxidative folding of 4 and 5 results in the formation of the expected isomers, with either 1-5, 2-3, 4-5 (4a and 5b) or 1-5, 2-4, 3-5 (5a) disulde connectivity, as the major products (see the ESI for the characterization of disulde pairing; Fig. S10-S14 †), which is very similar to that observed for 2. We further examined the ability of the oxidized peptides to block the adhesion of U87 glioblastoma cells (with surface-expressed integrin) to cell culture plates. All of the peptides were able to inhibit the adhesion of the cells (Fig. 5b and c). Although the activities of these RGD-containing peptides are lower compared to that of commercially provided cyclic RGD (c(RGDyK); as a positive control), this is not surprising as their cyclic structures have been thoroughly optimized. 12 In addition, the previously obtained 2a, as a negative control without an RGD motif, exhibits a negligible ability of inhibiting cell adhesion. We also observed an obvious change in the morphology of the cells on the plates when the cells were incubated with the RGD-containing peptides (4a,5a,and 5b;Fig. S18 †). Therefore, these results strongly suggested that the developed articial disulde-rich scaffolds are tolerant to extensive manipulation in the primary amino acid sequence and are amenable to the design of disulde-rich bioactive peptides by graing bioactive sequences.
Finally, to demonstrate if our strategy could be used to regulate the folding of peptides with different cysteine frameworks for the design and synthesis of novel articial disulderich scaffolds, a peptide with a C-CXC-C-CC framework was reengineered using Pen and Dtaa substitution, which generates a new peptide 6 with a unique Pen-PenXC-Pen-Dtaa framework (Fig. 6). We found that the oxidation of 6 in buffers leads to the formation of the two expected isomers (100% yield), in which the Pen residues are paired with either the Dtaa dithiol or the cysteine residue (6a and 6b; see the ESI for the characterization of disulde pairing; Fig. S15-S17 †). Thus, this result further validates the robustness and generality of the present strategy for the design and synthesis of articial peptide scaffolds with precisely dened disulde patterns and a minimized number of isomers. As the length of the peptide segments in the present scaffolds, the position of the PenXC motif and the Dtaa position can be changed at will, and a number of articial disulde-rich scaffolds could be designed and synthesized in the future.

Conclusions
In conclusion, we developed a combinatorial strategy to regulate the oxidative folding of peptides, which generates a novel class of articial peptide scaffolds with precisely dened disulde patterns and a minimized number of isomers. Despite the fact that the folding pathways of natural polypeptides (or proteins) can be strategically tuned through the use of disulde surrogates or selenocysteine, and that native folds might be obtained in high yields, 9a-c,14 the precise pairing of disulde bonds in these systems is considered, in essence, as a result of the primary sequence-specic prefolding or co-folding (i.e., the folding before the disulde formation or disulde-directed folding). 6b,15 It is still inconceivable to fundamentally reduce the complexity of disulde pairing in peptides containing up to three disuldes without the involvement of sequence-specic folding. In this work, we demonstrated that the total number of isomers formed aer the oxidative folding of a peptide containing six thiols can be decreased to a minimum of two (i.e., from 15 to 2). To our knowledge, such elegant precision in peptide folding has never been achieved. As a solely thiolpattern-based peptide folding strategy, standing out from the existing sequence-dependent ones, it would provide a valuable guide for designing novel bioactive disulde-rich peptides. Moreover, the articial disulde-rich scaffolds have been found to have high stability in redox buffers and should be more able to avoid problematic isomerization due to the presence of fewer isometric structures compared to normal six-cysteinecontaining peptides. We believe that articial disulde-rich scaffolds with an intrinsic and precise disulde pairing propensity would be more tolerant to sequence manipulation than natural peptide scaffolds. This feature would greatly benet the development of structurally constrained and multicyclic peptide therapeutics and ligands.