In general, disulfide bond formation by thiol-disulfide interchange is a reaction that proceeds through the nucleophilic attack of a thiol or thiolate at one of the sulfur atom of a disulfide. Directed disulfide formation by this method is desirable over other oxidation methods, majorly due to the simplicity and less complicated downstream processing (Andreu et al. 1995; Annis et al. 1997; Bulaj 2005; Schäfer and Barz 2018). In SPPS, the phenomenon of forming disulfide bonds by thiol-disulfide interchange has already been introduced and is associated with popularly known disulfide based thiol-activating protecting groups such as Npys (Isidro-Llobet et al. 2009; Rentier et al. 2017). In our previous work, we illustrated that SIT-protected peptides could also undergo thiol-disulfide interchange culminating in successfully synthesizing therapeutic peptides with one disulfide bridge, including oxytocin and somatostatin (Chakraborty et al. 2022). The SIT-directed disulfide formation required no oxidants but the use of a slightly basic medium that ultimately yielded the disulfide peptides in a very shorter time and with impressive HPLC purity profiles. Taking inspiration from this, we further explored the possibility of implementing the SIT-directed disulfide formation in synthesizing multi-disulfide peptides. Figure 2 illustrates different oxidation strategies for synthesizing multi-disulfide peptides, including the proposed SIT-directed one. In the proposed SIT-directed cyclization, we used SIT-protected Cys peptides as our substrate. Fmoc-Cys(SIT)-OH was used to incorporate the Cys closer to the N-terminal of each disulfide in the peptide sequence. So, for example, in a peptide with four Cys residues (rendering two disulfides), two would be SIT-protected, and the rest would be protected with acid labile groups like Trt.

Different oxidation strategies for constructing bisdisulfide peptides (X1, X2 = protecting groups)

First, we synthesized α-conotoxin SI, a 13-mer peptide (H-ICCNPACGPKYSC-NH2) with two disulfide bonds (Cys2-Cys7 and Cys3-Cys13) (Zafaralla et al. 1988; Benie et al. 2000). As shown in Scheme 1, two parallel syntheses of linear α-conotoxin SI were performed on Rink amide AM PS (polystyrene) resin following Fmoc/tBu based SPPS protocols (Trt protection for Asn, tBu for Tyr and Ser, and Boc for Lys). Here, it is important to note that for the first batch of synthesis (strategy I) Fmoc-Cys(Trt)-OH was used to introduce all four Cys residues in the sequence. For the second batch (strategy II), Fmoc-Cys(SIT)-OH was used to introduce Cys2 and Cys3 residues and Fmoc-Cys(Trt)-OH was used for the rest (Cys7 and Cys13). After synthesis, both peptidyl resins were cleaved and subjected to analysis by HPLC and LCMS. From cleavages, linear unprotected conotoxin amide (1c in Scheme 1) (Figure SI-1) and SIT-protected linear conotoxin amide (1d of Scheme 1) were obtained with purity profiles of 89% and 90% respectively (Fig. 3). Now, to afford the disulfide bridged derivatives, the crude peptides were then dissolved separately in 25 mM aqueous solutions of NH4HCO3 (pH ≤ 8), stirred vigorously and monitored timewise using HPLC and LCMS (Fig. 3A).

Syntheses of α-conotoxin SI from unprotected linear α-conotoxin SI (1c) (strategy I) and from SIT-protected (Cys2 and Cys3) linear α-conotoxin SI (1d) (strategy II)

Timewise monitoring of synthesis of α-conotoxin SI amide: HPLC chromatograms of A cyclization of unprotected linear α-conotoxin SI by air oxidation; cyclization condition—25 mM aqueous solutions of NH4HCO3 (pH ≤ 8), rt B cyclization of SIT-protected linear α-conotoxin SI under SIT-directed/activated method; cyclization condition—25 mM aqueous solutions of NH4HCO3 (pH ≤ 8), rt

In the case of strategy I, the unprotected linear α-conotoxin SI (1c in Scheme 1) rendered fully cyclized α-conotoxin SI in 12.5 h with the total disappearance of the starting material. The final product profile showed the presence of two different isomers of cyclized conotoxin peptide (Fig. 3A). Previously, Barany et al. attempted a similar non-orthogonal solution-phase cyclization of linear α-conotoxin SI under basic conditions where all four thiols of the Cys residues were free before being cyclized (Hargittai and Barany 1999). In that experiment, the major product of the cyclization was found to be the desired isomer [2–7 and 3–13] and the misfolded ‘nested’ isomer was the side product. This was also demonstrated in one of our group’s earlier works regarding the solution-phase cyclization of linear α-conotoxin SI (Postma and Albericio 2013a). In our solution-phase cyclization, it was evident that the result was similar to those in previous instances, producing the desired conotoxin with native disulfide connections as the major product (73.1% HPLC purity profile), alongside the misfolded entity (16% HPLC purity). In the case of strategy II, the cyclization of 1d in Scheme 1 took place in less than 30 min with no trace of the starting material (Fig. 3B). The final product profile, in this case, matched the one rendered by strategy I. The major product represented the disulfide-bridged conotoxin with native folding (76.4% purity in HPLC) and a misfolded isomer (17.4% HPLC purity). So, from the results, it was clear that strategy II showed similar efficiency as strategy I but within a shorter timeframe.

Encouraged by the successful synthesis of conotoxin, we further explored this new SIT-directed multi-disulfide formation strategy for synthesizing a peptide with three disulfide bonds. We synthesized linaclotide amide (H-CCEYCCNPACTGCY-NH2). Linaclotide is a commercially available 14-mer therapeutic peptide widely used to treat irritable bowel syndrome disease (Harris and Crowell 2007; Wensel and Luthin 2011). As derived from E. coli, the peptide sequence contains six Cys residues interconnected with three disulfide bonds (Cys1-Cys6, Cys2-Cys10 and Cys5-Cys13) (Góngora-Benítez et al. 2011). Similar to the synthesis of conotoxin, here, we performed two parallel synthesis for linaclotide on Rink amide AM PS resin following Fmoc/tBu based SPPS protocols (Trt protection for Asn, tBu for Tyr, Glu and Thr). As shown in Scheme 2, the first synthesis batch was performed using Fmoc-Cys(Trt)-OH to introduce all the Cys residues in the sequence (strategy I). For the second batch of synthesis (strategy II), Fmoc-Cys(SIT)-OH was used to introduce the N-terminal Cys1, Cys2 and Cys5 residues and Fmoc-Cys(Trt)-OH was used for the rest of the cysteines, i.e., Cys6, Cys10 and Cys13. After synthesis, both peptidyl resins were cleaved and analyzed by HPLC and LCMS. The linear deprotected linaclotide amide (2c in Scheme 2) and SIT-protected (Cys1, Cys2 and Cys5) linear linaclotide amide (2d in Scheme 2) were obtained with purity profiles of 84.5% and 90.2%, respectively (Fig. 4A and C). Next, both the linear derivatives were dissolved separately in 25 mM aqueous solutions of NH4HCO3 (pH ≤ 8), stirred vigorously at rt and monitored timewise using HPLC and LCMS.

Syntheses of linaclotide amide from unprotected linear linaclotide amide (2c) (strategy I) and from SIT-protected (Cys1, Cys2 and Cys5) linear linaclotide amide (2d) (strategy II)

Synthesis of linaclotide amide: HPLC chromatograms of A linear unprotected linaclotide amide, B cyclized linaclotide amide by air-oxidation (cyclization time 4.5 h), C SIT-protected linear linaclotide amide, and D cyclized linaclotide amide by SIT-directed/activated method (cyclization time < 30 min)

Previously, our group conducted a detailed study over the cyclization of linear linaclotide. In that study, it was shown that the air-oxidation of linear linaclotide performed under basic conditions with all free thiols produced the desired cyclic isomer of linaclotide (1–6, 2–10 and 5–13). Also, the air-oxidation method was superior to regioselective methods in rendering fully cyclized linaclotide (Góngora-Benítez et al. 2011). Likewise, here, the air-oxidation method (strategy I) produced fully cyclized linaclotide amide from 2c in over 4.5 h with the total disappearance of the starting material (Fig. 4B, Figure SI-2). The final product profile showed the presence of cyclized linaclotide amide with 62% HPLC purity profile.

In the case of SIT-directed cyclization (strategy II), it took place in less than 30 min with no trace of the starting material (Fig. 4D, Figure SI-3). The final product profile, in this case, matched the one rendered by strategy I. The product profile represented the disulfide-bridged linaclotide amide fold with 74.5% purity in HPLC. So, from the results, again, it was clear that strategy II showed similar efficiency as strategy I but within a shorter timeframe.

The rapidity observed in the SIT-directed cyclization (strategy II) could be explained from the reduction in number of intermediates that evolved during cyclization. Theoretically, the air-oxidation of an unprotected linear peptide with four Cys residues can proceed through nine intermediates (six 1-SS and three 2-SS). Likewise, an unprotected linear peptide with six Cys residues could generate 75 possible intermediates (15 3-SS) during cyclization (Bulaj 2005; Arai and Iwaoka 2021). As air-oxidation is a kinetically controlled process, all these intermediates undergo continuous thiol-disulfide interchanges and finally evolve into a thermodynamically stable motif with native disulfide connections. Although, this increases the chance of the formation of kinetically-trapped intermediates. This makes the typical air-oxidation method slower with low yield. On the other hand, the number of possible intermediates formed during air-oxidation of SIT-protected linear peptides (strategy II) reduced drastically. For example, in strategy II, the arrangement of SIT-protection in linear linaclotide allowed the formation of nine 3-SS intermediates. In contrast, fifteen intermediates (3-SS) could be formed during air-oxidation of unprotected linear linaclotide (strategy I). This drastic reduction in number of intermediates possibly reduced the chances of forming kinetically trapped disulfide intermediates during oxidation of SIT-protected linear peptides, thus expediting the desired cyclization.

Stepwise disulfide bond formation in peptides has remained an evolving field of study with continuous development in searching new protecting groups for Cys and in devising new strategies or methods for disulfide bond formation in solution or while the peptide was still anchored on-resin (Postma and Albericio 2014; Spears et al. 2021). In this aspect, orthogonal thiol protecting groups play essential roles as the disulfide bonds in peptides can be formed sequentially after the selective removal of each type of protection under specific conditions (Postma and Albericio 2014; Spears et al. 2021). Especially attractive are protecting groups like SIT or StBu, which could be removed by reducing agents such as DTT. Henceforth, an Atosiban analogue was synthesized to investigate the ease of disulfide bond formation in solution or on-resin starting from SIT containing Cys peptide (Scheme 3).

On-resin and solution syntheses of Atosiban analogue (DTDPA Dithiodipropionic acid, MPA 3-Mercaptopropionic acid)

Atosiban is an octapeptide with single disulfide bond and it acts as an inhibitor of hormones like oxytocin and vasopressin (Andersson et al. 2000; Imhof et al. 2020). Medicinally, it has been used to halt premature labour (Tsatsaris et al. 2004). The Atosiban analogue that we synthesized on Fmoc-Rink amide PS resin has its d-Tyr(4-Et) residue being replaced with l-Tyr (MPA-Tyr-Ile-Thr-Asn-Cys-Pro-Orn-Gly-NH2). The synthesis was carried out using conventional Fmoc/tBu SPPS, but the last residue was incorporated as 3,3′-dithiodipropionic acid (DTDPA) (3a in Scheme 3). DTDPA is a very affordable compound, composed of two MPA (mercaptopropionic acid) units connected by a disulfide bond. Since DTDPA has two carboxylic functions, the incorporation of DTDPA into the growing peptide chain could be achieved through one or both carboxylic functions, resulting in monomeric or dimeric peptidyl residues, respectively (Fig. 5A). This was verified by performing TFA cleavage and subsequent HPLC and LCMS analysis (Fig. 5B, trace a). However, after reduction both peptidyl residues of 3a would render the desired MPA attached free thiol peptide (3c) for further manipulation.

A schematic representation of DTDPA attached to one and two units of atosiban analogues. B HPLC chromatograms of SIT-protected linear atosiban analogue (3b in Scheme 3) (trace a) and the disulfide bonded atosiban analogues—after removal of SIT and cyclization on-resin using NCS (trace b) and linear deprotected atosiban analogue cyclized in solution by air oxidation at pH 8 (trace c); *DTDPA attached to two units of SIT-protected linear atosiban analogue through two carboxyl groups

First of all, we ventured into the on-resin cyclization, taking advantage of pseudo-dilution phenomena regarding solid phase. From a practical perspective, cyclization on-resin is always beneficial and desirable in the sense that the downstream purification process becomes hassle-free, because the solution oxidation is usually carried out a high dilution, and therefore the solid-phase oxidation does not require to concentrate the solution. Now for cyclization, linear Atosiban peptidyl resin (3c in Scheme 3) was divided into five equal portions and each portion was subjected to a particular cyclization condition, including DIEA in DMF, DIEA and DMSO in DMF, piperidine in DMF, potassium ferricyanide in DMF and N-chlorosuccinimide (NCS) in DMF. The volume of the cyclization mixture was maintained at 150 μL per mg of peptidyl resin in all cases. After each cyclization, the peptide was cleaved from the resin using TFA/TIS/H2O (95:2.5:2.5), and the resulting crude was analysed by both HPLC and LCMS. The cyclization was evident when the on-resin cyclization was attempted with 5% DIEA in DMF. Still, the kinetics of the reaction was very sluggish, yielding only 60% cyclized product even after leaving the reaction overnight (Figure SI-4). With 5% DIEA and 5% DMSO in DMF, the extent of cyclization was more or less the same as that of DIEA in DMF (Figure SI-5). At this stage, we thought that change of base could help increase the cyclization rate. Hence, we tried 5% piperidine in DMF. Practically, the use of piperidine did not improve the output at all, rather leading to the formation of several side products (Figure SI-6). So, all these mild basic conditions were not affording the complete cyclization of the crude and were taking an indefinite time. This situation prompted us to try the oxidation of free thiols using potassium ferricyanide. Potassium ferricyanide (0.2%) solution in DMF was added to the peptidyl resin while maintaining pH ̴ 8 adding aqueous ammonia. The mixture was stirred until it maintained a pale yellow colour. Analysis of the crude obtained after cleavage revealed that the percent conversion to the cyclized product was even poorer using this mild oxidant (Figure SI-7). Finally, we tried NCS in DMF. NCS activates free thiols, which is later attacked by an incoming thiol to form the disulphide (Postma and Albericio 2013a, 2013b). At first, the peptidyl resin was treated with NCS (2 equiv.) in DMF (200 μL) for 25 min at rt and then washed, cleaved and analyzed by HPLC and LCMS (Fig. 5B, trace b). From the results, it was clear that in this case we achieved efficient cyclization on-resin in very short time. NCS efficiently yielded the cyclized derivative with 81.2% of HPLC purity.

For solution phase cyclization, linear Atosiban peptidyl resin (3c in Scheme 3) was cleaved from the resin using TFA. The resultant crude showed the presence of the desired peptide and the S-tbutylated derivative as the main impurity (Figure SI-8). The crude peptide was then dissolved in 25 mM aqueous solution of NH4HCO3 with the pH adjusted to 8–9, and the mixture was kept under vigorous stirring overnight. Analysis of the reaction mixture by HPLC and LCMS showed the formation of cyclized Atosiban analogue (67% HPLC purity) with no trace of the unprotected linear analogue (Fig. 5B, trace c). S-tbutylated dimer having single disulfide bridge appeared as major side product of the cyclization (16.4% HPLC purity) and was originated from the cleavage step.

Interestingly, both approaches have pros and contras. The oxidation on solid-phase is easier and does not give the S-tbutylated side reaction, although it gives polymers, which can be detected in the HPLC after the main peak. On the other hand, the oxidation in solution, which is more tedious, does not show the presence of dimers, but it does show the S-tbutylated side reaction. However, the use of SIT protection in both cases was advantageous. Especially, the on-resin cyclization of a single disulfide peptide after efficient removal of SIT on solid-phase overcomes the problems associated with other similar Cys protecting groups like StBu.

The most feared reaction in peptide synthesis is racemization because it leads to the formation of diastereoisomer peptides with very similar structures to the target peptides. This poses difficulty in purification, which results in a decrease of the overall yield (Yang 2015; Lloyd-Williams et al. 2020). Although all protected amino acids can give racemization, it is well accepted that the derivatives of His, Ser, and Cys are most susceptible to this side reaction (Isidro-Llobet et al. 2009; Yang 2015). In the past, our group studied substantially the racemization of thiol protected Fmoc-Cys-OH derivatives using 1-hydroxybenzotriazole (HOBt) based coupling reagents, mainly N,N′-diisopropylcarbodiimide (DIC)-HOBt and N-[(1H-benzotriazol-1-yl)-(dimethylamino)methylene]-N-methylmethanaminium hexafluorophosphate N-oxide (HBTU) in the presence of the base (Han et al. 1997). This study’s main conclusion was that racemization minimization was reached when pre-activation of 5 min was carried out with DIC-HOBt. In contrast, pre-activation should be avoided in the case of HBTU-base (Han et al. 1997). During pre-activation, the carboxyl-activated intermediate is exposed to the base for an extended period of time in the absence of the amine. This facilitates the abstraction of the α-proton of Cys and results in racemization. Remembering these precedents, we studied racemization using OxymaPure-based coupling reagents, with and without pre-activation. OxymaPure gives higher coupling yields with lower racemization and is greener and safer than HOBt (Subirós-Funosas et al. 2009; Manne et al. 2020).

A model tripeptide amide, H-Gly-Cys-Phe-NH2, was synthesized on Fmoc-Rink amide PS resin to study the racemization in SIT-protected Cys peptides. In addition to using reducing agent labile Cys(SIT) and Cys(StBu) derivatives, Cys(Trt) was incorporated into the study as it is possibly the most widely used acid labile Cys protecting group. DIC was used as coupling reagent in conjunction with OxymaPure and OxymaB. The latter is an oxime derivative of the barbituric acid, and facilitated further minimizing racemization (Jad et al. 2014). As uronium salt derivative of OxymaPure, 1-[(1-(cyano-2-ethoxy-2-oxoethylideneaminooxy)-dimethylamino-morpholinomethylene)] methanaminium hexafluorophosphate (COMU) (El-Faham et al. 2009; El-Faham and Albericio 2010) was chosen for the study, which was compared with HBTU (Carpino et al. 2002). COMU- and HBTU-based coupling methods are more prone to racemization compared to DIC/Oxyma-based methods. Therefore, the evaluation of different Cys protecting groups under these slightly favourable conditions for racemization was necessary. In all cases, 4 equiv. of all reagents, including Fmoc-AA-OH were used, and in the case of COMU or HBTU-based couplings, 8 equiv. of DIEA was used. During the coupling, 0.2 M concentration of the reagents was maintained and each coupling continued for 30 min at rt, using DMF and N-butylpyrrolidone (NBP) as solvents. NBP is a green solvent with similar performances to DMF (Lopez et al. 2018; Kumar et al. 2020). Removal of Fmoc was achieved using either 20% piperidine in DMF (1 min + 7 min) or 20% piperidine in NBP (1 min + 7 min). In case of peptides having Cys(SIT) or Cys(StBu) residues, once the peptide was elongated, SIT and StBu were removed using dithiothreitol (DTT, 5 equiv.) in a solution of DMF/DIEA/H2O (95:2.5:2.5) at rt, three times for 10 min each. Peptides were cleaved from the resin using TFA/TIS/H2O (95:2.5:2.5) for 45 min at rt. The peptides were precipitated with ether and the resultant crudes were analysed by HPLC and LCMS. Percentage formation of H-Gly-D-Cys-Phe-NH2 was determined by integrating the corresponding peak in HPLC chromatograms over a linear gradient of 0–40% MeCN (0.1%TFA)/H2O (0.1% TFA) for 15 min. The results have been summarized in Table 1.

Conclusions from the racemization study (Table 1) are: DIC/OxymaPure and DIC/OxymaB render the peptide without racemization either with or without pre-activation and with both solvents. This confirms that Oxyma family is better coupling additive than HOBt in terms of minimizing racemization. On the other hand, when COMU and HBTU are used in the presence of DIEA, SIT is giving in all cases clearly less racemization than StBu. When SIT is compared with Trt, SIT in DMF also gave less racemization, while in NBP it gives slightly more racemization. As reported earlier, pre-activation gave more racemization than in situ activation. Overall, NBP gives less racemization than DMF. This aspect of NBP can also be attributed to its lower polarity compared to DMF and it was confirmed by our group’s previous findings in that regard (Kumar et al. 2020).