Rapid peptide synthesis using a methylimidazolium sulfinyl fluoride salt

The past two decades have witnessed a renaissance in the development of new peptide therapeutics¹. Between 2016 and 2023 alone, 31 new peptides have been approved for a wide variety of therapeutic areas², and there are hundreds more in clinical and preclinical development^3,4,5. This renewed focus on peptides has led to increased interest in improving existing protocols and developing new methods for their synthesis^6,7,8. Of particular interest for medicinal chemistry is the rapid syntheses of peptide libraries with enhanced pharmacokinetics compared to the lead compound. Solid phase synthesis can be used for these applications because of its facile purification, but the necessity of the anchor as well as the large excess of reagents required per step, and the aggregation of the growing chain, can be problematic^9,10. Solution phase synthesis using existing methods has the advantage of being more scalable with less waste, and works well for the preparation of dipeptides and tripeptides. However, the synthesis of short oligopeptides (>3 amino acids) can be laborious due to (1) flash column chromatographic purification of each step or (2) long reaction times¹¹. As an example, the synthesis of Leu-enkephalin using carbodiimide-based reagents typically takes a week of reaction time alone¹². There are many excellent traditional peptide coupling reagents that successfully address one of the aforementioned challenges¹¹, but even powerful modern methods do not meet both of these criteria^13,14. More recent non-traditional methods including mechanochemical couplings^15,16, light-assisted couplings¹⁷, and electrochemical couplings^18,19 are efficient and high-yielding, but still require column chromatography or specialized setups. Sulfur(IV) fluoride reagents are a promising option for rapid peptide coupling as the acyl-S(IV) intermediates are highly reactive. For example, diethylaminosulfur trifluoride (DAST) and Deoxofluor can convert carboxylic acids to the corresponding acyl fluorides in 30 min or less^20,21. Despite the high reactivity of these reagents and Xtalfluor derivatives^22,23, flash column chromatography is still required for purification. An intriguing alternative to DAST is thionyl fluoride because the acyl fluorosulfite intermediate (2, Fig. 1) can react with in situ-generated fluoride to form acyl fluorides, also in only 30 min. Furthermore, the sulfite by-products can be removed using aqueous washes^24,25,26. While this reaction approaches the ideal metrics (vide supra), the total reaction times are still too long as the key peptide coupling step takes 2 h. If we could harness the high reactivity of this sparsely studied intermediate through a direct coupling, then we may be able to shorten total reaction times while maintaining the benefits of facile purification.

a Previous work using fluoride to form an acyl fluoride as the reactive intermediate²⁴. b A novel approach for trapping acyl fluorosulfite intermediates.

Full size image

Given the importance of improving amide synthesis and our interest in sulfur(IV) fluoride reagents, we explored the direct capture of the highly reactive acyl fluorosulfite intermediates for peptide coupling. To this end, we report a sulfur(IV) fluoride-mediated protocol to access oligopeptides, up to five amino acids in length, without epimerization or column chromatography. This mild and rapid protocol affords the desired oligopeptides in good yields in a single day.

We first sought to minimize the concentration of fluoride in the reaction to explore alternative couplings by using our newly developed reagent, N-methylimidazolium sulfinyl fluoride hexafluorophosphate (MISF, 6)²⁷. The use of this reagent has the additional advantage over SOF₂ of greater concentration stability over time, as thionyl fluoride is slowly outgassed from the solution. Our investigations began with the coupling of N-Boc phenylglycine (1a) with alanine tert-butyl ester (2a). Treatment of a dichloromethane (DCM) solution of both amino acids and pyridine with MISF led to the formation of dipeptide 5a in 4% yield after 1.5 h (entry 1). As incomplete dissolution was observed during the reaction, we then examined both THF (entry 2) and acetonitrile (ACN, entry 3). While both reactions afforded comparable yields of 92% and 91% respectively, further screening focused on acetonitrile as only dilute solutions of MISF in THF could be generated (see Supplementary Information for more details on the generation of MISF). Decreasing the reaction time to only 15 min led to 90% yield (entry 4), which is considerably faster than the 2.5 h total reaction time required using previously described thionyl fluoride-mediated coupling. Notably, it is also faster than peptide couplings using acyl fluorides and amines, suggesting that this new reaction was proceeding through a different mechanism (vide infra).

Replacing pyridine with either triethylamine (entry 5) or 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) (entry 6) led to a decrease in the yield. We then examined the same reaction conditions that were previously reported for thionyl fluoride-mediated peptide coupling, namely formation of the acyl fluoride first followed by addition of the second amino acid (entry 7). Even after the reaction proceeded for 1.5 h, the yield was only 55%. Contrary to our initial hypothesis, using the optimized conditions from entry 3 with SOF₂ as the sulfur(IV) fluoride reagent instead of MISF resulted in the slightly higher yield (entry 8), which suggests that the amino acid effectively outcompetes the fluoride for reaction with intermediate 6 (Fig. 1). Despite the comparable yields for the two reagents, we selected the MISF conditions (entry 4) for continued studies as it is easier to handle than thionyl fluoride gas dissolved in solution. Aminosulfinyl fluorides, which could be generated by amine addition to the sulfur(IV) fluoride reagent, were not detected by ¹⁹F NMR spectroscopy.

N-Boc phenylglycine (1a) was specifically selected for early optimization as it is prone to epimerization^28,29 and, thus, it serves as a challenging benchmark³⁰. Furthermore, 1a is increasingly being incorporated into pharmaceuticals^31,32. Traditional coupling reagents, such as hexafluorophosphate benzotriazole tetramethyl uranium (HBTU), hexafluorophosphate azabenzotriazole tetramethyl uranium (HATU), and benzotriazol-1-yloxytripyrrolidinophosphonium hexafluorophosphate (PyBOP) afforded a dipeptide containing phenylglycine similar to 5a (Table 1) with a maximum of 76:24 dr³⁰. Thionyl fluoride-mediated acyl fluoride formation followed by coupling improved the selectivity to 98:2 dr²⁴, but epimerization was still observed. Under our optimized conditions (entry 4), the desired dipeptide was formed with no observed epimerization (>99:1), as determined by HPLC (see Supplementary Information for more details on the epimerization studies).

Substrate scope

With the optimized conditions in hand, we next explored the scope of the peptide coupling between alanine t-butyl ester and a wide range of N-protected amino acids (Fig. 2). Similar to phenyl glycine (1a), N-Boc glycine (1b) coupling with alanine afforded 5b in 86% yield with no observed epimerization and no need for flash column chromatography. Amino acids with aliphatic side chains, such as alanine, valine, leucine, and isoleucine afforded the corresponding dipeptides (5c–5f) in moderate yields (approximately 50–60%) under the optimized conditions. Further increasing the reaction time only led to small increases in the yields. However, using 2 equivalents of both the aliphatic N-Boc protected amino acids and MISF led to 5c, 5d, 5e, and 5f in 90%, 74%, 81%, and 91% yield, respectively with no observed epimerization. Other canonical amino acids (5g–5u) were successfully converted to the desired dipeptides in 54–94% yields without the need for flash column chromatography. Compared to standard peptide coupling methods with comparable canonical amino acids, this method affords the products in similar yields with lower amounts of epimerization, and with streamlined reaction times and purification procedures. 5t required a silica plug to remove residual starting material as it was not easily removed through aqueous washes.

All reactions were performed on a 0.6 mmol scale. Diastereomeric ratios were determined by ¹H NMR. ^aDiastereomeric ratio was determined by HPLC.

Full size image

Non-canonical amino acid 2-aminoisobutyric acid (Aib, 5v) was only formed in moderate yield under the reaction conditions, likely due to the increased steric hindrance of the substrate. When DMAP was used as the base instead of pyridine, 5v could be isolated in 40% yield. Unlike most of the other amino acids that were examined, the Aib-containing dipeptide required a silica plug for purification in addition to the standard aqueous washes.

Different protecting groups were then explored for both amino acid coupling partners. For the C-terminus, benzyl groups were tolerated, affording 5w and 5x in 74% and 79% yield, respectively. The Cbz protecting group was also tolerated for the N-terminus, affording 5y in 92% yield. Fmoc-His(Trt)-OH afforded 5z in 68% yield (after a silica plug to remove the residual free acid), with no detectable epimerization by ¹H NMR spectroscopy (see Supplementary Information for more details on the analysis of the ¹H NMR spectra of this product and its diastereomer). In contrast, electrochemical conditions gave a 3:1 mixture of diastereomers¹⁸.

To further explore the efficacy of this protocol, the scope of the amine coupling partner was investigated in a coupling with N-Boc alanine (Fig. 3). Under the optimized conditions, glycine was effectively coupled to afford 8a in 86% yield. Aliphatic amino acids afforded good isolated yields of 8b–8d after only an aqueous workup. Proline underwent successful coupling to afford 8e, albeit in a slightly lower yield (53%). Amino acids with aromatic side groups, such as phenylalanine (Phe) and tyrosine (Tyr), afforded desired dipeptides 8f and 8g with no observed epimerization in 79% and 92% yields. The amino acids containing sulfur, cysteine (Cys) and methionine (Met), were tolerated and afforded the dipeptide in 90% and 79% respectively. Hydroxylic amino acid threonine (Thr) underwent coupling successfully to give 8j in 61% yield. Amine, amide, and acid side chains also underwent successful coupling to afford 8k–8n in 54–78% yields. Overall, the yields are comparable to existing methods^{11,33,34,35,36} but column chromatography was not required for purification.

All reactions were performed on a 0.6 mmol scale. Diastereomeric ratios were determined by ¹H NMR. ^a2 equiv. of MISF and the free acid were used.

Full size image

With the efficacy of dipeptide coupling established, the next step was to investigate whether this method can be utilized in multi-step, column-free solution-phase synthesis. Using this protocol, the two amino acids can simply be combined with the solvent and the base. Addition of a MISF solution promotes the reaction to completion in 15 min (Fig. 4). Subsequent aqueous washes will remove excess amino acids, quench unreacted MISF, and remove sulfite salt by-products. A subsequent deprotection of the Boc group using an HCl solution followed by removal of volatiles in vacuo should afford the desired dipeptide that is ready for the subsequent coupling. These steps can then be repeated to afford the desired polypeptide.

All reactions were performed on a 0.6 mmol scale. Diastereomeric ratios were determined by ¹H NMR. ^aReaction was performed using MISF. ^bReaction was performed using SOF₂.

Full size image

Coupling beyond the dipeptide, regardless of the side group of the amino acid being introduced generally required two equivalents of the acid and MISF to achieve full conversion. Tetrapeptides 12a and 12b were obtained in 40% and 57% yields respectively over three steps. Pentapeptide 12c was similarly obtained in 24% yield over four steps. Notably, Leu-enkephalin derivative 12d was obtained in 32% yield, and was synthesized in 1 day without the need for extensive purification, compared to the week of reaction time alone required by traditional methods¹². This approach was also amenable to the use of thionyl fluoride and Fmoc protecting groups (Fig. 4), affording tripeptide 12e in 44% yield over two steps and tetrapeptide 12f in 42% yield over three steps. This protocol, however, was unable to be applied to efficiently synthesize peptides of seven or more amino acids as degradation of the nascent chain was observed, along with intramolecular cyclization of the activated acid that appears to be promoted in acetonitrile.

Mechanistic studies

The three most plausible reaction mechanisms for this reaction are depicted in Fig. 5a. The first pathway that we investigated was Mechanism A, in which the peptide coupling proceeds through acyl fluoride intermediate 3. As noted above, the optimized reaction times (15 min) are significantly faster than previous thionyl-fluoride coupling methods²⁴, which suggests that the current method proceeds through a unique mechanism and that acyl fluorides are not key intermediates. To rule out a predominant acyl fluoride pathway, in situ reaction monitoring via ReactIR was utilized to compare the rate of dipeptide formation between starting from the acyl fluoride and starting from the free acid under the optimized conditions (Fig. 5b). Thionyl fluoride in DCM was used rather than MISF in ACN due to solubility issues of the dipeptide in ACN. Formation of the dipeptide when starting from the free acid was significantly faster compared to starting with the acyl fluoride, suggesting that the major pathway under these conditions does not involve the acyl fluoride (Anhydride formation is also likely not the major pathway as the C=O carbonyl stretch of the anhydride was not observed when monitoring the reaction via ReactIR).

a Possible reaction pathways following acyl fluorosulfite formation. b Mechanistic studies conducted via ReactIR to compare the rate of dipeptide formation starting from the acyl fluoride to starting from the free acid. c Increased steric hindrance of the base in this reaction decreases the yield.

Full size image

While the acyl fluorosulfite is clearly the key reactive intermediate, the role of the base needed to be further elucidated. Pyridine may simply be a base in this transformation, but it (and derivatives) is well-known to serve as a nucleophilic catalyst to accelerate nucleophilic acyl substitution reactions³⁷. To elucidate whether the base was involved in the rate determining step (Mechanism C), the steric bulk was increased by using 2,6-lutidine and 2-6-di-tert-butyl pyridine (Fig. 5c). Thionyl fluoride was used in these experiments instead of MISF to avoid releasing N-methyl imidazole³⁸. The reaction was stopped at different times to monitor the yield. After 30 min, the yield when using 2,6-lutidine was 57%, and 26% with 2,6-di-tert-butylpyridine, while pyridine achieved 91% in half the amount of time (Table 1, entry 8). This suggests that the base is essential, either for activation of acyl fluorosulfite intermediate 2, activation of thionyl fluoride, or potentially both. Additionally, these mechanistic studies demonstrate that this reaction proceeds through a pathway distinct from that of our previous work²⁴.

Overall, capture of reactive acyl fluorosulfite intermediates with nucleophiles other than fluoride has been demonstrated. This carboxylic acid activation strategy was applied to rapid peptide coupling in 15 min at room temperature. Difficult substrates that readily epimerize with traditional coupling reagents were well-tolerated under these conditions. Canonical and non-canonical amino acids, with a wide range of protecting groups, were effective substrates. This protocol allowed facile synthesis of various oligopeptides through sequential solution-phase couplings. Reaction rates determined that acyl fluorides are not major intermediates in this reaction, and that this reaction proceeds through a nucleophilic catalysis pathway. This work provides another reaction in the toolbox for peptide chemists to achieve epimerization-free couplings, comparable to existing methods in the literature^13,14 with short reaction times but is one of the few protocols that obviates the need for column chromatography.

General procedure for peptide coupling

A 20 mL vial equipped with a magnetic stir bar was charged with N-protected amino acid (0.6 mmol, 1 equiv.), amino acid ester (0.6 mmol, 1 equiv.) and pyridine (96 µL, 1.2 mmol, 2 equiv.). MISF or SOF₂ in solution (0.6 mmol, 1 equiv.) was then added, and the reaction was allowed to stir at room temperature for 15 min. The reaction mixture was diluted with DCM, washed with saturated NaHCO₃ solution, 1.0 M HCl, and then brine. The organic layer was dried with Na₂SO₄, filtered, and concentrated in vacuo.