Peptide therapeutics play a role in drug development because of the unique properties of peptides, which include high specificity, good efficacy, and low immunogenicity when compared to small molecule drugs and biologics [1, 2]. To develop a peptide-based drug, chemical peptide synthesis and subsequent modifications on the sequence and terminal structures are steps that are essential to improve its pharmacological properties after the first step, the discovery of potential peptide therapeutics.

Modification of the C-terminus of a peptide can significantly affect its bioactivity, stability, or three-dimensional conformations [3]. Even relatively minor structural changes, such as α-amidation of the C-terminal moiety, often significantly affect the peptide properties. For this reason, many strategies to access C-terminally modified peptides have been developed. The most widely used strategy is based on the selection of linkers prior to solid-phase peptide synthesis (SPPS) [4, 5]. This is a reliable strategy but the desired functional groups must be incorporated at the beginning of the synthesis because the peptides are usually synthesized beginning at the C-terminus (Figure 1A). Therefore, this type of strategy requires a significant synthetic effort to create a peptide library containing diverse C-terminal residues.

Late-stage modification strategies, which can lead to new analogs of a parent structure, have been recognized as useful methods that may allow the performance of only a single SPPS to construct the peptide sequence and complete the subsequent transformation of its C-terminus to generate various analogs (Figure 1B). The key to the success of this strategy in peptides is achieving the chemoselective and site-selective transformation, particularly in the absence of any protecting groups on reactive functional groups. Consequently, hydrazide was used as a chemoselective handle at the C-terminus (Figure 1C). This study focused on transforming hydrazides to carboxylic acids and amides because both of these C-terminal structures are common in bioactive peptides, often significantly affecting their structure and activity. Oxidation of a hydrazide will convert it to a carboxylic acid [6–10]. Treatment with β-mercaptoethanol (BME) can generate peptide acids from the corresponding thioesters, which can be prepared from unprotected hydrazides [11–13]. Ammonolysis [14] or the Staudinger reaction [15] of the peptide azides which in turn can be prepared from hydrazides are two potential methods for the preparation of peptide amides. Here a systematic and comprehensive evaluation of the relevant transformations is described and their utility by synthesizing modelin-5 derivatives including an artificial antimicrobial peptide is demonstrated [16].

To examine the feasibility of converting peptide hydrazides to carboxylic acids, the model peptide (H-LYRAG-NHNH₂, 1a) was treated with several oxidants (Figure 2 and Table 1). Although copper (II) sulfate (CuSO₄) was effective in converting an N-alkyl hydrazide peptide to carboxylic acid [6, 7], it produced the corresponding carboxylic acid (2a) only in a low yield (entry 1). Oxone oxidation proceeded smoothly within 30 min (entry 2). Treatment with IBX or NBS consumed the starting material but failed to generate the desired product (entries 3 and 4). The plausible mechanism of Oxone-mediated transformation of hydrazides to acids is given in Figure 3 [22].

Formation of azides from hydrazides followed by a reaction with BME is another approach for the conversion of hydrazides to carboxylic acids. Unprotected hydrazides can be transformed into peptidyl azides with the aid of NaNO₂ under acidic conditions. In the presence of thiols [19], azides readily afford the corresponding thioesters. Thioesters of BME are formally hydrolyzed through the intramolecular S-to-O acyl transfer followed by thiirane formation, releasing the appropriate carboxylate (Figure 4) [11]. This reaction sequence has been used for the epimerization-free synthesis of peptide acids with a C-terminal Cys [12], but has not been systematically and comprehensively evaluated.

To evaluate the scope of these two methods to generate carboxylic acids, peptide hydrazides bearing amino acids at the C-terminus were prepared. The exceptions were aspartic acid (Asp), asparagine (Asn), and glutamine (Gln) because of the synthetic difficulty due to intramolecular cyclization [19]. Using oxidation by Oxone, the desired carboxylic acids were obtained with 63–87% purity (Figure 5, Table 2). Oxidation-prone amino acids, namely Cys, Met, and Trp, were however incompatible with this condition. A separate protocol using BME after azide formation succeeded in the transformation of hydrazides to the corresponding carboxylic acids in 60–95% yield. Ile and Pro required a longer reaction time to completely convert the intermediate thioester to the acid (entries 4 and 7). Unlike the reaction conditions of the Oxone reaction, the BME-mediated reaction was applicable to the peptides with C-terminal Cys, Met, or Trp (entries 11, 13, and 18). The epimerization of C-terminal amino acids (Ala) was suppressed during the reactions (epimer ratio > 99:1, Figure S110).

To convert a peptide hydrazide to the peptide amide, two methods were used: 1) Azidation followed by ammonolysis, and 2) azidation followed by the Staudinger reaction (Figure 6). After preliminary experiments, peptide amides were found to be formed by ammonolysis of peptidyl azides using 6 mol/L NH₄OAc in a slightly basic buffer. Similarly, the Staudinger reaction of peptidyl azides using 0.1 mol/L TCEP·HCl as a water-soluble phosphine gave the desired amide within 30 min.

With the optimal conditions in hand, the scope of these conditions with respect to various C-terminal amino acids was examined (Figure 7 and Table 3). In the NH₄OAc-mediated ammonolysis, the desired amides (4) were obtained together with the hydrolyzed products (2) and compounds with molecular weight 17 less than these, probably because of the intramolecularly cyclized peptides (8). In the case of C-terminal Glu and His, significant amounts of the hydrolyzed peptide (2) were obtained (entries 10 and 15). The peptide with a C-terminal Cys residue afforded a low yield because of the formation of a disulfide-related byproduct, which was suppressed by using Acm protection for the thiol (entries 11 and 12).

The procedure using the Staudinger reaction also proceeded smoothly, yielding the peptide amides (4) in acceptable yields (80–94%) except for sequences containing Glu-, His-, or Cys-. Glu- or His-containing peptides afforded a low yield of the desired products, but 87% and 28% of the hydrolyzed product (2) were observed (entries 10 and 15). When performing the azide formation in a TFA cocktail [20] instead of under aqueous conditions, the hydrolysis was suppressed, giving the desired amides in 74% for Glu and 83% for His. The thiol group of Cys-containing peptide was oxidized to the sulfinic acid while the side reaction was suppressed using Cys(Acm) as with the ammonolysis condition (entries 11 and 12). Azidation in TFA followed by the Staudinger reaction was also effective in generating the corresponding amides from hydrazides, but in some cases, amines, oxazolidinones, or thiazolidinones were observed as a main byproduct formed through Curtius rearrangement followed by hydrolysis or cyclization of the resulting isocyanate (Table S1). The peptides having Ala at the C-terminus showed no epimerization in the transformation to amides (epimer ratio > 99:1, Figure S111).

The applicability of the developed late-stage diversification strategies was validated by synthesizing modelin-5 derivatives (H-KLAKKLAKLAKLAKAL-X; X = OH, NH₂, or NHNH₂), an artificial antimicrobial peptide (Figure 8A). Modelin-5 is known to have varied antimicrobial potencies depending on its C-terminal structure; in particular, modelin-5-CONH₂ is more potent than modelin-5-COOH [23]. The modelin-5 sequence was elongated on a hydrazine-incorporated resin through the conventional Fmoc SPPS. The standard cleavage of the elongated peptide afforded the modelin-5-CONHNH₂ (9) in 34% isolated yield (Figure 8B). The global deprotection of the peptidyl resin followed by oxidation by Oxone smoothly yielded the modeline-5-COOH (10) in 32% isolated yield from the resin loading (Figure 8C). The modelin-5-CONH₂ (11) was synthesized in 37% yield through the azidation of the corresponding hydrazide followed by the Staudinger reaction (Figure 8D).

Finally, the antimicrobial potency against E. coli of the synthesized modelin-5 derivatives was evaluated. The cells were incubated with 500 μmol/L of each peptide derivative for 1 h. The treated cells were plated and then incubated at 37℃ for 14 h, after which the number of colonies was counted (Figure 8E). As has been reported [23], the potency of the amide derivative (11) was higher than that of the carboxylic acid (10). Surprisingly, the hydrazide-type derivative (9) showed the most potent antimicrobial activity against E. coli but the mechanism that underlies such potency differences was unclear.

Derivatization of peptides is an important method to control and improve their properties. Modification of the C-terminus of bioactive peptides has important implications for their potency. Late-stage diversification is a practical concept to create libraries with diversity, although it requires a chemoselective toolbox to apply it to peptides because peptides have various types of functional groups. In this study, simple and practical protocols were developed for the late-stage transformation of peptide C-terminus residues from hydrazides to the corresponding peptide acids and amides. Oxone oxidation yielded carboxylic acids directly from hydrazides without any additional transformations although some oxidation-prone amino acids showed low compatibility with this reaction. Chemoselective azidation of hydrazides followed by the addition of BME also succeeded in yielding the carboxylic acids via a thioester intermediate. This method is compatible with oxidation-prone amino acids. Different kinds of reaction sequences have been attempted for some specific examples. The same procedure, hydrazide activation by acetylacetone followed by BME addition [13], azidation followed by hydrolysis in basic aqueous conditions have been used [12, 24], but to the best of our knowledge, such a systematic and comprehensive evaluation has not been completed to date.

Chemoselective azidation of hydrazides and subsequent ammonolysis or a Staudinger reaction also succeeded in the late-stage amide synthesis although a C-terminal Cys must be protected. C-terminal Glu hampers the efficient amide synthesis probably due to the intramolecular anhydride formation followed by hydrolysis. This undesired reaction was suppressed by using a protocol other than a low pH aqueous buffer, to generate a peptide azide in TFA, but the versatility of the TFA condition was not so strong and side reactions involving Curtius rearrangement occurred more frequently. These results suggest that it is important to choose appropriate reaction conditions according to the components and sequence of peptides.

A general strategy to obtain both acid and amide peptides requires double SPPS using a resin with different linkers. In contrast, our protocols allow the performance of only a single SPPS to obtain both peptides. For example, the synthesis of modelin-5-COOH and -CONH₂ using different linkers requires a total of 64 steps (16 couplings and 16 removals of N_α protection for each SPPS). In comparison, late-stage transformation reduces the number of synthetic steps to 34 (32 steps for peptide elongation and two steps for the transformation) and additionally produces a hydrazide as another component to enrich the library. This benefit was confirmed by evaluating the antimicrobial activity of the modelin-5 derivatives, leading to the unexpected discovery that modelin-5-CONHNH₂ showed the most potent activity. Considering the antimicrobial peptide with C-terminal hydrazide, Li et al. [24] reported that the antimicrobial spectrum of Chex1-Arg20, an antimicrobial peptide, was broadened by incorporation of the hydrazide. They hypothesized that the broadened spectrum of activity may be due to the stability against bacterial enzyme attacks being improved in the hydrazide. Since modelin-5-CONH₂ has a potent ability to kill Bacillus subtilis via membranolytic modes of action [25], the hydrazide might also improve the stability against bacterial enzymes. Efforts to exploit the mechanism and influence of C-terminal hydrazide modification into the antimicrobial activity are in progress in our laboratory.