Diverse thioether macrocyclized peptides through a radical SAM maturase

Significance

Chemical routes to thioether-containing cyclic peptides rely on preinstalled electrophiles and often complex syntheses, restricting scaffold diversity and limiting applications. We found that the radical S-adenosyl-L-methionine enzyme, PapB, possesses an unanticipated ability to macrocyclize peptides by joining the thiol moiety of a cysteine residue directly to the C-terminal carboxylate residue, even when that residue is D-configured, β-amino acid derived, or N-methylated. This sequence-independent promiscuity unlocks significant chemical space and offers a one-step enzymatic strategy for the preparation of metabolically robust macrocycles that are not accessible synthetically.

Abstract

Disulfide bonds stabilize many bioactive peptides, but their susceptibility to reduction under physiological conditions limits broad applicability in biotechnology. PapB is a promiscuous radical S-adenosyl-L-methionine enzyme that is involved in the maturation of PapA, which is a ribosomally produced and posttranslationally modified polypeptide. PapB introduces six thioether linkages between internal Cys residues and carbon atom that is α to the side-chain carboxylate of Asp/Glu residues C-terminal to the Cys residues. Herein, we show that PapB also efficiently couples an internal Cys thiol to the C-terminal carboxylate of peptides terminating in D- or β-amino acids, forming α- or β-thioether macrocycles. Moreover, PapB tolerates β- and N-methyl amino acids within the peptide, resulting in the formation of macrocycles that are comprised entirely of unnatural amino acids, such as peptides containing all β-residues. These findings establish PapB as a sequence-agnostic thioether ligase for efficient C-terminal macrocyclization. Our work expands the enzymatic toolbox for constructing conformationally constrained peptides for therapeutics and chemical biology.

Protein and peptide disulfide bonds have been integral in advancing our understanding of structure–function relationships throughout the 20th century. The discovery of insulin by Banting and Best (1) served as the basis of the first peptide-based therapeutic agent, and later, the identification of disulfide bonds in insulin by Sanger (2) motivated the appearance of disulfide in many clinically approved peptide therapeutic agents (3–7). Disulfide-based technologies have also driven method development in many areas, underscoring the importance of disulfide bonds in biology and medicine.

Despite their widespread use, disulfides are inherently unstable in the reducing environments typically encountered in biological fluids. Anfinsen showed the lability of disulfide-containing proteins to reduction, demonstrating the critical role disulfides play in preserving protein structure and function (8). Thioether crosslinks have emerged as stable alternatives to disulfides. In nature, certain ribosomally produced and posttranslationally modified polypeptides (RiPPs) are modified by specific maturases that crosslink Cys to various positions on the peptide. The point of attachment generally involves the α-carbon (sactipeptides) or the β/γ-carbon (ranthipeptides) (9). Nisin is a well-characterized example of a natural product with multiple thioethers. Notably, in nisin, the thioether crosslink occurs to a Ser/Thr that has been modified to be a Michael acceptor. In recent decades, synthetic strategies to prepare stable alternatives to disulfides that derive from modified sidechain functionalities have emerged (10). Peptide stapling strategies include Ru-catalyzed ring-closing metathesis (11–13) and Cu-catalyzed azide–alkyne cycloaddition to access cyclic peptides with alkenyl- and 1,4-disubstituted 1,2,3-triazole-based linkages, respectively (Fig. 1A) (14–16). Other cyclization strategies involve reacting peptides with biselectrophiles to insert an inert chemical spacer between two cysteine residues (Fig. 1A) (17, 18). The orthogonal nature of these methods makes them a predictable means to access stable cyclic peptides. However, their sp²-hybridized linkages are geometrically dissimilar from disulfides, motivating the development of methods to access thioether-, ether-, and 1,5-disubstituted 1,2,3-triazole-based linkages as improved, nonreducing disulfide-bond mimics (Fig. 1A) (19–22). Among these, stable thioether crosslinks that derive from the intramolecular alkylation of cysteine-containing sequences are accessible and commonly used to identify cyclic peptide therapeutic leads; however, not all derivative crosslinks can appropriately mimic a native disulfide (23–26). While polymacrocyclization can improve the metabolic stability of a cyclic peptide therapeutic lead, using these strategies to regiospecifically install multiple thioether crosslinks is considerably challenging. However, nature produces polythioether-containing peptides with ease, motivating our efforts to understand the substrate requirements and mechanistic pathways involved in the activation and cyclization of (poly)cysteine-containing peptides.

Fig. 1.

Disulfide mimics and disulfide-inspired crosslinks. (A) Native disulfides are compared with thioether linkages, other disulfide mimics, and disulfide-inspired crosslinks. This work details the installation of α-carboxythioether crosslinks via the promiscuous metalloenzyme PapB. (B) PapB installs six thioether crosslinks in CXXXD motifs in the native PapA substrate by using S-adenosyl-L-methionine (SAM) and an electron source. (C) SAM binds to the RS cluster via its amino and carboxylate moieties. Reductive cleavage forms, dAdo· and Met. (D) During the catalytic cycle of PapB, the peptide Cys residue is bound to AC1 (27).

PapB from Paenibacillus polymyxa was originally reported to form six thioether crosslinks in repeated CXXXD motifs in its substrate peptide, PapA (Fig. 1B) (28, 29). PapB is part of the expansive 700,000+ member radical SAM (rSAM) superfamily (30–33). Members of this superfamily are metalloenzymes that contain one or more iron–sulfur clusters (34–37). A nearly universally conserved C₃CX₂C motif that coordinates a [4Fe-4S] cluster is a defining feature of rSAM enzymes (32, 36–38). This cluster is crucial for radical initiation of the catalytic cycle. Notably, the unique Fe in the cluster that is not bound to a Cys thiolate binds to the amino acid carboxylate moieties of the cofactor SAM (37). The subsequent reductive cleavage of SAM generates a 5′-deoxyadenosyl radical (dAdo·) (Fig. 1C), which initiates the transformation of the substrate, typically through H-atom transfer (39–41). In addition to the radical SAM (RS) cluster which binds SAM, PapB also binds two auxiliary [4Fe-4S] clusters (ACs) (29, 42). H-atom abstraction from the methylene carbon neighboring the carboxylate functionality of an Asp or Glu acceptor residue in the PapA peptide is followed by thioether crosslink formation with a pendent Cys residue (Fig. 1D) (29, 42, 43).

Previous in vitro studies with purified PapB and synthetic PapA analogs have established that the acceptor Asp or Glu residue can be located adjacent to the donor Cys residue (i + 1 position) or separated by up to six residues (i + 7 position). The amino acid residues flanked by the required Cys and Asp/Glu residues can be changed to accept steric bulk (e.g., Phe), cationic charge (e.g., Lys), or D-configured residues at every position in the thioether motif (42, 43). Moreover, we demonstrated that selenol-containing substrates (e.g., SeCys) are processed to selenoethers, and demonstrated that a tetrazole-bearing sidechain functions as an acceptor residue mimic, forming an unnatural stable thioether crosslink with the donor Cys residue (27, 43). To gain insight into the H-atom abstraction event, we found that methylaspartate (Me-Asp) isomers, (2S,3S)-2-amino-3-methylsuccinic acid ((2S,3S)Me-Asp) and (2S,3R)-2-amino-3-methylsuccinic acid ((2S,3R)Me-Asp), are differentially processed by the enzyme (43). Notably, only the (2S,3R)-MeAsp-containing substrate is preferentially converted to the product, suggesting a requirement for a pro-S configured H-atom at the methylene position adjacent to the acceptor carboxylate.

The site-selective installation of thioether crosslinks in peptides remains a long-standing challenge in both chemical and enzymatic synthesis. Traditional chemical approaches often rely on prefunctionalized residues or the incorporation of noncanonical amino acid residues via solid phase peptide synthesis (SPPS) or engineered tRNAs—strategies that complicate synthesis and limit compatibility with in vivo systems. While enzymatic methods offer a compelling alternative, particularly for ribosomally synthesized peptides, most enzymes are presumed to have narrow substrate scopes. Although recent work has demonstrated the feasibility of constructing novel RiPPs using enzymes from multiple biosynthetic classes, the current enzymatic toolkit remains constrained by limited residue compatibility, strict spacing requirements, and a dependence on defined leader motifs (44). In the case of PapB, while the enzyme is agnostic to the residues positioned and the number of residues between the thiol and carboxylate moieties, crosslinking has thus far been restricted to acidic side chains such as Asp, Glu, homoglutamate (^homoGlu), and a tetrazole bioisostere. This regiochemical constraint imposes a significant limitation on the general applicability of PapB for building structurally diverse macrocyclic scaffolds. To unlock PapB’s full potential as a versatile biocatalyst, its activity must extend beyond acidic side-chain acceptors and remain effective even when the peptide backbone is built entirely from noncanonical residues. In this work, we demonstrate both capabilities, vastly enlarging the chemical space accessible to enzymatic macrocyclization.

Results

Formation of a C-Terminal Thioether Crosslink to the Peptide Backbone.

To explore whether the promiscuity of PapB can be expanded to C-terminal macrocyclization, we synthesized a peptide leader(ℓ)-CSANG, wherein the C-terminal residue (Gly) replaces the native Asp and presents the only available carboxylate moiety C-terminal to the Cys. The design of this substrate was motivated by prior findings that demonstrate the acceptor carboxylate-bearing residue must be positioned C-terminal relative to donor thiol-containing residue. The mass spectra of ℓ-CSANG before (black) and after (red) incubation with PapB in the presence of SAM and sodium dithionite (NaDT) are shown in Fig. 2A. NaDT is used by PapB as a source of reducing equivalents to reductively cleave SAM and generate the dAdo· to initiate catalysis. The unreacted substrate exhibits a monoisotopic m/z of 801.1055 (z = 3), which is within 0.87 ppm of the expected value. Incubation with PapB results in a 2 Da mass shift, consistent with the formation of a single thioether crosslink (observed monoisotopic m/z of 800.4330, ppm error = −0.75).

Fig. 2.

PapB installs a thioether to a C-terminal Gly residue to form a macrocyclized peptide. (A) Incubation (5 min) of ℓ-CSANG with PapB (5:100 enzyme:peptide ratio) leads to a 2 Da loss compared to the unreacted peptide (compare red and black spectra). (B) With ℓ-CSANG(2,2-d²), a 3 Da loss, corresponding to the loss of one hydrogen and one deuterium is observed after a similar treatment with PapB. The incomplete conversion to the product suggests a substantial kinetic isotope effect is associated with the reaction. (C) The dAdo generated in reactions with unlabeled ℓ-CSANG exhibits the expected monoisotopic m/z for unlabeled dAdo of 252.1090. (D) With the labeled substrate, the major species is singly deuterated dAdo. (E) A ℓ-CSANG peptide which is terminated by a carboxamide is not turned over by PapB even after 18 h incubation (see SI Appendix, Fig. S1 for MS traces). (F) In the presence of iodoacetamide (IAM), the peptide that is not treated with PapB is fully carbamidomethylated (see the black isotopic distribution, observed monoisotopic mass = 820.1135, ppm error = 0.97). When PapB is present, the modified peptide (observed monoisotopic mass = 800.4331, ppm error = −0.26) is unreactive to IAM, and only the peptide that remains unmodified and retains the free thiol is carbamidomethylated (observed monoisotopic mass = 820.1142, ppm error = 1.83).

Previous studies have shown that the thioether crosslinking reaction carried out by PapB entails H-atom abstraction from methylene or methine carbons that are α to the acceptor carboxylate moiety of the sidechain. To confirm that the modification of ℓ-CSANG proceeds by a similar pathway to form a C-terminal macrocycle, we prepared ℓ-CSANG(2,2-d²), bearing a C-terminal Gly(2,2-d²) residue. The mass spectrum of ℓ-CSANG(2,2-d²) before (black) and after (red) incubation with PapB in the presence of SAM and NaDT is shown in Fig. 2B. The unmodified peptide exhibits a monoisotopic m/z of 801.7761 (z = 3), which is within 1.12 ppm of the expected value. Under catalytic conditions, a new envelope is observed that displays a 3 Da mass shift of the monoisotopic peak to m/z 800.7700 (z = 3, ppm error = −0.12). This shift is consistent with the loss of one H-atom (1 Da) from the donor Cys residue and one D-atom (2 Da) from the acceptor Gly(2,2-d²) residue. PapB initiates crosslinking by H-atom transfer from peptide to dAdo. Indeed, analysis of the dAdo provides clear evidence that H- (Fig. 2C) or D-atom (Fig. 2D) transfer occurs from the position adjacent to the carboxylate in each substrate. The mass spectrum of the dAdo forming reaction with protiated substrate exhibits a peak m/z of 252.1090, and an additional peak at m/z of 253.1123 from the natural abundance ¹³C. With the deuterated substrate, the pattern is different, with a large peak at m/z of 253.1154, which is consistent with singly deuterated dAdo (0.4 ppm error from expected value). The natural abundance peak is still present but not resolved because of the larger deuterated dAdo peak. rSAM enzymes catalyze reductive cleavage of SAM to generate dAdo, in a reaction that is uncoupled from H-atom transfer from substrate leading to formation of unlabeled dAdo. As a result, some of the dAdo formed in the course of the incubation with the deuterated substrate is unlabeled, as evidenced by a peak at 252.1052. The reaction with ℓ-CSANG(2,2-d²) does not go to completion, whereas ℓ-CSANG is quantitatively converted to the crosslinked product (compare Fig. 2 A and B) in the same timeframe. Control experiments show PapB demonstrates no reactivity with a C-terminal carboxamide-terminated peptide (Fig. 2E), which confirms that a carboxylate moiety is essential for recognition (see SI Appendix, Fig. S1 for mass spectra).

The loss of the thiol in peptides processed by PapB was supported by treatment of the product with 2-iodoacetamide (IAM) (Fig. 2F). Treatment of the unmodified peptide with IAM leads to loss of the parent peak and formation of a new envelope with monoisotopic m/z of 820.1135 (z = 3), consistent with a single carbamidomethylation. Since the reaction of the peptide with PapB is nearly quantitative, only a small amount of unreacted peptide is available to be carbamidomethylated, as evidence by appearance of a minor envelope at m/z of 820.1142 in the red trace (SI Appendix, Fig. S2).

To quantify the rate of C-terminal macrocyclization and probe the chemical step that limits turnover, we monitored the PapB reaction with both the native peptide W-ℓ-CSANG and its Gly(2,2-d²) analog (Fig. 3). In the unlabeled substrate (Fig. 3A), a 2 Da mass loss, consistent with thioether formation, appears within seconds and approaches near-quantitative conversion on the minute time scale. Substituting Gly(2,2-d²) at the C-terminal position produces a 3 Da mass decrease (Fig. 3B), reflecting the loss of one protium and one deuterium during crosslink formation. Direct comparison of the progress curves (Fig. 3C) reveals that turnover of the deuterated peptide is slowed at least 6.8-fold, establishing a large primary kinetic isotope effect. This provides clear evidence that H-atom transfer is substantially rate-determining in catalysis. The data for each run and the triplicate analysis are shown in SI Appendix, Figs. S3–S5 and Table S2.

Fig. 3.

Kinetic analysis of PapB processing of W-ℓ-CSANG and W-ℓ-CSANG(2,2-d²) peptides. (A) A representative mass spectrum of the unlabeled peptide WLKQINVIAGVKEPIRAYGCSANG collected at t = 0 s (black) and after incubation with PapB for the indicated times. A 2 Da decrease, consistent with Cys thioether formation, is observed at every time point beyond t = 0, with the ppm error of every peptide < |3.71|. (B) Analogous spectra for the Gly(2,2-d²) substrate. The reaction now yields a 3 Da mass decrease at every time point beyond t = 0, with a ppm error of every peptide < |2.43|, consistent with the net loss of one hydrogen and one deuterium during crosslink formation. (C) Progress curve summarizing triplicate experiments in (A) and (B) for assays with unlabeled (black circles) and Gly(2,2-d²) (red squares) peptides. A primary kinetic isotope effect of ~7 is apparent by comparing the slope of the points in the progress curve between 0 and 30 s.

To further probe substrates that are tolerated for C-terminal macrocyclization, several ℓ-CSANX substrates where X = Ala, Lys, Phe, Asn, Ile, and Gln, were synthesized and evaluated by incubation with PapB in the presence of SAM and NaDT. None of these ℓ-CSANX substrates were processed by PapB (SI Appendix, Fig. S6) over an 18 h incubation period. We reasoned that the lack of activity may result from side chains occupying the position typically reserved for H-atom abstraction by dAdo·. In fact, our previous studies with ℓ-CSANXA substrates, where X is either an (2S,3S)Me-Asp or (2S,3R)Me-Asp acceptor residue, have shown that only the (2S,3R)Me-Asp-containing epimer is efficiently processed. Because of this diasteromeric preference exhibited by PapB, we hypothesized that PapB may process a substrate peptide, ℓ-CSANa (lower case amino acid abbreviations are used to denote D-amino acid residues), bearing a C-terminal D-Ala residue.

To test this hypothesis, the ℓ-CSANa peptide, which terminates D-Ala residue was synthesized. When incubated with PapB, the ℓ-CSANa was crosslinked (SI Appendix, Fig. S7). Moreover, an isotopically labeled substrate, ℓ-CSANa(2-d¹), containing a ²H at the α-carbon reacts to form a product with a 3 Da mass shift, unambiguously establishing that crosslinking occurs at the a-carbon of the D-Ala residue (SI Appendix, Fig. S8). Building on this initial experiment, several ℓ-CSANx substrates where x = D-Cys, D-Phe, D-His, D-Ile, D-Lys, D-Leu, D-Met, D-Asn, D-Pro, D-Gln, D-Arg, D-Ser, D-Thr, D-Val, D-Trp, and D-Tyr were synthesized and evaluated by incubation with PapB in the presence of SAM and NaDT. Apart from the D-Ile, D-Pro, and D-Thr-bearing substrates, all evaluated ℓ-CSANx substrates show a 2 Da loss upon incubation with PapB (Fig. 4 A–C). Consistent with the formation of a thioether, none of the corresponding products can be modified by IAM (SI Appendix, Figs. S9–S25). Substrates bearing C-terminal D-Pro residue are likely unreactive due to steric clashing and restricted conformationally flexibility. However, in contrast to the appreciable turnover we observe with the related D-Val and D-Leu variants, substrates terminating in D-Ile consistently yield no detectable product across multiple independent PapB preparations, suggesting that orientation of the D-Ile sidechain imposes a constraint that abolishes catalysis. Fascinatingly, the substrate with a C-terminal D-Thr residue (ℓ-CSANt) is processed to yield a product with a −46 Da mass shift (SI Appendix, Fig. S22). This may be the result of initial H-atom abstraction from the β-position followed by thioether formation to generate an intermediate O,S-acetal, which subsequently collapses culminating with a β-keto decarboxylation event (see Fig. 4D for the proposed mechanism). Indeed, evidence for the thioether-containing intermediate is present in the iodoacetamide-treated mass spectrum of the peptide (SI Appendix, Fig. S22). Subsequent treatment of this presumed ketone species with sodium borohydride results in the appearance of a species with a +2 Da shift, which is consistent with a reduction to the corresponding alcohol (SI Appendix, Fig. S26). This unique chemical transformation yields a biorthogonal chemical handle in peptides and can be potentially exploited for a myriad of applications, ranging from site-specific labeling to novel peptide conjugation strategies (45).

Fig. 4.

PapB crosslinks to C-terminal D-amino acid residues. Selective deuteration, lack of carbamidomethylation in the presence of IAM, and MS/MS give evidence for the crosslinking location in each example. (A) D-configured nonpolar side chains show differential reactivity with PapB. C-terminal D-configured sidechains cross-link at Cα (D-Ala, D-Leu, and D-Met), Cβ (D-Val, D-Phe, and D-Trp), or undergo no reaction (D-Ile and D-Pro). (B) D-configured polar side chains crosslink at either Cα (D-Cys, D-Ser, D-Asn, and D-Gln) or Cβ positions (D-Tyr). (C) The positively charged side chains all crosslink at Cα. (D) D-Thr undergoes modification after addition of PapB, forming a modified C-terminal amino acid. Evidence for the thioether crosslinked intermediate is observed in the iodoacetamide-treated sample (SI Appendix, Fig. S22). (E) Sactipeptide-like peptides demonstrate patterns that show fragments between the Cys and acceptor-residue, whereas ranthipeptide-like patterns show no fragmentation between the donor Cys and C-terminal acceptor residue.

To assign the regiochemical outcome(s) of experiments with ℓ-CSANx substrates, several isotopologs were prepared where the deuterated amino acid residues were commercially available and evaluated by incubation with PapB in the presence of SAM and NaDT. Similar to the ℓ-CSANa(2-d¹), the processing of ℓ-CSANm(2-d¹) results in the formation of a product with a 3 Da mass shift (SI Appendix, Fig. S27), consistent with the loss of one deuterium and a protium from the peptide. By contrast, when the ℓ-CSANv(2-d¹) peptide was incubated with PapB under catalytic conditions, only a 2 Da mass shift was observed, suggesting that H-atom transfer occurs from a different location (SI Appendix, Fig. S28). Indeed, with ℓ-CSANv(3-d¹), a 3 Da mass shift was observed, which is consistent with formation of a thioether to the C-3 position (SI Appendix, Fig. S29).

These results using deuterated isotopologs of C-terminal D-Val-bearing substrates are consistent with D-atom abstraction from the β-position. Together with the results from the experiments with ℓ-CSANt, these observations suggest that apart from D-Ile, C-terminal residues that are D-configured and β-branched undergo preferential H-atom abstraction at the β-position. Furthermore, substrates with sterically encumbered side chain functionalities appear to favor H-atom abstraction from the β-position. For example, ℓ-CSANf and ℓ-CSANy are processed to products with a 3 Da mass shift, consistent with thioether formation at the C3 (SI Appendix, Figs. S30–S33). These observations blur the distinction between sactipeptides and ranthipeptides, which are arbitrarily assigned based on physiological function. But rather, the actual outcome is likely dictated by a combination of substrate fit and enzyme function.

For ℓ-CSANw, the commercially available selectively deuterated amino acid included deuterium enrichment at both the α-and β-carbons (SI Appendix, Fig. S34). Therefore, tandem mass spectrometry was used to determine the regiochemistry of the crosslink. There is a significant difference in bond stability between sactipeptide crosslinks and ranthipeptide crosslinks (25). If the crosslinking occurred at the α-position, then the resulting peptide would fragment to reveal a 2 Da loss in each b-fragment after the C-terminal w-residue. However, if β-crosslinking occurred, then no fragmentation would be expected due to the increased stability of the β-crosslinked macrocycle (see Fig. 4E for the differences in fragmentation between sactipeptides and ranthipeptides). These differences allow MS/MS analysis to be used to distinguish between the two regichemical outcomes of the reaction. The tandem mass spectrometry results consistently revealed the presence of stable β-thioether crosslinks for ℓ-CSANw. No fragmentation occurs between the Cys and the C-terminal residue, mirroring the patterns we observe with ℓ-CSANf, ℓ-CSANv, and ℓ-CSANy (see Fig. 4 A and B in the blue dotted cutout for the resulting connectivity). In all cases, tandem mass spectrometry corroborates sactipeptide-like linkages formed for ℓ-CSANa, ℓ-CSANc, ℓ-CSANh, ℓ-CSANk, ℓ-CSANi, ℓ-CSANm, ℓ-CSANn, ℓ-CSANq, ℓ-CSANr, and ℓ-CSANs substrates (SI Appendix, Figs. S35–S44), and stable ranthipeptide-like processing for substrates that bear bulky C-terminal D-configured residues (SI Appendix, Figs. S45–S48).

The findings from the D-amino acid experiments are summarized in Fig. 4. It is notable that the identity of the substrate appears to drive regiochemistry of the crosslink, C2 or C3, underscoring to some extent the arbitrary nature of ranthi- versus sactisynthase designations.

PapB Effectively Crosslinks to C-Terminal Residues that Derive from β-Amino Acids.

The promiscuity exhibited by PapB for substrates bearing D-configured C-terminal residues inspired us to probe the tolerance for substrates that derive from commercially available β-amino acids. We reasoned that the one carbon extension might offer flexibility to build substrates from both L- and D-configured β-amino acids at the C-terminus. In an initial experiment, ℓ-CSAN^βA was prepared using the corresponding Fmoc-^βAla and incubated with PapB in the presence of SAM and NaDT. In the 5 min incubation, the peptide is quantitatively processed (SI Appendix, Fig. S49), as evidenced by a 2 Da loss in the product relative to the substrate. This substrate presents two methylenes that are positioned α- and β- relative to the C-terminal carboxylate. To unambiguously establish the site of attachment, we prepared two dimethylated substrates, ℓ-CSAN^βA(2,2-Me,Me) and ℓ-CSAN^βA(3,3-Me,Me) and evaluated their reactivity. The substrate ℓ-CSAN^βA(2,2-Me,Me), where the α-position is effectively blocked by methyl groups, did not react even after 18 h incubation (SI Appendix, Fig. S50). By contrast, ℓ-CSAN^βA(3,3-Me,Me) is quantitatively crosslinked by PapB in a 5 min incubation (SI Appendix, Fig. S51). These observations strongly point to crosslinking occurring at the α-methylene, which is consistent with previously reported ℓ-CSANDA, ℓ-CSANEA, and ℓ-^homoCSAN^homoEA substrates. To better understand processing diastereospecificity, we prepared and evaluated diastereomeric substrates, ℓ-CSAN^βA(2-Me) (from Fmoc-(S)-3-amino-2-methylpropanoic acid) and ℓ-CSAN^βa(2-Me) (from Fmoc-(R)-3-amino-2-methylpropanoic acid). The results were distinct, with only ℓ-CSAN^βA(2-Me) being crosslinked (compare SI Appendix, Figs. S52 and S53). This diastereoselective reactivity is reminiscent of the preference of PapB for ℓ-CSANa, as well as previously reported data with where (2S,3R)Me-Asp is processed and (2S,3S)Me-Asp is not (Fig. 5A). Collectively, these observations provide strong support for PapB catalyzing the abstraction of a proS-configured H-atom of the carbon. This specificity persists even when the chiral center is transposed due to the introduction of an extra methylene (see Fig. 5A for the summary of stereochemical outcomes).

Fig. 5.

PapB installs crosslinks to C-terminal β-amino acids. (A) Each of the stereochemical outcomes of specific methylations adjacent to the carboxylate moiety are shown. No crosslinking is observed if the position α-to-the-carboxylate is dimethylated (SI Appendix, Fig. S50). PapB crosslinks the peptide if the position that is β-to-the-carboxylate is dimethylated (SI Appendix, Fig. S51). (B) With few exceptions, peptides with nonpolar C-terminal amino acids are crosslinked by PapB, including N-methylated and D-configured variants. (C) Polar C-terminal β-amino acids are crosslinked by PapB. (D) C-Terminal basic β-amino acids are crosslinked by PapB.

The ability of PapB to crosslink substrates bearing D-configured and β-amino acid-derived C-terminal residues suggested that other noncanonical amino acids could participate in crosslinking. We systematically probed the ability of PapB to process substrates derived from many commercially available L- and D-configured β-amino acids, as well as those bearing N-methylated β-amino acids (see Fig. 5B for nonpolar β-amino acids, 5C for polar β-amino acids, and 5D for basic β-amino acids). A uniform 2 Da mass shift is consistently observed in most cases upon incubation with PapB (SI Appendix, Figs. S54–S69) to form IAM-insensitive products. The only exception was ℓ-CSAN^β-homoP (SI Appendix, Fig. S64), which as with ℓ-CSANp, may not undergo H-atom abstraction due to limited conformational flexibility. As anticipated, a single carbamidomethylation is observed in ℓ-CSAN^β-homoC (SI Appendix, Fig. S59), and tandem mass spectrometry results are consistent with the formation of a stable ranthipeptide linkage (SI Appendix, Figs. S70–S83). The results for substrates derived from D-configured β-amino acids, as well as those bearing N-methylated β-amino acid residues, follow the same pattern (SI Appendix, Figs. S84–S87). Unsurprisingly, the C-terminal β-homothreonine- and β-homoserine-containing peptides are the predominant forms after solid phase peptide synthesis and HPLC purification. Since these peptides do not have a free C-terminal carboxylate, they are not processed by PapB (SI Appendix, Figs. S65 and S66). However, a minor species corresponding to the expected linear mass of the parent peak was observed and underwent the characteristic 2 Da shift forming a species that is not reactive toward IAM.

PapB as Sequence Agnostic Thioether Crosslink Catalyst.

The most frequently employed current strategy for the formation of macrocyclized peptides entails the use of electrophilic groups (e.g., 2-chloroacetamide), which can chemoselectively react with Cys residues to form a thioether linkage. However, the in vitro incorporation of alkyl-halogenated amino acid residues can be challenging, demanding a re-engineered tRNA translation system, and the use of evolved tRNA synthetases or flexizymes to achieve site-specific aminoacylation on the tRNA of interest (23–25, 46, 47). Synthetically, it is difficult to install alkyl halides into complex peptides in a regiospecific and stereoselective manner. Accordingly, synthetic residues that can be incorporated into peptides have been shown to serve as reactive intermediate precursors. Baran and coworkers showed that cyclic peptides bearing cyclic N,S-acetals can be formed via the in situ generation of electrophilic cyclic iminium intermediates (48). Ting and coworkers recently demonstrated that a related iminium ion-based strategy enables access to the sactipeptide, enteropeptin-A (49, 50). Skrydstrup and coworkers have developed strategies to reductively generate reactive glycyl radicals from their α-linked 2-pyridyl sulfide precursors to enable carbon–carbon bond-forming peptide modifications (51). Malins and coworkers have shown that sactipeptide linkages can be assembled via electrophilic glycine synthons that derive from oxidized Gly residues (e.g., α-bromo Gly, α-hydroxy Gly) (21). Outside of sactipeptide-type linkages, libraries of thioether macrocycles have been prepared by the intramolecular alkylation of embedded thiol-containing residues (e.g., Cys) (52–54). These types of strategies share the need to assemble inherently reactive oxidized precursors (e.g., alkyl halides, alkyl sulfides), prompting the development of carefully planned routes and protection schemes that would be difficult to control in more advanced scenarios (55). Considering these challenges, the demonstrated versatility of PapB supports its further development as a tool for the preparation of macrocycles with stereo-defined thioether linkages. Additionally, given the demonstrated promiscuity of PapB in processing a wide range of carboxylate-presenting substrates, we wondered whether artificial substrates composed entirely of noncanonical residues within the thioether motif would be tolerated.

We designed the peptide LRKQINVIAGVKEPIRAYENLYFQG(^β-homoC)(^β-homoS)(^β-homoA)(^β-homoN)(^β-homoD)A to test the applicability of PapB in macrocyclization workflows. The peptide has multiple unnatural β-homoamino acids, which are linked to a leader sequence via a tobacco etch virus (TEV) recognition sequence. The TEV protease could be used to liberate the core after cyclization (Fig. 6A). Incubation of the peptide with PapB leads to quantitative conversion to a modified species exhibiting a loss of 2 Da (Fig. 6B). Subsequent TEV protease incubation yielded fragments consistent with the core sequence (Fig. 6C).

Fig. 6.

PapB introduces thioether motifs in peptides that are entirely composed of nonproteinogenic amino acids. (A) Strategy for synthesis of cyclic peptide composed of entirely nonproteinogenic amino acids. The insertion of a TEV protease recognition sequence between Y17 and G18 of the leader sequence allows for liberation of the macrocyclic core. (B) Mass spectra of the peptide in (A) before (black, [M + 4H]⁴⁺ expected monoisotopic m/z 895.9710, observed monoisotopic m/z 895.9735, ppm error = 2.79) and after (red, [M + 4H]⁴⁺ expected monoisotopic m/z 895.4671, observed monoisotopic m/z = 895.4660, ppm error = -1.2) treatment with PapB. (C) Addition of TEV protease liberates the macrocyclic core from the leader peptide. The TEV protease treated peptide without PapB treatment is shown in black ([M + 1H]¹⁺ expected monoisotopic m/z = 707.3029, observed monoisotopic m/z = 707.3036, ppm error = 0.99) while the PapB and TEV protease treated peptide is shown in red ([M + 1H]¹⁺ expected monoisotopic m/z = 705.2872, observed monoisotopic m/z = 705.2866, ppm error = −0.85). (D) An analog of a peptide reported by Novartis to have oral availability and passive permeability in rats (53) was generated using PapB. The peptide contains norleucine (Nle) and N-methylated Nle residues within the thioether motif. (E) The mass spectrum of the linear analog is shown in black ([M + 3H]³⁺ expected monoisotopic m/z = 905.2116, observed monoisotopic m/z = 905.2128, ppm error = 1.32), while the PapB crosslinked peptide is shown in red ([M + 3H]³⁺ expected monoisotopic m/z = 904.5397, observed monoisotopic m/z = 904.5397, ppm error = 0).

Recently, Novartis reported a series of orally bioavailable macrocyclic peptides with the sequence XBZBZC, where B and Z denote norleucine and N-methylnorleucine, respectively (56). The terminal amino acid, X, was a chlorinated phenylalanine, allowing cyclization by crosslinking with the Cys sidechain. As proof-of-concept for the use of PapB in thioether crosslink workflows, we synthesized ℓ-CBZBZ-(^β-homoF). Upon incubation of the peptide with PapB, a 2 Da shift in mass of the peptide is clearly visible (Fig. 6 D and E). These data underscore the utility of PapB as a general macrocyclization catalyst.

Discussion

Enzymes are widely regarded as ideal catalysts due to their ability to promote challenging chemical transformations under mild, physiologically relevant conditions. Throughout evolution, substrate-specific enzymes have arisen to meet precise metabolic and regulatory needs, typically exhibiting high fidelity and selectivity for their native substrates. As a result, the canonical view of enzymatic specificity, first articulated by Fischer’s lock-and-key model and later refined by Koshland’s induced-fit hypothesis, has been shaped by studies of metabolic enzymes acting on small-molecule substrates. These paradigms emphasize complementarity and dynamic adaptability between enzyme and substrate as the basis for specificity.

PapB, however, presents a departure from this framework. Despite being a relatively compact 467-residue enzyme, PapB catalyzes thioether crosslinking across a remarkably broad range of peptide substrates. In contrast to expectations derived from the metabolic enzyme literature, PapB efficiently modifies peptides containing β-amino acids, D-configured residues, and even N-methylated backbones. Normally, these types of modifications would render a substrate unrecognizable by the cognate enzyme. In effect, PapB, through no directed evolution or engineering, has inherent promiscuity that is completely unanticipated.

From an evolutionary standpoint, the behavior of PapB suggests that some RiPP-modifying enzymes may have evolved not for tight substrate discrimination but rather for modular interaction with recognition elements that allow flexible catalytic activity. In the case of PapB, substrate recognition appears to be dominated by binding of the leader peptide to the RiPP recognition element (RRE), which likely functions as a docking module (29). This interaction guides the core region of the peptide into the catalytic domain with sufficient fidelity to position key reactive groups, while tolerating wide variability in the remainder of the sequence. In this model, the crosslinking chemistry is dictated by local recognition of a cysteine thiol and an acceptor residue bearing an acidic or otherwise reactive side chain, with only minimal tertiary contacts between the enzyme and the intervening sequence. The active site positions the thiol and target C–H bond in proximity to the SAM cofactor, which is bound to the rSAM cluster. Reductive cleavage of SAM generates the dAdo·, which abstracts the appropriate hydrogen atom and initiates crosslink formation. The substrate radical is resolved by the formation of a thioether crosslink and reduction of AC1. PapB also has an additional auxiliary cluster, AC2, which has been proposed to be involved in recycling the reducing equivalent from AC1 to the RS cluster in a process that entails intermolecular transfer (55).

The precise control over catalytic geometry paired with tolerance for variation in unreacted regions may be an emergent feature of RiPP biosynthetic enzymes more generally. Because RiPP substrates are genetically encoded and often highly variable in nature, the evolutionary pressure may favor enzymes that maintain reactivity through modular leader-mediated recognition while remaining permissive to core diversity. The behavior observed with PapB is thus likely not an exception but rather a hitherto underappreciated property of many RiPP-tailoring enzymes. The broad substrate scope of PapB highlights the promise of enzymatic macrocyclization as a versatile alternative to limited metal-catalyzed C–H functionalization strategies (57).

If this model proves general, it has significant implications for the use of RiPP enzymes as biocatalysts. Enzymes like PapB could serve as powerful tools for site-selective modification of synthetic peptides, including those bearing unnatural features that are difficult to accommodate through traditional chemistry or enzyme engineering. Broadening our understanding of how such enzymes balance specificity and promiscuity will be essential to fully unlocking their potential for biocatalysis, therapeutic development, and peptide engineering.

Materials and Methods

PapB was overexpressed in Escherichia coli and purified by the previously reported Ni-NTA workflow (42). Linear msPapA variants were synthesized on a Chorus synthesizer (0.025 mmol, 2-chlorotrityl resin) by standard FMOC procedures and purified by preparative HPLC on a C12 column. Anaerobic macrocyclization assays (200 µL) were carried out for 15 min at 22 °C in 50 mM PIPES-KOH pH 7.4, 150 mM KCl, 15% glycerol, 2 mM SAM, 2 mM NaDT, 200 µM peptide, and 2 µM PapB. Before TCA quenching, aliquots were treated, as appropriate, with 10 mM IAM to alkylate unreacted cysteines or digested with TEV protease after spin-filtration. Reaction mixtures and purified peptides were analyzed on a Vanquish UHPLC in-line with a Q-Exactive HRMS (70,000 resolution) using a C18 column. Crosslinked peptides yields were calculated from monoisotopic peak areas and modification sites were verified by targeted CID MS/MS. Detailed protocols for enzyme expression and purification, peptide synthesis and purification, macrocyclization assays, mass spectrometry analysis, and tandem MS data interpretation are provided in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.