Bioconjugate Platform for Iterative Backbone N-Methylation of Peptides

Abstract

N-methylation of peptide backbones has often been utilized as a strategy towards the development of peptidic drugs. However, difficulties in the chemical synthesis, high cost of enantiopure N-methyl building blocks, and subsequent coupling inefficiencies have hampered larger-scale medicinal chemical efforts. Here, we present a chemoenzymatic strategy for backbone N-methylation by bioconjugation of peptides of interest to the catalytic scaffold of a borosin-type methyltransferase. Crystal structures of a substrate tolerant enzyme from Mycena rosella guided the design of a decoupled catalytic scaffold that can be linked via a heterobifunctional crosslinker to any peptide substrate of choice. Peptides linked to the scaffold, including those with non-proteinogenic residues, show robust backbone N-methylation. Various crosslinking strategies were tested to facilitate substrate disassembly, which enabled a reversible bioconjugation approach that efficiently released modified peptide. Our results provide general framework for the backbone N-methylation on any peptide of interest and may facilitate the production of large libraries of N-methylated peptides.

Graphical Abstract

INTRODUCTION

Peptides have long attracted interest in diverse pharmaceutical applications due to ease of synthesis, predictive biophysical properties, and chemical diversity of the building blocks, which are then amenable to further modifications ^1–3. Starting with the first therapeutic uses of insulin in diabetics nearly a century ago ⁴, the application of peptides as drugs has continued to evolve owing to enabling advances in peptide synthesis and characterization ^5–7. In recent years, the discovery of venoms from gastropods and arthropods with bioactive properties against defined eukaryotic receptors has driven targeted therapeutic efforts ^8–11. Despite the therapeutic promises of several peptides isolated from nature, wider efforts for de novo production of peptide-based therapeutics have been limited due to many drawbacks such as lack of oral bioavailability and proteolytic susceptibility, among others.

Numerous strategies have been adapted by medicinal chemists to overcome the liabilities of in vivo stability and bioavailability with the goal of turning candidate peptides into drug-like compounds ^{2, 12–13}. Examples of successful approaches include the incorporation of peptide bond isosteres, peptidomimetics, peptoids, among others ^14–16. However, such strategies are often synthetically challenging and require attention to design for high-level production. Despite these challenges, there are many successful endeavors in designed peptide therapeutics such as the type 2 anti-diabetes drugs dulaglutide (trade name: Trulicity) and liraglutide (trade name: Victoza), which consists of a fragment of the glucagon-like peptide-1 covalently linked to either an Fc fragment of human IgG4, or palmitic acid, respectively ^17–18. Other examples of peptidic drugs include the anti-cancer drug carfilzomib (trade name: Kyprolis), a synthetic analog of the proteasome inhibitor epoxomicin ¹⁹, and romiplostim (trade name: Nplate), an analog of thrombopoietin used for the treatment of blood clotting disorders ²⁰.

While most of the ~80 peptide drugs that are currently approved world-wide are synthetic analogs that target extracellular hormone receptors ², newer efforts have extended focus towards the design of peptidic protein kinase inhibitors ²¹, and peptides that can disrupt protein-protein interactions ^22–23. The use of bio-panning methods, such as surface display, has enabled the identification of new peptide scaffolds that can be further optimized through medicinal chemistry to improve bioavailability and in vivo stability ^24–25. N-methylation of the amide backbone has been used in medicinal chemistry for years as a strategy to endow peptides with drug-like properties ²⁶. These approaches may have been inspired by the number of naturally occurring N-methylated bioactive products, such as cyclosporin (1), the lagunamides (2) ²⁷, and the talaropeptides (3) ²⁸, among others, all of which show desirable pharmacological profiles (Figure 1A). For example, the cyclic peptide cyclosporin demonstrates excellent oral bioavailability (~30%) despite violating Lipinski’s rules ²⁹. Synthetic efforts have devised several strategies for the efficient production of N-methylated peptides on solid phase and in solution and have been effectively used to demonstrate improvements in oral bioavailability ^30–33. Despite these remarkable outcomes, synthetic and semi-synthetic designs are low throughput and likely require careful design considerations with multiple iterations for success.

Figure 1:

(A) Chemical structures of N-methylated peptides of natural origin (clockwise from left cyclosporin (1), lagunamide A (2), and talaropeptide A (3). (B) Arrow diagram showing RiPP biosynthetic logic for fused borosin methyltransferases. The chemical structure of omphalotin (4), the first characterized borosin is shown. Cartoon schematic of a homodimeric fused borosin methyltransferase complex. Noting how each methyltransferase domain acts in trans on its peptide substrate. (C) C-terminal regions of borosin methyltransferases initially investigated in this study. Residues after the red dotted line represent the peptides analyzed by MALDI-TOF MS. Blue circles represent residues that are methylated as determined by MS/MS analysis. Circles with dashed borders represent likely positions of methylation. MALDI-TOF MS traces validating in vivo (black) and in vitro (red) for the corresponding methyltransferases. (D) C-terminal regions of the MroM-cyclosporin like peptide (CspL) and MroM-A2 constructs (top) and MALDI-TOF MS traces validating in vivo (black) and in vitro (red) for the corresponding constructs (bottom). Residues after the red dotted line represent the peptides analyzed by MALDI-TOF MS. (E) Core peptides were appending to MroM_1–388. The backbone amide methylations were analyzed by MALDI-TOF MS. Details of MS analysis can be found in Figure 1C, Figure 1D and Figure S7. The expected and observed m/z are summarized in the Table S4.

The borosins (Figure 1B), a class of cyclic ribosomally synthesized and post-translationally modified peptide (RiPP) natural products ^34–35 that contain characteristic α-N-methylated backbone amides (4). Methyltransferases involved in borosin biosynthesis of are viable catalysts for the biocatalytic modification of peptide substrates ^36–37. Borosin biosynthesis involves an enzyme consisting of a single chain of an N-terminal S-adenosylmethionine (SAM) dependent catalytic domain (designated M) fused in frame with a C-terminal core sequence (designated as A). Residues in the A core are post-translationally modified by the catalytic M domain to introduce backbone α-N-methylation ^38–39. The α-N-methylated core is then excised and cyclized by a protease to elaborate the modified peptide product ³⁷. In the common fused systems, as the M catalytic domain only modifies the covalently fused core sequence, proteolytic removal of the modified A core produces a spent catalytic domain.

Recent reports have identified naturally occurring ‘split’ borosin systems, in which the methyltransferase and the core peptide are encoded separately ^40–42. However, the core peptide in this system is α-helical and this secondary structure may be needed to drive the iterative reaction ^{40, 42}. Hence, the ‘split’ borosin system may not be applicable for biocatalytic conversion of a broad range of peptide substrates. In a second study, an intein-based approach has been reported, in which short peptide substrates are conjugated to the C-terminus of the methyltransferase catalytic core by thiol-thioester exchange ⁴³. This ligation approach is powerful and can effectively catalyze backbone methylation on peptides incorporating non-natural amino acids. However, ligation of the peptide substrate is limited by the slow second order rate constant for native chemical ligation (~0.067 M⁻¹ s⁻¹) ⁴⁴. To prevent the hydrolysis of thioester in the catalytic domain, millimolar concentrations of peptide substrate must be added to the ligation reaction. Consequently, the thiol-thioester exchange-mediated ligation approach is limited to peptide substrates that are soluble at millimolar concentrations, and even then, much of the peptide will be free in solution not ligated to the scaffold.

Here, we present a one-pot biocatalytic platform for the chemoenzymatic backbone N-methylation on any peptide substrate. We identified a suitable enzymatic scaffold by testing various homologs of borosin-like α-N-methyltransferases. High-resolution crystal structures of the enzyme from Mycena rosella allowed for the rational demarcation of a catalytic domain that retains activity independent of the core peptide. We tested various covalent tethering methods and identified viable strategies that allow for the facile linking of non-native peptides to the catalytic domain, allowing for backbone α-N-methylation on the linked peptide. Judicious choices of covalent linking strategies identified a process wherein a peptide synthon could be linked to, α-N-methylated by, and then removed from the catalyst which can be reused in principle. This strategy represents a reversible and iterative bioconjugative platform for the installation of methyl groups on the amide backbone of any peptide of interest.

RESULTS

Selection criteria for a biocatalytic scaffold

To develop a general methodology for backbone N-methylation that is broadly tolerant of substrate peptides includes those of low solubility, we sought to identify a viable catalyst along with an expedient ligation strategy. The catalytic efficiency for autocatalysis varies greatly among different borosin methyltransferase homologs and most are poor catalysts ^{39, 45}. Prior genome mining efforts have identified borosin biosynthetic clusters consisting of nearly identical catalytic domains but divergent core sequences suggesting that these homologs may demonstrate some degree of substrate tolerance ⁴⁶. Synthetic genes corresponding to homologs of the full-length autocatalytic borosin α-N-methyltransferases (containing both the catalytic domain and the core sequence; designated M-A) were cloned into suitable vectors for expression in E. coli. We screened a total of 5 different catalytic domains with 8 different peptide substrates (see Table S1 and S2 for accession codes and sequences). Selection of a suitable catalytic scaffold focused on enzymes that showed high expression levels in this heterologous host and demonstrated efficient autocatalytic activity. Effort was devoted to identifying homologs that were fusions of near-identical catalytic domains with variant core peptides (Figure S1). We reasoned that a catalyst that can function on different core peptides would be the ideal homolog for our biocatalytic platform to modify peptides of diverse sequences.

Turnover efficiencies were determined by monitoring the extent to which each purified methyltransferase-core fusion homolog could carry out complete modification on the core peptide sequences. We first monitored the number of methylations installed on the core peptide in vivo. Core peptides from candidate homologs were excised using GluC protease, and the extent of N-methylation was assessed using matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) mass spectrometry. Each homolog was then purified, incubated with exogenous SAM, and the proteolytic core A peptide was assessed for the number of methylations installed in vitro. The identity of the modified residues was determined using tandem (MS/MS) mass spectroscopy (Figures S2-S4).

Several candidate homologs were identified as robust catalysts both in vivo (CmaM-A from Coprinopsis marcescibilis), and in vitro (SveM-A from Serendipita vermifera ssp. bescii). We directed further attention to the enzymes from Mycena rosella (MroM-A1 and MroM-A2) for several reasons (Figure 1C). First, the core peptide that is proteolytically cleaved from the recombinantly purified MroM-A1 shows a mixture of intermediates, but these intermediates converge to a single species with six backbone N-methylations when the purified enzyme is incubated with exogenously provided SAM. Hence, the purified protein demonstrates robust catalytic activity in vitro, which was not observed for certain borosin methyltransferases ^37–38. Second, Mycena rosella contains two borosin methyltransferases in which the catalytic domains are near identical in sequence (97% sequence identity between MroM catalytic domains) but contain different core peptides A1 and A2 (Figures 1C & D). This divergence of the core peptides suggests that MroM catalytic domain must have some degree of promiscuity for the sequence of the peptide substrate.

We next tested the substrate tolerance of the MroM catalytic domain by appending the sequence of the A2 peptide in place of the core sequence in MroM-A1 (Figure 1D). Upon purification of the fusion protein, incubation with SAM, and GluC cleavage, up to six backbone N-methylations could be observed on the peptide (Figure 1D, Figure S5). As a further test for substrate tolerance, we replaced the core peptide with a sequence derived from cyclosporin (designated as CspL for cyclosporin-like) ⁴⁷. Tandem mass spectrometric analysis of the GluC-proteolytic fragment of this fusion, following in vitro incubation with SAM showed up to seven backbone N-methylations on this peptide (Figure 1D, Figure S6). To further demonstrate that the MroM catalytic domain can accept divergent substrate peptides, the C-terminal substrate peptide was replaced individually with sequences derived from omphalotin (MroM-Oph), lagunamide (MroM-Tag), talaropeptide (MroM-Tal) and three random peptides (MroM-R_n; n = 1–3). All the fusion proteins were expressed in E. coli for 18 hr. After purification and GluC cleavage of the fusion proteins, the extent of methylation was determined by MALDI-TOF analysis. Methylation could be observed on all substrate peptides (Figure 1E, Figure S7). Interestingly, despite the presence of a turn-inducing Pro embedded in the middle of the core sequence, the talaropeptide-like peptide was still modified by up to six backbone amide methylations. In contrast, prior studies demonstrated that neither of the homologs from Omphalotus olearius (OphM-A) or Dendrothele bispora (dbOphM-A) could tolerate Pro residues in the core peptide ^{38, 45}. These data are consistent with the broad tolerance of MroM catalytic domain for sequence diverse substrates, and hence suitable for our bioconjugate approaches.

Structural basis for the substrate tolerance of MroM catalytic domain

To understand how the MroM catalytic domain can function on divergent substrate peptides, we determined the high-resolution crystal structures of full-length MroM with peptide A1 (MroM-A1 to 1.79 Å resolution), peptide A2 (MroM-A2 to 1.86 Å resolution), and the CspL core sequence (MroM-CspL to 1.93 Å). Each structure was determined in complex with product S-adenosylhomocysteine (SAH). Clear electron density can be visualized for SAH and for the entire polypeptide in the A1 cocrystal structure, and all but residues Asp384-Thr397 in the A2 and cyclosporin-like core structures. Each of the crystal forms contains four copies of the respective methyltransferase-core fusion in the crystallographic asymmetric unit, ruling out crystal packing artifacts. The statistics of data collection and refinement are summarized in Table S3.

The structures of MroM bound to each peptide recapitulates the architecture common amongst borosin α-N-methyltransferases and consists of a core α/β fold methyltransferase domain (Lys5-Gln276), an extended region (Gly283-Val304), a helical bundle domain (Ala310-Ala383), and the core/follower peptides (Figure 2A) ^38–39. Two molecules of MroM-peptide form an intertwined homodimer wherein the core peptide from one monomer is bound to the active site of the second monomer (Figure 2B). The overall structure is similar to that of OphM-A (RMSD of 1.6 Å over 364 α-carbon residues aligned) ³⁹ and dbOphM-A (RMSD of 1.5 Å over 363 α-carbon residues aligned) ³⁸, with the exception that residues between the helical bundle domain and the core peptide are now either partially or fully resolved in the MroM-peptide structures. Unambiguous electron density corresponding to amide N-methylation can be observed at residues Ile391, Ile392 and Tyr393 in the A1 structure (Figure 2C), at residues Leu391 and Val392 in the A2 structure (Figure 2D), and Leu394 and Leu395 in the cyclosporin structure (Figure 2E). The α-N-methyl group of the last residue modified in each structure is between 3.2–3.4 Å away from the sulfur atom of SAH consistent with a direct S_N2-like nucleophilic attack mechanism. Lastly, as observed in the structures of OphM-A and dbOphM-A, there are no side-chain specific hydrogen bonding interactions between the enzyme and the peptide substrate. In each of the three MroM-peptide structures, only residues involved in amide bond deprotonation (Tyr64 Arg70 and Tyr74) are within hydrogen bonding distance with the peptide (Figure S8).

Figure 2.

(A) Structure of the MroM-A1 monomer and (B) interlocked dimer with a diagram of the monomer domain architecture (below, centered). (C-E) Difference Fourier difference maps (F_o-F_c), calculated with the substrate peptide (contoured at 3σ above background) omitted for one round of refinement prior to map calculation, for (C) A1, (D) A2, and (E) CspL. Methylated residues are indicated with an asterisk (*) and are also boxed. Refined atomic coordinates are superimposed and active site residues shown as sticks, methylated residues are colored orange, and SAH is shown in tan. (F) MroM-A2 monomer (green) superimposed onto the MroM-A1 dimer (pale blue and pale pink). The A1 substrate peptide (cyan) is shown, highlighting the position of Tyr389 (yellow) and how Tyr96 in MroM-A1 (pink) shifts away to accommodate the phenyl group (red arrow). Tyr96 in MroM-A2 (green) would clash with Tyr389.

A superposition of the three MroM structures revealed that there is some degree of flexibility among the orientations of the substrate peptides (Figure S9). Notably, the A1 peptide contains a central Tyr393 residue. In the MroM-A2 structure, Tyr96 shifts away from the active site to accommodate the phenyl group of the substrate (Figure 2F), providing a rationale for the tolerance of MroM for substrates with bulky aromatic residues at this position. Two factors likely account for the broad tolerance of MroM. First, the active site entrance of MroM is flanked by extended loops (Figure S10) that likely promote flexibility for substrate peptide accommodation. The analogous regions in OphM-A, dbOphM-A, and the recently characterized split borosin-like system SonM-SonA ⁴² are surrounded by more compact loops or are devoid of additional loops (Figure S10). Secondly, the fully resolved A1 structure reveals that MroM contains an additional short α-helix (Figure S10) (Gln377 – Ala381) that allows for core peptide anchoring, which may increase the residence time of substrates in the active site such that even less favorable substrates are processed. In contrast, electron density is absent for the region immediately preceding the core sequences in OphM-A and dbOphM-A. Together, these factors promote core peptide substrate tolerance in MroM, consistent with the tolerance for sequence distinct substrates.

Utility of the MroM scaffold for backbone amide N-methylation on peptide conjugates

Given the broad substrate tolerance observed for MroM, we focused on utilizing this scaffold as a generic catalyst for backbone modification. Guided by our structural data, full-length MroM-A1 was decoupled into two parts using the helical bundle as a break point. We generated constructs in which the catalytic domain was split at the junctures between Lys354 and Pro355 (MroM_1–354), between Arg363 and Gln364 (MroM_1–363), and between Gln373 and Glu374 (MroM_1–373) and tested each for in vitro activity against core peptides provided in trans. However, in each case only the unmodified core peptide could be observed (Figure S11). These data demonstrate that MroM can only modify substrates that are covalently linked.

We next focused on tethering strategies to attach diverse peptides to the catalytic domain utilizing each of the truncated catalytic domains but with the addition of a Cys residue at the C-terminus (MroM_1–354Cys, MroM_1–363Cys, MroM_1–373Cys; hereafter collectively referred to as MroM_TruncantCys). Various C-terminal peptide fragments (5-10 in Figure 3A) were purified from E. coli cells as poly-histidine tagged maltose binding protein (MBP) fusions. Following proteolytical removal of the MBP tag, the isolated peptides were conjugated at the α-amino group to the heterobifunctional crosslinker 3-maleimidopropionic acid N-hydroxysuccinimide (BMPS), which is used to generate a stable maleimide-activated substrate that can spontaneously react with sulfhydryls (Figure 3B) ⁴⁸. Incubation of the MroM_1–373Cys scaffold with the BMPS-peptide in aqueous buffer (Tris, pH=8.0) resulted in the facile production (~15 minutes) of mostly single species as the major product (Figures S12-S13). To confirm the C-terminal Cys as the site of modification, the MroM_1–373Cys-BMSP-5 conjugate was extensively digested using trypsin and subjected to MALDI-TOF analysis. The major fragments show lack of modifications on lysines or the other two cysteines in MroM (Figure S14). However, we were unable to observed masses corresponding to BMPS-5 attached to the C-terminal Cys residue, likely due to insolubility of this peptide. To further identify the modification site, we synthesized a short hydrophilic peptide sequence derived from peptide 5 (sequence: ADDATAFII), modified this peptide with maleimide, and conjugated it to MroM_1–373Cys. The conjugate was digested by GluC MALDI-TOF-MS analysis show that the modification occurs mainly at the C-terminal Cys residue (Figure S15-16).

Figure 3:

(A) Sequences of peptides that were modified with a maleimide group for subsequent ligation to the MroM catalytic domain. CT indicates the carboxy-terminus of the sequence. The GluC digestion sites are colored in green. (B) Schematic demonstrating the attachment of BMPS-peptides to MroM_TruncantCys scaffolds. (C) MALDI-TOF MS spectra of MroM-BMPS conjugates that were modified and then subject to GluC digestion. The regions shown correspond to mass ranges of C-terminal peptide sequences of interest. The expected and observed m/z are summarized in the Table S4. (D) Structures of the unnatural amino acids D-Ala (5) and α-aminobutyric acid (6) that are incorporated into peptide 8.

These conjugated scaffolds were subsequently incubated with SAM under aqueous conditions, following which the core peptides were liberated by GluC and tested for the presence of modifications using MALDI-TOF mass spectrometry. While the peptides conjugated to MroM_1–354Cys or MroM_1–363Cys did not show any modification, methylation could be observed on peptides conjugated to MroM_1–373Cys. The lack of methylation activity of MroM_1–354Cys and MroM_1–363Cys can be attributed to disruption of interaction within helical bundle, which is critical to position the substrate into the active site (Figure 2A). Covalent conjugates of MroM_1–373Cys with core peptide 5, 6 or 7 showed up to 3, 5 and 4 backbone amide methylations, respectively (Figure 3C). Of note, conjugation reactions were conducted at high micromolar concentrations of peptide and required only 1.5-fold excess MroM_1–373Cys relative to peptide to ensure attachment to the C-terminal Cys. We further tested the substrate tolerance of MroM_1–373Cys by providing it with a peptide containing two non-proteinogenic residues (Figure 3A & D). A synthetic peptide 8 consisting of a core peptide with D-Ala, and α-aminobutyric acid was linked to MroM_1–373Cys using the bifunctional BMPS linker, and subsequently incubated with SAM under aqueous conditions. Work-up of the reaction revealed up to three methylations (Figure 3C). Hence, MroM_1–373Cys is a suitable catalytic scaffold to install backbone amide methylations on a range of peptide substrates at stoichiometric concentrations.

A strategy for reversible bioconjugation of peptides to MroM_1–373Cys

While use of the BMPS linker for bioconjugation demonstrated proof of concept for our approach to generate backbone N-methylated peptides, there are some limitations. The generally irreversible nature of the Michael addition reaction between a Cys thiolate and the electrophilic maleimide moiety of BMPS renders this a single use application. After GluC release of the core peptide, the BMPS conjugate remains attached to the MroM_1–373Cys scaffold, and the unmodified protein cannot be regenerated. To overcome this limitation, we tested several thiols as suitable reagents for controlled disassembly of BMPS via a retro-thiol conjugate addition reaction (Figure S12) ⁴⁹. However, these efforts proved to be unsuccessful.

Next, we focused on the use of dynamic covalent thiol reactive agents as a strategy for disassembly to afford a modified substrate and regenerated enzyme ⁵⁰. Prior studies have demonstrated that the thiol reactivity of electron-deficient olefins can be tuned by the α-substitution of one or more electron-withdrawing groups ⁵¹. Di-substituted olefins can be modulated to react with cysteine thiols in a rapid and reversible manner ⁵². To test the suitability of such substituted olefin for our approach, we synthesized three Michael acceptors based on a scaffold of the electrophilic 4-vinylbenzoic acid (Figure S17). This scaffold was chosen as it is bifunctional, allowing for thiol reactivity at the olefin and peptide attachment by functionalization of the acid. Additionally, the cyanoacrylate moiety enables us to monitor the integrity and stability of the synthesized compounds by UV-visible spectroscopy (λ_max 310–320 nm).

We synthesized three acrylonitriles with different electron-withdrawing substituents attached at the electrophilic β-carbon of 4-vinylbenzoic acid. The geminal dinitrile (11) and the cyanoacrylate (12) derivatives could be produced from 4-formylbenzoic acid in high yields (between 85–92%) and were stable in methanol (see SI for details). However, both compounds underwent rapid hydrolysis in aqueous buffers with half-life values of 1.8 min (11) or 32.7 min (12), as determined spectroscopic analysis (Figure S17). Hence, these compounds were not suitable for use as bioconjugation linkers in biological buffers.

In contrast, the 4-pyridyl acrylonitrile (13) demonstrated robust stability in aqueous buffers and remained intact beyond 24 hours when stored in 20 mM Tris-HCl, pH=8.5. Encouraged by this result, we next tested the reactivity of this doubly substituted Michael acceptor by monitoring the reduction in the 320 nm absorption band with increasing concentrations of the model thiol, glutathione (GSH). Data fitting assuming pseudo-first order binding kinetics yielded a K_d of 6.7 mM (Figure S17). While this value is within the range observed with other reversible acrylonitriles ⁵¹, the reaction could not be saturated and stalled at ~75% completion. Hence, we sought other strategies for the reversible assembly of substrates to MroM_1–373Cys while continuing to develop and optimize acrylonitrile-based linkers.

Prior studies have reported on the utility of mono- and di-bromo maleimides as viable reagents for the reversible modification of thiols ⁵³. We reasoned that a reversible bioconjugation strategy that utilized a linker akin to that found in BMPS would be viable given our success with the use of this irreversible linker. To this end, peptide substrates were modified with di(N-succinimidyl) 3,3’-dithiodipropionate (DSP or Loman’s reagent), a crosslinker that can be cleaved with a thiol. Treatment of the resultant peptide with the reductant dithiothreitol (DTT) yielded peptides with a 3-mercaptopropionic acid attached to the N-terminus of the peptide. This strategy allows for the facile incorporation of a reactive thiol on any peptide and avoids the difficulties and additional efforts necessary to either synthesize or recombinantly produce a peptide with an N-terminal Cys. Treatment of the 3-mercaptoproprionate peptide with 2,3-dibromomaleimide resulted in the facile formation of the monobromo adduct and confirmed that the maleimide had added exclusively to the peptide N-terminal thiolate as determined by MALDI-TOF analysis.

Monobromomaleimide (MBM) adducts of both the peptides 5 and 6 with the 3-mercaptoproprionate linker were produced as described above (Figure 4A). Incubation of either peptide with MroM_1–373Cys under aqueous conditions (20 mM Tris-HCl, pH=8.0) resulted in facile (~30 minutes) formation of the desired covalent species. Each conjugated scaffold was incubated with SAM under aqueous conditions for 24 hours. Treatment of these conjugates with an excess of thiol (20 mM 3-mercaptoethanol, MSH) resulted in the clean recovery of unmodified MroM_1–373Cys and release of the peptide (Figures 4B & C and Figure S18). Remarkably, MALDI-TOF mass spectrometric analysis of the liberated peptide revealed that up to 2 methylations could be observed on either peptide (Figures 4B & C). The small number of methylations suggest that the MBM conjugation strategy was not as robust as the use of the irreversibly attached BMPS. Nonetheless, the results here speak to the utility of our catch and release approach as a developmental tool to install backbone N-methylations on peptides.

Figure 4:

(A) Schematic of the catch and release strategy for enzymatic N-methylation of peptides using monobromomaleimide (MBM) peptide conjugates, and subsequent ligation of MBM-peptides to MroM_1–373Cys. (B) MALDI-TOF MS spectra of the MBM-5 conjugate (top), after incubation with MroM_1–373Cys and SAM (middle), after incubation with 2-mercaptoethanol (MSH) to release the methylated peptide (bottom). (C) MALDI-TOF MS spectra of the MBM-6 conjugate (top), after incubation with MroM_1–373Cys and SAM (middle), after incubation with 2-mercaptoethanol (MSH) to release the methylated peptide (bottom). The expected and observed m/z were summarized in the Table S4.

CONCLUSIONS

In this work, we present a bioconjugation strategy for the installation of backbone N-methylation on peptides of diverse sequences. Using a truncated variant of the borosin-type methyltransferase MroM_1–373Cys, we have developed multiple strategies for the covalent attachment of peptide substrates which can function on a range of peptide substrates at sub-millimolar concentrations. Each of these approaches generates the requisite covalent bioconjugate in a facile manner (~15–30 minutes) and can be conveniently carried out in aqueous conditions. Each of the peptide substrates can be produced either via solid phase synthesis or as heterologous MBP fusions in an E. coli expression host. Moreover, peptides with non-proteinogenic components are also viable substrates for this approach, though additional comprehensive testing is required to determine the full substrate scope. The reagents necessary for the bioconjugation strategies are either commonly available or require minimal synthetic efforts to afford the requisite heterobifunctional crosslinkers in high yields (greater than 80%).

While naturally occurring ‘split’ borosins can efficiently catalyze α-N-methylation of peptides in trans, the need for an α-helical core peptide limits broad use ^{40, 42}. Likewise, the native chemical ligation approach to attach short peptides to the C-terminus of the catalytic core by thiol-thioester exchange requires peptide substrates that are soluble at low-millimolar concentration due to the slow rate of ligation. The reaction rate of maleimide with thiols is about 3 to 4 orders of magnitude faster than the reaction rate of thiol-thioester exchange reaction ^43–44. The strategy developed in the current work can function with equal component stoichiometries at high micromolar concentrations and is amenable for modification of peptides that are poorly soluble.

While conjugation using the BMPS is irreversible, the approach is robust and can yield products with backbone methylations approaching levels observed for the naturally occurring covalently linked peptide substrates. The modified peptide can be liberated from the spent catalyst by judicious incorporation of an Arg or Glu residue that would allow for proteolytic cleavage using trypsin or Glu-C, respectively. We also present a catch and release strategy that allows for the reversible, covalent attachment of peptide substrates onto the MroM_1–373Cys scaffold and removal of the modified peptide. This latter method is still under development and optimization of linker length, choice of buffer conditions and other parameters may result in further improvement in the extent of backbone modifications.

While the proof of concept of this bioconjugation method had been shown, there are numerous concerns to overcome to enable large scale biotechnological usage. First, the principles that govern site selectivity of modification are not yet known. To date, it has not been possible to predict a priori the site and number of methylations on a given peptide substrate. Hence, regioselective backbone N-methylation is still not possible. Secondly. peptide substrates containing charged residues are generally not tolerated by borosin methyltransferases. Lastly, it is unclear what level of tolerance MroM would have for substrates with non-natural residues and if the method can be used to catalyze α-N-methylation at these residues.

With these caveats in mind, the approaches described herein provide a possible framework for further optimization of crosslinking strategies. Biocatalyst that may enable modifications on peptide substrates containing charged residues can be identified either through structure-based redesign of the active site of known enzymes or by identification of broader tolerant enzymes through genome mining efforts. Rules for specificity may emerge with wider uses of this and similar approaches to enable the development of backbone modified peptide libraries. Lastly, residue tolerance can be further explored through combinatorial biosynthesis with other RiPP enzymes that install orthogonal modifications on peptide backbones, such as members of the YcaO superfamily ^54–55.