Applying promiscuous RiPP enzymes to peptide backbone N-methylation chemistry

Abstract

The methylation of peptide backbone amides is a hallmark of bioactive natural products, and it also greatly modifies the pharmacology of synthetic peptides. Usually, bioactive N-methylated peptides are cyclic. However, there is very limited knowledge about how post-translational enzymes can be applied to synthesis of designed N-methylated peptides or peptide libraries. Here, driven by the established ability of some RiPP enzymes to process diverse substrates, we sought to define catalysts for the in vivo and in vitro macrocyclization of backbone-methylated peptides. We developed efficient methods in which short, synthetic N-methylated peptides could be modified using sidechain and mainchain macrocyclases, PsnB and PCY1 from plesiocin and orbitide biosynthetic pathways respectively. Most significantly, a strategy for PsnB cyclase was designed enabling simple in vitro methods compatible with solid-phase peptide synthesis. We show that cyanobactin N-terminal protease PatA is a broadly useful catalyst that is also compatible with N-methylation chemistry, but that cyanobactin macrocyclase PatG is strongly biased against N-methylated substrates. Finally, we sought to marry these macrocyclase tools with an enzyme that N-methylates its core peptide: OphMA from the omphalotin pathway. However, instead we reveal some limitations of OphMA and demonstrate that it unexpectedly and extensively modified the enzyme itself in vivo. Together, these results demonstrate proof-of-concept for enzymatic synthesis of N-methylated peptide macrocycles.

Graphical abstract

INTRODUCTION

Peptide therapeutics represent an ever-growing field of drug development as they effectively combine the physiological advantages of small molecules and large biologics, while meeting challenging medical needs.^{1, 2} Metabolic instability and low membrane permeability of peptides often overshadow their synthetic accessibility, low toxicity, and selectivity. The two structural modifications predominantly present in potent peptide drugs and bioactive natural products are macrocyclization and backbone N^α-methylation.³ Macrocyclization reinforces the structural stability of peptides by limiting solvent exposure and preventing non-specific proteolysis. The N-methylation of amide linkages in peptides has been shown to improve bioactivity, as in the case of somatostatin analogues, melanocortin receptor agonists, and α_Vβ₃ integrin antagonists.^4–6 Additionally, N-methylation also alters the conformation of peptides, sometimes making them more membrane permeable.^{7, 8} Therefore, the incorporation of N-methylation in peptides can assist the development of orally bioavailable and metabolically stable potent drug candidates.⁹ A limiting factor is that synthesis of selectively N-methylated cyclic peptides can be challenging, especially in the context of a drug discovery library.^{8, 10, 11}

Inspired by the recent discovery of ribosomally synthesized and post-translationally modified peptide (RiPP) enzymes that catalyze amide N-methylation,^12–14 we sought to lay the groundwork for the rational enzymatic design and synthesis of N-methylated peptides and peptide libraries (Figure 1).^{14, 15} Several RiPP enzymes have the advantage of being promiscuous in terms of substrate scope, potentially enabling their broad adoption in synthetic strategies. In RiPP biosynthesis, the mature peptide is first translated as part of a larger precursor peptide. Enzymes modify a portion of the precursor peptide known as the core peptide, which encodes the mature peptide natural product. In promiscuous RiPP pathways, enzymes bind to conserved recognition sequence (RS) elements embedded in the precursor peptide. Because binding energy originates largely in the RSs, the core peptide can be hypervariable, encoding a library of diverse substrates. In turn, the RSs are proteolytically removed from the core peptide, affording a scarless product. Thus, by correctly positioning RSs around a core peptide, it is possible to direct a multistep biosynthesis of designed peptides.

Figure 1.

Overview of enzymatic reactions individually investigated in this study to introduce N-methylation to diverse scaffolds. Synthetic substrates for in vitro enzymatic studies were obtained through solid-phase peptide synthesis. N-terminal proteolysis is catalyzed by PatA from patellamide biosynthesis. PatA regioselectively cleaves the amide bonds immediately following a conserved recognition sequence (RS). PatA activity is not affected by the presence of N-methylated residues. Macrolactonization is carried out using ATP-grasp ligase, PsnB from plesiocin biosynthesis. PsnB catalyzes the formation of ω-ester linkages between The and Glu residues. PsnB is highly tolerant of N-methylated residues with at least one macrolactone ring forming in all cases. PCY1 from segetalin biosynthesis is used for head-to-tail cyclization. PCY1 is a dual-function protease/macrocyclase that cleaves off a C-terminal RS whilst forming a head-to-tail cyclized peptide product. PCY1 accepts N-methylated substrates but the enzymatic activity is highly dependent upon the length of RS and the position of N-methylated residues. The SAM-dependent methyltransferase, OphMA from omphalotin biosynthesis could potentially be harnessed to install backbone methylation. Sequential incubation of the OphMA constructs with PatA and PCY1 would then generate cyclic peptides.

In this study, we aimed to discover tools for amide N-methylated peptide synthetic biology by adapting well-studied, promiscuous RiPP enzymes as well as investigating recently discovered RiPP methyltransferases (Figure 1). Since N-methylation greatly alters the conformation and hydrogen bonding of substrate peptides, it is not known whether these substrates will be widely accessible to promiscuous tailoring enzymes, such as proteases and macrocyclases. Moreover, although promiscuous RiPP pathways have been widely studied, few studies investigate how to efficiently incorporate enzymes from several pathways, leaving fundamental principles unknown.^{16, 17} Here, we probe how validated, broad-substrate RiPP cyclases and proteases tolerate amide N-methylation (Figure 1). Furthermore, we sought to develop the N-methyltransferase, OphMA from omphalotin biosynthesis as a functional synthetic tool by integrating it with other RiPP proteases and macrocyclases.

Our overall strategy had three parts: methylation of a core peptide, selective N-terminal cleavage of the leader peptide from the methylated core, and macrocyclization of the methylated core peptide. PatA is among the best and most specific proteases used to cleave the N-terminus of polypeptides. PatA is recruited by conserved recognition sequences (RS) such as ‘GVDAS/X’ (Figure 1),¹⁸ where X can be any amino acid sequence. Additionally, PatA also recognizes AVLAS/X, GLEAS/X, and GVEPS/X. This is a strong combination because PatA RS-like sequences are rare, so that PatA is relatively selective and will not cleave most core peptides. Moreover, the lack of requirement at the P1’ position makes PatA a scarless protease. Together, these factors make PatA particularly useful in peptide engineering. However, we envisioned that a RiPP amide methyltransferase might block the N-terminal proteolysis site by generating a highly methylated and sterically hindered substrate, thereby rendering PatA inactive. Therefore, we aimed to test the hypothesis that PatA tolerates amide N-methylation. Additionally, we present various bioengineering examples to demonstrate the universal applicability of PatA.

Macrocyclization is an attractive modification for drug development as it improves the membrane permeability and proteolytic stability of peptides.¹⁹ However, there have been only a few examples that probe the use of biosynthetic enzymes to cyclize non-native and diverse N-methylated substrates.^{20, 21} Here, we investigate the substrate tolerance rules to facilitate two different types of cyclization reactions, side- chain cyclization and head-to-tail cyclization. We aimed to investigate whether N-methylated core peptides can be efficiently macrocyclized through their side chains, using the ATP-grasp ligase, PsnB from the plesiocin RiPP biosynthetic pathway.²² Plesiocin contains four repeats of the TTxxxxEE core peptide, with each repeat containing two macrolactone linkages between Thr and Glu residues, which are installed in sequential reactions by PsnB. Here, we identified the recognition sequence that enables PsnB reaction to occur in trans. Furthermore, we examined the substrate tolerance of PsnB with synthetic N-methylated substrates.

Additionally, we aimed to introduce head-to-tail cyclization in N-methylated peptides, using two broad-substrate RiPP enzymes, PCY1 from orbitide biosynthesis and PatG from cyanobactin biosynthesis.^{18, 23} Both PCY1 and PatG are serine proteases that recognize C-terminal RSs: (A/S)YD in PatG substrates and FQAKDVENASAPV in the substrates of PCY1.²¹ The enzymes cleave off the C-terminal RSs to form a core peptide-acyl enzyme intermediate, which is subsequently circularized to yield macrocyclic products.^{21, 24–26} Both PCY1 and PatG have been characterized to produce an exceptionally large array of macrocycles with both proteinogenic and nonproteinogenic amino acids. PCY1 reactivity has previously been investigated using a substrate with a single amide methyl group, while the functionality of PatG with methylated amides was unknown.²¹ Moreover, the design principles to obtain facile macrocyclization with N-methylated substrates were not determined.

We show that PatA, PCY1 and PsnB are broadly tolerant of methylation, while the usually highly promiscuous PatG did not accept amide methylation. Additionally, we integrated protease and macrocyclase RSs in OphMA to harness its autocatalytic methyltransferase activity. This strategy again shows that PatA can be seamlessly incorporated in hybrid precursor peptides to liberate modified core peptides. In this study, OphMA exhibits a narrow substrate range and in the current state of knowledge, is likely not optimal for engineering studies. These results provide a framework for using PatA, PsnB, and PCY1 in the rational design of amide-methylated peptides and peptide libraries. With better understanding of OphMA and its homologs, our strategies can potentially be implemented to synthesize a library of diverse N-methylated peptides. The goal of this study to exploit the available biosynthetic enzymes to introduce drug-like scaffolds to synthetic peptides.

RESULTS AND DISCUSSION

General experimental strategy.

Short, unmethylated or N^α-methylated peptides were prepared using solid-phase peptide synthesis (SPPS) followed by purification to homogeneity. Longer substrates were expressed in E. coli and purified, as were all enzymes used in this study. Enzymes and substrates were co-incubated, and their products were analyzed using LC-MS and LC-MS/MS protocols specifically designed for each type of enzymatic reaction. These methods enable precise determination of the position and type of post-translational modification on each substrate.

Proteolytic strategy to liberate N-methylated peptides.

PatA is a highly substrate tolerant N-terminal protease involved in patellamide biosynthesis (Figure 1).²⁷ To test the effect of N-methylation on PatA proteolytic activity, we synthesized a set of two sterically hindered N-methylated substrates (Figure S2). We designed substrates that included the native RS element (GVDAS) required for PatA activity. In previous experiments PatA was shown to accept virtually any sequence as long as that RS was in place,²⁸ although never before with N-methylated substrates; therefore, because of the broad substrate tolerance of PatA, we selected flanking elements that we believed would afford easy handling and solubility (many short peptides are poorly soluble in enzyme assays, so solubility must be included as a design element). These sequences were also related to those used in the context of the whole OphMA protein (see below), aiming to test whether methylation across this region might affect cleavage in the OphMA context.

SPPS afforded peptides GFPGVDAS/TVAT (1a) and GFPGVDAS/TVAT (1b). (Underlined residues indicate amide N-methylation and italicized residues represent PatA RS.) These substrates contain the PatA RS and N-methylated residues on the N- and C- terminal side of the cleavage site. Overnight incubation of 1a or 1b with PatA resulted in regioselective cleavage at the desired position with 65% and 32% yield respectively, demonstrating that PatA still functions in the presence of substantial amide methylation (Supporting figure S2). In extensive previous work in our group, PatA is usually nearly quantitative in converting non-methylated peptides.^{29, 30} Therefore, PatA was employed in OphMA designs described later, in which many further methylated peptides could still be efficiently cleaved by PatA although with slightly reduced efficiency in comparison to unmethylated peptides (Figure 2).

Figure 2.

Incorporating N-terminal protease, PatA with diverse substrates containing multiple PTMs. The N-terminal protease, PatA selectively cleaves after the conserved RS ‘G(L/V)E(A/P)S’. The enzyme has no requirement at the P1’ position, and hence the prudent addition of PatA RS simplifies design of modified peptides. In this study, PatA RS ‘GVDAS’ was introduced N-terminal to the core peptide, enabling PatA to regioselectively liberate the modified N-methylated core at later stages.

Furthermore, to demonstrate the universal applicability of PatA for leader peptide removal, we integrated PatA RS with constructs containing recognition elements for the lanthionine synthetase, ProcM.³¹ We genetically engineered hybrid precursor peptide constructs such that the ‘GLEAS’ sequence was sandwiched between the ProcM RS and core peptide (Supporting figure S3). PatA regioselectively cleaved the hybrid precursor before being modified by ProcM, resulting in the long N-terminal leader and the short C-terminal core peptide. In an alternative approach, we introduced a lanthionine bridge in the precursor by incubating with ProcM and ATP. In this case, PatA cleavage was performed after the precursor had been modified by ProcM, and as a result the PatA reaction yielded a lanthionine bridge containing core peptide. The presence of lanthionine and the consequent absence of free cysteine in the final product was confirmed through LC-MS/MS and through labeling with iodoacetamide.³¹ We also engineered a precursor containing multiple core peptides/cassettes, each flanked by PatA RS sequences (Supporting figure S3). PatA recognized the multiple RSs scattered through the precursor, thereby liberating multiple core peptides through a single reaction. These results demonstrate that PatA is broadly compatible with modified, sterically hindered, and unmodified polypeptide substrates, making it an excellent tool for scarless synthetic biology approaches.

Side-chain cyclization of N-methylated peptides.

PsnB catalyzes the formation of macrolactone bridges between Thr and Glu residues in plesiocin (Figure 1).³² To determine whether PsnB could use its RS in trans, simplifying all further experiments, we used SPPS to synthesize the RS LRDLFIEDL (2) and a part of the plesiocin core peptide (3a) containing one TTxxxxEE motif (Table 1 and Supporting figure S4). We selected the native core peptide sequence, KGGPYTTLAIGEE for this study because substantial previous work has demonstrated that ATP-grasp ligases have very broad amino acid sequence tolerance, but it has not been investigated for their tolerance of peptide methylation.³³ Here, we hypothesized that PsnB would accept all N-methylated substrates because the macrocyclization acts on the sidechain.

Table 1.

Macrolactonization efficiency of PsnB using N-methylated peptide substrates^[a]

^[a]N-methylated residues are denoted with underlined residues.

^[b] denotes ω-ester bond between Thr and Glu.

^[c]Percent conversions are calculated using area under the curve from MS chromatograms.

^[d]The % conversion at 5 h and 16 h are indicated. For all other substrates, only the % conversion at 5 h is indicated.

When peptides 2 and 3a were co-incubated with ATP and PsnB for 5 hours, the substrate underwent the loss of two water molecules, concomitant with the formation of two macrolactone bridges (Figure 3A and Table 1). Our work represents the first demonstration of leader-free in trans activation of PsnB. Previously, the leader peptide was covalently attached to PsnB for macrolactonization.³²

Figure 3.

in trans PsnB activity with truncated core peptide containing N-methylated residues. Extracted ion chromatograms of substrates (A) 3a, (B) 3b and (C) 4a and the corresponding ω-ester ring containing product. The EIC of substrates are shown in black, while those of cyclized products are colored. The LC-MS/MS of the enzymatic products is used to determine the ring position. In case of (A) 3a and (B) 3b, the product is formed after loss of two water molecules and contains two macrolactone rings. Substrate (C) 4a undergoes the loss of one water molecules and the ω-ester linkage forms between the internal Thr and Glu residues. The fragment ions indicated in yellow are important in determining the macrolactone ring topology in each case.

With a simplified system in hand, we synthesized a series of short pleisocin core peptide substrate analogs of 3a containing various degrees of N-methylation (Table 1 and Supporting figure S4). Synthetic N-methylated substrates were used for in vitro reactions with PsnB. Substrates 3b-d contained N-methylated residues in the extended N-terminal region (KGGPY), upstream of potential macrolactone bridges. In the presence of PsnB and the RS, these substrates underwent the formation of both macrolactone bridges (Table 1 and Figure 3B). This indicates that methylation outside of the bridged region does not impede macrolactonization (Supporting figure S5).

Substrates 3e-f contained N-methylated residues on the terminal positions within the TTLAIGEE core peptide, which is enclosed by macrolactone rings in the natural product (Table 1 and Supporting figure S4). PsnB accepted amide methylations at the terminal Leu and Gly residues, installing both macrolactone linkages, albeit with lower yield that slowly increased over a 16-h reaction period. Mixtures of singly and doubly macrocyclized products were observed (Supporting figure S5).

Substrates 4a-c contained internal N-methylated Ala residue in the TTLAIGEE. PsnB only synthesized a single macrolactone ring between the innermost Thr and Glu residues in these substrates as observed through LC-MS/MS (Figure 3C and Table 1). Collectively, these data support our initial hypothesis, showing that PsnB has broad tolerance for N-methylated peptides, although a mixture of products or only partial macrolactonization is observed for disfavored substrates (Table 1 and Supporting figure S4and S5).

Head-to-tail cyclization of N-methylated peptides.

In order to find a promiscuous macrocyclase to synthesize N-methylated cyclic peptide, we used PatG from patellamide biosynthesis and PCY1 from segetalin biosynthesis (Figure 1).^{23, 27} Although PatG is an exceptionally broad substrate-tolerant macrocyclase, in pilot studies PatG could not process the initial series of N-methylated peptides tested (Supporting table S1 and Supporting figure S6). Therefore, although PatG might macrocyclize some N-methylated substrates, no further experiments were performed with PatG.

PCY1, the dual-function protease/macrocyclase from segetalin/orbitide biosynthesis, presented a good alternative to PatG (Figure 1).²³ Using SPPS, we first investigated the RS requirements by synthesizing a series of N-methylated substrates containing the full-length RS FQAKDVENASAPV (series 5) or the shorter RS FQA (series 6) (Table 2 and Supporting figure S7). Series 5 substrates containing full-length RS were more favorable substrates as they underwent completed conversion to a mixture of linear and cyclic core peptide products. The truncated RS in series 6 substrates were convenient because of its ease of synthesis, but often exhibited lower yield (Table 2). A series of native, non-native, and N-methylated synthetic core peptides were used in the presence of PCY1 (Table 2). Except in a few cases, PCY1 relatively efficiently proteolyzed all substrates. However, in many substrates only the linear hydrolyzed product was observed, while in others a mixture of linear and cyclic products was seen. Thus, PCY1 effectively cyclizes many N-methylated substrates, but at the cost of selectivity for transamidation versus hydrolysis.

Table 2.

Head-to-tail cyclization efficiency of PCY1 using N-methylated peptide substrates

	Substrate[a]	Products	% Conversion[b]
5a	GVAWAFQAKDVENASAPV	Cyclic [GVAWA] GVAWA	60.0 40.0
5b	GVAWAFQAKDVENASAPV	Cyclic [GVAWA] GVAWA	94.7 5.3
5c	GVAWAFQAKDVENASAPV	Cyclic [GVAWA] GVAWA	20.6 79.4
5d	FSASYSSKPFQAKDVENASAPV	Cyclic [FSASYSSKP] FSASYSSKP	70.5 29.5
5e	FSASYSSKPFQAKDVENASAPV	Cyclic [FSASYSSKP] FSASYSSKP	71.5 28.4
5f	QAYLGIPLPFQAKDVENASAPV	Cyclic [QAYLGIPLP] QAYLGIPLP	41.7 58.2
5g	QAYLGIPLPFQAKDVENASAPV	Cyclic [QAYLGIPLP] QAYLGIPLP	14.4 85.6
5h	TSIAPFPFQAKDVENASAPV	TSIAPFP	100
5i	TSIAPFPFQAKDVENASAPV	TSIAPFP	100
6a	GVAWAFQA	Cyclic [GVAWA] GVAWA	48.7 8.7
6b	GVAWAFQA	Cyclic [GVAWA] GVAWA	41.2 28.4
6c	GVAWAFQA	GVAWA	79.3
6d	FSASYSSKPFQA	Cyclic [FSASYSSKP] FSASYSSKP	89.3 10.7
6e	TSIAPFPFQA	TSIAPFP	100
6f	TVPTLPFQA	TVPTLP	99.1
6g	TVPTLPFQA	NR[c]	-
6h	QAYLGIPLPFQA	QAYLGIPLP	2.5
6i	QAYLGIPLPFQA	NR[c]	-
6j	QAYLGIPLPFQA	Cyclic [QAYLGIPLP] QAYLGIPLP QAYLGIP	58.9 26.9 9.4
6k	QAYLGIPLFQA	QAYLGIP	68.1
6l	INPYLYPFQA	INPYLYP YLYP	63.2 5.2

^[a]N-methylated residues are denoted with underlined residues.

^[b]Percent conversions are calculated using area under the curve from MS chromatograms.

^[c]NR, no reaction.

We began by using substrates 5a-e that contain the identical amino acid sequences to wild-type presegetalin precursors.^{34, 35} To investigate whether PCY1 could modify native substrates that are N-methylated, 5a-e were synthesized with varying degrees of methylation and with the full-length RS (Table 2). After overnight incubation with PCY1, these substrates were converted to a mixture of cyclic core peptides formed by transamidation and linear core peptides formed by hydrolysis (Table 2). Similarly, substrates 6a-d mimicked N-methylated presegetalin precursors but with the truncated RS. All substrates except 6c resulted in partial cyclization. Overall, these results indicate that PCY1 tolerates significant methylation and may be useful in generating backbone methylated peptides (Supporting figure S7).

To determine whether PCY1 can modify N-methylated substrates that are completely sequence different to its wildtype cores, we synthesized substrates 5f-i and 6e-i derived from the cyanobactin family of RiPPs. While 5f-i had the full-length RS, 6e-i had the truncated RS. These sequences were chosen because they are known to be cyclizable (there is no intrinsic disfavoring of cyclization), but they are extremely distant from the native peptides normally processed by PCY1. PCY1 proteolyzed most substrates but only efficiently cyclized substrates 5f-g (Table 2). The cyclizable substrates had the full-length RS, demonstrating that this feature will be necessary in the synthesis of cyclic peptide libraries using PCY1. Moreover, PCY1 cyclized the cyanobactin substrate QAYLGIPLP, but not TSIAPFP, demonstrating that the substrate scope of PCY1 will have to be thoroughly investigated before being applied to library synthesis. PCY1 readily handled N-methylation within the peptide framework. Additionally, PCY1 digested non-methylated substrates, 6j-l containing Pro in multiple fragments. PCY1 is a broad-substrate enzyme that will be widely useful in generating cyclic, N-methylated peptides, but further catalysts will be required with complementary substrate selectivities to more completely cover desired sequence spaces.³⁶

Methylation of artificial peptides with omphalotin methyltransferase.

We sought to integrate enzymes from multiple RiPP biosynthetic pathways as modules to synthesize diverse N-methylated peptides (Figure 1). The omphalotin methyltransferase provided a potentially ideal enzyme to test our design strategy (Figure 4).^{12, 13} It consists of a methyltransferase domain that is covalently fused to the precursor peptide, where amide N-methyl groups are appended during expression within E. coli cells. Initial data with omphalotin methyltransferase OphMA suggested that it may have a relatively wide substrate tolerance, although only single-site point mutations or substrates very similar to the native substrate have been employed.^{12, 37, 38} In order to fuse OphMA with broader-substrate RiPP enzymes, our design strategy aimed to incorporate different RSs needed to free the N-terminus and to circularize the resulting peptides (Figure 4 and Table 3). This is in part because the native OphP macrocyclase has only been co-expressed with OphMA in a fungal host, making it unsuitable for an in vitro scheme. Furthermore, the substrate scope of OphP, with diverse substrates remains unknown.¹³

Figure 4.

Strategy for generating diverse N-methylated peptides using OphMA from omphalotin biosynthesis. Artificial peptides mimicking the core peptide sequences of other RiPPs, flanked with RSs of N-terminal protease and C-terminal macrocyclase were embedded into the OphMA C-terminus at the genetic level. The mutated OphMA variant was expressed in E. coli for in cis methylation under in vivo conditions. The hypermethylated OphMA protein was purified and subjected to other tailoring enzymes for proteolytic cleavage, cyclization or other late-stage modification of the core peptide.

Table 3.

Number of methylations in the C-terminal region of OphMA variants with non-cognate core peptides

	C-terminal sequence (OphMA400-..)[a]				Number and positions of methylations[b]	Products of macrocyclization reaction
7a	GVDAS	QAYLGIPLP	SYD	SVMSTE	-	Cyclic [QAYLGIPLP] [d]
7b	GVDAS	TSIAPFP	SYD	SVMSTE	1 (T405)	NR
7c	GVDAS	TVPTLP	SYD	SVMSTE	-	NR
8a	GVDAS	QAYLGIPLP	F QA	S VMSTE	3 (Q415, A416, S417)	Cyclic [QAYLGIPLP] [e]
8b	GVDAS	GVAWA	F QA	S VMSTE	3 (Q411, A412, S413)	Cyclic [GVAWA] + GVAWA[e]
8c	(GSG) 2 GVDAS	GVAWA	F QA	S VMSTE	3 (Q417, A418, S419)	Cyclic [GVAWA] + GVAWA[e]
8d	GVDAS	TSIAPFP	F QA	S VMSTE	3 (Q413, A414, S415)	Cyclic [TSIAPFP] + TSIAPFP[e]
8e	GVDAS	GVPVWA	F QA	S VMSTE	3 (Q412, A413, S414)	Cyclic [GVPVWA] [e]
8f	GVDAS	TSIAPFC	FQA	SVMSTE	NR	NR
8g	GVDAS	TAPYP	F QA	SVM STE	5 (Q411, A412, S413, V414, M415)	TAPYP[e]
8h	GVDAS	VPTTP	F QA	S VMSTE	3 (Q411, A412, S413)	VPTP[e]
8i	GVDAS	FFPP	FQA	SVMSTE	5 [c]	FFPP[e]

^[a]The C-terminal sequence of full length OphMA proteins. The engineered cleavage site for N-terminal protease PatA is shown in red, the recognition sequence for macrocyclase is shown in blue, and the core peptide is shown in black. The follower peptide sequence ‘SVMSTE’ is brown.

^[b]Number of methylations observed in the engineered sequence after N-terminal proteolysis. The amino acid residues modified through N-methylation are underlined.

^[c]The site of methylation could not be determined due to poor solubility of cleavage product.

^[d]Macrocyclization reaction was carried out with PatG.

^[e]Macrocyclization reactions were carried out with PCY1.

Because OphMA has previously been investigated with sequences that are very similar to the wild-type core peptide, we used a broad suite of core peptide sequences that greatly differ from the native substrates (Table 3). The non-native core peptides used in our OphMA constructs were derived from various natural cyanobactin and orbitide core peptide sequences that are compatible with the macrocyclases characterized above (Table 3).^{34, 35, 39–42} We flanked these modified core peptide sequences with the RSs for PatA and for macrocyclases (either PatG or PCY1).^{18, 21, 36} We also left the C-terminal domain from the native OphMA (SVMSTE) in the constructs. Previous reports have suggested that removal of this follower peptide leads to incomplete methylation.¹² Furthermore, the follower peptide should not interfere with PatG or PCY1 macrocyclization based upon current knowledge of those enzymes.^{12, 43} In total, 12 variants (7a-c and 8a-f) were successfully engineered and functionally expressed (Table 3).

OphMA variants were expressed in E. coli for 3 days based upon previous precedents,^{12, 13, 37, 38} and purified (Supporting Figure S1). The purified OphMA variants were treated with PatA, demonstrating that all OphMA variants were efficient substrates for selective hydrolysis by PatA, except for 8f, which contains a free Cys residue (Figure 5A). This reinforces the usefulness of PatA in synthetic biology pipelines (Table 3 and Supporting figure S8). It should be noted that OphMA variants (~47 kDa) are starkly different from native PatA substrates from patellamide/trunkamide precursors (~7 kDa) in overall size and sequence.⁴⁰ This again demonstrates the remarkable substrate tolerance of PatA, with limited dependence on amino acid sequences flanking the short, conserved recognition sequence ‘GVDAS’.

Figure 5.

Mass spectrometry of PatA proteolysis products obtained from OphMA variants (A) Scheme for generating PatA products for LC-MS/MS (B) LC-MS and LC-MS/MS of 7b variant after PatA N-terminal cleavage showed N-methylation localized at Thr405. (C) LC-MS and LC-MS/MS of 8b variant after PatA N-terminal cleavage showed 3 N-methylations localized at Gln411, Ala412, Ser413. In the MS/MS data, important fragment ions are highlighted in yellow.

Following liberation of the C-terminal peptides by PatA, the reaction mixtures were analyzed by LC-MS/MS to determine the number and positions of amide N-methylation reactions (Figure 5B, 5C and Supporting figure S9). While two variants (7a and 7c) were not methylated in the C-terminal region, and one (8f) could not be cleaved by PatA for analysis, the remaining 9 mutants experienced between 1–5 methylations in the C-terminal region. Unfortunately, only one of these was observed in the core peptide region (7b), while the remainder were seen in the region C-terminal to the core peptide (Table 3 and Figure 5B).

A few new things were learned about OphMA in this study. Variants 8a-e and 8g-h were methylated at Gln and Ala in the PCY1 tripeptide recognition sequence (FQA) and the first Ser in the hexapeptide follower sequence (SVMSTE) (Figure 5C and Supporting figure S9). Additionally, 8g was also methylated at Val and Met residues in the follower peptide (Table 3). To the best of our knowledge, this is the first example showing OphMA methylates the neutral and hydrophilic Gln residue. The variant 8c contained a six-residue GS linker region connecting the engineered C-terminal region to the OphMA protein. However, the Phe, Gln and Ala residues in the macrocyclase recognition and follower region were still selectively methylated suggesting that the position of the residues is irrelevant for methylation (Table 3).

The methylated PatA cleavage products were further treated with the cognate C-terminal protease/macrocyclase based on the RS following the core peptide (Supporting figure S10). The results replicated what was found with the synthetic substrates, reinforcing the fact that PatG could not accept methylated substrates, while PCY1 was broadly accepting of methylation, with only linear product formation in some cases (Table 3, Supporting table S9 and Supporting figure S10). Furthermore, it was observed that PCY1 functions with N-methylated RS residues; the salt bridge interactions formed by α-nitrogen atoms in PCY1 RS are not essential for cyclization. These results reveal that, once the rules of methylation of OphMA or other methyltransferase are better defined, the application of PatA and PCY1 will enable the synthesis of libraries or designed products.

Unexpected extensive self-methylation of OphMA enzyme region.

In addition to obtaining PatA cleavage products and measuring the position of methylation toward the C-terminus, we also measured the high-resolution, intact masses of the full-length OphMA variants 7a-c, 8a-b (Figure 6A). We expected to find that each protein had a few methyl groups added, mainly at the C-terminus after the PatA cleavage site. However, in the event we were surprised to find that each variant comprised a mixture of a large number of N-methylated protein variants (Figure 6B and Supporting figure S11). In all cases, a small amount of the unmethylated parent protein could be observed, but in addition, a series of further methylated proteins dominated the spectra. The most abundant peaks had 13–20 methyl groups in comparison with the unmodified protein, but in addition a major fraction of peaks indicated proteins were methylated up to 35–38 times (Supporting figure S11).

Figure 6.

Mass spectrometry of full-length OphMA core peptide variants. (A) Scheme for generating LC-MS spectra of methylated OphMA variants (B) LC-MS of 8b variant reveals that the protein has 35 methyl groups. Experimental mass spectrum is in grey, and the MS pattern simulated with chemical formula is shown in red.

The use of a high-resolution intact mass spectrometer enabled us to confirm that the methyl groups were present on the same proteins. They eluted at the same point in the mass spectrum, and the calculated exact masses of the intact proteins matched the observed masses to <10 ppm mass accuracy in all cases even though the proteins are >48,000 kDa in size (Supporting table S10). The methylation pattern was modeled, and the resulting spectra very closely match the prediction for proteins with varying methylation states (Figure 6B and Supporting figure S11). To ensure that the proteins consisted of OphMA variants, proteins 7b-c and 8b was digested with trypsin, and MS/MS experiments were performed on the resulting enzymatic digest. In all cases, only fragments resulting from the OphMA variants were observed (Supporting table S11). Fragments with additional N-methylation could be observed, but the high degree of variability in the proteins and the computational complexity of the problem made it quite challenging to nail down the precise sites of methylation. The best example was protein 9, which was otherwise not used in this study because it contained the PagG RS elements (Supporting table S6, S11 and Supporting figure S12). PagG is another broad-spectrum macrocyclase from the cyanobactin natural product family.⁴⁴ Trypsinization of protein 9 afforded several fragments in which peptide backbone N-methylation could be assigned to specific positions within the enzyme core and clasp region of OphMA (Supporting figure S12).^{12, 43}

Thirty-eight methylations are many more than can be accommodated by the C-terminal RS-core peptide regions that we engineered in the OphMA variants. To determine how significantly the C-terminus contributed to the observed N-methylation pattern, proteins 7a-c and 8a-b were treated with PatA, and the methylation was measured using tandem MS (Figure S8 and S9). These showed that only 1–3 methylations could be accounted for in the C-terminal regions of the proteins (Table 3). This unexpected methylation of internal positions in the proteins made it challenging to design evolutionary strategies to improve core peptide modification. It was not previously observed because in previous work, the intact proteins were not directly measured. Until autocatalysis in OphMA is better understood, this will likely complicate future efforts to engineer variants.

DISCUSSION

To harness ribosomal biosynthetic machinery, most synthetic biology approaches engineer long precursor peptides containing elements from multiple biosynthetic pathways, using a protease to free the core peptide after modification.^{16, 17} The ideal protease should facilitate traceless cleavage by selectively recognizing a specific sequence that lies outside of the core peptide (Figure1 and 2). Commonly used proteases such as trypsin and Glu-C have very simple recognition requirements, constraining the possible core peptide sequences. Alternatively, some proteases are more selective, but require residues on the P1’ side. For example, tobacco-etch virus (TEV) protease requires a Gly or Ser at the P1’ position, so that the cleaved product retains an undesirable residue.⁴⁵ We envisioned that PatA might be an excellent protease that fulfills these requirements, and here we demonstrate its utility across a broad array of highly modified substrates in addition to the previously known linear and azoline-containing substrate families (Figure 2). PatA functions in presence of multiple N-methylated residues (Supporting figure S2),¹⁸ and regioselectivity was not affected by the number or position of N-methylation. Additionally, PatA activity could be harnessed in three different bioengineering examples for leader peptide removal, including substrates in which a complex macrocycle directly abuts the cleavage site (Supporting figure S3). This expands the substrate tolerance of PatA and indicates that it can be integrated seamlessly in numerous biotechnological designs.

We used PsnB to investigate the effect of N-methylation on side-chain cyclization (Figure 1).²² Recently, the minimal peptide containing the partial leader and one core peptide repeat was recognized.^{32, 46} Driven by this discovery, we identified the PsnB RS (2) that enables the leader-free in trans processing of plesiocin core peptide-like substrates (3a) (Figure 3). We show that PsnB is highly tolerant of N-methylation, catalyzing macrolactone ring formation between the internal Thr and Glu residues in all cases (Table 1 and Figure 3). We defined the relative efficiency of PsnB when N-methylation is present in different positions, paving the way for engineering of N-methylated peptides that are macrocyclized through the side chains (Supporting figure S4 and S5). Our findings expand the substrate tolerance of PsnB to enable the pharmacokinetic modulation of similar bioactive natural products. Moreover, our findings are applicable to different topological classes of ω-ester containing peptides recently discovered through genome mining.³³

Two macrocyclases were investigated for their ability to catalyze head-to-tail cyclization of N-methylated peptides.^{18, 21} While PatG did not accept N-methylated substrates (Supporting table S1 and Supporting figure S6), PCY1 was broadly tolerant of multiple N-methylated residues (Figure 1). Furthermore, we screened a series of substrates to characterize the substrate rules for efficient cyclization with PCY1 (Supporting figure S7). We found the following rules (Table 2). Rule 1: The full-length PCY1 RS is required for complete conversion of substrates to products as seen in case of substrates 5c vs 6b. Rule 2: The yield of cyclic core peptide decreases with increase in the number of N-methylated residues as seen for 5b vs 5c and 5f vs 5g. Rule 3: The presence of N-methylated residues closer to the point of scission leads to unsuccessful reaction or compromises the yield severely (6g-i). Rule 4: The presence of Pro before the C-terminal RS is required to suppress non-specific proteolysis. These rules suggest that head-to-tail cyclization is a complex chemical transformation and is affected by many factors. It is arguable that the inherent conformation of the methylated cyclic peptide is also crucial for the success for the reaction. As a part of this series, we synthesized N-methylated analogs of segetalin B, segetalin F and prenylagaramide derivatives. Segetalin B and F have been implicated to show estrogen-like and vasodilatory activity, respectively.⁴⁷ Hence, producing N-methylated analogs of these cyclic peptides might be of therapeutic importance.

Omphalotin biosynthetic pathways have an unusually long precursor peptide OphMA, encoding a SAM-dependent methyltransferase at the N-terminus and the omphalotin core peptide at the C-terminus, linked by a clasp region in the same frame (Figure 1).¹² In our study, we replaced the core peptide completely with highly dissimilar and much longer artificial peptides to truly probe the substrate tolerance of OphMA methyltransferase (Figure 4). Using the strategy employed in our study, in cis methylation of artificial substrates, followed by sequential proteolysis and macrocyclization could be achieved (Table 3 and Supporting table S9). In our designs, PatA RS seamlessly worked with different constructs to facilitate traceless removal of the core peptide from the rest of the precursor (Supporting figure S8 and S9). Moreover, macrocyclization of different core peptides with PCY1 was observed (Supporting figure S10). Moreover, extensive methylation within the enzyme core and clasp region was observed (Supporting figure S11 and S12). So far, OphMA core peptide could be successfully swapped with closely related homologs, such as cyclosporin A and dictyonamide A sequence without affecting the methyltransferase activity.¹² Our study indicated that the N-terminal SAM-dependent methyltransferase is not very tolerant of diverse substrates, so that the methylation activity can only be harnessed for closely related core peptide sequences at present. A deeper understanding of the molecular complexities that initiate and deactivate the methyltransferase would greatly advance engineering efforts.

CONCLUSION

In this study, we demonstrate the ability of three promiscuous biosynthetic enzymes —protease PatA, ATP-grasp ligase PsnB, and macrocyclase PCY1 — to accept N-methylated substrates. Additionally, we describe the design principles required to use these enzymes efficiently. Furthermore, we integrate PatA and PCY1 with omphalotin methyltransferase, OphMA to build a pipeline to obtain N-methylated peptide library. Although our efforts led to proof-of-concept demonstration for these integrated designs, the primary obstacle in this approach was directing OphMA to methylate core peptide residues. We conclude that OphMA is not optimal to produce N-methylated peptide libraries. However, with the continuous discovery of new RiPPs biosynthetic enzymes, these strategies can be universally applied to generate diverse drug-like peptides.

MATERIALS AND METHODS

Materials

Plasmids encoding the wild-type OphMA, PCY1, and PsnB were synthesized by the Joint Genome Institute (JGI). Core peptide variations in OphMA were introduced through site-directed mutagenesis. Primers for site directed mutagenesis were synthesized by the University of Utah DNA and Peptide Synthesis Core Facility. For some constructs, gBlocks were purchased from Genewiz. The peptides for in vitro assays were synthesized in house using solid-phase peptide synthesis, using materials from Aapptec, Fisher Scientific, and Sigma Aldrich. Enzymes for PCR and cloning reactions were purchased from NEB. Amplified DNA was purified using QIAquick gel extraction kit, and plasmid DNA was purified using QIAprep spin miniprep kit.

Molecular Cloning

The synthetic ophMA plasmid contained the gene encoding an N-terminal hexahistidine tag, embedded in pET28b backbone at the NdeI/XhoI restriction sites. Genetic mutations in OphMA to encode diverse core peptides at the C-terminus were introduced using site-directed mutagenesis. Non-overlapping primers were designed with the 5’-end of both primers incorporating half of the desired insertion (see Table S4.2). Template was amplified using Q5 high-fidelity DNA polymerase to yield a linear DNA product containing the new insertion.⁴⁸ Incubation of the amplified DNA fragment with T4 polynucleotide kinase, T4 DNA ligase and DpnI for 1 h at room temperature resulted in circularization and concomitant removal of template DNA.⁴⁹ The reaction mixture was transformed using chemically competent Escherichia coli DH10β cells and isolated. The sequence was confirmed using Sanger sequencing.

Protein/precursor peptide expression or purification

Expression plasmids encoding the OphMA variants were transformed with chemically competent E. coli BL21(DE3) cells. Transformed single colonies were used to inoculate Terrific Broth (TB) media (10 mL) containing kanamycin (50 μg/ml). The overnight culture was diluted 100-fold into fresh TB medium with kanamycin (50 μg/ml) in a 2.5 L baffled bottom flask. The cultures were grown with continuous shaking at 37 °C until an optical density (OD₆₀₀) measuring 0.8–1 was obtained. The cultures were chilled at 4 °C for 20 min, followed by the addition of isopropyl β-D-1-thiogalactopyranoside (IPTG, 0.2 mM). After induction, the cultures were further shaken at 16 °C for 3 d. The cells were harvested by centrifuging the cultures at 4000 × g at 4 °C for 20 min, and the resultant pellets were stored at −80 °C until further use.

The thawed pellet was resuspended in lysis buffer (25 mM Tris, 500 mM NaCl, 10% glycerol, pH 8.0) and lysozyme (600 μg/mL). Cells were stirred at 4 °C and lysed by sonication with 30 s long pulses. The lysed sells were incubated with deoxyribonuclease I (20 μg/mL) and MgCl₂ (10 mM) for 15 min and centrifuged at 35,000 × g at 4 °C for 30 min. The supernatant was applied to pre-equilibrated Ni-NTA agarose beads. After 30 min of incubation at 4 °C, the flowthrough was collected under gravity. The column was washed with approximately 20 column volumes of wash buffer (25 mM Tris, 1 M NaCl, 30 mM imidazole, pH 8.0). The Ni-NTA beads in the column were incubated with elution buffer (25 mM Tris, 500 mM NaCl, 300 mM imidazole, pH 8.0) for 15 min. The eluant was collected and analyzed using SDS-PAGE. Proteins were concentrated using an Amicon 30kDa centrifugal filter, aliquoted for single use and flash-frozen with liquid nitrogen.

Plasmids encoding PatA protease domain (residues 1–306), PatG, or PsnB were transformed into E. coli Rosetta2(DE3) using electroporation, and single colonies were grown for 16 h in LB medium (10 mL) supplemented with 50 μg/mL of kanamycin sulfate and 25 μg/mL of chloramphenicol. The cultures were added to LB medium (1 L) and 2xYT (1 L) in the case of PsnB, while maintaining the chloramphenicol and kanamycin concentration. The cultures were grown until OD₆₀₀ 0.6–0.8, induced with IPTG (0.1 mM), and grown for a further 16 h at 18 °C. Cells were spun down, and soluble protein was extracted with the same protocol as described earlier for OphMA variants. In case of PsnB, the eluant fractions containing soluble protein were dialyzed with 3.5 kDa molecular weight cut-off (MWCO) dialysis membranes in presence of dialysis buffer (NaCl 500mM, HEPES 25mM 10% glycerol, pH 8.0).

Plasmids encoding PCY1 were transformed into chemically competent E. coli BL21 (DE3) cells. Single colonies containing the desired plasmids were transferred to LB medium (10 mL) containing kanamycin (50 μg/ml). The overnight cultures were diluted 100-fold in fresh LB medium and grown until measured OD₆₀₀ was in the range 0.6–0.8. Protein expression was induced by the addition of IPTG (0.5 mM). Using the method described above for OphMA, induced cultures were grown overnight at 16 °C, spun down, and soluble proteins were extracted.

Precursors were expressed and purified with the same method except that 1 mM of IPTG was added during induction. Lysate was passed through a Ni-NTA column following Qiagen’s denaturing protocol that uses urea (6 M). Bio-gel P-2 gel (Bio-rad) was rinsed with Tris-MgCl₂ buffer (Tris pH 7.5 (50 mM) MgCl₂ (5 mM)) three times, and the peptide was loaded and further with Tris-MgCl₂ buffer to obtain the desalted precursor peptides. The precursors were further purified using High-Performance Liquid Chromatography (HPLC). Alternatively, precursors larger than 8 kDa are dialyzed using a 2k MWCO dialysis cassette (Thermo-Fisher) with the same Tris-MgCl₂ buffer. Precursors were then aliquoted and frozen at −80 °C.

Solid phase peptide synthesis of non-methylated and N-methylated peptides

Peptides were made on a Syro I peptide synthesizer. The Fmoc-protected analogue of the C-terminal amino acid was loaded onto 2-chlorotrityl chloride resin (1.17 mmol/g). The resin (22 mg) was weighed in a SPE column and swelled in a 1:1 mixture of DMF and DCM for 30 min. The swelling solution was discarded, and the swelled resin was vigorously shaken for 16 h with the amino acid solution (4 equiv. with respect to the resin) containing DIPEA (8 equiv.). In the case of a-N-methylated amino acid-containing peptides, a lower substitution 2-chlorotrityl chloride resin (0.69 mmol/g) was used to avoid steric hindrance. Briefly, the resin (36 mg) was swelled and loaded with a mixture of Fmoc-amino acid (0.6 equiv.) and DIPEA (2.5 equiv.) with overnight shaking.

After manually loading the C-terminal amino acid, the unloaded sites were capped with methanol. The resin was transferred to Syro I for automated SPPS. The coupling reaction was carried out twice for 40 min at room temperature using Fmoc-amino acid (4 equiv.), HATU (4 equiv.) and DIPEA (8 equiv.) in DMF. After the coupling step, Fmoc was removed with 40% piperidine for 15 min at room temperature. The completed peptide was removed from the resin with a cocktail of TFA/TIS/H₂O (95:2.5:2.5). After a 2 h incubation with the cleavage cocktail, the deprotected peptide was filtered and concentrated to a minimal volume (200 μL). The peptide was precipitated slowly into cold ethyl ether (15 mL). The ether was carefully decanted, and the peptide was purified on a CombiFlash system via reversed-phase chromatography.

Esterification reaction with PsnB

Synthetic peptides mimicking core peptides (50 μM) were incubated in Tris buffer (100 mM, pH 8.0) for 5 h at 37 °C with the recognition sequence peptide LRDLFIEDL (100 μM) and recombinantly expressed PsnB (10 μM). The reaction mixtures also contained ATP (5 mM), KCl (50 μM), MgCl₂ (10 mM) and DTT (1 mM). Product formation was monitored with LC-MS.

C-terminal proteolytic cleavage and macrocyclization

PatG reactions were carried out using synthetic substrates at 37 °C for 16 h with PatG (20 μM), substrate peptide (100 μM), and MgCl₂ (5 mM) in Tris (50 mM, pH 8.0). The PCY1 reaction mixtures contained PCY1 (1 μM), substrate peptide (100 μM), NaCl (100 mM), DTT (5 mM), Tris (20 mM, pH 8.0) and were incubated at 30 °C for 16 h. Three or more replicates of the enzymatic reactions were performed for each substrate and condition and were analyzed using LC-MS.

ProcM lanthionine bridge formation and iodoacetamide (IAA) labeling

For ProcM reactions, substrate (50 μM), ProcM (10 μM), HEPES (50 mM), MgCl₂ (5 mM), ATP (5 mM), TCEP (0.2 mM) were incubated at 25 °C in a thermocycler for 24 h.

ProcM modified substrates (50 μM) were further digested with PatA (0.5 μM) in Tris-HCl (pH 7.5, 50 mM) and MgCl₂ (5 mM) at 37 °C for 2 h. To alkylate the unreacted Cys residue that did not form a lanthionine bridge with the Dha/Dhb residues, IAA was incubated with the products under the following conditions: enzyme reaction mixture (16 μL), IAA (10 mg/mL), NH₄HCO₃ (200 mM, pH 8.8), TCEP (10 mM) in a 40 μL scale reaction.² The reaction was incubated at ~ 7 °C for 3 h.

N-terminal proteolytic cleavage using PatA

The N-terminal proteolysis reaction was carried out in presence of N-His₆-OphMA or synthetic substrate (100 μM), PatA (10 μM) and 5 mM of MgCl₂ in Tris buffer (50 mM, pH 8.0) at 37 °C for 16 h. PatA proteolytic reactions with precursor peptides containing ProcM RS were carried out using following conditions: precursor peptide (50 μM), PatA (1 μM) and 5 mM of MgCl₂ in Tris buffer (50 mM, pH 8.0). In constructs where the core peptide was flanked by an N-terminal TEV cleavage site, the proteolysis reaction was performed using TEV protease (5 μM) in the supplied TEV buffer. Three or more replicates of the enzymatic reactions were performed and were analyzed using LC-MS.

HPLC-MS and HPLC-MS/MS analysis

The results of the enzymatic reaction were monitored with high-resolution LC/ESI-MS on a Waters Xevo G2-XS qTOF. The reactions were quenched by boiling for 5 min or by the addition of an equivalent amount of methanol. The resulting mixtures were further diluted to a substrate/product concentration of approximately 1 μg/mL, and the diluted solutions (2 μL) were applied to a ACQUITY UPLC BEH C18 (130Å, 1.7 μm, 2.1 mm X 50 mm) column. A linear gradient of 5% to 100% B over 10 min was used where the mobile phases A and B were LC-MS grade water and acetonitrile respectively, each containing 0.1% formic acid. An ACQUITY UPLC BEH C4 column (130Å, 1.7 μm, 2.1 mm X 100 mm) was used instead for longer peptides containing ProcM/PsnB RSs to minimize the peptide binding. The gradient conditions were the same. To elucidate the sites of methylation in the cleaved products, tandem mass spectrometry was used. A collision energy ramp of 40 V to 50 V with a 1 s scan time was used.

Intact mass analysis of OphMA variants

Intact mass LC-MS of OphMA variants were performed on an Eksigent Ekspert nanoLC 425 system (SciEx) coupled to a Bruker MAXIS ETD II QToF mass spectrometer. Samples were diluted to 5 μM using a 1:1 ratio of sample:aqueous 0.1% formic acid. Aliquots (5 μL) were injected onto the LC-MS using buffer A (0.2% formic acid in water) and buffer B (0.2% formic acid in acetonitrile) at a flow rate of 10 μL/min at 60 °C. The 70-min run began with a gradient from 20% to 90% buffer B over 25 min. A 10 cm long/1 mm inner diameter Waters Acquity UPLC BEH column was used.

The resultant mass spectrums were analyzed using Compass data analysis software. The OphMA variants eluted around 51–56 minutes. The mass spectrum was deconvoluted to show the neutral mass of the multiply charged ions present in the spectrum. The protein sequence was used to generate the chemical formula of the non-methylated and various methylated variants, which was used to simulate their mass spectrum which was overlaid on the experimental spectra for comparison. The simulated masses matched the isotopic distribution and observed masses (within an error limit of 10 ppm) of the unmethylated and methylated proteins in the chromatogram.

Trypsin digestion of OphMA variants and MASCOT analysis of LC-MS/MS data

Proteins were treated with DTT (5 mM) for 45 min at 60 °C followed by 10 mM IAA (10 mM) for 30 min at room temperature in the dark. Trypsin/GluC (for variants 7b-c, 8b) or Trypsin/LysC (for variant 9) were added to the reaction mixtures in a 1:100 ratio, left overnight at 38 °C, quenched with 1% formic acid to a pH of 2–3, and concentrated to 5 μL. Samples were diluted 1:1 with 0.1% aqueous formic acid, and aliquots of the resulting solution (5 μL) were run on a nano-LC/MS/MS UltiMate 3000 RSLCnano system (Dionex) with a BEH C18 3.0 μm nanocolumn (40 cm x100 μm)coupled to a ThermoScientific QExactive-HF mass spectrometer with a nanoelectrospray source. The same buffers as for intact mass were useda t a flow rate of 150 μL/min at 60 °C. An 83-min run initiated with 5% buffer B increasing to 55% over 53 min, followed by an increase to 95% over 63 minutes. Mascot (v 2.6) was used to analyze Mascot generic format (MGF) MS/MS data.