Dissecting Bottromycin Biosynthesis Using Comparative Untargeted Metabolomics

Abstract Bottromycin A2 is a structurally unique ribosomally synthesized and post‐translationally modified peptide (RiPP) that possesses potent antibacterial activity towards multidrug‐resistant bacteria. The structural novelty of bottromycin stems from its unprecedented macrocyclic amidine and rare β‐methylated amino acid residues. The N‐terminus of a precursor peptide (BtmD) is converted into bottromycin A2 by tailoring enzymes encoded in the btm gene cluster. However, little was known about key transformations in this pathway, including the unprecedented macrocyclization. To understand the pathway in detail, an untargeted metabolomic approach that harnesses mass spectral networking was used to assess the metabolomes of a series of pathway mutants. This analysis has yielded key information on the function of a variety of previously uncharacterized biosynthetic enzymes, including a YcaO domain protein and a partner protein that together catalyze the macrocyclization.

Abstract: Bottromycin A 2 is astructurally unique ribosomally synthesized and post-translationally modified peptide (RiPP) that possesses potent antibacterial activity towards multidrugresistant bacteria. The structural noveltyofbottromycin stems from its unprecedented macrocyclic amidine and rare bmethylated amino acid residues.The N-terminus of aprecursor peptide (BtmD) is converted into bottromycin A 2 by tailoring enzymes encoded in the btm gene cluster.H owever,l ittle was knowna bout key transformations in this pathway,i ncluding the unprecedented macrocyclization. To understand the pathway in detail, an untargeted metabolomic approacht hat harnesses mass spectral networking was used to assess the metabolomes of aseries of pathway mutants.This analysis has yielded key information on the function of av ariety of previously uncharacterized biosynthetic enzymes,i ncluding aY caO domain protein and ap artner protein that together catalyze the macrocyclization.
Ribosomally synthesized and post-translationally modified peptides (RiPPs) are natural products that are prevalent throughout nature, [1] and their biosynthetic pathways are capable of transforming simple proteinogenic amino acids into structurally complex compounds that have potent bioactivities. [2][3][4] However,e lucidating the biosynthesis of RiPPs can be hindered by the difficulty of isolating intermediates,asthe biosynthesis takes place on alarger precursor peptide,a nd intermediates may be rapidly proteolyzed. Therefore,i mproved methods for the identification of RiPP intermediates are desirable.Bottromycin A 2 (1,Scheme 1) [5][6][7][8] possesses potent antibacterial activity towards multidrugresistant bacteria, [9] and is structurally unique owing its unprecedented macrocyclic amidine,r are b-methylated amino acids residues,and aterminal thiazole.Nature employs av ariety of strategies for peptide macrocyclization, [10][11][12] but amidine formation has only been observed for bottromycin. Initial studies on bottromycin biosynthesis showed that its amino acids were b-methylated by radical SAM methyltransferases [5,7] (RSMTs), but the rest of the bottromycin pathway represented abiosynthetic black box, where little was known about key steps in the pathway,including the unprecedented macrocyclization. In this study,weemploy untargeted metabolomics and mass spectral networking to deduce the biosynthetic route to bottromycins in Streptomyces scabies.T his analysis identifies the enzymes responsible for macrocyclization, thiazole formation, and aspartate epimerization, thereby demonstrating the utility of an untargeted metabolomic approach for elucidating atargeted biosynthetic pathway.
To assess the role of the putative tailoring genes in the bottromycin pathway (Supporting Information, Figure S1), we had previously generated S. scabies DbtmC, DbtmE, DbtmF, DbtmI,a nd DbtmJ,b ut were unable to identify bottromycin-like compounds in these mutants. [5] We therefore established that these deletions did not lead to deleterious polar effects on the pathway by successfully complementing each mutant strain with acopy of the deleted gene (Supporting Information, Table S2 and Figure S2). Furthermore,a n RT-PCR analysis of wild-type and DbtmD strains showed that transcription still occurs in the absence of the precursor peptide at ac omparable level to the wild-type (WT) strain (Supporting Information, Figure S3), indicating that there is no essential regulatory feedback mechanism associated with the production of ap athway intermediate. [13] Therefore,ac omparative untargeted metabolomic analysis was carried out using WT S. scabies alongside the DbtmC, DbtmD, DbtmE, DbtmF, DbtmG, DbtmI,a nd DbtmJ deletion strains.Untargeted metabolomics is frequently used to assess the total metabolome of an organism, [14] for example to prioritize strains and compounds for drug discovery, [15,16] or to identify novel natural products, [17] but has rarely been used to assess as ingle pathway.H igh-resolution liquid chromatog-raphy-mass spectrometry (LC-MS) data for triplicate threeday production cultures of each strain (Supporting Information, Figure S4) were analyzed using two untargeted comparative metabolic methods.F irst, global LC-MS metabolomic profiles for each strain were used to generate an aligned data matrix that indicated significant differences between each mutant (Supporting Information, Table S5). This dataset was filtered to remove any metabolites that appeared in either DbtmD or the production medium. Mature bottromycins (1-5,F igure 2) were clearly absent in every mutant, but the complexity of the data hampered the detailed characterization of metabolites.T herefore,t his was followed by mass spectral network analysis, [18] which is ap owerful tool that identifies similarities in MS 2 fragmentation data and builds anetwork of species with related MS 2 spectra, thus identifying structurally-related molecules within acomplex mixture. [18][19][20][21][22][23][24] This has been used to assess the global metabolic profiles of as ingle organism, either in isolation [19] or when interacting with neighboring species, [18] to compare the metabolomes of related organisms, [20][21][22] to assess the metabolic potential of anew bacterial taxon, [23] and to identify metabolites related to the colibactin pathway. [24] Mass spectral network analysis of WT, DbtmC, DbtmD, DbtmF, DbtmI,a nd DbtmJ strains revealed an extensive metabolic network (Supporting Information, Figure S5). An analysis of the metabolomes of DbtmE and DbtmG was used to map molecules produced by these mutants onto this network. Nodes representing species that were not present in DbtmD were manually assessed using MS 2 to identify molecules related to the btm pathway.T his global metabolomic analysis showed that the bottromycin pathway contributes much more to the total metabolite profile of S. scabies than was previously understood, [25] and identified 14 distinct molecules in the wild-type strain, and at least 6a dditional molecules across the mutant strains,w ith masses and fragmentation patterns that are entirely consistent with bottromycin-like molecules (1-20;F igure 1, Figure 2; Supporting Information, Figures S6-S23, Table S3). Theo nly significant molecule that was not revealed by network analysis,owing to al ack of MS 2 fragmentation homology,w as an abundant species with m/z 406.27 (17;F igure 2; Supporting Information, Figure S17), which was identified by the initial comparative analysis of LC-MS data. Thea bundance of various bottromycin-like metabolites in WT S. scabies (Figures 1; Supporting Information, Figure S23) indicates that there are significant bottlenecks in the biosynthetic pathway that preclude the efficient processing of BtmD into bottromycin. Instead, partially processed BtmD can be proteolyzed, and the data show that there are multiple points at which the pathway stalls.T he diversity of bottromycin-like molecules produced by the WT could explain why it was difficult in prior studies to identify novel metabolites from mutants.
Them acrocyclic amidine of bottromycin is unique in nature,a nd ap lausible biosynthetic route involves the nucleophilic attack of Gly 1o nto the amide bond between Va l4 and Va l5,w hich could require the activation of the amide carbonyl. YcaO domain proteins activate backbone amide bonds by phosphorylation [26,27] or adenylation [28] of the carbonyl oxygen, and all YcaO domain proteins with ac haracterized activity have ap artner cyclodehydratase that aids catalysis of cyclization to oxazolines or thiazolines. [29] Theb ottromycin gene cluster encodes two YcaO domain proteins,B tmE and BtmF,b ut no cyclodehydratases.T herefore,w eh ypothesized that one participates in macrocyclization and the other is involved in the formation of the terminal thiazole.
Analysis of the comparative metabolomic and MS 2 network datasets revealed two new molecules (m/z 873.45 and m/ z 887.47) produced by both DbtmF and the amidohydrolase mutant DbtmI,but not by WT S. scabies.Masses of 873.45 Da and 887.47 Da correspond to the addition of H 2 Ot o carboxylated O-desmethylated bottromycins A 2 (1)a nd C 2 , [25] respectively,w hich indicated that one of the cyclodehydrations does not occur in DbtmF and DbtmI.M S n revealed that these molecules are not macrocyclized but do feature the thiazoline ring (10 and 11;F igure 3a;S upporting Information, Figure S10), thus indicating that BtmF and BtmI cooperate to catalyze amidine ring formation, but are not required for thiazoline formation. Both mutant strains also produced ar ange of other bottromycin derivatives that contain at hiazoline ring but no macrocycle (Figure 1, Figure 2; Supporting Information, Figure S23). DbtmI did produce trace amounts of macrocyclized 7 and 13,w hich could reflect inefficient spontaneous cyclization following BtmF-catalyzed amide activation. Acyclization mechanism is proposed ( Figure 3b)w here BtmF activates the amide bond Figure 1. Bottromycin mass spectral network from WT S. scabies and aseries of pathway mutants. Each node represents one metaboliteand edge thickness between nodes reflects the relative similarity of MS 2 data. The WT network is annotated with all observed m/z data and is enlarged 2 for clarity.Gray nodes indicate an absence of acompound and the area of the node is partially proportionalt ot he abundance of the metabolite.
using ATPa nd BtmI catalyzes cyclization. Further experiments with purified proteins will be needed to verify this, especially in relation to timing of ATPactivation. Cyclization is contingent on the removal of the N-terminal methionine, which is usually catalyzed by endogenous aminopeptidases, but these do not function efficiently with an MGP sequence. [30] In vitro analysis of the M17 peptidase [31] BtmM with BtmD demonstrated that BtmM catalyzes this reaction when either Zn 2+ or Co 2+ are used as co-factors (Supporting Information, Figures S25 and S26).
In contrast to DbtmF,the only abundant species that could be confidently assigned as aB tmD-derived metabolite in DbtmE was 17 (Supporting Information, Figure S16), which is at rimethylated tripeptide that is also found in the WT, DbtmF, DbtmI,a nd DbtmJ strains (Supporting Information, Figure S23). This assignment is consistent with the absence of 17 in DbtmD and in the RSMT mutants DbtmC and DbtmG. Unfortunately,this provided no evidence on Cys8 cyclization; the absence of cysteine-containing peptides could reflect rapid peptide degradation when cyclization does not occur. Thelack of any thiazole or thiazoline-containing metabolites does imply that BtmE catalyzes thiazoline formation, although further in vitro characterization is required to confirm this.T he absence of macrocyclized metabolites suggests that thiazoline formation is an early step in the pathway.BtmH, the only uncharacterized hydrolytic enzyme in the pathway,i sp roposed to remove the follower peptide, although it is possible that it could also participate in heterocyclization.
The btm cluster lacks af lavin-dependent dehydrogenase that is required for the biosynthesis of all other thiazole/ oxazole-containing RiPPs. [29] Instead, aP 450 enzyme,B tmJ, was predicted to catalyze the oxidative decarboxylation of the thiazoline into at hiazole. [5][6][7] This is an uncommon role for aP 450, although it has been reported for thiazole formation in the biosynthesis of the plant alkaloid camalexin [32] and could be mechanistically similar to the fatty acid P450 decarboxylase OleT. [33] Analysis of DbtmJ revealed two abundant compounds with m/z 841.43 and 855.44 (Figure 4), which were confirmed to be carboxylated O-desmethyl bottromycins B 2 and A 2 ,r espectively (12 and 13)u sing MS 2 (Supporting Information, Figure S11).
Interestingly,t wo distinct peaks are observed by LC-MS for both m/z 855.44 and 841.43 (Figure 4a), and each pair of Predicted stereochemistry is based on the ribosomalo rigin of each amino acid, although Asp 7stereochemistry is not provided for some compounds owing to the potential for epimerization and corresponding double peaks in their LC traces. peaks with the same mass have identical MS 2 fragmentation patterns (Supporting Information, Figure S11). This could reflect am ixture of epimers at the aspartate residue,w hich has an on-proteinogenic d-stereocenter in bottromycin A 2 . Therefore,w eh ypothesized that aspartate epimerization occurs after thiazoline formation, when the pK a of the aspartate a-proton is lowest owing to imine-enamine tautomerization that is disfavored once the aromatic thiazole is formed (Figure 4b). This is consistent with previous reports of epimerization of amino acids adjacent to carboxylated thiazolines, [34] and we could observe spontaneous interconversion of these peaks at pH 7.5 (Supporting Information, Figure S12).
To further assess whether this proton is exchangeable,we carried out ad euterium labeling experiment. Here,a ll exchangeable protons were replaced with deuterium in D 2 O, the thiazoline was then hydrolyzed back to Cys in dilute aq. DCl, and the sample was finally treated with H 2 O. Theoretically,t his would trap ad euterium in the Asp aposition as back exchange would be prevented following loss of the thiazoline.This indeed showed specific incorporation of one deuterium into 13 at Asp 7( Supporting Information, Figures S13 and S14), indicating that this position can readily undergo non-enzymic epimerization. Thedrop in abundance of both forms of m/z 855.44 in the WT compared to DbtmJ (Supporting Information, Figure S23) implies that adynamic kinetic resolution converts this mixture of epimers into stereochemically pure mature bottromycins.
To investigate whether any of the metabolites reported are authentic pathway intermediates,each mutant strain was co-cultivated with DbtmD,w hich is unable to produce the precursor peptide.A ny diffusible molecules produced by mutants that are genuine intermediates should be converted into 1 by the functional enzymes in DbtmD.Only the DbtmJ + DbtmD co-cultivation resulted in the production of 1 (Supporting Information, Figure S27), which implies that 12 and 13 are true intermediates and supports the proposed roles and substrate specificities of BtmJ and BtmB.I nc ontrast, the failure of the DbtmF and DbtmI co-cultivation experiments suggests that the linear compounds 10 and 11 are shunt metabolites rather than authentic intermediates,a nd that BtmF and BtmI require as ubstrate that contains af ollower peptide.However,wecannot rule out the possibility that 10/ 11 are not exported/imported as effectively as 12/13.The lack of an O-methyl group on the d-aspartyl residue in any of the metabolites identified from mutant strains indicates that O-methylation is the last step in the pathway,thereby generating an active antibiotic. [35] This was supported by the in vitro Omethylation of 4 using recombinant BtmB (Supporting Information, Figure S28).
Three RSMTs catalyze four C-methylations in the btm pathway. [5,7] In S. scabies,b ottromycin production is either severely reduced or entirely abolished when either RSMT gene, btmC or btmG,i sd eleted. BtmG methylates Va l4 and Va l5,and BtmC methylates Phe 6, [5,7] but it is unclear why the pathway stalls when either step is missed. Them etabolic datasets showed that both DbtmC and DbtmG have highly similar metabolite profiles,a nd the production of macrocyclized shunt metabolites 6 and 7 indicates that C-methylation is not ap rerequisite for cyclization.H owever,t he fully C-methylated metabolites produced by DbtmF and DbtmI demonstrate that macrocyclization is not aprerequisite for Cmethylation either.A lso,t he production of methylated tripeptides by DbtmC and DbtmG indicates that the pathway can stall before cyclization when C-methylation is disrupted. Thed ata are consistent with incomplete C-methylation reducing the efficiencyo fv arious downstream modification steps.
There has been widespread recent interest in both the biosynthesis [5][6][7][8] and biological activity [9,35] of bottromycin owing to its unusual structure and potent antimicrobial activity.I nt his study,w ehave harnessed untargeted metabolomics to elucidate the biosynthetic pathway to bottromycin  Table S4). Our analysis identified aw ide array of metabolites related to bottromycin, and the untargeted metabolomic data matrix (Supporting Information, Table S5) indicates that there may be further, currently uncharacterized, metabolites produced by this pathway.T his study also reveals the first example of YcaO domain-catalyzed macrocyclization, which provides the foundation for detailed mechanistic investigations into this step.