Pyridoxal-5′-phosphate as an oxygenase cofactor:Discovery of a carboxamide-forming, α-aminoacid monooxygenase-decarboxylaseYing Huanga,b, Xiaodong Liua, Zheng Cuia, Daniel Wiegmannc, Giuliana Niroc, Christian Duchoc, Yuan Songb,Zhaoyong Yangd,1, and Steven G. Van Lanena,1

aDepartment of Pharmaceutical Sciences, University of Kentucky, Lexington, KY 40536; bDepartment of Microbiology, College of Biological Sciences, ChinaAgricultural University, Beijing 100094, People’s Republic of China; cDepartment of Pharmacy, Pharmaceutical and Medicinal Chemistry, Saarland University,66123 Saarbrücken, Germany; and dInstitute of Medicinal Biotechnology, Chinese Academy of Medical Sciences and Peking Union Medical College,Beijing 100050, People’s Republic of China

Edited by Jerrold Meinwald, Cornell University, Ithaca, NY, and approved December 18, 2017 (received for review October 25, 2017)

Capuramycins are antimycobacterial antibiotics that consist of amodified nucleoside named uridine-5′-carboxamide (CarU). Previ-ous biochemical studies have revealed that CarU is derived fromUMP, which is first converted to uridine-5′-aldehyde in a reactioncatalyzed by the dioxygenase CapA and subsequently to 5′-C-glycyluridine (GlyU), an unusual β–hydroxy-α-amino acid, in a re-action catalyzed by the pyridoxal-5′-phosphate (PLP)-dependenttransaldolase CapH. The remaining steps that are necessary tofurnish CarU include decarboxylation, O atom insertion, and oxi-dation. We demonstrate that Cap15, which has sequence similarityto proteins annotated as bacterial, PLP-dependent L-seryl-tRNA(Sec) selenium transferases, is the sole catalyst responsible forcomplete conversion of GlyU to CarU. Using a complementarypanel of in vitro assays, Cap15 is shown to be dependent uponsubstrates O2 and (5′S,6′R)-GlyU, the latter of which was unex-pected given that (5′S,6′S)-GlyU is the isomeric product of thetransaldolase CapH. The two products of Cap15 are identified asthe carboxamide-containing CarU and CO2. While known enzymesthat catalyze this type of chemistry, namely α-amino acid 2-mono-oxygenase, utilize flavin adenine dinucleotide as the redox cofac-tor, Cap15 remarkably requires only PLP. Furthermore, Cap15 doesnot produce hydrogen peroxide and is shown to directly incorpo-rate a single O atom from O2 into the product CarU and thus is anauthentic PLP-dependent monooxygenase. In addition to these un-usual discoveries, Cap15 activity is revealed to be dependent uponthe inclusion of phosphate. The biochemical characteristics alongwith initiatory mechanistic studies of Cap15 are reported, whichhas allowed us to assign Cap15 as a PLP-dependent (5′S,6′R)-GlyU:O2 monooxygenase-decarboxylase.

natural products | biosynthesis | antibiotic | enzyme function

Pyridoxal-5′-phosphate (PLP)-dependent enzymes account formore than 230 distinct catalytic functions, most of which in-

volve transamination, decarboxylation, deamination, or racemiza-tion of α-amino acids (1). Although ostensibly less common, othertypes of PLP-dependent enzymes are known, including those cat-alyzing β- or γ-elimination or -substitution, aldol- and Claisen-typereactions, and even reactions involving O2-dependent oxidation (2–4). Collectively, PLP-dependent enzymes currently occupy five ofthe six classes used by the Enzyme Commission, with synthetasesthe lone category yet to be reported. The power and versatilityinherent in PLP is chiefly due to the ability to form an imine—firstwith the side-chain amine of a conserved Lys of the enzyme to forman internal aldimine and then with a substrate to form an externalaldimine—and subsequently serve as an electron sink during thechemical transformation of the substrate. Despite steady advancesin our understanding of the factors dictating the chemistry down-stream of aldimine formation (2), the prediction of a PLP-dependent enzyme activity solely based on sequence remains—forthe most part—an inexact science.

The A-500359s from Streptomyces griseus SANK 60196 (5), the A-503083s from Streptomyces sp. SANK 62799 (6), and the A-102395from Amycolotopsis sp. SANK 60206 (7) are capuramycin-typesecondary metabolites (8) that are potent inhibitors of bacterialtranslocase I and have promising antimycobacterial activity. Thisgroup of nucleoside antibiotics is characterized by two rarely ob-served structural components: a uridine-5′-carboxamide (CarU)nucleoside core and an unsaturated hexuronic acid appended tothe core through a 5′-O-glycosidic bond (Fig. 1). The hexuronic acidis subsequently modified with a third component, an L-amino-caprolactam in the case of A-500359s/A-503083s and an unusualarylamine-containing polyamide in the case of A-102395, both ofwhich are attached to the hexuronic acid via an amide bond (Fig. 1).The biosynthetic gene clusters for all three capuramycin-type me-tabolites have been cloned and characterized (9–11). Comparativebioinformatic analysis revealed 18 homologous open reading frames(orfs) between the capuramycin clusters, consistent with an impor-tant role in resistance, regulation, and the biosynthesis of the sharedCarU-hexuronic acid disaccharide core (11).Of the 18 shared orfs in the gene clusters for the capuramycin-

type metabolites, two encode sequence-unrelated proteins thatare each predicted to be PLP-dependent enzymes: CapH, with

Significance

Enzymes that activate dioxygen typically rely on flavins ortransition metals as redox cofactors. We describe the discoveryof the enzyme catalyst Cap15 that converts 5′-C-glycyluridine(GlyU), an unusual β–hydroxy-α-amino acid, to the modifiednucleoside uridine-5′-carboxamide. In contrast to expectations,pyridoxal-5’-phosphate (PLP) is the sole cofactor that serves asthe initiating reducing agent to activate dioxygen for in-corporation into GlyU prior to decarboxylation. Thus, Cap15 isnow classified as an O2- and PLP-dependent monooxygenase-decarboxylase. Also in contrast to expectations, phosphate andthe (5′S,6′R) isomer of GlyU—and not (5′S,6′S)-GlyU, theestablished pathway intermediate—are essential forCap15 activity. In total, Cap15 utilizes an enzymatic strat-egy for oxygen incorporation during conversion of anα-amino acid to a carboxamide.

Author contributions: Y.H., X.L., D.W., G.N., C.D., Y.S., Z.Y., and S.G.V.L. designed re-search; Y.H., X.L., Z.C., D.W., G.N., and S.G.V.L. performed research; Y.H., X.L., C.D., andS.G.V.L. analyzed data; and Y.H., X.L., G.N., C.D., and S.G.V.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence may be addressed. Email: zhaoyongy@163.com or svanlanen@uky.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1718667115/-/DCSupplemental.

974–979 | PNAS | January 30, 2018 | vol. 115 | no. 5 www.pnas.org/cgi/doi/10.1073/pnas.1718667115

primary sequence similarity to proteins annotated as serinehydroxymethyltransferases, and ORF15, with sequence similarityto bacterial proteins annotated as L-seryl-tRNA(Sec) seleniumtransferases, SelA. CapH—the discovery of which added to thecatalog of reported PLP-dependent catalytic activities—wasassigned as a transaldolase that generates acetaldehyde and 5′-C-glycyluridine (GlyU) from L-Thr and uridine-5′-aldehyde (U5′A), the last of which is generated from UMP in a reaction cat-alyzed by a nonheme Fe(II)-dependent α-ketoglutarate:UMPdioxygenase CapA (Fig. 1) (11, 12). The uncovering of the tan-dem reactions catalyzed by CapA and CapH revealed GlyU to bethe likely biosynthetic precursor of CarU, which was furthersupported by isotopic enrichment studies that were consistentwith L-Thr as the source of the carbon and nitrogen atoms of thecarboxamide of CarU and, significantly, that the C-N bond ofGlyU remains intact during the transformation to CarU (11).The totality of the results suggested that the remaining steps tofurnish CarU minimally involve (i) decarboxylation of GlyU, (ii)incorporation of an oxygen atom via a hydratase or oxygenaseactivity, and (iii) oxidation to the carboxamide. In lieu of bio-chemical precedence of PLP-dependent enzymes, it seemedreasonable that ORF15 might serve as the catalyst for the de-carboxylation step, despite inference from sequence analysis.In our attempt to delineate the remaining steps that “trim” GlyU

to CarU, we now report the function of ORF15, herein namedCap15. As anticipated, Cap15 catalyzes decarboxylation of GlyU,but also has an unexpected oxygenase activity and is identified as thesole catalyst responsible for the complete conversion of GlyU to thesignature CarU nucleoside of the capuramycin-type metabolites.Specifically, the data are consistent with the functional assignmentof Cap15 as a PLP-dependent (5′S,6′R)-GlyU monooxygenase-decarboxylase. Additional biochemical studies have enabled us topropose a mechanistic framework for this PLP-dependent enzyme-

catalyzed transformation, adding to the already remarkable catalyticdiversity of PLP-dependent enzymes found in living systems.

ResultsBioinformatic Analysis of Cap15. The gene product of cap15 hasmoderate sequence identity (45–50%) to proteins annotated asputative L-seryl-tRNA(Sec) selenium transferase, SelA. Sec-ondary structure-based predictions mirrored this analysis withthe highest similarity to Aquifex aeolicus SelA (13), althoughsequence identity by pairwise alignment was significantly lower(24%, SI Appendix, Fig. S1A). Bacterial SelA catalyzes the PLP-dependent production of selenocysteinyl-tRNA(Sec) and phos-phate from substrates seryl-tRNA(Sec) and selenophosphate (SIAppendix, Fig. S2). Studies with AaSelA have revealed that thisenzyme functions as a pentamer of dimers, with residues fromeach dimer subunit (termed A and B subunits) contributing tocatalysis (13), and the reaction is initiated by formation of aninternal aldimine between Lys285A and PLP. Based on sequencecomparison, Lys230 is predicted to be the respective residue inCap15 (SI Appendix, Fig. S1A). Only two additional Lys residuesare conserved (Lys318 and Lys362 of AaSelA corresponding toLys265 and Lys303 of Cap15, respectively). As part of the B sub-unit, the former Lys is located within ∼12 Å of the PLP andselenophosphate binding site of the A subunit; in contrast, thelatter Lys is located distant from the active site. The importanceof either residue has not been established. The B subunit of theAaSelA dimer also has two essential Arg residues that are pro-posed to establish an important electrostatic interaction forbinding of PLP and the substrate selenophosphate (SI Appendix,Fig. S1B). While one of the Arg residues is conserved in Cap15(Arg259), the second is substituted with a Lys (Lys262) (SI Ap-pendix, Fig. S1A).

Functional Assignment of Cap15. Recombinant His6-Cap15 wassoluble upon production in Escherichia coli and purified to nearhomogeneity using immobilized metal affinity chromatography(SI Appendix, Fig. S3A). Cap15(K230A) was also prepared as apotential control (SI Appendix, Fig. S3A). Both purified proteinsolutions were colorless and lacked a UV/Vis spectrum in-dicative of a PLP cofactor. The hypothetical substrate (5′S,6′S)-GlyU, which was identified as the stereoisomer produced by theCapH-homolog LipK (47% sequence identity via pairwisealignment) that is involved in the biosynthesis of A-90289 (Fig. 1)(14), was synthesized (SI Appendix, SI Materials and Methodsand Fig. S4) and tested with Cap15 in Hepes buffer with PLP.Following overnight reactions, HPLC analysis did not reveal anynew peaks compared with the controls. In contrast to LipK (14),the stereochemistry of the CapH product was not previouslyassigned (11). Therefore, two additional stereoisomers of GlyU—

(5′R,6′S) and (5′S,6′R)—that could hypothetically arise fromthe transaldolase reaction (14) were tested for activity withCap15. We have detailed the stereocontrolled synthesis andanalytic characterization of these two diastereomers elsewhere(15–17). Once again, however, no new peaks were observedwith Cap15.We next explored a variety of reaction conditions including

different pH and buffer systems. When reactions were per-formed in phosphate buffer with the supposedly unnatural (5′S,6′R)-GlyU isomer, a new peak was observed by HPLC, andformation of the new peak was dependent upon the inclusion ofCap15 and PLP (Fig. 2A). A trace amount of the new peakappeared following reactions with Cap15(K230A) (Fig. 2A). Nonew peaks were detected using the (5′S,6′S) or (5′R,6′S) isomersof GlyU as potential substrates under the phosphate-bufferedconditions (SI Appendix, Fig. S5). The new peak was analyzedby LC-MS yielding an (M +H)+ ion atm/z = 288.0, unexpectedlyconsistent with the molecular formula for CarU [expected(M + H)+ ion at m/z = 288.1 and (M + Na)+ ion at m/z = 310.0].

Fig. 1. Structure and biosynthesis of capuramycin-type metabolites. Pro-posed biosynthesis of the shared CarU core of capuramycin-type metabolitesA-500359s and A-503083s starting from UMP. CapA and CapH, along withthe homologs LipL and LipK from the biosynthetic pathway for A-90289 thatcontains an intact (5′S,6′S)-GlyU in the final metabolite, have been previouslyassigned in vitro. *LipK was previously shown to stereospecifically generate(5′S,6′S)-GlyU. αKG, α-ketoglutarate.

Huang et al. PNAS | January 30, 2018 | vol. 115 | no. 5 | 975

BIOCH

EMISTR

Y

To simplify product identification, CarU was synthesized andpurified, and authentic CarU coeluted with the Cap15 productand had identical mass and NMR spectroscopic properties (Fig.2A and SI Appendix, Figs. S6–S11). HPLC analysis of theCap15 reaction revealed the enzyme was inactive under anaerobicconditions (Fig. 2A) and inclusion of EDTA had no effect (SIAppendix, Fig. S12). Thus, Cap15 catalyzes a PLP-, phosphate-,and O2-dependent conversion of (5′S,6′R)-GlyU to CarU.The unanticipated requirement of O2 and phosphate for

Cap15 activity led us to more closely examine their role in thereaction. Using a sealed vessel and monitoring the reaction withan O2 electrode, dissolved O2 was steadily consumed over time

(SI Appendix, Fig. S3B), consistent with Cap15 functioning as anoxygenase. We subsequently probed the origin of the oxygenatom of the carboxamide of CarU, which hypothetically could bederived from phosphate, O2, or H2O. Three independent reac-tions were performed with isotopically enriched H3P

18O4, H218O,

or 18O2, and CarU production was monitored by LC-MS (Fig. 2B).An (M + H)+ ion atm/z = 290.1 was detected only in the presenceof 18O2, which corresponds to a 2-amu increase compared withunlabeled CarU. An estimated 93% enrichment was calculatedbased on peak intensities under these conditions. Finally, hydro-gen peroxide was not detected under optimized reaction condi-tions. Thus, the data are consistent with the classification ofCap15 as a monooxygenase and also suggest that phosphate doesnot participate directly in the chemistry of the Cap15 reaction.The identification of CarU as the product suggested that CO2

was concomitantly formed during the reaction. To confirm this,we used a commercial enzyme-linked bicarbonate detectionsystem for in situ detection of CO2. Using reactions without PLPor phosphate as controls, CO2 was confirmed as the secondproduct of the Cap15-catalyzed reaction (SI Appendix, Fig. S3C).Contrastingly, reactions under anaerobic conditions did notproduce detectable CO2. Furthermore, reactions with Cap15(K230A)resulted in only trace amounts of CO2. In total, the data areconsistent with the functional assignment of Cap15 as a PLP-dependent (5′S,6′R)-GlyU:O2 monooxygenase-decarboxylasethat generates carboxamide-containing CarU and CO2.

Biochemical Properties.Cap15 was active only in phosphate buffer,wherein the pH profile resembled a bell-shaped curve with op-timal activity at pH 7.5 (Fig. 3A). The specific activity ofCap15 was (5.2 ± 0.2) × 10−3 μmol/min/mg under the optimizedconditions. The specific activity with Cap15(K230A) was (2.4 ±0.1) × 10−4 μmol/min/mg, an approximately 20-fold decreasecompared with the wild-type enzyme. Using HPLC to detectCarU formation, single-substrate kinetic analysis with variable(5′S,6′R)-GlyU revealed Michaelis–Menten kinetics yielding aKm = (56 ± 7) × 101 μM and kcat = (9.3 ± 0.5) × 10−1 min−1 (Fig.3B). Increasing the concentration of phosphate from 50 mM to1 M had little impact with respect to the overall catalytic efficiency[Km = (42 ± 4) × 101 μM and kcat = (6.2 ± 0.2) × 10−1 min−1](SI Appendix, Fig. S13). Furthermore, nearly identical kineticparameters were calculated when activity was measured by CO2detection instead of CarU detection by HPLC (Fig. 3B).As previously noted, the activity of Cap15 was stereoselective

for the (5′S,6′R)-GlyU diastereomer. Thus, the reaction cata-lyzed by CapH was re-examined to identify the stereochemistryof the GlyU product. Initially, a one-pot reaction with CapH andCap15 starting from U5′A and L-Thr was analyzed by HPLC,revealing the formation of a peak coeluting with GlyU (SI Ap-pendix, Fig. S14). However, no CarU was observed, suggestingthat the CapH product—presumably (5′S,6′S)-GlyU—is not asubstrate for Cap15. Subsequently, the stereochemistry of theCapH product was directly assessed. As described elsewhere(14), phosgene modification of the β–hydroxy-amino acid ofGlyU enabled the analytical separation of the three syntheticdiastereomers in hand. Following the CapH reaction using syn-thetic U5′A and L-Thr as substrates, the phosgene-modifiedGlyU coeluted with authentic, phosgene-modified (5′S,6′S)-GlyU (Fig. 3C). Modification of the LipK product as a controlconfirmed this result (14). Thus, CapH stereospecifically gen-erates (5′S,6′S)-GlyU, which must undergo epimerization beforethe Cap15-catalyzed reaction.

Initiatory Mechanistic Analysis. The K230A mutation of Cap15 didnot abolish activity as initially predicted, and thus site-directedmutants of the two remaining conserved Lys residues (Lys265 andLys303) were prepared, and the mutant proteins analyzed foractivity. Cap15-Lys262 was also targeted for mutagenesis due to

Fig. 2. Biochemical characterization of Cap15. (A) HPLC traces of the re-action catalyzed by Cap15 using substrate (5′S,6′R)-GlyU with (i) exclusion ofenzyme; (ii) complete reaction mixture containing phosphate, (5′S,6′R)-GlyU,O2, PLP, and Cap15; (iii) exclusion of PLP; (iv) substitution of Cap15(K230A)for the wild-type enzyme; (v) exclusion of O2; and (vi) synthetic CarU. A260,absorbance at 260 nm. (B) LC-MS analysis of the CarU product followingCap15-catalyzed reactions in the presence of the indicated isotopically la-beled molecule. Data shown are representative of duplicates. RA, relativeabundance.

976 | www.pnas.org/cgi/doi/10.1073/pnas.1718667115 Huang et al.

the importance of the corresponding Arg residue in AaSelA (SIAppendix, Fig. S1). The three additional Lys → Ala mutantproteins were soluble when expressed in E. coli (SI Appendix, Fig.S3A), and activity analysis by HPLC revealed that Cap15(K303A)retained activity comparable to the wild-type enzyme (SI Appendix,Fig. S15). Contrastingly, the K262A and K265A mutant proteinsdid not produce detectable levels of CarU.The UV/Vis absorption spectra for the wild-type and mutant

proteins were examined with the goal of establishing a mecha-nistic framework for Cap15 (2, 18, 19). Wild-type Cap15 and themutant variants did not have an obvious absorbance at wave-lengths greater than 300 nm, consistent with the production ofthe recombinant enzyme in an apo-form (Fig. 4A, trace I, and SIAppendix, Fig. S16A). The lone exception was Cap15(K265A),which displayed a λmax = 411 nm that is consistent with someprotein copurifying with PLP as the internal aldimine (SI Ap-pendix, Fig. S16A). Identical results were obtained when in-cluding excess PLP before protein purification by affinitychromatography. Following the addition of substoichiometricPLP, which has a λmax at 388 nm representing the aldehyde form(Fig. 4A, trace II), a subtle red shift to λmax at 391 nm was de-tected for wild-type Cap15 (Fig. 4A, trace III), Cap15(K262A),and Cap15(K303A) (SI Appendix, Fig. S16B); the spectrum for

Cap15(K230A) was essentially identical to free PLP. The λmax =411 nm for Cap15(K265A) did not shift but increased in intensityfollowing the addition of PLP (SI Appendix, Fig. S16B). Uponaddition of excess (5′S,6′R)-GlyU to the wild-type Cap15-PLPsolution, a red shift was observed to λmax = 415 nm (Fig. 4A,trace IV). Cap15(K230A) and Cap15(K303A), both active ascatalysts, had similar spectral shifts (SI Appendix, Fig. S16C). Incontrast, the UV/Vis spectra for Cap15(K262A) and Cap15(K265A)did not change (SI Appendix, Fig. S16C). As expected, nonsubstrates(5′S,6′S)-GlyU and (5′R,6′S)-GlyU did not significantly alter thespectral profile when added to the wild-type Cap15-PLP mixture(SI Appendix, Fig. S17). Closer examination of the Cap15 UV/Visspectrum revealed a second λmax at 503 nm, which appeared im-mediately following the addition of (5′S,6′R)-GlyU and remainednearly constant over the course of 24 min (Fig. 4B, trace V). Incontrast to the other mutant proteins, only Cap15(K303A) displayedthe identical λmax at 503 nm (SI Appendix, Fig. S16D).

DiscussionThe functional assignment of Cap15 has revealed unprecedentedenzyme chemistry that involves PLP as an oxygenase cofactor, anunusual role for an already well-known and highly versatile co-factor. Several enzymes involved in primary metabolism that

Fig. 3. Activity analysis of Cap15 and CapH. (A) Dependence of Cap15 activity on buffer and pH. (B) Single-substrate kinetic analysis with variable (5′S,6′R)-GlyU by detection of CarU product by HPLC (○) or by detection of coproduct CO2 (■). (C) HPLC traces of (i) phosgene-modified, synthetic (5′S,6′R)-GlyU (●);(ii) phosgene-modified, synthetic (5′S,6′S)-GlyU (♦); (iii) phosgene-modified product of CapH-catalyzed reaction; and (iv) negative control without CapH.

Fig. 4. Mechanism of Cap15. (A) UV/Vis spectra of PLP in phosphate pH 7.5 (dotted line); Cap15 before (dashed line) and after (black solid line) the additionof PLP; and the ternary complex after addition of (5′S,6′R)-GlyU (rainbow lines). (B) UV/Vis spectra for Cap15-PLP (black solid line) and ternary complex(rainbow lines) at longer wavelength revealing the λmax at 503 nm. Spectral datasets were zeroed at 600 nm. (C) Proposed mechanism for Cap15. The romannumerals under the chemical species are proposed in part based on interpretation of the spectra highlighted in A and B.

Huang et al. PNAS | January 30, 2018 | vol. 115 | no. 5 | 977

BIOCH

EMISTR

Y

generate carbanionic intermediates have previously been reportedto catalyze off-pathway side reactions—so-called paracatalyticreactions—that involve oxidation of the carbanionic interme-diate with O2 (20, 21). Given that most members of the PLP-dependent transferase superfamily are characterized by theirability to stabilize a carbanion through electron delocaliza-tion within PLP, it is not surprising that five distinct PLP-dependent enzymes are included in the paracatalytic reactionlist (20). These five PLP-dependent enzymes use an aminoacid substrate to catalyze oxidative deamination in additionto decarboxylation, the former of which was not detected withCap15. For three of these PLP-dependent enzyme “paracatalysts,”H2O2 is stoichiometrically produced, and thus they are classifiedas oxidases (20, 22–25). The fate of the oxygen atoms for theother two paracatalysts, however, remains unknown.In contrast to PLP-dependent enzymes that catalyze para-

catalytic reactions, a few enzymes have recently been reported tocatalyze PLP-dependent oxidations as their primary function (3, 4,26, 27). Ind4, involved in indolymycin biosynthesis, was shown tocatalyze tandem 2,3- and 4,5-dehydrogenation of L-Arg, concomi-tantly reducing O2 to H2O2 (3). CcdF, involved in celesticin bio-synthesis, was shown to catalyze an oxidative decarboxylation anddeamination of an α–amino-acid component of a celesticin pre-cursor, also concomitantly reducing O2 to H2O2 (26). Thus, bothInd4 and CcdF are bona fide PLP-dependent oxidases. Contrastingly,MppP involved in L-enduracidine biosynthesis was shown to cata-lyze an O2 and PLP-dependent α-deamination and 4-hydroxylationof L-Arg, the latter suggesting that this enzyme is an oxygenase (4).However, neither the fate of the oxygen atoms of O2 nor the origin ofthe oxygen atom in the hydroxylated product was determined, andthus the enzyme classification of MppP is not clear. Our results de-finitively establish the use of PLP as a redox cofactor that enablesCap15 to function as a true monooxygenase, incorporating oneO atom of O2 directly into the carboxamide-containing productCarU. Interestingly, a similar PLP-dependent, α–amino-acid-to-carboxamide transformation has been proposed during pyoverdinebiosynthesis, although the activity of the probable catalyst (PvdN)remains to be demonstrated in vitro (27).Bioinformatically, Cap15 has primary sequence homology to

bacterial proteins annotated as SelA, which is a fold-type I PLP-dependent enzyme that catalyzes the transformation of seryl-tRNA(Sec) to selenocysteinyl-tRNA(Sec) using selenophosphate as theselenium donor (13). The discovery that the Cap15 reaction notonly requires O2 but also depends on phosphate, in this case thestandard oxygenated variant, gave rise to the possibility that theO atom incorporated into CarU could originate directly fromphosphate in a reaction analogous to that catalyzed by SelA.However, the isotopic enrichment studies with 18O-labeledphosphate, compared with 18O-labeled O2, clearly revealed thatthis was not the case, thus leaving the role of phosphate in theCap15-catalyzed reaction unknown. Interestingly, the dioxyge-nase CapA, which initiates the CarU biosynthetic pathway,produces phosphate via an unusual oxidative dephosphorylationmechanism that could potentially participate as a feed-forwardallosteric modulator (11). Preliminary analysis suggests thatCap15 does in fact differentially oligomerize in the presence ofphosphate (SI Appendix, Fig. S18), although ongoing kinetic andstructural studies will be essential to clarify the quaternary structureand role of phosphate.Wild-type Cap15 (I) does not copurify with a detectable amount

of PLP (Fig. 4B) and is inactive without an exogenous supply ofPLP. Furthermore, the addition of PLP to Cap15 (III) resulted ina UV/Vis spectrum that is nearly indistinguishable from the al-dehyde form of PLP (II), which is in contrast to most other PLP-dependent enzymes that display a significant red shift (λmax of410–440 nm) that is indicative of an internal aldimine (19). As aresult, we initially considered a mechanism that bypasses the in-ternal aldimine. However, studies with PLP-dependent aspartate

aminotransferase (28) and 1-aminocyclopropane-1-carboxylatesynthase (29), among others (2, 19), have revealed an internalaldimine that—depending upon the orientation and protonationstate of PLP in the active site—can display a UV/Vis spectrumwith a λmax ranging from 390 nm (corresponding to an unproto-nated imine) to 420 nm (corresponding to a protonated imine).Additionally, mutational analysis of Cap15 revealed that the in-active mutant Cap15(K265A) displays a spectrum with a λmax of411 nm following purification and addition of exogenous PLP,which is consistent with formation of a protonated internal aldi-mine (30, 31). In lieu of the literature precedent along with thesedata, we propose a mechanism wherein Cap15 does indeed form aprotonated internal aldimine (III) with an unidentified Lys that ischaracterized by a λmax of 391 nm (Fig. 4B). Mutation of Lys265 isproposed to promote orientation of the internal aldimine to favorprotonation, hence displaying the more traditional spectrum as-sociated with PLP-dependent enzymes. Regardless, the red shiftto λmax at 416 nm following the addition of (5′S,6′R)-GlyU to theCap15-PLP complex strongly suggests the formation of an exter-nal aldimine (IV) that undergoes Cα-deprotonation to form aresonance-stabilized quininoid (V), which is characterized by theappearance of a second λmax at 503 nm in the spectrum for wild-type Cap15 and Cap15(K303A). Similar to the results with MppPand Ind4 (3, 4), the quininoid appears to reach a steady stateduring the reaction that would suggest that downstream chemistryis rate-limiting (2).The most intriguing aspect of the Cap15 reaction is the oxi-

dative steps that follow aldimine formation. Other enzymes thatcatalyze oxidative decarboxylation of α-amino acids to formcarboxamides—namely Trp (32), Lys (33), Arg (34), and Phe(35) 2-monooxygenases–require flavin adenine dinucleotide(FAD) to initiate electron transfer to dioxygen (36). For Trp-2-monooxygenase, the best characterized of the carboxamide-forming flavoproteins, the reaction begins by hydride transferfrom Trp to FAD to generate an imine and reduced FAD, re-spectively (32). The reduced FAD subsequently donates anelectron to O2 to yield a caged radical pair that, following spininversion, recombines to form a C(4a)-hydroperoxide interme-diate (34). Contrastingly, the oxidative steps of the Cap15-cata-lyzed reaction do not require a flavin cofactor, nor does it appearthat a redox active metal cofactor is involved. Consequently, wepropose a hydroperoxide mechanism for Cap15 (Fig. 4B),wherein the carbanion resonance form of the quininoid (V)serves a role that is comparable to the reduced FAD for Trp-2-monooxygenase. Thus, an electron is first transferred from thecarbanion to O2 to form a resonance-stabilized GlyU-PLP rad-ical and superoxide, the latter of which can undergo protonationand rebound with the carbon-centered radical to form the hy-droperoxide species. Subsequently, decarboxylation occurs withelimination of the distal hydroperoxyl O atom in the form ofwater to generate a CarU-PLP aldimine that is ultimately hy-drolyzed to release CarU. Although it remains possible that theinitial carbanion is generated by decarboxylation instead ofdeprotonation of the aldimine, we were unable to detect CO2under anaerobic conditions, thus suggesting that O2 reductionprecedes decarboxylation.In addition to the unusual role of PLP as a redox cofactor, the

finding that Cap15 is stereospecific for the (5′S,6′R)-diastereomerof GlyU was unexpected. This led us to reexamine the stereo-chemical outcome of the transaldolase CapH. Although it ismechanistically feasible for CapH to produce (5′S,6′R)-GlyU, theproduct was clearly identified as (5′S,6′R)-GlyU upon compari-sons with synthetic standards and the product of the well-char-acterized transaldolase LipK (14). Thus, the discovery of Cap15activity—despite filling a major gap regarding the biosyntheticmechanism of CarU—introduces another biochemical step in thepathway and suggests the involvement of an additional, un-identified epimerase. Intriguingly, a putative NAD-dependent

978 | www.pnas.org/cgi/doi/10.1073/pnas.1718667115 Huang et al.

NDP-hexose epimerase is among the few remaining unassignedproteins within the gene cluster; however, other possibilities can-not be excluded at this time.In summary, we have provided evidence to support the func-

tional assignment of Cap15 as a PLP-dependent (5′S,6′R)-GlyU:O2 monooxygenase-decarboxylase that generates the carboxamidefunctionality of CarU, the signature nucleoside core of thecapuramycins. Generally, enzymes known to activate O2 do so byemploying a flavin or transition metal cofactor or, in some in-stances, are substrate-assisted (also known as cofactor indepen-dent). Indeed, enzymes that catalyze a comparable α-aminoacid→carboxamide conversion require a reduced FAD to achievethis chemical outcome. Characterization of Cap15 now provides adefinitive example of an enzyme whose native function requiresPLP as an oxygenase cofactor. With the mechanistic frameworkprovided here, it is now possible to compare the role of the flavinand PLP cofactors as it relates to the oxidative decarboxylation ofamino acids to produce the respective carboxamide-containingproducts. Furthermore, the abundance of predicted PLP-dependent

enzymes per genome with as-of-yet-unknown function hints atthe possibility that other PLP-dependent oxygenases not unlikeCap15 remain to be discovered.

Materials and MethodsCloning, mutatgenesis, heterologous production, and activity analysis forCap15 were performed using standard procedures and are described in the SIAppendix, SI Materials and Methods. Reactions were monitored by C18 re-verse phase HPLC, LC-MS, a Carbon Dioxide Enzymatic Assay kit (DiazymeLaboratories), and a Profiling Oxygen Microsensor PM-PSt7 equipped withMicrox 4 oxygen meter (PreSens Precision Sensing), and UV-visible spectros-copy was performed with a Shimadzu UV/Vis-1800 spectrophotometer. Thesynthesis of (5′R,6′S)-GlyU, (5′S,6′R)-GlyU, and U5′A was previously described(14–17). The syntheses and analytical analyses of (5′S,6′S)-GlyU and CarU aredescribed in the SI Appendix, SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Dr. Koichi Nonaka (Daiichi-Sankyo Co.) forintellectual contributions and for providing strains and cosmids. This workwas supported in part by National Institutes of Health Grants AI128862 andAI087849 (to S.G.V.L.).

1. Percudani R, Peracchi A (2009) The B6 database: A tool for the description and clas-sification of vitamin B6-dependent enzymatic activities and of the correspondingprotein families. BMC Bioinformatics 10:273.

2. Toney MD (2011) Controlling reaction specificity in pyridoxal phosphate enzymes.Biochim Biophys Acta 1814:1407–1418.

3. Du Y-L, et al. (2016) A pyridoxal phosphate-dependent enzyme that oxidizes an un-activated carbon-carbon bond. Nat Chem Biol 12:194–199.

4. Han L, Schwabacher AW, Moran GR, Silvaggi NR (2015) Streptomyces wadayamensisMppP is a pyridoxal 5′-phosphate-dependent L-arginine α-deaminase, γ-hydroxylasein the enduracidine biosynthetic pathway. Biochemistry 54:7029–7040.

5. Muramatsu Y, et al. (2003) Studies on novel bacterial translocase I inhibitors, A-500359s. I. Taxonomy, fermentation, isolation, physico-chemical properties andstructure elucidation of A-500359 A, C, D and G. J Antibiot (Tokyo) 56:243–252.

6. Muramatsu Y, et al. (2004) A-503083 A, B, E and F, novel inhibitors of bacterial trans-locase I, produced by Streptomyces sp. SANK 62799. J Antibiot (Tokyo) 57:639–646.

7. Murakami R, et al. (2007) A-102395, a new inhibitor of bacterial translocase I, pro-duced by Amycolatopsis sp. SANK 60206. J Antibiot (Tokyo) 60:690–695.

8. Yamaguchi H, et al. (1986) Capuramycin, a new nucleoside antibiotic. Taxonomy,fermentation, isolation and characterization. J Antibiot (Tokyo) 39:1047–1053.

9. Funabashi M, et al. (2009) Identification of the biosynthetic gene cluster of A-500359sin Streptomyces griseus SANK60196. J Antibiot (Tokyo) 62:325–332.

10. Funabashi M, et al. (2010) An ATP-independent strategy for amide bond formation inantibiotic biosynthesis. Nat Chem Biol 6:581–586.

11. Cai W, et al. (2015) The biosynthesis of capuramycin-type antibiotics: Identification ofthe A-102395 biosynthetic gene cluster, mechanism of self-resistance, and formationof uridine-5′-carboxamide. J Biol Chem 290:13710–13724.

12. Goswami A, et al. (2017) Evidence that oxidative dephosphorylation by the nonhemeFe(II), α-ketoglutarate:UMP oxygenase occurs by stereospecific hydroxylation. FEBSLett 591:468–478.

13. Itoh Y, et al. (2013) Decameric SelA•tRNA(Sec) ring structure reveals mechanism ofbacterial selenocysteine formation. Science 340:75–78.

14. Barnard-Britson S, et al. (2012) Amalgamation of nucleosides and amino acids inantibiotic biosynthesis: Discovery of an L-threonine:uridine-5′-aldehyde transaldolase.J Am Chem Soc 134:18514–18517.

15. Spork AP, Ducho C (2013) Stereocontrolled synthesis of 5′- and 6′-epimeric analoguesof muraymycin nucleoside antibiotics. Synlett 24:343–346.

16. Spork AP, Koppermann S, Dittrich B, Herbst-Irmer R, Ducho C (2010) Efficient synthesisof the core structure of muraymycin and caprazamycin nucleoside antibiotics basedon a stereochemically revised sulfur ylide reaction. Tetrahedron Asymmetry 21:763–766.

17. Spork AP, et al. (2014) Lead structures for new antibacterials: Stereocontrolled syn-thesis of a bioactive muraymycin analogue. Chemistry 20:15292–15297.

18. Johnson RJ, Metzler DE (1970) Analyzing spectra of vitamin B6 derivatives. MethodsEnzymol 18:433–471.

19. Kallen RG, et al. (1985) Chemical and spectroscopic properties of pyridoxal and pyr-idoxamine phosphates. Transaminases, eds Christen P, Metzler DE (Wiley, New York),pp 37–108.

20. Bunik VI, Schloss JV, Pinto JT, Dudareva N, Cooper AJL (2011) A survey of oxidativeparacatalytic reactions catalyzed by enzymes that generate carbanionic intermedi-

ates: Implications for ROS production, cancer etiology, and neurodegenerative dis-

eases. Adv Enzymol Relat Areas Mol Biol 77:307–360.21. Abell LM, Schloss JV (1991) Oxygenase side reactions of acetolactate synthase and

other carbanion-forming enzymes. Biochemistry 30:7883–7887.22. Sakai K, et al. (1997) L-ornithine decarboxylase from Hafnia alvei has a novel L-ornithine

oxidase activity. J Biochem 122:961–968.23. Kaminaga Y, et al. (2006) Plant phenylacetaldehyde synthase is a bifunctional ho-

motetrameric enzyme that catalyzes phenylalanine decarboxylation and oxidation.

J Biol Chem 281:23357–23366.24. Bertoldi M, Cellini B, Montioli R, Borri Voltattorni C (2008) Insights into the mecha-

nism of oxidative deamination catalyzed by DOPA decarboxylase. Biochemistry 47:

7187–7195.25. Bertoldi M, Borri Voltattorni C (2000) Reaction of dopa decarboxylase with L-aromatic

amino acids under aerobic and anaerobic conditions. Biochem J 352:533–538.26. Wang M, Zhao Q, Zhang Q, Liu W (2016) Differences in PLP-dependent cysteinyl

processing lead to diverse S-functionalization of lincosamide antibiotics. J Am Chem

Soc 138:6348–6351.27. Ringel MT, Dräger G, Brüser T (2016) PvdN enzyme catalyzes a periplasmic pyoverdine

modification. J Biol Chem 291:23929–23938.28. Goldberg JM, et al. (1993) Structure of the complex between pyridoxal 5′-phosphate

and the tyrosine 225 to phenylalanine mutant of Escherichia coli aspartate amino-

transferase determined by isotope-edited classical Raman difference spectroscopy.Biochemistry 32:8092–8097.

29. Li Y, Feng L, Kirsch JF (1997) Kinetic and spectroscopic investigations of wild-type and

mutant forms of apple 1-aminocyclopropane-1-carboxylate synthase. Biochemistry36:15477–15488.

30. Zhang J, Cheltsov AV, Ferreira GC (2005) Conversion of 5-aminolevulinate synthase

into a more active enzyme by linking the two subunits: Spectroscopic and kineticproperties. Protein Sci 14:1190–1200.

31. Vacca RA, Giannattasio S, Capitani G, Marra E, Christen P (2008) Molecular evolutionof B6 enzymes: Binding of pyridoxal-5′-phosphate and Lys41Arg substitution turn ri-

bonuclease A into a model B6 protoenzyme. BMC Biochem 9:17.32. Emanuele JJ, Fitzpatrick PF (1995) Mechanistic studies of the flavoprotein tryptophan

2-monooxygenase. 1. Kinetic mechanism. Biochemistry 34:3710–3715.33. Matsui D, et al. (2014) Mutational and crystallographic analysis of l-amino acid oxi-

dase/monooxygenase from Pseudomonas sp. AIU 813: Interconversion between oxi-

dase and monooxygenase activities. FEBS Open Bio 4:220–228.34. Hong H, Fill T, Leadlay PF (2013) A common origin for guanidinobutanoate starter

units in antifungal natural products. Angew Chem Int Ed Engl 52:13096–13099.35. Ida K, et al. (2008) Structural basis of proteolytic activation of L-phenylalanine oxidase

from Pseudomonas sp. P-501. J Biol Chem 283:16584–16590.36. Massey V (1994) Activation of molecular oxygen by flavins and flavoproteins.

J Biol Chem 269:22459–22462.

Huang et al. PNAS | January 30, 2018 | vol. 115 | no. 5 | 979

BIOCH

EMISTR

Y