Structural basis for nonribosomal peptide synthesisby an aminoacyl-tRNA synthetase paralogLuc Bonnefonda, Taiga Araib, Yuriko Sakaguchib, Tsutomu Suzukib, Ryuichiro Ishitania, and Osamu Nurekia,1

aDepartment of Biophysics and Biochemistry, Graduate School of Science, University of Tokyo, 2-11-16 Yayoi, Bunkyo-ku, 113-0032 Tokyo, Japan; andbDepartment of Chemistry and Biotechnology, Graduate School of Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, 113-8656 Tokyo, Japan

Edited by Paul Schimmel, The Skaggs Institute for Chemical Biology, La Jolla, CA, and approved January 14, 2011 (received for review December 23, 2010)

Cyclodipeptides are secondary metabolites biosynthesized bymany bacteria and exhibit a wide array of biological activities.Recently, a new class of small proteins, named cyclodipeptidesynthases (CDPS), which are unrelated to the typical nonribosomalpeptide synthetases, was shown to generate several cyclodipep-tides, using aminoacyl-tRNAs as substrates. The Mycobacteriumtuberculosis CDPS, Rv2275, was found to generate cyclodityrosinethrough the formation of an aminoacyl-enzyme intermediateand to have a structure and oligomeric state similar to those ofthe class Ic aminoacyl-tRNA synthetases (aaRSs). However, the poorsequence conservation among CDPSs has raised questions aboutthe architecture and catalytic mechanism of the identified homo-logs. Here we report the crystal structures of Bacillus licheniformisCDPS YvmC-Blic, in the apo form and complexed with substratemimics, at 1.7–2.4-Å resolutions. The YvmC-Blic structure also exhi-bits similarity to the class Ic aaRSs catalytic domain. Our mutationalanalysis confirmed the importance of a set of residues for cyclodi-leucine formation among the conserved residues localized in thecatalytic pocket. Our biochemical data indicated that YvmC-Blicbinds tRNA and generates cyclodileucine as a monomer. We werealso able to detect the presence of an aminoacyl-enzyme reactionintermediate, but not a dipeptide tRNA intermediate, whoseexistence was postulated for Rv2275. Instead, our results supporta sequential catalytic mechanism for YvmC-Blic, with the successiveattachment of two leucine residues on the enzyme via a conservedserine residue. Altogether, our findings suggest that all CDPSenzymes share a common aaRS-like architecture and a catalyticmechanism involving the formation of an enzyme-bound inter-mediate.

X-ray crystallography ∣ pulcherrimin ∣ diketopiperazine ∣ divergentevolution ∣ thioesterase domain

Cyclodipeptides belong to the diketopiperazine (DKP) familyof secondary metabolites that are produced in many bacteria.This class of compounds has recently gained interest due to itswide array of biological activities, including antibacterial, anti-fungal, antiviral, antiprion, antitumor, and immunosuppressivefunctions (1, 2). Most cyclodipeptide biosynthesis pathways in-volve nonribosomal peptide synthetases (NRPSs) (3, 4). NRPSsare large multifunctional modular enzymes in which each modulecatalyzes a single step of peptide synthesis by incorporating oneamino acid. In the standard NRPS module, the peptide bond-forming condensation domain (C) and the amino acid adenyla-tion domain (A) are linked through a phosphopantetheine-tethered domain (T) to synthesize the peptidyl chain as a growingthioester. The enzyme usually ends with a thioesterase domain(TE), which releases the peptidyl chain by hydrolysis or cycliza-tion (5).

However, in 2002 Lautru et al. identified the albC gene in thegenome of the bacterium Streptomyces noursei, and its productAlbC, which lacked similarity to the NRPSs, was able to catalyzethe formation of a cyclic dipeptide (6). AlbC is the first enzyme inthe biosynthetic pathway of albonoursin, a secreted antibacterialDKP bearing a cyclodipeptide structure. Recently, AlbC wasshown to use aminoacylated tRNAs as substrates instead of free

amino acids (7), thus allowing the enzyme to skip the amino acidactivation step usually performed by the adenylation domainin NRPSs. Mining of bacterial genomes identified seven otherpotentially homologous proteins sharing only 20 to 26% sequenceidentity with AlbC. They were all shown to use aminoacyl-tRNAsto generate a variety of cyclodipeptides (7) and have been namedcyclodipeptide synthases (CDPSs). Although other enzymes havebeen reported to divert aminoacylated tRNAs from their primaryuse in the translation process [e.g., the FemXAB family, partici-pating in peptidoglycan synthesis (8); the L/F-transferases, in-volved in protein turnover (9)], they lack sequence similaritywith the CDPS enzymes. In addition, CDPSs are peculiar becausethey bring tRNA into the secondary metabolism.

The structure of the Mycobacterium tuberculosis CDPS,Rv2275, was recently solved (10), revealing a striking structuralsimilarity to the catalytic domains of the class Ic aminoacyl-tRNA synthetases (aaRSs; i.e., tyrosyl- and tryptophanyl-tRNAsynthetases, TyrRS, and TrpRS), although they share less than15% sequence identity between them. Rv2275 forms a dimer andcatalyzes the formation of cyclodityrosine (cYY), by the transientattachment of the tyrosine moiety to a serine residue in the activesite. The authors proposed a catalytic mechanism that involvesthe formation of a dityrosine-tRNA intermediate.

Bacillus licheniformisYvmC-Blic is another CDPS enzyme thatcatalyzes the formation of cyclodileucine (cLL), by employingtwo Leu-tRNALeu molecules (7). cLL is a precursor of pulcher-rimin (Fig. 1), a red extracellular pigment formed by severalspecies of bacteria (including B. licheniformis) after growth inFe3þ ion-enriched media. This pathway was best characterizedin Bacillus subtilis (11). First, YvmC-Bsub catalyzes the formationof cLL from Leu-tRNALeu (7). The cytochrome P450 CypXthen oxidizes cLL into pulcherriminic acid (12), which is releasedin the medium and reacts with an Fe3þ ion to produce pulcher-rimin. In B. subtilis, B. licheniformis, and other related bacteria,the cypX gene is located downstream of the yvmC gene (7).

We now report the structure of YvmC-Blic, which also resem-bles those of the catalytic domains of the class Ic aaRSs, despitethe poor sequence similarity. Our biochemical data indicatedthat cLL formation by YvmC-Blic did not require either ATP orfreestanding amino acids. Mutational analyses revealed a set ofCDPS-specific residues that recognize the aminoacyl moiety andcatalyze its transient attachment to a conserved serine residue.Analyses by mass spectrometry failed to detect the dileucine-tRNA intermediate, prompting us to propose a different catalyticmechanism from that postulated for cYY formation by Rv2275

Author contributions: L.B., T.S., and O.N. designed research; L.B. and T.A. performedresearch; L.B., T.A., Y.S., T.S., and R.I. analyzed data; L.B., R.I., and O.N. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and structure factors for the YvmC-Blic apo form,the CHES, and the CAPSO complexes have been deposited in the Protein Data Bank,www.pdb.org (PDB ID codes 3OQH, 3OQI, and 3OQJ, respectively).1To whom correspondence should be addressed. E-mail: nureki@ims.u-tokyo.ac.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental.

3912–3917 ∣ PNAS ∣ March 8, 2011 ∣ vol. 108 ∣ no. 10 www.pnas.org/cgi/doi/10.1073/pnas.1019480108

Dow

nloa

ded

by g

uest

on

July

9, 2

021

www.pdb.orgwww.pdb.orgwww.pdb.orghttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental

(10). Altogether, our findings suggest that all CDPS enzymesshare a common aaRS-like architecture and a catalytic mechan-ism involving the formation of an enzyme-bound intermediate.

Results and DiscussionThe Architecture of YvmC-Blic Is Similar to That of a Class Ic aaRSCatalytic Domain. The crystal structures of B. licheniformisCDPS YvmC-Blic, in the apo form and in the complexes with2-(N-cyclohexylamino)ethane sulfonic acid (CHES) and 3-(cyclo-hexylamino)-2-hydroxy-1-propanesulfonic acid (CAPSO), weredetermined at 1.9-, 1.7-, and 2.4-Å resolutions, respectively(PDB IDs 3OQH, 3OQI, 3OQJ) (Table S1). In all structures,the 11 N-terminal and 14 C-terminal residues are disordered.The asymmetric unit contains one or two monomers of YvmC-Blic, depending on the crystal form (Fig. S1). The YvmC-Blicstructure exhibits striking structural similarity with those of theclass Ic aaRS catalytic domains, consisting of an α/β fold withsix strand-helix units, but it lacks the last helix and the helicalinserts before and after the fourth strand, as compared withthe aaRS catalytic domain (Fig. 2A and B). The YvmC-Blic struc-ture superimposed well on those of several eukaryotic TyrRSsand TrpRSs, despite the low sequence identity (less than 15%).For instance, the superimposition of the YvmC-Blic structure onthose of Saccharomyces cerevisiae TyrRS (PDB ID code 2DLC)and TrpRS (PDB ID code 3KT6) yielded rms deviations of theCα atoms of 3.4 and 3.6 Å over 182 and 176 residues, respectively(Fig. S2A).

However, a structural comparison of YvmC-Blic and yeastTyrRS revealed several important differences. First, both classI aaRS signature motifs (i.e., HIGH and KMSKS) are absentin YvmC-Blic, and most of the residues specific to the class IcaaRSs are not conserved (13,14). Second, the YvmC-Blic cataly-tic pocket lacks the residues that bind ATP in TyrRSs and istoo small to accommodate ATP (Fig. S2A). Third, YvmC-Bliccontains a long, flexible loop at the end of the CP1 insertion ofthe class Ic aaRSs. The loop adopts a different conformation ineach structure, partly due to the intermolecular crystal contacts,and is always disordered in one of the monomers in the asym-metric unit (Fig. S2B). The loop is localized at the top of thecatalytic pocket, suggesting a role analogous to that of theKMSKS loop in classI aaRSs (i.e., covering the catalytic pocketduring the two-step reaction mechanism). Fourth, the long helixat the “back” of YvmC-Blic is rich in positively charged residuesand might form the tRNA-binding surface (Fig. S2C). CDPSsequences usually contain a number of basic residues in thisloop (Fig. S3), suggesting a similar tRNA-binding mode for allCDPSs. These differences may account for the functional differ-ences observed in the reactions catalyzed by CDPSs and aaRSs.

Despite the moderate sequence identity (27%) (Table S2),YvmC-Blic and Rv2275 exhibit high structural similarity, with anrms deviation of 2.2 Å for 211 Cα atoms (Fig. S2D). Both CDPSstructures also share a similarly twisted β-sheet, as compared tothat of the TyrRSs (Fig. S2E). This observation suggests a con-served architecture for all CDPSs. The major difference betweenboth CDPSs is the conformation of the helix α6 and the followingflexible loop covering the catalytic pocket, which might reflecttheir respective substrate specificities.

CHES and CAPSO Bind in the Catalytic Pocket. Intriguingly, the buffermolecules CHES and CAPSO, which were critical for growing theYvmC-Blic crystals, were both clearly visible in the respectiveelectron density maps (Fig. 2 C and D). Two CHES moleculeswere bound to each monomer of CDPS, one in the catalyticpocket and the other one at the border of the pocket, mediatingthe crystallographic contacts with a symmetric CDPS molecule.

Fig. 1. The biosynthetic pathway of pulcherrimin in B. licheniformis. The formation of pulcherriminic acid from Leu-tRNALeu requires two enzymes, the CDPSYvmC-Blic and the cytochrome P450 CypX. YvmC-Blic first generates cyclodileucine, which is then oxidized by CypX. A nonenzymatic reaction with Fe3þ ions inthe extracellular medium transforms pulcherreminic acid into pulcherrimin.

Fig. 2. YvmC-Blic structure and ligand binding. (A) Overall structure ofYvmC-Blic. Each type of secondary structure is colored differently: Loopsare white, helices are light blue, and strands are deep blue. (B) Overall struc-ture of the yeast TyrRS catalytic domain (PDB ID code 2DLC), in the same re-presentation as A. The tyrosyl-adenylate analog bound in the catalytic site isdrawn as a sphere model. (C) Recognition of the two CHES molecules. Watermolecules are shown as red spheres. Hydrogen bonds are represented bydotted lines. The F0-FC difference maps (4.0σ) for the CHES molecules are co-lored yellow. (D) Recognition of the CAPSO molecule. The F0-FC differencemap (3.0σ) for the CAPSO molecule is colored green. (E) Locations of con-served residues around the catalytic pockets of YvmC-Blic and Rv2275(PDB ID code 2X9Q), colored light and dark blue, respectively. For compari-son, the yeast TyrRS residues that hydrogen bond with the tyrosyl amidegroup are depicted as semitransparent green stick models, together withthe tyrosyl-adenylate analog. Residue names and/or numbering correspondto YvmC-Blic, Rv2275, and yeast TyrRS.

Bonnefond et al. PNAS ∣ March 8, 2011 ∣ vol. 108 ∣ no. 10 ∣ 3913

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

July

9, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=ST1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=ST2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2

The cyclohexane ring of the first CHES molecule forms van derWaals contacts with the main chains of Gly35 and Leu65, while itsmain chain nitrogen atom hydrogen bonds with the side chains ofTyr180 and Glu184. The sulfate oxygen atoms of the CHES mo-lecule hydrogen bond with the Arg206 and Asn40 amide groupsand the Tyr204 hydroxyl group. The cyclohexane ring of the sec-ond CHES molecule interacts hydrophobically with the sidechains of Arg206 and Pro207 and with Met131, Glu134, andAla135 from the symmetric molecule. The oxygen atoms ofthe sulfate group hydrogen bond with the main chains ofLys209 and Leu210. In the CAPSO-bound structure, the CAPSOcyclohexane ring interacts hydrophobically with the Gly35 mainchain and the Leu65 and Phe188 side chains, while its main chainnitrogen and oxygen atoms hydrogen bond with the Glu184carboxyl groups. One oxygen atom of the sulfate group hydrogenbonds with the Arg206 amino group. Notably, the positions of thecyclohexane rings in the first CHES and the CAPSO moleculescorrespond to the positions of the side chains of the bound Tyrand Trp molecules in the catalytic sites of the correspondingaaRSs. Furthermore, Phe is one of the amino acids used byYvmC-Blic to generate cyclodipeptides. These observations sug-gest that the two buffer molecules mimic the aminoacyl moietiesof the physiological substrates.

The hydrophobic residues at the bottom of the pocket areloosely conserved, which may explain the relaxed specificity ofthe CDPSs. A superimposition of the YvmC-Blic and yeast TyrRSstructures suggests the functional conservation of Yvmc-BlicTyr180 and Glu184 for recognizing the amide group of the Leumoiety in Leu-tRNALeu (Fig. 2E). A comparison of the residuesaround the catalytic pocket between Rv2275 and YvmC-Blicrevealed the conservation of several residues located at theentrance of the pocket, which are likely to be involved in the cat-alytic mechanism (see below). In the Rv2275 catalytic pocket,Asn251 was proposed to recognize the hydroxyl group of thesubstrate tyrosyl moiety (10). Asn251 is replaced by Leu202 inYvmC-Blic (Fig. 2E), and the Leu residue is conserved amongthe cLL-producing CDPSs (Fig. S3) (7). This observation sug-gests that the residue at this position is involved in definingthe CDPS substrate specificity. Despite repeated crystallizationtrials in the presence of small ligands, we were unable to obtainany consistent electron density for the free amino acids or ami-noacyl-tRNA mimics in the pocket. Rv2275 was also crystallizedonly in the absence of ligands. The narrow access to the pocketin the solved structures suggests that they represent a “closed”state of the enzyme, which might require tRNA binding to open.

YvmC-Blic Is Active as a Monomer. Previous experiments definedaminoacyl-tRNAs as the substrates for CDPSs (7) and suggesteda sequential mechanism for tRNA binding and cYY formationfor Rv2275 (10). To assess the catalytic mechanism of YvmC-Blic,we performed biochemical assays to define the origins of theincorporated amino acids. Because YvmC-Blic preferentiallyproduces cyclo-L-leucyl-L-leucine, we employed stable isotope13C-labeled Leu and tRNALeu. By reacting the labeled [13C]Leu (Leu*) with purified B. licheniformis tRNALeu and LeuRS,we produced Leu*-tRNALeu, which was subsequently purified toremove traces of impurities, such as ATP and free Leu*. Thispurified Leu*-tRNALeu was then reacted with YvmC-Blic in thepresence of a 10-fold excess of unlabeled free Leu, to prevent theincorporation of Leu* liberated by the spontaneous deacylationof Leu*-tRNALeu. The ester bond of aminoacyl-tRNAs is un-stable and can be quickly hydrolyzed in in vitro assays. Our massspectrometry analysis of the reaction revealed the incorporationof two Leu* residues in the final product (cL*L*). ATP was notnecessary for cLL formation. The experiment using unlabeledLeu-tRNALeu in the presence of a 10-fold excess of free Leu*produced unlabeled cLL (Fig. 3 A and Fig. S4). These results

established that YvmC-Blic uses Leu-tRNALeu as the sole sub-strate, with no free amino acids derived thereof.

Rv2275 reportedly forms a dimer, similar to the class Ic aaRSs,which must dimerize for tRNA binding (with the tRNA boundacross both monomers) (15) (Fig. S5). In contrast, CDPS dimer-ization does not seem to be required for cYY formation, becausethe catalytic serine residues in both active sites are 50 Å apartand located on diametrically opposite surfaces of the dimer.The dimerization mode is drastically different from that of theclass Ic aaRSs (even considering the atypical case of the mimi-virus TyrRS) and thus is incompatible with a similar tRNA-bind-ing mode (Fig. S5). Therefore, we analyzed the oligomeric statesof the B. licheniformis (YvmC-Blic) and B. subtilis (YvmC-Bsub)CDPSs, which share 70% sequence identity (Table S2) by gelfiltration. Both enzymes eluted as a single peak, with estimatedMr of 31 and 29 kDa, respectively (Fig. 3B). These values are veryclose to their theoretical Mr of 29.6 kDa and clearly indicate amonomeric state. The ability of YvmC-Blic to crystallize with dif-ferent intermolecular contacts also supports this view (Fig. S1).We further determined the stoichiometry of tRNA binding forboth enzymes by using the cognate tRNALeu. The B. licheniformisand B. subtilis tRNALeu molecules share the same 89-nucleotide-long sequence, with a Mr of 30 kDa. The tRNA migrated on thecolumn with an apparent Mr of 46 kDa, probably due to itsL-shape. The YvmC-Blic∶tRNALeu and YvmC-Bsub∶tRNALeucomplexes migrated with estimated Mr of 80 and 71 kDa, respec-tively, compatible with a 1∶1 complex of CDPS and tRNA(Fig. 3B). These results revealed that the aminoacylated stateof the tRNA is not critical for the binding by CDPSs. The inabilityof YvmC-Blic and YvmC-Bsub to dimerize and to bind twotRNALeu molecules simultaneously implies a sequential mechan-ism for cLL production from Leu-tRNALeu.

Critical Residues for Catalysis.Based on the structure of YvmC-Blicand the sequence conservation among the CDPSs, we designed aseries of point mutants and a loop-deletion mutant to elucidate

Fig. 3. YvmC-Blic substrates and stoichiometry. (A) The cyclodipeptideformation reaction was performed with stable-isotopic leucyl(L*)-tRNALeu,in the presence of free leucine (L) (Left). Detection of proton adducts ofcLL (m∕z 227.18; blue), cLL* (m∕z 234.19; red), and cL*L* (m∕z 241.21; green)is shown on the mass chromatograms. The same experiment was performedwith leucyl(L)-tRNALeu, in the presence of free, stable-isotopic leucine (L*)(Right). (B) Gel filtration analyses of YvmC-Blic and YvmC-Bsub, in thepresence (Right) or absence (Left) of tRNALeu. The estimated Mr values werecalculated from the elution volumes. In the presence of tRNALeu, YvmC-Bsubgave a single peak of the complex, whereas YvmC-Bsub gave two peaks,one corresponding to free tRNALeu and the other to the complex (indicatedwith an arrow).

3914 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1019480108 Bonnefond et al.

Dow

nloa

ded

by g

uest

on

July

9, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=ST2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF1

the catalytic mechanism (Fig. 4A). We first tested the tRNA-bind-ing ability of the mutants by a qualitative gel-shift assay. Most ofthe mutants were still able to bind the tRNA, albeit with loweraffinity than the wild-type enzyme (Fig. S6A). The tRNA-bindingabilities of the Y180F and H205E mutants, the K88A/R98A dou-ble mutant and the loop-deletion mutant were significantly de-creased, suggesting the involvement of these residues in tRNAbinding. We next measured the cLL biosynthetic activity of themutants by thin layer chromatography (Fig. 4B and Fig. S6B).The S37A, N40D, E184A, and E184Q mutants, which are all ableto bind tRNALeu, had almost no activity, suggesting their criticalrole in the catalytic mechanism. The Y204A and Y204F mutantsshowed strong defects, while the K88A/R98A, R158A, Y180A,H205, and the loop-deletion mutants showed only moderate de-fects. The negative impact of the Tyr180 and Glu184 mutationssupports their role in the recognition of the leucyl amide group.The mild effects of the tRNA binding impaired mutants on theenzyme catalytic activity confirmed their function in tRNA recog-nition. In addition, the reaction mixtures of the Y180A, E184A,Y204A, and Y204F mutants contained residual Leu-tRNALeumolecules after a 30 min reaction, probably due to the protectionof the ester bond from spontaneous deacylation (Fig. S6B).These results suggest that Tyr180, Glu184, and TyrY204 areinvolved in the latter step of catalysis rather than the initial sub-strate binding.

Ser37 and Asn40, and to a lesser extent, His205 and Tyr204,are close together in the catalytic pocket (Fig. 2E), and theirmutations impaired the cLL formation activity of YvmC-Blic(Fig. 4B). This residue arrangement reminded us of the catalyticSer, Glu, His triad found in the TE domain of NRPS (4). Thisdomain is located in the terminal module of the NRPSs andcatalyzes the final peptide release, which is covalently attachedto the triad Ser residue by either hydrolysis or macrocyclization.Furthermore, Ser37 is located within a GxSxP/G motif, in a loopbetween the β1 strand and the α2 helix, which is a similar patternto that of the catalytic serine residue from the hydrolases, locatedwithin a GxSxG consensus motif (16). These analogies led usto hypothesize that the aminoacyl moiety of Leu-tRNALeu istransferred to the Ser37 hydroxyl group. Indeed, we detected

an aminoacyl-CDPS intermediate by SDS/PAGE using radiola-beled aminoacyl-tRNAs and short incubation times (Fig. 4C).The reaction intermediate was not detected with the Ser37 andAsn40 mutants. Furthermore, the enzyme activity was maintainedafter the replacement of Ser37 by a cysteine residue, supportingthe assignment of this serine residue as the acylation site (17).The lack of intermediate formation by the S37C mutant aftera 30-s incubation time is probably due to the low reactivity ofthe sulfhydryl group in this structural context.

However, several differences were observed between YvmC-Blic and the NRPS TE domain. First, in YvmC-Blic, the His205mutation has only a mild effect on the reaction, whereas the Hisresidue in the catalytic triad of the TE domain is critical forthe activity. Second, the Asn40 of YvmC-Blic cannot be replacedby Asp, suggesting a CDPS-specific role for its amide group(Fig. 4B), whereas the triad Asp from the TE domain is strictlyconserved and critical for the enzyme activity. Third, PMSF, awell-known serine protease inhibitor that also inhibits the TEdomain activity (18), does not inhibit the cyclodipeptide forma-tion by YvmC-Blic (Fig. 4B). Fourth, YvmC-Blic uses aminoacyl-tRNAs as acyl-carriers, in contrast to the phosphopantetheinylarm of the NRPS TE domain. Fifth, the YvmC-Blic structureis unrelated to the α/β hydrolase fold of the TE domains (19)(Fig. S7). Thus, although the CDPSs and the NRPS TE domainscan form similar aminoacyl-enzyme intermediates through aconserved Ser residue, the catalytic mechanism of cyclodipeptideformation by CDPSs should be rather different from that of theNRPS TE domains for cyclization and product release.

Proposed Catalytic Mechanism. Vetting et al. proposed a catalyticmechanism for cYY formation by Rv2275, in which the catalyticserine hydroxyl group attacks the first Tyr-tRNATyr to form anRv2275-Tyr intermediate (10). The active site is then rearranged,with the rotation of the serine-tyrosyl ester from the catalyticpocket to a so-called “second surface depression” (Fig. S2F). Theemptied catalytic pocket accommodates the second Tyr-tRNATyr,which is then coupled with the first tyrosine to generate thetRNA-bound dityrosine. Finally, the enzyme assists in deproto-nating the 3′ hydroxyl group of the tRNA and the dipeptidecyclize when it adopts the cis conformation.

The present YvmC-Blic structure and our mutational analysissupport a similar mechanism for the first reaction step (i.e., thebinding of Leu-tRNALeu and the transfer of its aminoacyl moietyonto the conserved Ser37 residue). However, the following threeobservations led us to propose a different chemistry for thesecond step. First, the second surface depression in both Rv2275and YvmC-Blic is positively charged (Fig. S2 F and G) andthus seems more suitable for binding the terminal nucleotide ofthe tRNA. Second, the cyclodipeptides formed by a given CDPSenzyme almost always have one amino acid in common [e.g.,YvmC-Blic generates mostly cLL, cLF(phenylalanine) andcLM(methionine), with all three cyclodipeptide having one leu-cine residue in common] (7). Third, our mass spectrometric trialsfailed to detect a dileucine-tRNA intermediate in the reactionmixture, although it was predicted to exist as an intermediatein the catalytic mechanism proposed by Vetting et al. (10). There-fore, we believe that only the “common” aminoacyl group bindsand remains in the catalytic pocket as the first amino acid,whereas the second aminoacyl group is accommodated at a dif-ferent site with less stringent recognition. Indeed, sufficient spaceis available above the first aminoacyl moiety-binding site in theYvmC-Blic structure (Fig S2G).

Altogether, we propose the following catalytic mechanism(Fig. 5): Upon Leu-tRNALeu binding, the leucine amino groupis appropriately positioned to hydrogen bond with the side chainsof Glu184 and Tyr180 (step 1). The ester carbonyl is polarizedby the Asn40 amide group, which facilitates the nucleophilic at-tack by the Ser37 hydroxyl group. The initial transesterification

Fig. 4. Product and enzyme-bound intermediate formation by YvmC-Blicmutants. (A) Overall structure of YvmC-Blic, with the mutated residues de-picted by dark blue stick models. (B) cLL cyclodipeptide formation activityof wild-type YvmC-Blic (WT) and its mutants. Error bars represent SD ofthree independent experiments. (C) Autoradiogram of aminoacyl-enzymeintermediate formation, detected by SDS/PAGE using a [14C]Leu-tRNALeu

substrate. The thin bands in the WT, E184A, and E184Q lanes correspondto an Mr of 30 kDa, as estimated by the migration profile of a prestainedladder on the same gel.

Bonnefond et al. PNAS ∣ March 8, 2011 ∣ vol. 108 ∣ no. 10 ∣ 3915

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

July

9, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF6http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF6http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF6http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF6http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF7http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF7http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF2

does not require Glu184 (step 2). The inability of the S37Aand N40Dmutants to form the Leu-CDPS covalent intermediate,in contrast to the E184 and E184Q mutants, supports the firststep of the reaction mechanism. Subsequently, the second Leu-tRNALeu binds, and its amino group interacts with the Glu184carboxyl group (step 3). The amino group of the first leucineattacks the carbonyl ester of the second Leu-tRNALeu, to gener-ate an enzyme-bound dileucine (step 4). Finally, the dileucineadopts a cis conformation, probably due to the small volumeof the catalytic pocket, which allows rapid and spontaneouscyclization by the attack of the second leucine’s amide groupon the first leucine’s ester carbonyl (step 5). The absence ofcLL formation by the E184A and E184Q mutants, despite theformation of the Leu-CDPS intermediate, supports the secondstep of our proposed reaction mechanism.

Perspectives. Several aspects remain unclear. The impact ofaminoacyl-tRNA binding on the enzyme structure is unknown,but some remodeling seems to be required to allow access to thecatalytic pocket. The CDPS specificity for the tRNA sequencesalso remains to be determined. AaRS domain paralogs perform-ing functions other than the canonical aminoacylation have beenidentified in various organisms (20); however, these paralogshave generally retained the ability to activate their cognateamino acids. In contrast, the CDPS enzymes have lost the aminoacid activation activity and have acquired broader or differentamino acid specificities. Indeed, they can bind and process Tyr-(Rv2275) and Trp-tRNAs, as suggested by their structuralsimilarity to the class Ic aaRSs, as well as Leu- (YvmC-Blic),Met-, and Phe-tRNAs (7). With so few examples of CDPSspresently available, the evolutionary history of this protein familyis obscure. The apparently scarce distribution of CDPS genes inbacterial genomes, and potentially in other phyla, is also puzzling.A structure-based phylogenetic analysis of the YvmC-Blic andRv2275 enzymes placed them at the junction between the TyrRSsand TrpRSs, suggesting an ancient origin preceding the classIc aaRS differentiation event (Fig. S8). It will also be interestingto test whether the gene products identified in archaea andeukaryotes (e.g., the sea anemone Nematostella vectensis) are in-deed active CDPSs.

Such small proteins are easy to produce and manipulate andthus could be used for the synthesis of cyclodipeptides. This couldprovide access to a larger repertoire of scaffolds for the develop-ment of new, potentially therapeutic molecules. Understandingthe catalytic mechanism of CDPS is the first step toward theengineering of enzymes producing desired cyclodipeptides.

Materials and MethodsYvmC-Blic Cloning, Expression, and Purification. The B. licheniformis geneyvmCwas cloned in the pET22b vector and was used to transform Escherichiacoli BL21 (DE3) Rosetta 2 cells. The protein expression was induced for 16 h at20 °C, with 1 mM IPTG in LB medium. The cells were harvested by centrifuga-tion and resuspended in buffer A (100 mM Tris-HCl, pH 8.8, 300 mM NaCl,10 mM β-ME, 10% glycerol, 20 mM imidazole) supplemented with 1 mMPMSF. After sonication, the lysate was centrifuged at 40;000 × g for 20 minat 4 °C. The supernatant was mixed with 1 ml of NiNTA superflow resin(Qiagen) per liter of culture, incubated for 30 min at room temperature,and then loaded on an Econo column (Bio-Rad). After washing the columnwith 20 column volumes (c.v.) of buffer A, the bound recombinant proteinwas eluted with 5 c.v. of buffer A, supplemented with 500 mM imidazole.The fractions containing the recombinant protein were detected by a Brad-ford assay, pooled and supplemented with 0.5 volumes of buffer B (50 mMTris-HCl, pH 8.8, 10 mM MgCl2, 5% glycerol, 10 mM β-ME), thus reducingthe NaCl concentration to 200 mM. The diluted sample was loaded onto a10 ml Hitrap Heparin HP column (GE Healthcare), equilibrated in buffer Bsupplemented with 200 mM NaCl. The protein was eluted with a lineargradient of 0.2 to 1 M NaCl. The fractions containing the protein were thenpooled and fractionated on a HiLoad 16∕60 Superdex 75 prep grade column(GE Healthcare), equilibrated in buffer C (15 mM Tris-HCl, pH 8.8, 150 mMNaCl, 10mMMgCl2, 2 mMDTT, 5% glycerol). The recombinant protein elutedwith an apparent Mr of 31 kDa, consistent with the monomeric form insolution. The protein was finally concentrated with an Amicon Ultra-410,000 MWCO filter (Millipore) to a 12 mg∕ml final concentration. Sampleswere either directly used for crystallization or flash-frozen as 25 μl aliquotsin liquid nitrogen and stored at −80 °C.

Mass Spectrometric Analysis of Cyclodipeptide Formation. The cyclodipeptideformation was performed basically according to the literature (7), with slightmodifications. The reaction mixture, consisting of 100 mM Tris-HCl, pH 8.8,150 mM NaCl, 10 mM MgCl2, 2.5 μM ([13C6,

15N]-labeled or nonlabeled)Leu-tRNALeu, 25 μM Leu, and 5 nM YvmC-Blic, was incubated at 30 °C for30 min. Formic acid was added to a final concentration of 2% (v∕v) tostop the reaction. The product was desalted by passage through a NuTipC-18 column (Glygen Corp.), dried and dissolved in 0.1% TFA. To analyzethe cyclodipeptide, we employed a capillary LC nano ESI/mass spectrometrysystem, consisting of a quadrupole time-of-flight (QqTOF) mass spectrometer(QSTAR®XL, Applied Biosystems) and a splitless nanoflow HPLC system (DiNa,KYA Technologies). The product was loaded onto a trap column packedwith C18 reverse-phase resin, for desalting with 0.1% TFA, and then wasdirectly injected into a C18 capillary column (HiQ Sil, 5 μm C18, 100 Å poresize; Φ0.1 × 100 mm, KYA Technologies) and fractionated at a flow rate of300 nl∕min, with a linear gradient of 10–98% Solvent B for 30 min. Thesolvent system consisted of 0.1% formic acid and 2% acetonitrile (Solvent A)and 0.1% formic acid and 80% acetonitrile (Solvent B). The applied voltageon the sprayer was set at þ2.2 kV, to generate a nanoelectrospray. TOFMS and product ion spectra were acquired in the positive mode, using theinformation-dependent data acquisition (IDA) program of the Analyst QSsoftware (Applied Biosystems). The mass ranges for the full scan and thecollision-induced dissociation (CID) were set to m∕z 200–500 and m∕z50–500, respectively. The collision energy for CID was set at þ20–25 V, toobtain the product ions of the cyclodipeptide. The CID spectra were assignedaccording to the literature (7).

YvmC-Blic Crystallization. Screening of crystallization conditions was per-formed at 20 °C and 4 °C using the sitting drop vapor diffusion method.The drops were prepared by mixing equal amounts (100 nlþ 100 nl) of theprotein and crystallization solutions using a Mosquito robot (TTP Labtech).Initial crystals were obtained from drops with the crystallization solutioncontaining 100 mM CHES, pH 9.5, and 20% PEG 8,000. Large crystals wereobtained using the hanging drop vapor diffusion method with 2 μl drops(1 μlþ 1 μl). The crystals grew to maximum dimensions of 0.5 × 0.2×0.2 mm3 within 4 days. Cocrystallization and soaking were performed withthe small ligands L-leucine, L-phenylalanine, puromycin, L-leucyl-L-leucine,L-leucyl-N-2′-adenylate, L-leucyl-L-leucyl-N-2′-adenylate, and cyclo-L-leucyl-L-leucine. For soaking, the CHES buffer was replaced by either CAPSO orBis-Tris, pH 9.5.

Structure Determination and Refinement. A SeMet-substituted crystal ofYvmC-Blic was soaked in the crystallization buffer, cryoprotected with20% glycerol and flash-cooled in a nitrogen stream at 100 K. A SeMet deri-vative dataset was collected at 0.97904 Å wavelength at the SPring-8 BL41XUbeamline, using an ADSC Q315 CCD detector and was processed at 2.8-Å

Fig. 5. Proposed catalytic mechanism for cyclodileucine formation byYvmC-Blic. The first and second incorporated leucyl moieties of Leu-tRNALeu

are highlighted in red and blue, respectively.

3916 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1019480108 Bonnefond et al.

Dow

nloa

ded

by g

uest

on

July

9, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF8http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=SF8

resolution with HKL2000 (21). Among the 18 potential selenium sites, 10were found with SHELXC/D (22) and were used for phase calculations at4.0-Å resolution with the SHARP program (23). The initial phases wereimproved by density modification with RESOLVE (24). The initial model, con-taining 75% of all residues, was automatically built with RESOLVE. The modelwas further built manually using Coot (25) and improved by iterative cyclesof refinement with Phenix (26). The structures of the native, CHES- andCAPSO-complexed CDPSs were solved by molecular replacement at 1.9-,1.7-, and 2.4-Å resolutions, respectively, using the model determined bysingle-wavelength anomalous diffraction phasing as a search model. Modelbuilding and structure refinement were performed with Coot and Phenix,respectively. The quality of the final models was assessed with MolProbity(27). A summary of the structure refinement statistics is provided in Table S1.

Blic-YvmC mutants, Bsub-YvmC, B. licheniformis LeuRS and aminoacyl-tRNA preparations, gel-shift assays, gel filtration assays, thin layer chromato-graphy, SDS/PAGE analysis of reaction intermediates, and phylogenetic treeconstruction are described in SI Text.

ACKNOWLEDGMENTS. We thank the staff at the BL41XU beamline at SPring-8(Harima, Japan) for assistance during data collection, and H. Nishimasu fordiscussions and critical comments on the manuscript. This work was sup-ported by a grant for the National Project on Protein Structural andFunctional Analyses from the Ministry of Education, Culture, Sports, Scienceand Technology (MEXT) (to O.N.), by grants from MEXT (to R.I. and O.N.),and by grants from The Uehara Memorial Foundation (to O.N.). L.B. wassupported by a postdoctoral fellowship for foreign researchers from theJapan Society for the Promotion of Science.

1. Martins MB, Carvalho I (2007) Diketopiperazines: Biological activity and synthesis.Tetrahedron 63:9923–9932.

2. Bolognesi ML, et al. (2010) Discovery of a class of diketopiperazines as antiprioncompounds. ChemMedChem 5:1324–1334.

3. Sattely ES, Fischbach MA, Walsh CT (2008) Total biosynthesis: In vitro reconstitutionof polyketide and nonribosomal peptide pathways. Nat Prod Rep 25:757–793.

4. Strieker M, Tanovic A, Marahiel MA (2010) Nonribosomal peptide synthetases:Structures and dynamics. Curr Opin Struc Biol 20:234–240.

5. Kopp F, Marahiel MA (2007) Macrocyclization strategies in polyketide and nonriboso-mal peptide biosynthesis. Nat Prod Rep 24:735–749.

6. Lautru S, Gondry M, Genet R, Pernodet JL (2002) The albonoursin gene cluster ofS. noursei: Biosynthesis of diketopiperazine metabolites independent of nonriboso-mal peptide synthetases. Chem Biol 9:1355–1364.

7. Gondry M, et al. (2009) Cyclodipeptide synthases are a family of tRNA-dependentpeptide bond-forming enzymes. Nat Chem Biol 5:414–420.

8. Fonvielle M, et al. (2009) Aminoacyl-tRNA recognition by the FemXWv transferasefor bacterial cell wall synthesis. Nucleic Acids Res 37:1589–1601.

9. Watanabe K, et al. (2007) Protein-based peptide-bond formation by aminoacyl-tRNAprotein transferase. Nature 449:867–871.

10. Vetting MW, Hegde SS, Blanchard JS (2010) The structure and mechanism of theMycobacterium tuberculosis cyclodityrosine synthetase. Nat Chem Biol 6:797–799.

11. Uffen RL, Canale-Parola E (1972) Synthesis of pulcherriminic acid by Bacillus subtilis.J Bacteriol 111:86–93.

12. Cryle MJ, Bell SG, Schlichting I (2010) Structural and biochemical characterization ofthe cytochrome P450 CypX (CYP134A1) from Bacillus subtilis: A cyclo-L-leucyl-L-leucyldipeptide oxidase. Biochemistry 49:7282–7296.

13. Eriani G, Delarue M, Poch O, Gangloff J, Moras D (1990) Partition of tRNA synthetasesinto two classes based on mutually exclusive sets of sequence motifs. Nature347:203–206.

14. Moras D (1992) Structural and functional relationships between aminoacyl-tRNAsynthetases. Trends Biochem Sci 17:159–164.

15. Tsunoda M, et al. (2007) Structural basis for recognition of cognate tRNA by tyrosyl-tRNA synthetase from three kingdoms. Nucleic Acids Res 35:4289–4300.

16. Brenner S (1988) The molecular evolution of genes and proteins: A tale of two serines.Nature 334:528–530.

17. Li J, Szittner R, Derewenda ZS, Meighen EA (1996) Conversion of serine-114 tocysteine-114 and the role of the active site nucleophile in acyl transfer by myris-toyl-ACP thioesterase from Vibrio harveyi. Biochemistry 35:9967–9973.

18. Pazirandeh M, Chirala SS, Wakil SJ (1991) Site-directed mutagenesis studies onthe recombinant thioesterase domain of chicken fatty acid synthase expressed inEscherichia coli. J Biol Chem 266:20946–20952.

19. Bruner SD, et al. (2002) Structural basis for the cyclization of the lipopeptide antibioticsurfactin by the thioesterase domain SrfTE. Structure 10:301–310.

20. Schimmel P, Ribas de Pouplana L (2000) Footprints of aminoacyl-tRNA synthetasesare everywhere. Trends Biochem Sci 25:207–209.

21. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in oscilla-tion mode. Methods Enzymol 276:307–326.

22. Sheldrick GM (2008) A short history of SHELX. Acta Crystallogr A 64:112–122.23. de La Fortelle E, Bricogne G (1997) Maximum-likelihood heavy-atom parameter

refinement for multiple isomorphous replacement and multiwavelength anomalousdiffraction methods. Methods Enzymol 276:472–494.

24. Terwilliger TC (2003) SOLVE and RESOLVE: Automated structure solution, densitymodification and model building. J Synchrotron Radiat 11:49–52.

25. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.Acta Crystallogr D 66:486–501.

26. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-molecular structure solution. Acta Crystallogr D 66:213–221.

27. Chen VB, et al. (2010) MolProbity: All-atom structure validation for macromolecularcrystallography. Acta Crystallogr D 66:12–21.

Bonnefond et al. PNAS ∣ March 8, 2011 ∣ vol. 108 ∣ no. 10 ∣ 3917

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

July

9, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=ST1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019480108/-/DCSupplemental/pnas.1019480108_SI.pdf?targetid=STXT