Crystal structures of the PLP- and PMP-bound forms of BtrR, a dual functional aminotransferase involved in butirosin biosynthesis

Crystal Structures of the PLP- and PMP-Bound Forms ofBtrR, A Dual Functional Aminotransferase Involved inButirosin BiosynthesisBojana Popovic, Xiao Tang, Dimitri Y. Chirgadze, Fanglu Huang, Tom L. Blundell, and Jonathan B. Spencer*1Department of Biochemistry, University of Cambridge, Cambridge, United Kingdom2Department of Chemistry, University of Cambridge, Cambridge, United Kingdom

ABSTRACT The aminotransferase (BtrR),which is involved in the biosynthesis of butirosin, a2-deoxystreptamine (2-DOS)-containing aminoglyco-side antibiotic produced by Bacillus circulans, ca-talyses the pyridoxal phosphate (PLP)-dependenttransamination reaction both of 2-deoxy-scyllo-inosose to 2-deoxy-scyllo-inosamine and of amino-dideoxy-scyllo-inosose to 2-DOS. The high-resolu-tion crystal structures of the PLP- and PMP-boundforms of BtrR aminotransferase from B. circulanswere solved at resolutions of 2.1 Å and 1.7 Å withRfactor/Rfree values of 17.4/20.6 and 19.9/21.9, respec-tively. BtrR has a fold characteristic of the aspar-tate aminotransferase family, and sequence andstructure analysis categorises it as a member ofSMAT (secondary metabolite aminotransferases)subfamily. It exists as a homodimer with two activesites per dimer. The active site of the BtrR protomeris located in a cleft between an � helical N-terminus,a central �� sandwich domain and an �� C-terminal domain. The structures of the PLP- andPMP-bound enzymes are very similar; however BtrR-PMP lacks the covalent bond to Lys192. Further-more, the two forms differ in the side-chain confor-mations of Trp92, Asp163, and Tyr342 that are likelyto be important in substrate selectivity and sub-strate binding. This is the first three-dimensionalstructure of an enzyme from the butirosin biosynthe-sis gene cluster. Proteins 2006;65:220–230.© 2006 Wiley-Liss, Inc.

Key words: aminotransferase; secondary metabo-lite aminotransferases (SMAT); buti-rosin; aminoglycoside antibiotics; X-raycrystallography

INTRODUCTION

The aminoglycoside antibiotics represent an importantclass of antibiotics used in the treatment of bacterial andsome protozoal infections. These molecules comprise asignificant component of our antimicrobial arsenal and,like all antibiotics, are under pressure from the spread ofresistance. They are classified into two groups dependingon the structure of the aglycone. One class, which includesstreptomycin and fortimicin, is characterized by a fullysubstituted aminocyclitol streptamine. The other has 2-de-oxystreptamine (2-DOS) as the common aglycone and

includes many of the clinically important aminoglycosides,for example, kanamycin, neomycin, gentamicin, and buti-rosin (Fig. 1).1 The butirosin gene cluster has recentlybeen identified and sequenced,2 providing the foundationfor a detailed investigation of the biosynthesis of this classof aminoglycosides.

In order to study the biosynthesis of the aglycone,2-DOS, we synthesised two of the proposed intermediates2-deoxy-scyllo-inosose and 2-deoxy-scyllo-inosamine by aconvenient route from myo-inositol.3 Using these com-pounds as standards we identified BtrR (also designatedBtrS by Tamegai et al.4) as the aminotransferase thatconverts 2-deoxy-scyllo-inosose to 2-deoxy-scyllo-inosamine,the second step of 2-DOS biosynthesis [Fig. 1(b)]. Evidencewas also obtained that BtrR could catalyse the last step inthe 2-DOS biosynthetic pathway, suggesting it was adoubly functional aminotransferase.5 Subsequent experi-ments using gene disruption studies have provided furtherevidence that BtrR is involved in the second transamina-tion of 2-DOS biosynthesis.6

In this article, we describe the structures of the PLP-and PMP-bound forms of BtrR solved to resolutions of 2.1Å and 1.7 Å respectively, and we analyze possible sub-strate and inhibitor binding. The structures reveal thatBtrR has a fold characteristic of the aspartate aminotrans-ferase family and is a member of the new class of amino-transferases that function primarily in the biosynthesis ofsecondary metabolites (SMAT).

Abbreviations: 2DOI, DOI, 2-deoxy-scyllo-inosose; AHBA, 3-amino-5-hydroxybenzoic acid synthase; DOS, 2-deoxystreptamine; PID, per-centage identity; PLP, pyridoxal-5�-phosphate; PMP, pyridoxamine-5�-phosphate; SMAT, secondary metabolite aminotransferases.

Grant sponsor: Darwin Trust; Grant sponsor: BBSRC; Grant spon-sor: Wellcome Trust.

Bojana Popovic, Xiao Tang, and Dimitri Y. Chirgadze contributedequally to this work.

*Correspondence to: Jonathan B. Spencer, University of Cambridge,Lensfield Road, Cambridge, CB2 1EW, UK. E-mail: [email protected]

Received 6 December 2005; Revised 14 March 2006; Accepted 30March 2006

Published online 7 August 2006 in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/prot.21076

PROTEINS: Structure, Function, and Bioinformatics 65:220–230 (2006)

© 2006 WILEY-LISS, INC.

MATERIALS AND METHODSSequence Analysis

The sequence of BtrR was compared with protein se-quences deposited in SWISS-PROT by using the Blast v2.2program.7 Homologous proteins with known structureswere identified using the FUGUE8 homology recognitionserver.

Protein Expression and PurificationPLP-bound and PMP-bound forms

The btrR gene was amplified by PCR from Bacilluscirculans genomic DNA and cloned into a pET28 expres-sion vector (Novagen) with six N-terminal His tags andtransformed into E. coli BL21 (DE3) cells.9 The cellsharboring the btrR/pET28 plasmid were grown overnightin 10 mL LB broth in the presence of kanamycin (50 �g/L)at 37°C. The overnight culture was inoculated into 1 Lfresh LB with kanamycin and grown at 37°C until OD600nm

of approximately 0.6 to 0.8 was reached. Expression of therecombinant protein was induced with 0.1 mM isopropyl-�-D-thiogalactopyranoside (IPTG) and culture growth contin-ued at 28°C for approximately 4 h. Cells were harvested bycentrifugation at 7500 rpm for 10 min at 4°C, resuspendedin 10 mL of binding buffer (5 mM imidazole, 0.5M NaCl, 20mM Tris-HCl, 10% glycerol, pH 7.9), with 0.1 mM PLP forthe PLP-bound form and without PLP for PMP-boundform. The cells were disrupted by sonication (4 min with2 s bursts and 9.9 s pauses). All steps were carried out onice. After sonication the solution was centrifuged at 18000rpm for 30 min to remove cell debris. The cell-free extractwas applied to a Ni-NTA resin column that was chargedwith charging buffer, prewashed with water and binding

buffer. The column with bound proteins was washed withbinding buffer and wash buffer (60 mM imidazole, 0.5MNaCl, 20 mM Tris-HCl, 10%glycerol, pH 7.9) followed byelution buffer (250 mM imidazole, 0.5M NaCl, 20 mMTris-HCl, 10% glycerol, pH 7.9) to elute the His-taggedprotein. The eluted protein was concentrated to 2 mLusing a 6 mL concentration cell (Vivaspin6, Sigma-Aldrich) and subjected to FPLC purification on a SuperdexS-200 HiLoad 16/60-gel filtration column at 1 mL/min(Amersham Pharmacia Biotech). The purified protein wasconcentrated to 100 �L using a 6 mL concentration cell(Vivapsin6). The PLP-bound protein was concentrated to12 mg/mL and buffer exchanged to 250 mM NaCl, 20mMTri-HCl, pH 8.0. The PMP-bound protein in 20 mMTris-HCl and 100 mM KCl at pH7.5 was concentrated to 10mg/mL. Both samples were used for crystallization trials.

Crystallization and X-Ray Data Collection

The crystallization drops normally contained 1 �L ofprotein solution mixed with 1 �L of crystallisation solutionand were equilibrated against 1 mL of the latter in 24-wellplates (Molecular Dimensions Ltd.) at room temperatureof 19°C. In both cases, the initial crystallisation trials wereperformed with sparse crystallization matrices. The PLP-bound form produced microcrystals in condition #26 ofHampton Research Corp., Crystal Screen II™. The crystal-lization condition was slightly modified by varying theconcentrations of crystallizing agents, and this yieldedcrystals suitable for X-ray diffraction analysis. The bestcrystals were bi-pyramidal in shape and grew in 0.1Mammonium sulphate, 15% PEG5000, and 50 mM MES atpH 6.5. In contrast, the PMP-bound form generated just a

Fig. 1. Butirosin and 2-DOS biosynthesis. Structures of (a) butirosin A and (b) butirosin B. 2-DOSbiosynthetic pathway.

PROTEINS: Structure, Function, and Bioinformatics DOI 10.1002/prot

CRYSTAL STRUCTURES OF BTRR 221

few crystals in condition #26 of Molecular DimensionsLtd., Structure Screen I that were already suitable for theX-ray diffraction analysis without any need of furtheroptimization of crystallization conditions. The correspond-ing crystallization condition formulation is 0.8M K, Natartrate tetrahydrate, and 0.1M Na HEPES at pH 7.5.

The fully grown crystals of both forms of BtrR weresoaked for 5 to 10 s in a cryoprotectant solution consistingof 26% to 28% of ethylene glycol together with the motherliquor as defined by the crystallization conditions, mountedinto the cryo-loops (Hampton Research, Ltd) and theneither placed into the liquid nitrogen or into the stream ofnitrogen gas at 100 K provided by the CryoStream device(Oxford Instruments, Ltd.). X-ray diffraction data collec-tion experiments were performed at cryogenic tempera-ture using either an in-house copper-rotating anode (gen-erator RU-H3R, Rigaku-MSC Ltd. equipped with Max-Flux confocal multilayer optics, Osmic Inc.) or synchrotron(beam station PX 9.6 of Daresbury SRS, Warrington, UK)radiation sources. The diffraction data were recordedusing either Raxis IV�� image plate (Rigaku-MCS Ltd.)or Quantum4 CCD (Area Detector Systems Corp.) detec-tors. In all cases, the raw diffraction data were collectedusing a single crystal at 1° oscillation steps and wereindexed, integrated, scaled, and reduced using HKL diffrac-tion data processing suite10 with the use of various crystal-lographic programs from the CCP4 suite11 for all subse-quent calculations.

Structure Solution and Refinement

Calculation of Matthews’ coefficient11 suggested thepresence of one molecule of BtrR in the asymmetric unit,resulting in a solvent content of approximately 50%.Initially, the structure of the PLP-bound form of BtrR wasdetermined by molecular replacement using the crystalstructure of AHBA synthase (PDB accession code: 1b9h) asthe search probe. AHBA shares 29% sequence identitywith BtrR over a stretch of 416 of residues used in themolecular replacement study and is the closest homologueof known structure. Molecular replacement calculationswere performed using AMoRe program.12 The rotationfunction produced a peak with a signal-to-noise ratio of0.7� (resolution range 8–3 Å was used for all calculations).This solution produced a clear peak in the translationfunction with a correlation coefficient and Rcryst betweenobserved and calculated structure factor amplitudes of25.4% and 54.9%, respectively. The rigid body refinementperformed in AMoRe improved both the correlation coeffi-cient and the Rfactor to 37% and 52.7%, respectively, thusconfirming the correctness of the solution. Subsequently,the structure of PMP-bound form was also determined bythe molecular replacement. However, in this case therefined structure of the PLP-bound form of BtrR, withoutthe bound cofactor was used as the search probe.

The initial model thus obtained was then subjected toseveral rounds of crystallographic refinement and manualrebuilding. The atomic coordinates were refined usingCNS.13 Simulated annealing protocols as implemented inCNS were utilized in the first rounds of refinement and

replaced by the Powell minimization protocol in the lastrounds. The temperature factor refinement included therestrained individual Bfactor refinement. Manual rebuild-ing was performed in XtalView suite14 using sigmaAweighted 2Fo-Fc, Fo-Fc and annealed omit maps. An auto-mated model-rebuilding program ARP/WARP15 was alsoemployed in the early stages of refinement to aid themanual rebuilding procedure. Most water molecules werepicked using the XtalView internal subroutine and addi-tional ones were placed manually using the followingcriteria: a peak of at least 2.5� for a Fo-Fc map, a peak of atleast 1� for a 2Fo-Fc map, and reasonable intermolecularinteractions. The PLP or PMP molecules were includedonly in the last stages of refinement.

Small Molecule Docking

A three-dimensional model of the 2DOI (2-deoxy-scyllo-inosose) substrate was built in silico and minimized. Themolecule of 2DOI was docked into the refined structure ofBtrR-PMP and covalently linked to the PMP. The dockingwas conducted using InsightII, a molecular modeling tool(Accelrys).

RESULTS AND DISCUSSIONSequence Analysis of BtrR

PSI-BLAST searches of protein sequences in SWISS-PROT identified BtrR as a member of the aspartateaminotransferase family, a major group of enzymes thatbelong to the superfamily of PLP-dependent enzymes.

The enzymes of this superfamily catalyse a diverserange of chemical reactions including transamination,decarboxylation, and racemization. The closest homo-logues of BtrR are aminotransferases involved in thebiosynthesis of aminoglycoside antibiotics, such as genta-micin, kanamycin, tobramycin, and streptomycin. Theseenzymes are members of the DegT/DnrJ/EryC1/StrS ami-notransferase family (Pfam accession code: PF01041) whosemembers are involved in the biosynthesis of secondarymetabolites and are known as secondary metabolite amin-otransferases (SMATs) subfamily. SMATs were originallycharacterised through studies of stsC, the gene encodingthe L-glutamine:scyllo-inosose aminotransferase fromstreptomycin-producing Streptomycetes.16 Sequence align-ments revealed conserved sequence motifs that distantlyresembled those in the family of aspartate aminotrans-ferases with a conserved lysine residue. With the exceptionof lysine that is strictly conserved in all PLP-dependentenzymes, the motif characteristic for aspartate aminotrans-ferases family (PROSITE PS00105: GS-LIVMFYTAC-GSTA-K-x2-GSALVN-LIVMFA-x-GNAR-x-R-LIVMA-GA)is not conserved in SMATs subfamily. However, analysis ofSMAT enzyme sequences identified motif G-D-E-x77-E-D-x10-G-x3-G-x8-S-x4-K-x5-6-(E,Q)-G-G that is strictly con-served amongst the members of SMATs.16 This motif isconserved in BtrR and it therefore categorizes the enzymeas a member of the SMAT subfamily.

This is confirmed by sequence-structure comparisons.FUGUE identified two high-resolution structures for sub-family members: 3-amino-5-hydroxybenzoic acid (AHBA)


222 B. POPOVIC ET AL.

synthase from Amycolatopsis mediterranei and ArnB ami-notransferase from Salmonella typhimurium. AHBA (PDBcode: 1B9H) is the homologue of known structure withhighest percentage sequence identity (29% PID) to BtrR.The sequences of AHBA synthase and BtrR are not toodissimilar in regions away from the active site. However,the central cofactor-binding regions align best around theconserved lysine that covalently binds the PLP cofactor.

Structure Determination for BtrR

Crystallization of the PLP-bound and PMP-bound formsyielded trigonal bipyramidal crystals suitable for X-raydiffraction analysis. The crystals of PLP bound formusually appeared after 1 to 2 days and reached theirmaximum dimensions of 0.25 � 0.2 � 0.2 mm3 in 2 weeks.In contrast the crystals of PMP bound form grew to 0.3 �

0.3 � 0.2 mm3 in 7 days. The crystals of PLP-boundenzyme diffracted to a resolution of 2.1Å, whereas crystalsof PMP-bound form diffracted to 1.7Å. The crystallo-graphic data collection statistics are given in Table I. Therefined model of the BtrR-PLP yielded a Rfactor of 17.4 andRfree 20.6 for all data to 2.1 Å. In addition, the BtrR withPMP yielded an Rfactor of 19.9 and Rfree 21.9 for all data to1.7Å (Table I). The final models contain 365 and 362 aminoacid residues and 394 and 396 ordered water molecules,respectively, in addition to one cofactor molecule perprotomer. The structures contain a cis-proline at residue12. Analyses by PROCHECK17 showed that all residueshave expected allowed conformations, except F140 anamino acid that is 9Å away from the active site cofactorand it is present in the active site cavity. It could perhapsbe involved in catalysis.

TABLE 1. Summary of Crystallographic Statistics of PLP- and PMP-Bound Forms ofBtrR Aminotransferase Structures

PLP-bound form PMP-bound form

X-ray diffraction dataX-ray source SRS, Daresbury In-houseSpace group P3221 P3221Unit cell (Å) a � b 73.74 73.75

C 162.52 162.36Resolution range (Å) 40.0-2.10 30.0-1.70

highest resolution shell 2.15-2.10 1.73-1.70Rsym

a 7.2 5.1highest resolution shell 35.9 37.3

Completeness (%) 99.2 91.5highest resolution shell 96.0 98.9

No. of unique reflections 30455 52335Multiplicity 6.5 4.0Average intensity, I/(I)� 11.0 19.9% reflections with I/(I)�3 in the highest resolution shell 57.7 46.1Wilson B-factor (Å2) 25.8 20.8RefinementRcryst

b highest resolution shell 17.4 (19.0) 19.9 (25.3)Rfree

c highest resolution shell 20.6 (25.3) 21.9 (27.6)No. of reflections

working 28868 49670test 1536 2637

No. of non-hydrogen atomsprotein 3185 3196water 343 396

Model qualityEstimated coordinate errord (Å) 0.20 0.20R.m.s. deviation bonds (Å) 0.005 0.005R.m.s. deviation angles (°) 1.2 1.2Overall mean B-factor (Å2) 25.2 21.9Ramachandran plot analysise

No. of residues in:most favoured regions 328 327additionally allowed regions 36 34disallowed regions 1 1

aRsym � h�Ih I��/hIh, where Ih is the intensity of reflection h, and I� is the mean intensity of allsymmetry-related reflections.bRcryst � � �Fobs� �Fcalc� �/�Fobs�, Fobs and Fcalc are observed and calculated structure factor amplitudes.cRfree as for Rcryst using a random subset of the data (5%) excluded from the refinement (Brunger 1992).dEstimated coordinate error based on the R-value as calculated by CNS (Brunger 1998).eAs calculated by PROCHECK (Laskowski et al., 1993).



Overview of the BtrR Structure

The BtrR structure has the characteristic aspartateaminotransferase fold consisting of three distinct regions,an N-terminal domain, a central cofactor-binding domainand a C-terminal domain. The cofactor binds at theinterface of the central and C-terminal domains (Fig. 2).The asymmetric unit of BtrR contains one protomer of theenzyme, and a crystallographic two-fold axis generates adimer that has many close contacts between the subunits,especially at the active site (Fig. 3).

The dimer interface consists of 80 residues including 36hydrogen bonds and buries an extensive, predominantlyhydrophobic interface area of � 3000 Å.2 The observationof a dimer in the crystal structure is in agreement with thefinding that BtrR behaves as a functional dimer in solu-tion. Together these observations provide strong evidencethat the enzyme functions as a dimer. The majority of theaspartate aminotransferases exist as similar dimers18–23

and the hydrophobic nature of the interface is conserved,although the individual residues are not.

The C� RMSDs for BtrR when superimposed withAHBA synthase (PDB:1B9H) and ArnB aminotransferase(PDB:1MDO) are 1.08 Å and 2.00 Å respectively, confirm-

ing that AHBA synthase is a closer homologue than ArnBaminotransferase. When the large cofactor-binding do-mains are overlaid it is observed that the orientations ofthe C-terminal domains differ. In addition, the C-terminaldomains differ in their arrangement of �-strands, �-heli-ces, and loops. Together these differences are likely tocontribute to the specificity for the substrate and reactiontype.

Binding and Comparison between PLP- and PMP-Bound Forms

The resting state of BtrR, in the absence of substrate,contains PLP bound through an imine to Lys192. Theconversion of 2-deoxy-scyllo-inosose to 2-deoxy-scyllo-in-osamine and of amino-dideoxy-scyllo-inosose to 2-DOS cata-lyzed by BtrR proceeds by two half-reactions. First, PLPreacts with glutamic acid to form PMP and �-ketoglutarate.Second, PMP reacts with either 2-deoxy-scyllo-inosose oramino-dideoxy-scyllo-inosose to yield the respective ami-nated product and regenerate PLP bound to BtrR. Compari-son of the two three-dimensional structures reported hereshows that the binding interactions of PLP and PMP to BtrRaminotransferase are very similar, but PMP, unlike PLP, isnot covalently bound to Lys192 (Fig. 4). Other than this, all ofthe interactions of the cofactor with the enzyme are virtuallythe same with both hydrogen bonds and hydrophobic interac-tions being important. The cofactor is situated in the centralgroove of each protomer (Fig. 2) as in homologous aspartateaminotransferases.18 Figures 3 and 5 show that the Ntermini of �3 and �4 as well as �8 (from protomer B) and �10provide the scaffold for the cofactor binding while the connect-ing loops 187 to 200 and 166 to 171 mediate the binding ofPLP and PMP.

In helix �3, Gly66 NH and Ser67 NH and �OH coordi-nate the phosphate moiety of PMP by binding O2P andO1P, respectively. The phosphate oxygen O2P is alsohydrogen bonded to a water molecule and the last residueof �8, Ser187 �OH. The phosphate O1P is also hydrogenbonded to a water molecule. Asp163 �OH of �10 interactswith the N1 of the PMP pyrimidine ring. In addition,Gln166 �NH2 is interacting with the O3 from PMP and two

Fig. 2. The overall fold of BtrR protomer. Ribbon diagram of BtrR protomer with PMP, showing theN-terminal helix (red), central cofactor binding domain (blue), and C-terminal (brown) domain.

Fig. 3. Ribbon representation of the BtrR dimer. The two protomersare colored green and blue with the two molecules of PMP shown in redand the central Lys192 in orange. The two active sites of the dimer aresituated on the two faces hereby accommodating the cofactor PMP.



water molecules. In summary, residues Gly66, Ser67,Asp163, Gln166, and Ser187 and a network of watermolecules mediate hydrogen-bonding interactions withthe bound PMP. Additionally, mainly hydrophobic interac-tions are provided by Trp92, Val137, Ala94, Leu70, andAla165 [Fig. 4(b)].

PLP, like PMP, is bound by hydrogen bonds that aremediated by residues Gly66, Ser67, Asp163, Gln166,Ser187, but also has a covalent bond to Lys192. As in the

PMP structure, Trp92, Val137 and Ala165 provide mainlyhydrophobic interactions [Fig. 4(a)].

Although PLP- and PMP-BtrR structures are very similar,there are subtle differences between the two, particularly inthe positions and orientations of the side chains of Trp92,Tyr342, and Asp163. Upon PLP to PMP conversion theTrp92 ring undergoes a 45° flip around C�-C�, the cofactortilts by 30° towards Trp92 and Asp163 flips its side-chainaround �1 C�-C� by 15° away from the cofactor (Fig. 4). In

Fig. 4. Diagram of PLP and PMP bound to BtrR showing hydrogen bonding and the differences betweenthe two forms. The figures illustrate the residues that form charged hydrogen bonds (red) or other interactions(colored by atom). (a) Stereo image of PLP bound BtrR showing no electron density for the side-chain ofTyr342 and (b) stereo image of PMP bound BtrR showing Tyr342, flip of Trp92, and rotation of Asp163side-chain.

Fig. 5. Possible substrate binding site. (a) A model of putative BtrR-PMP-gabaculine complex from superposi-tion with AHBA synthase. (b) Sulphate binding in BtrR-PLP indicating the residues that are close to its binding site.



addition, the PLP bound form shows no electron density forthe side-chain Tyr342 [Fig. 4(b)]. The conformational changesof Trp92, Asp163, and Tyr342 could be concerted with thetilting of the cofactor and possibly with accommodation of thesubstrate.

The residues that are conserved between aspartate amin-otransferases are in the cofactor-binding region and includethe catalytic Lys192, Asp163, and Ala165 (Fig. 7). Ala165together with Trp92 provide stacking interactions on eitherface of the pyridoxal ring. The residues that are conserved inthe SMAT subfamily are Val137 and Ser187, which providethe cofactor with hydrophobic and hydrogen-bonding interac-tions (Fig. 4). Gln166 is not conserved but has previouslybeen identified as important in determining the reactionpathway of the SMATs.24 In BtrR it plays a role in cofactorbinding, but its role in catalysis still needs to be confirmed.

Soaking and Inhibitor Studies

Soaking of the PLP form of BtrR crystals with eitherthe substrate 2-deoxy-scyllo-inosamine or inhibitors ofPLP-dependent aminotransferases, such as gabaculin(3-amino-2,3-dihydrobenzoic acid) and (amino-oxy) ace-tic acid, resulted in the enzyme turning over rather thanforming an enzyme substrate/inhibitor complex. Whileneither substrate nor inhibitors can be seen in theelectron density, the structures show that BtrR is in thePMP-bound form after soaking with these three com-pounds. Although the enzyme is an aminotransferase, itis an atypical aminotransferase because it is not inhib-ited with classical inhibitors such as gabaculine.

In order to investigate possible substrate binding, struc-tures of the BtrR homologue — AHBA synthase with gabacu-line — and BtrR were superimposed using molecular super-imposition software, COMPARER25 (Fig. 5). Superposition ofthe structures demonstrates that the pocket occupied by asulphate ion in BtrR-PLP is equivalent to the pocket occu-pied by gabaculine in AHBA synthase, suggesting that this isthe putative substrate-binding site. Furthermore, the super-position suggests that Gln189 and Trp92 in BtrR play a rolein substrate binding, and possibly in substrate specificitybecause they are not conserved in the subfamily (Fig. 5).

The sulphate ion is hydrogen bonded in the BtrR-PLPstructure via residues Trp92, Tyr304, and a water mole-cule [Fig. 5(b)]. Phe140 and His339 are also within 4 Å ofthe sulphate ion, lining the groove within which it lies. TheTyr342 that is not visible in BtrR-PLP but present inBtrR-PMP structure is also within the hydrogen bondingdistance of the sulphate ion, suggesting its involvement insubstrate binding. It is possible that these residues areinvolved in coordinating the substrate through hydrogenbonding interactions.

In Silico Substrate Docking

To analyze possible substrate binding, the 2DOI moleculewas manually docked into the pocket where the sulphate ion

Fig. 6. Possible substrate binding site in BtrR. (a) Electrostatic surface of BtrR active site cavity within 7 Å ofthe cofactor (created with CCP4mg).27,28 (b) Binding of 2DOI showing all the residues within 3 Å distanceinvolved in possible hydrogen bonding.

Fig. 7. Putative active site of BtrR aminotransferase. The chains Aand B are colored purple and yellow. The residues shown are possiblesubstrate binding sites for 2DOI (Tyr342A, Tyr304A, and Gln166A) andfor L-Glu (Met235B and Asn247B).



is found in the crystal structure of BtrR and a covalent bondwas created between PMP and 2DOI, simulating the interme-diate in the catalytic mechanism. To achieve binding of theintermediate, the side-chain of Trp92 would have to flipcreating access to the cofactor so that the substrate couldbind, as observed in BtrR-PMP structure.

The volume of the putative BtrR active site is 690 Å (ascalculated by castP26). On docking it is clear that theproposed binding site can accommodate the substrate 2DOIand the binding site is lined with residues that could interactwith the substrate (Fig. 6). The residues that are within 3 Åof 2DOI are Trp92, Tyr304, Tyr342, and Gln166 [Fig. 6(b)].All the residues except Gln166 were previously identified onsuperimposition with gabaculine-AHBA synthase.

BtrR accepts four different carbocyclic substrates, 2-deoxy-scyllo-inosose, 2-deoxy-scyllo-inosamine, amino-dideoxy-scyllo-inosose, and 2-deoxysteptamine, requiring the bindingsite in the enzyme to have some flexibility. It is interestingthat BtrR combines this flexibility with high specificitybecause the enzyme has been shown to distinguish betweenthe enantiotopic amino groups of DOS.6 Because a triol is aconserved feature in all these compounds it seems possiblethat selective binding is affected primarily by hydrogen-bonding to these hydroxy groups. In particular, these threehydroxyl groups of the substrates could be bound by Tyr304OH, Tyr342 OH, and Gln166 εNH2 of BtrR [Fig. 6(b)]. Thiscould explain how BtrR can bind different substrates butretain high specificity. Furthermore, binding of glutamic acidas the second substrate in catalysis of BtrR is not incompat-ible with this model.

Putative Active Site

The electron sink nature of the cofactor is enhanced by aclose interaction of the pyridinium nitrogen with an aspar-tate residue (Asp163), which acts to maintain the cofactor inthe protonated form. The displaced Schiff base lysine residue(Lys192) is positioned to transfer proton to and from C� andC4�. Because the substrate of BtrR — L-glutamate — cannotundergo �- or �- elimination, the primary side reaction thatmust be minimized is racemization. This could be achievedby closure of the enzyme active site around the substrate, sothat the solvent water molecules do not access the quinonoidintermediate.29 However, in the structures of BtrR-PMP andBtrR-PLP no large-scale rearrangements between the struc-tures were observed. In a similar way, no large-scale rear-rangements between PMP and PLP structures were ob-served either in ArnB aminotransferase, which wascocrystallized with �-ketoglutarate and soaked with L-cycloserine,30 or in AHBA synthase, which was cocrystallizedwith gabaculine.18 Additionally, the binding of inhibitorscaused only small changes at the active site and dimerinterface with the overall tertiary structure remaining essen-tially unchanged. It is possible that BtrR aminotransferasebehaves in a similar way.

The four types of PLP-dependent enzymes do not have acommon fold but there is a limited conservation of the PLPbinding residues. The type I fold (aspartate aminotransfer-ase family) consists of large �� sandwich and small ��domains. Type I enzymes have been divided into sub-

groups on the basis of conformational changes upon thebinding of substrate or inhibitors: subgroup Ia, largeconformational changes; subgroup Ib, absence of conforma-tional changes. Although it is not clear if the possibleconformational changes contribute to substrate specificity,the existence of the closed conformation certainly contrib-utes to reaction type specificity within the aspartateaminotransferase family.32 Even though the BtrR-sub-strate complex structure is unavailable all the structuresproduced by turning over of substrate and inhibitors thatwere solved showed no conformational change. The crystalstructures of BtrR and its comparison to SMAT subfamilyperhaps suggest that this enzyme belongs to the subgroupIb rather than to Ia.31 Although it is possible that theenzyme belongs to the type Ib subgroup, the evidence formthe structures is insufficient to draw this conclusion.

The basis for the dual specificity of aminotransferaseshas been elucidated by recent structural studies. Becausethe aminotransferase reaction requires two different sub-strates to bind in succession to the same cofactor activesite, these enzymes must be able to accommodate bothstructures and discriminate against all others. One pos-sible way of achieving this is by cofactor movement be-tween the two different binding sites but such movementhas never been observed. An alternative is to position thefunctional groups of the substrates into exclusive bindingsites.32 Most known aminotransferases adopt strategiesfor binding both substrates in the same site and this is alsolikely for BtrR not least because the volume of active sitecavity is considerable. Because BtrR, like many otheraminotransferases, uses the common substrate glutamate,the problem of dual specificity is generally that of accommo-dating the negatively charged �-carboxylate of glutamatein the same site that must also accept a neutral orpositively charged side-chain.

Comparison of BtrR with ArnB aminotransferase identi-fied residues that could be determinants of L-glutamineselectivity in BtrR. The presence of �-ketoglutarate prod-uct in the ArnB structure allowed the identification ofresidues involved in the binding. The side-chain of Lys241from the opposite protomer was found hydrogen bonded tothe terminal carboxyl group of �-ketoglutarate whilst theguanido group of Arg229 also from the opposite protomerwas found hydrogen bonded to the �-carboxylate of �-keto-glutarate.30 In the structure of BtrR the equivalent resi-dues to Lys241B and Arg229B of ArnB are Met235B andAsn247B. Together, Met235B and Asn247B could consti-tute the selectivity filter for the amino donor — L-glutamate — of BtrR aminotransferase (Fig. 7).

The proposed absence of large conformational changebetween active and inactive BtrR forms and the absence ofarginine residues in the active site cavity — preventing anarginine switch mechanism32 — suggest that the mecha-nism of this enzyme is mediated by subtle rearrangementsof hydrogen bonding network in the active site.

CONCLUSION

Multiple scaffolds have evolved to bind PLP and to assistin catalysis. In no case does the fold type dictate reaction



type, as each fold type catalyses multiple reaction types.Recently the number of PLP-dependent enzyme structuresdetermined has increased significantly allowing compari-sons and further assessments of how different proteinscaffolds may be used in binding a common cofactor butcatalyzing diverse reactions. BtrR aminotransferase is afunctional dimer with one active site per protomer whereboth protomers are involved in contributing to the activesite. The enzyme shows closest structural similarity to theaspartate aminotransferase family within the superfamilyof PLP-dependent enzymes. Here we show that the BtrRenzyme should be assigned as a member of the secondarymetabolite aminotransferase subfamily.

On analysis of cofactor and substrate binding we haveidentified residues that might be important in substratebinding and catalysis, such as Trp92, Tyr304, Tyr342 andthe catalytic loops containing Ser187, Gln189, Q166 andLys192. From subfamily analysis and small moleculedocking of 2DOI some of the observations above wereverified and additional residues identified that could playa part in L-glutamine amino donor substrate selectivity,such as residues Met235B and Asn247B.

The structure reported here, for the dual functional amino-transferase BtrR, is the first for an enzyme from the 2-DOSbiosynthetic pathway. This provides an initial step towardsthe detailed structural investigation of the enzymes respon-sible for biosynthesis of this class of aminoglycosides.

ACKNOWLEDGMENTS

All the images were created using PyMOL molecularvisualization software.33 We thank Dr. Ricardo Nunez forhelp with bioinformatic analyses and the beamline scientistsat PX 9.6 of Daresbury SRS, Warrington, UK. Atomiccoordinates have been deposited in the Protein Data Bankunder accession codes 2c7t and 2c81.

REFERENCES

1. Vicens Q, Westhof E. Molecular recognition of aminoglycosideantibiotics by ribosomal RNA and resistance enzymes: an analysisof x-ray crystal structures. Biopolymers 2003;70:42–57.

2. Ota Y, Tamegai H, Kudo F, Kuriki H, Koike-Takeshita A, EguchiT, Kakinuma K. Butirosin-biosynthetic gene cluster from Bacilluscirculans. J Antibiot (Tokyo) 2000;53:1158–1167.

3. Yu J. Convenient synthesis of 2-deoxy-scyllo-inosose and 2-deoxy-scyllo-inosamine: two key intermediates on the biosynthetic pathwayto aminoglycoside antibiotics. Tetrahedron Lett 2001;42:4219–4221.

4. Tamegai H, et al. Identification of L-glutamine: 2-deoxy-scyllo-inosose aminotransferase required for the biosynthesis of buti-rosin in Bacillus circulans. J Antibiot (Tokyo) 2002;55:707–714.

5. Huang F, Li Y, Yu J, Spencer JB. Biosynthesis of aminoglycosideantibiotics: cloning, expression and characterisation of an amino-transferase involved in the pathway to 2-deoxystreptamine. ChemCommun (Camb) 2002;2860–2861.

6. Yokoyama K, et al. Stereochemical recognition of doubly func-tional aminotransferase in 2-deoxystreptamine biosynthesis. J AmChem Soc 2005;127: 5869–5874.

7. Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, MillerW, Lipman DJ. Gapped BLAST and PSI-BLAST: a new generationof protein database search programs. Nucleic Acids Res 1997;25:3389–3402.

8. Shi J, Blundell TL, Mizuguchi K. FUGUE: sequence-structure homol-ogy recognition using environment-specific substitution tables andstructure-dependent gap penalties. J Mol Biol 2001;310:243–257.

9. Studier WF, Dunn JJ, Dubendorff JW. Use of T7 RNA polymeraseto direct expression of cloned genes. Methods Enzymol 1990;185:60–89.

10. Otwinowski Z, Minor W. Processing of X-ray diffraction data col-lected in oscillation mode. Methods Enzymol 1997;276:307–326.

11. Project CC. The CCP4 suite: programs for protein crystallography.Acta Crystallogr D Biol Crystallogr 1994;50:760–763.

12. Navaza J. AMoRe: an automated package for molecular replace-ment. Acta Crystallogr D Biol Crystallogr 1994;50:157–163.

13. Brunger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J, Nilges M, Pannu NS, ReadRJ, Rice LM, Simonson T, Warren GL. Crystallography & NMRsystem: a new software suite for macromolecular structure deter-mination. Acta Crystallogr D Biol Crystallogr 1998;54:905–921.

14. McRee DE. XtalView/Xfit — a versatile program for manipulatingatomic coordinates and electron density. J Struct Biol 1999;125:156–165.

15. Morris RJ, Perrakis A, Lamzin VS. ARP/wARP’s model-buildingalgorithms. I. The main chain. Acta Crystallogr D Biol Crystallogr2002;58:968–975.

16. Ahlert J, Distler J, Mansouri K, Piepersberg W. Identification ofstsC, the gene encoding the L-glutamine:scyllo-inosose aminotrans-ferase from streptomycin-producing Streptomycetes. Arch Micro-biol 1997;168:102–113.

17. Laskowski RA, MacArthur MW, Moss DS, Thornton JM. PRO-CHECK: a program to check the stereochemical quality of proteinstructures. J Appl Cryst 1993;26:283–291.

18. Eads JC, Beeby M, Scapin G, Yu TW, Floss HG. Crystal structureof 3-amino-5-hydroxybenzoic acid (AHBA) synthase. Biochemistry1999;38:9840–9849.

19. Jeffery CJ, Barry T, Doonan S, Petsko GA, Ringe D. Crystalstructure of Saccharomyces cerevisiae cytosolic aspartate amino-transferase. Protein Sci 1998;7:1380–1387.

20. Kack H, Sandmark J, Gibson K, Schneider G, Lindqvist Y. Crystalstructure of diaminopelargonic acid synthase: evolutionary rela-tionships between pyridoxal-5�-phosphate-dependent enzymes. JMol Biol 1999;291:857–876.

21. Malashkevich VN, Strokopytov BV, Borisov VV, Dauter Z, WilsonKS, Torchinsky YM. Crystal structure of the closed form ofchicken cytosolic aspartate aminotransferase at 1.9 A resolution. JMol Biol 1995;247:111–124.

22. Nakai T, Okada K, Akutsu S, Miyahara I, Kawaguchi S, Kato R,Kuramitsu S, Hirotsu K. Structure of Thermus thermophilus HB8aspartate aminotransferase and its complex with maleate. Bio-chemistry 1999;38:2413–2424.

23. Rhee S, Silva MM, Hyde CC, Rogers PH, Metzler CM, Metzler DE,Arnone A. Refinement and comparisons of the crystal structures ofpig cytosolic aspartate aminotransferase and its complex with2-methylaspartate. J Biol Chem 1997;272:17293–17302.

24. Yano T, Mizuno T, Kagamiyama H. A hydrogen-bonding networkmodulating enzyme function: asparagine-194 and tyrosine-225 ofEscherichia coli aspartate aminotransferase. Biochemistry 1993;32:1810–1815.

25. Sali A, Blundell TL. Definition of general topological equivalencein protein structures. A procedure involving comparison of proper-ties and relationships through simulated annealing and dynamicprogramming. J Mol Biol 1990;212:403–428.

26. Binkowski TA, Naghibzadeh S, Liang J. CASTp: Computed Atlasof Surface Topography of proteins. Nucleic Acids Res 2003;31:3352–3355.

27. Potterton E, McNicholas S, Krissinel E, Cowtan K, Noble M. TheCCP4 molecular-graphics project. Acta Crystallogr D Biol Crystal-logr 2002;58:1955–1857.

28. Potterton L, et al. Developments in the CCP4 molecular-graphicsproject. Acta Crystallogr D Biol Crystallogr 2004;60:2288–2294.

29. Kochhar S, Christen P. Mechanism of racemization of amino acidsby aspartate aminotransferase. Eur J Biochem 1992;203:563–569.

30. Noland BW, et al. Structural studies of Salmonella typhimuriumArnB (PmrH) aminotransferase: a 4-amino-4-deoxy-L-arabinoselipopolysaccharide-modifying enzyme. Structure (Camb) 2002;10:1569–1580.

31. Jakubowicz M. Structure, catalytic activity and evolutionaryrelationships of 1-aminocyclopropane-1-carboxylate synthase, thekey enzyme of ethylene synthesis in higher plants. Acta BiochimPol 2002;49:757–774.

32. Eliot AC, Kirsch JF. Pyridoxal phosphate enzymes: mechanistic,structural, and evolutionary considerations. Annu Rev Biochem2004;73:383–415.

33. DeLano WL. The PyMOL Molecular Graphics System. 2002.Available from: http://www.pymol.org.



APPENDIX





Documents

Crystal structures of the PLP- and PMP-bound forms of BtrR, a dual functional aminotransferase involved in butirosin biosynthesis