proteinsSTRUCTURE O FUNCTION O BIOINFORMATICS

Crystallographic and mass spectrometricanalyses of a tandem GNAT protein from theclavulanic acid biosynthesis pathwayAman Iqbal, Haren Arunlanantham, Tom Brown Jr., Rasheduzzaman Chowdhury,

Ian J. Clifton, Nadia J. Kershaw, Kirsty S. Hewitson, Michael A. McDonough,*

and Christopher J. Schofield

Chemistry Research Laboratory, University of Oxford, Oxford OX1 3TA, United Kingdom

INTRODUCTION

Clavulanic acid (CA) cannot be produced efficiently by chemical

synthesis, hence it is produced by fermentation of Streptomyces clavuli-

gerus (for review see Refs. 1 and 2). The biosynthetic pathway to CA is

of interest because it contains unusual biochemistry and its engineer-

ing may enable methods for producing modified CA analogs which are

more potent b-lactamase inhibitors or have broader spectra of activity.

Considerable progress has been made in elucidating the steps leading

to the bicyclic intermediate (3S,5S)-clavaminic acid. It is also known

that (3R,5R)-clavaldehyde is reduced to CA in the final step of the

pathway.3 However, the process by which (3S,5S)-clavaminic acid

undergoes the requisite ‘‘double-epimerization’’ to yield (3R,5R)-clav-

aldehyde is unknown (Supporting Information Scheme 1).

It is proposed that the products of the ‘‘later’’ genes of the CA bio-

synthesis gene cluster (orf10-orf18) are involved in the conversion of

(3S,5S)-clavaminic acid to (3R,5R)-clavaldehyde.1,2,4 Acylated

(3S,5S)-derivatives of (3S,5S)-clavaminic acid, N-acetyl-clavaminic

acid, N-acetyl-glycyl-clavaminic acid and N-glycyl-clavaminic acid,

have been isolated from mutant strains of S. clavuligerus deficient in

CA production.4 ORF17 (NGCAS) from S. clavuligerus has been

reported to catalyze the addition of glycine to (3S,5S)-clavaminic

acid to give N-glycyl-clavaminic acid5; It is unclear whether N-glycyl-

clavaminic acid is an intermediate between clavaminic acid and CA.

Given the role of amines and potentially N-acetylated intermedi-

ates in CA biosynthesis, it is of interest that the product of orf14,

CBG (CA biosynthesis GNAT protein, 339 amino acids, 37 kDa), is

predicted to belong to the GNAT superfamily.6 GNAT enzymes cata-

lyze the transfer of an acetyl group from AcCoA (or in some cases

other acyl-CoA derivatives) to an acceptor amine; their consensus

mechanism proceeds via a ‘‘tetrahedral’’ intermediate which is stabi-

lized by binding in an ‘‘oxyanion hole’’ (for reviews see Refs. 6 and 7).

GNAT enzymes employ a range of ‘acceptor’ substrates including

Additional Supporting Information may be found in the online version of this article.

Grant sponsor: Biotechnology and Biological Sciences Research council (BBSRC).

*Correspondence to: Michael A. McDonough, Department of Chemistry, Oxford Centre for

Integrative Systems Biology, Chemistry Research Laboratory, University of Oxford, Mansfield Road,

Oxford OX1 3TA, United Kingdom. E-mail: michael.mcdonough@chem.ox.ac.uk.

Received 1 September 2009; Revised 7 October 2009; Accepted 14 October 2009

Published online 6 November 2009 in Wiley InterScience (www.interscience.wiley.com).

DOI: 10.1002/prot.22653

ABSTRACT

(3R,5R)-Clavulanic acid (CA) is a clinically

important inhibitor of Class A b-lactamases.

Sequence comparisons suggest that orf14 of the

clavulanic acid biosynthesis gene cluster encodes

for an acetyl transferase (CBG). Crystallographic

studies reveal CBG to be a member of the

emerging structural subfamily of tandem

Gcn5-related acetyl transferase (GNAT) proteins.

Two crystal forms (C2 and P21 space groups)

of CBG were obtained; in both forms one mole-

cule of acetyl-CoA (AcCoA) was bound to the

N-terminal GNAT domain, with the C-terminal

domain being unoccupied by a ligand. Mass

spectrometric analyzes on CBG demonstrate

that, in addition to one strongly bound AcCoA

molecule, a second acyl-CoA molecule can bind

to CBG. Succinyl-CoA and myristoyl-CoA dis-

played the strongest binding to the ‘‘second’’

CoA binding site, which is likely in the C-termi-

nal GNAT domain. Analysis of the CBG struc-

tures, together with those of other tandem

GNAT proteins, suggest that the AcCoA in the

N-terminal GNAT domain plays a structural role

whereas the C-terminal domain is more likely to

be directly involved in acetyl transfer. The avail-

able crystallographic and mass spectrometric

evidence suggests that binding of the second

acyl-CoA occurs preferentially to monomeric

rather than dimeric CBG. The N-terminal

AcCoA binding site and the proposed C-termi-

nal acyl-CoA binding site of CBG are compared

with acyl-CoA binding sites of other tandem

and single domain GNAT proteins.

Proteins 2010; 78:1398–1407.VVC 2009 Wiley-Liss, Inc.

Key words: acetyl coenzyme A; acetyl transfer-

ase; antibiotic biosynthesis; clavulanic acid;

Gcn5 related N-acetyl transferase (GNAT); tan-

dem-GNAT protein.

1398 PROTEINS VVC 2009 WILEY-LISS, INC.

proteins (e.g., histones) and small molecules such as the

b-lactam tabtoxin.8 Most studied GNAT proteins have a

characteristic ab fold that is responsible for binding a sin-

gle AcCoA molecule.7 Recently a structural subfamily of

GNAT proteins has emerged, which contain two stereotyp-

ical GNAT ab folds in tandem linked through a shared b-

strand. Reported crystal structures for this subfamily

include myristoyl transferase (MYST), FemX, and myco-

thiol synthase (MSHD).6,9,10

Here we describe crystal structures for CBG in two

space groups, C2 and P21, to 2.30 A and 2.38 A resolu-

tion, respectively. The structures reveal CBG as a member

of the tandem GNAT subfamily. In both crystal forms, a

single molecule of AcCoA was observed bound to the N-

terminal domain. Mass spectrometric analyzes reveal that

a second acyl-CoA derivative can bind to CBG, likely via

its C-terminal domain. Comparison of the two proposed

acyl-CoA binding sites in CBG with those of other tan-

dem and single domain GNAT proteins reveal differences

in the proposed structural and catalytic acyl-CoA binding

sites.

MATERIALS AND METHODS

Cloning, expression and purification studies

The orf14 gene was amplified by the polymerase chain

reaction using genomic DNA from S. clavuligerus and

inserted into the pET24a(1) expression vector via NdeI

and BamHI restriction sites, at the 50 and 30 termini

respectively. The primers used were as follows: Forward:

50 – GGT GGT CAT ATG AAC GAC ACC G – 30; Reverse:

50 – GGT GGT GGA TCC TCA GTC GGA ACG – 30.Escherichia coli cells transformed with the orf14/

pET24a(1) plasmid were grown at 378C in 2TY media

containing the antibiotic kanamycin at 30 lg mL-1.

When the cells reached an OD600 of 0.6–0.8, the temper-

ature was reduced to 288C and expression was induced

by the addition of 0.5 mM isopropyl thio-ß-D-galacto-

side. After a further four hours, cells were harvested by

centrifugation and were stored at 2808C.

To purify CBG, thawed cells were re-suspended in

buffer (50 mM tris-HCl pH 7.5) and lysed by sonication.

Chromatographic purification of CBG protein from the

cell lysate involved the sequential use of Q-Sepharose,

Phenyl-Resource1, and S75 gel filtration columns (Amer-

sham Biosciences (now GE healthcare)), to give a final

yield of �11 mg CBG from �30 g of cells (�90% purity

by SDS-PAGE analysis). CBG was eluted from the

Q-Sepharose column over eight column volumes at a

NaCl gradient of 0–1 M. Fractions (as judged by SDS-

PAGE) were combined and an equal volume of 2 M

ammonium sulphate was added. The protein was loaded

onto a Phenyl-Resource1 column (13 mL) pre-equili-

brated with 50 mM Tris-HCl pH 7.5, 10% glycerol, and

1 M ammonium sulphate. The protein was eluted over

five column volumes using a 1.0–0.0 M gradient of am-

monium sulphate in 50 mM Tris-HCl pH 7.5 and 10%

glycerol. Fractions containing CBG were combined and

purified further using an S75 gel filtration column pre-

equilibrated with 50 mM Tris-HCl pH 7.5. Native PAGE

and analytical gel filtration analyses (data not shown)

indicated that CBG can exist as a mixture of monomeric

and dimeric (predominantly monomeric) forms in

solution.

Seleno-methionine CBG production

E. coli BL21 (DE3) cells that had been transformed with

the orf14/pET24a(1) vector, were grown at 378C to an

O.D.600 of 0.6–0.8 in L-seleno-methionine containing

medium (Molecular Dimensions, UK). The cell growth

and purification conditions for the seleno-methionine

CBG were as for the wild type CBG. Mass spectrometric

analysis of the purified seleno-methionine CBG indicated

>95% seleno-methionine incorporation (data not shown).

CBG crystallization

High-throughput crystallization screening was carried

out at the Oxford Protein Production Facility.11 The ini-

tial optimisation of hits was carried out using the hang-

ing-drop vapor-diffusion method in 24-well VDXTM

plates (Hampton Research).

Optimization of P21 crystal form

Small crystals were observed after about 1 week in

crystallization drops consisting of 2 lL protein solution

(12 mg mL21) mixed with 2 lL of reservoir solution

containing 30% PEG 8000, 0.1 M Na cacodylate pH 6.5,

0.2 M sodium acetate; large amounts of precipitate were

obtained in these drops with some crystalline plate for-

mations. This condition was further optimized to 10%

PEG 8000, 0.1 M cacodylate pH 6, and 0.2 M sodium

acetate at 48C to give rectangular plate-shaped crystals

having dimensions of �0.10 3 0.15 3 0.20 mm, after

2 weeks. These crystals were macro-seeded into pre-equi-

librated drops containing 5% PEG 8000, 0.1 M cacodyl-

ate pH6, 0.2 M sodium acetate and CBG (final concen-

tration of 3 mg mL21). Incubation at 48C for a further

two weeks led to larger crystals with dimensions of

�0.15 3 0.25 3 0.30 mm. Crystals were flash-cooled in

liquid nitrogen after rapid sequential transfer by cryo-

loop from 5, 10, 20, 30% glycerol in well solution.

Optimization of C2 crystal form

From the initial screen, thick clusters of needle-shaped

crystals were observed after three days at 178C in 4 lL

hanging-drops consisting of 2 lL of protein solution

(3 mg mL21) with a three-fold molar excess of AcCoA,

mixed with 2 lL of well solution, over 500 lL of well

Analyzes of a Tandem GNAT Protein

PROTEINS 1399

solution containing 1 M sodium citrate and 0.07 M caco-

dylate (pH 6.5). The crystals were crushed and streak-

seeded into pre-equilibrated drops at 178C containing

CBG (3 mg mL21), a threefold molar excess of AcCoA, 1

M sodium citrate and 0.07 M cacodylate (pH 6.5). Rod-

shaped crystals with dimensions of 0.15 3 0.20 3 0.35

mm grew after 1 week. All crystals were mounted in

nylon loops and were flash-frozen by rapid plunging in

liquid nitrogen using 30% glycerol as a cryoprotectant.

Data collection and structure determination

MAD data were collected using seleno-methionine

derivatized CBG crystals at the SRS, Daresbury (Beamline

14.2). Data were indexed, integrated and scaled using

HKL200012 (Table I). The calculated Matthews coeffi-

cient using CCP4 indicated four molecules per asymmet-

ric unit were probable. Twenty four selenium sites were

identified and used to obtain initial phases to 3.0 A using

the program SOLVE.13 Density modification and phase

extension to 2.4 A using the peak wavelength data was

then carried out using RESOLVE.13 The initial model

was built using the program O14 and refined with strict

non-crystallographic symmetry (NCS) restraints using

CNS.15 Further cycles of model fitting and refinement,

included individual B-factor refinement and the addition

of solvent molecules while NCS restraints were gradually

relaxed over of the course of refinement and removed for

the final rounds.

There are two CBG molecules per asymmetric unit in

the C2 crystal form. The structure was solved by molecu-

lar replacement using C2 with P21 structure as the search

model using EPMR.16 Refinement of the C2 structure

was performed using CNS by the same methods as for

the P21 crystal form. Atomic coordinates and structure

factors have been deposited in the RCSB protein data

bank (Accession codes 2wpw (P21) and 2wpx (C21)).

Mass spectrometric analyzes

Electrospray ionization-mass spectrometry (ESI-MS)

analyzes used an ESI time of flight mass spectrometer

(Q-Tof micro, Micromass, Altrincham, UK) interfaced

with a NanoMateTM HD chip-based nano-ESI source

(Advion Biosciences, Ithaca, NY). The instrument was

equipped with a standard Z-spray source block. Clusters

of Na(n11)In (1 mg mL21 NaI in 50:50 water/isopropa-

nol) were used for calibration. The instrument settings

were as follows: spraying voltage 1.70 kV; sample cone

voltage 50–200 V (only spectra acquired at 80 V are

shown); source temperature 408C; acquisition time 15 s;

scan time 0.5 s; acquisition range m/z 500–10,000; posi-

tive ion mode. The pressure at the interface between the

atmospheric source and the high vacuum region was

fixed at 6.60 mbar (measured with the roughing pump

Pirani gauge) by throttling the pumping line using an

Edwards Speedivalve to provide collisional cooling.

Table IData Statistics of CBG Crystal Structures

Native SeMet MAD data single crystal

X-ray source Daresbury SRS 14.1 May 2003 Daresbury SRS 14.2 9th Sept 2003

Wavelength (�) 1.488 0.97976 (peak) 0.98033 (inflection) 0.96988 (remote)PDB acquisition code 2wpx 2wpwResolution (�) (outer shell) 2.32 (2.40–2.32) 2.40 (2.49–2.40) 2.40 (2.49–2.40) 2.60 (2.69–2.60)Space group C2 P21 P21 P21

Unit cell dimensions(a �, b �, c �, b8)

151.98, 70.09, 106.29, 134.38 75.45, 93.50, 126.43, 97.86 75.47, 93.43, 126.45, 97.84 75.41, 93.46, 126.62, 97.80

Total number of reflections 117844 317235 238914 192737Number of unique reflections 34742 68,427 (6427) 67,153 (6312) 54,695 (5315)Redundancy �3.5 �4.5 �4 �4Completeness (%) 99.3 (95.3) 98.0 (92.4) 97.9 (91.9) 98.5 (95.9)I/r(I) 13.8 (5.1) 10.2 (3.7) 8.0 (2.8) 7.1 (2.8)Rmerge (%)a 8.7 (24.3) 12.8 (35.8)b 12.6 (37.3)b 12.6 (37.3)b

Rcryst (%)c 19.1 21.2Rfree (%)d 25.3 26.6RMS deviation (�/8) 0.019/2.3 0.019/2.1Average B factors (�2)e 21.2, 23.7, 21.1, 23.8, 23.2 28.5, 31.0, 28.9, 31.6,

28.7, 31.5, 29.0, 31.7, 28.5Number of water molecules 590 1159

aRmerge 5P

j,h| Ih,j 2 Ih| /P

j,hIh 3 100.bBijvoet pairs unmerged.cRcryst 5

PkFobs| 2 |Fcalck/|Fobs| 3 100.

dRfree, based on 10% of the total reflections.eC2 structure: Main chain A, side chain A, main chain B, side chain B, solvent. P21 structure: Main chain A, side chain A, main chain B, side chain B, main chain C,

side chain C, main chain D, side chain D, solvent.

A. Iqbal et al.

1400 PROTEINS

Before analyses, CBG was desalted using a Micro Bio-

Spin 6 column (Bio-Rad, Hemel Hempstead, UK) in 15

mM ammonium acetate pH 7.5. The CBG solution was

diluted in ammonium acetate buffer to a stock concen-

tration of 50 lM. Coenzyme A derivatives were dissolved

in ammonium acetate at 100 lM. Samples of CBG and

coenzyme A derivatives were prepared in a 96-well plate

at 158C (final concentrations: CBG 10 lM; CoA deriva-

tives 20 lM) and introduced into the mass spectrometer

via the NanoMate device (estimated flow rate 100 nL

min21).

ESI-MS data were analyzed using MassLynx software

(Waters, Milford, MA). Raw data were processed by

smoothing (mean method; smooth windows: 10; number

of smooths: 2), subtracting the background (polynomial

order: 15; % below curve: 10; tolerance 0.01) and cen-

tring peaks (minimum peak width at half-height: 2; %

centroid top: 80).

RESULTS

Overall structure of CBG

The S. clavuligerus orf14 gene was cloned and recombi-

nant CBG was produced in E. coli. The protein was puri-

fied to near >95% purity (by SDS-PAGE analysis). Opti-

mization of high throughput crystallization hits led to

two distinct crystal forms having P21 and C2 space

groups, which diffracted to 2.40 and 2.32 A, respectively.

The structure of the P21 crystal form was solved by the

MAD method and the structure of the C2 form was sub-

sequently solved by molecular replacement. Electron den-

sity maps revealed one AcCoA molecule bound to the

N-terminal GNAT domain in both the P21 and C2 crystal

forms suggesting that one AcCoA per molecule acquired

during expression remains tightly bound during the puri-

fication and crystallization procedures.

The structure of CBG P21 crystal form reveals an

apparent homo-dimer with four molecules in the asym-

metric unit. The C2 crystal form exists as a crystallo-

graphic homo-tetramer with two molecules per asym-

metric unit. In both the P21 and C2 crystal forms the a6

helix of the C-terminal domain is positioned to form

dimer contacts [Fig. 1(c–e)]. In the C2 crystal form, only

half of the residues that form the full a6 helix observed

in the P21 form are structured; the other half of these

residues form tetramer contacts by extending into a

neighboring cavity, stabilizing the tetramer [Fig. 1(b,c)].

The observation that the same C-terminal residues form

dimer interactions, albeit in a different manner, in the

two crystal forms suggests that the observed dimer-dimer

interface in the crystal form is relevant in solution. Ana-

lytical gel filtration and native SDS-PAGE studies revealed

that CBG exists predominantly as a monomer with some

homodimer present. Non-denaturing ESI-MS studies

revealed that, as purified, CBG is distributed between

monomeric [37.0 kDa, corresponding to the CBG mono-

mer 1 1 molecule of AcCoA (809 Da)] and dimeric

(73.2 kDa corresponding to the CBG dimer 1 1 mole-

cule of AcCoA) forms [Fig. 3(a)]. When ESI-MS analyses

were carried out after the addition of CoA derivatives,

only the monomeric form was observed (see later). This

is notable because, in addition to the role of a6 in ena-

bling dimer formation, the loop immediately proceeding

a6 is involved in CoA binding as observed for other

GNAT family members. We propose that binding of a

second CoA derivative weakens dimer formation via

ligand induced-fit conformational changes, including to

a6 and nearby regions; it also seems likely that the

catalytically active form is monomeric.

Both the N- and C-terminal domains of CBG possess

the characteristic GNAT superfamily mixed a,b-fold. The

root mean square deviation (r.m.s.d.) of the N-terminal

domain of CBG relative to its C-terminal domain is 1.58

A over 94 Ca atoms. The GNAT fold comprises a central

b-sheet (N-terminal domain: b1-b4 and C-terminal do-

main: b7-b10) flanked by helices (N-terminal domain:

a1 and a2 on one face of the b-sheet and a3 and a4 on

the other face; C-terminal domain: a7-a8 on one face of

the b-sheet and a5- a6 on the other face) (Fig. 1) and

(Supporting Information Fig. 2). The two GNAT

domains of CBG are linked by a b-strand (b6).

Both the N-terminal and C-terminal domains of CBG

display similarity with several single domain GNAT

enzymes that accept small molecule substrates, including

the tabtoxin resistance protein (TTR)8 (r.m.s.d. of 1.15 A

over 118 Ca atoms; 67% of the CBG N-terminal domain

residues), and serotonin N-acetyltransferase (AANAT)17

(r.m.s.d. of 1.58 A over 100 Ca atoms; 57% of the CBG

N-terminal domain residues). The predicted C-terminal

domain AcCoA binding site of CBG, which we propose

is directly involved in acetyl transfer, is unoccupied in

the crystal structures. A comparison with AANAT17,18

suggests it is possible that, upon AcCoA binding to the

C-terminal domain of CBG, the apparently disordered

region between a5 and a6, [Fig. 1(g,h)] stabilizes, possi-

bly as part of an induced-fit mechanism.

Acetyl-CoA binding sites

N-terminal domain

One molecule of AcCoA was observed bound to each

of the N-terminal domains of the two molecules forming

the CBG homodimer. The acetyl group of AcCoA in the

N-terminal domain of CBG is apparently somewhat

more deeply buried than acyl-CoA molecules observed in

other reported structures of GNAT proteins, including

the N-terminal domain of MSHD in which it is proposed

that a ‘‘structural’’ AcCoA is bound.19 Notably, in the N-

terminal domain of CBG there are no clear hydrogen-

bonding interactions with the carbonyl of the S-acetyl

Analyzes of a Tandem GNAT Protein

PROTEINS 1401

Figure 1Views from crystal structures of CBG. (a) View from the structure of the P21 crystal form of CBG showing the tetramer. (b) View from the CBG

P21 crystal form showing the dimer-forming interactions (red and blue), highlighting the role of a6 and subsequent disordered residues (one

monomer is in green, with the other in purple). (c) Superposition of the C-terminal GNAT domains of C2 (blue) and P21 (dark pink) crystal

forms, showing the restructuring of the a6 helix in the P21 form. (d) View from the CBG C2 crystal form showing the dimer and the interactions

between a6 helices. (e) Surface representation of CBG dimer interaction, showing the extensive dimer contacts. (f) CBG monomer showing the

N-terminal (beige) and C-terminal (marine blue) domains. Acetyl-CoA is in yellow sticks. (g) CBG N-terminal domain with bound AcCoA.

(h) CBG C-terminal domain. The secondary structural elements (a-helices and b-sheets) are labeled.

A. Iqbal et al.

1402 PROTEINS

group, that is, there is no clear oxyanion hole, [Fig. 2].

The acetyl group is bound in a predominantly hydropho-

bic pocket (formed in part by the side chains of Trp-104

and Phe-142), with no residues nearby that could be

obviously involved in general acid/base catalysis.

C-terminal domain

Although, the C-terminal domain of CBG does not

have AcCoA or CoA bound in the crystal structures,

there is sufficient space for AcCoA or a CoA derivative to

bind to the C-terminal domain. A potential oxyanion

hole is formed by the backbone amides of Met-268 and,

possibly Thr-269 [Fig. 4(b)]. Superimposition of N-ter-

minal domains of CBG and the N-myristoyl transferase

(NMT) onto the C-terminal domain of CBG supports

the proposal that the backbone NH of Thr-269 of CBG,

and with a lower probability (because of its orientation)

that of Met-268, form part of the oxyanion hole in

the C-terminal domain of CBG [Fig. 4(b), Supporting

Figure 2Interactions of AcCoA bound in the N-terminal domain of CBG. (a) Stereo view showing the AcCoA binding interactions in CBG. (b) Stereoview of the

refined 2Fo–Fc electron density for AcCoA contoured to 1r. (c) Schematic showing the interactions of AcCoA with the CBG N-terminal domain.

Analyzes of a Tandem GNAT Protein

PROTEINS 1403

Figure 3Electrospray ionization mass spectrometry (ESI-MS) spectra of CBG under nondenaturing conditions (pH 7.5). (a) m/z spectrum showing

individual charge states for the monomeric and the dimeric CBG; (b) m/z spectrum showing the apparent disruption of the CBG dimer into a

monomer upon incubation and binding of succinyl CoA (sample cone voltage: 80 V). Deconvoluted ESI-MS spectra under nondenaturing

conditions (pH 7.5) showing the complexes formed between (CBG) and (CoA derivatives): (c) CBG as purified; (d) CBG 1 CoASH; (e) CBG 1

acetyl CoA; (f) CBG 1 palmitoyl CoA; (g) CBG 1 lauroyl CoA; (h) CBG 1 myristoyl CoA; (i) CBG 1 succinyl CoA (j) competitive binding

experiment (CBG 1 CoASH 1 AcCoA 1 palmitoyl CoA 1 lauroyl CoA 1 myristoyl coA 1 succinyl CoA). Peak A: CBG; Peak B: CBG:CoA

complex. Concentrations: CBG (10 lM); coenzyme A derivatives (20 lM). Sample cone voltage: 80 V. Mass shifts relative to Peak A are indicated in

brackets after Peak B masses.

Information Fig. 4]. The predicted CoA binding pocket

in the C-terminal domain also contains hydrophobic resi-

dues (Leu-289, Val-292, Leu-293 and Val-254) suggesting

that it may accommodate CoA analogs with hydrophobic

character.

Non-denaturing ESI-MS analyzes were then conducted

to investigate the binding of CBG to other CoA deriva-

tives. All of the CoAs tested appeared to bind to the

CBG monomer, without displacing the already-bound

AcCoA (presumably in the N-terminal domain) [Fig.

3(d–i)]. These observations support the proposal that the

C-terminal GNAT domain of CBG is the ‘‘catalytic’’

domain. Notably, binding of the CoAs resulted in the ob-

servation of near complete disruption of the dimer (at

least in the MS analysis), resulting in the observation of

the monomeric complex only (e.g., with succinyl CoA:

[Fig. 3(b)]. The extent of binding varied with CoASH

and the acyl-CoA derivatives tested. The ratios of com-

plexed:uncomplexed forms of monomeric CBG were 4:1

CoASH (1764 Da) [Fig. 3(d)]; 4.5:1 (palmitoyl

CoA,11002 Da), [Fig. 3(f)]; 4.5:1 (lauroyl CoA, 1946

Da), [Fig. 3(g)]; 8:1 (myristoyl CoA, 1972 Da), [Fig.

3(h)]; and 9:1 (succinyl CoA, 1864 Da), 95% [Fig. 3(j)].

A competitive binding experiment, in which an equimo-

lar mixture of all the above CoAs was incubated with

CBG, revealed that succinyl CoA bound most strongly to

CBG, although small peaks corresponding to complexes

with myristoyl CoA and palmitoyl CoA were also

observed [Fig. 3(j)].

DISCUSSION

The crystallographic and biochemical analyzes identify

CBG as a tandem GNAT protein that copurifies with

a single AcCoA molecule bound in the N-terminal

GNAT domain. This AcCoA molecule and the N-terminal

domain are unlikely to be directly involved in

Figure 4Variations in the CoA binding sites and (proposed) oxyanion holes of GNAT enzymes, as indicated by crystallographic analyzes. (I) Tandem GNATs

(a) CBG N-terminal domain showing the AcCoA binding site; No oxyanion hole residues are apparent. (b) CBG C-terminal domain showing the

probable CoA binding site with the proposed oxyanion hole residues Thr-269 and Met-268. (c) N-Myristoyl transferase (NMT) (PDB id: 2nmt)

N-terminal domain bound to myristoyl CoA, showing the proposed oxyanion hole residues. (d) N-Terminal domain of mycothiol synthase

(MSHD) (PDB id:1ozp) showing the apparent lack of oxyanion hole residues. (e) MSHD C-terminal domain bound to a molecule of AcCoA

showing the oxyanion hole residues. (II) Proposed oxyanion hole residues of single domain GNATs. (f) Rv0802c (PDB id: 2vzz) bound to succinyl

CoA. (g) Tabtoxin resistance (TTR) protein (PDB id:1ghe) bound to AcCoA. (h) Histone acetyl transferase (HPA2) (PDB id: 1qsm) bound toAcCoA. (i) Aminoglycoside N-acetyl transferase (AACIB) (PDB id: 2vqy) bound to AcCoA. (j) Spermine acetyl transferase (SSAT) (Pdb id:2jev)

bound to spermine-AcCoA. (k) Glucosamine N-acetyl transferase (GNA1) from Saccharomyces Cerevisiae (PDB id: 1i12) bound to AcCoA. (The

CoAs are shown in light blue sticks and the oxyanion hole residues in green).

Analyzes of a Tandem GNAT Protein

PROTEINS 1405

catalysis because of the lack of an oxyanion hole in the

N-terminal domain and because the AcCoA molecule is

apparently tightly bound and buried. The C-terminal

domain of CBG is thus more likely to be directly

involved in acyl transfer activity.

It is plausible that tandem GNAT proteins likely

evolved by duplication of single GNAT-encoding genes.

However, the relative position (N- or C-termini) of the

structural and catalytic GNAT domains varies. Mycothiol

synthase (MSHD)20–22 copurifies with a tightly bound

AcCoA (as does CBG) in its N-terminal domain which

lacks an oxyanion hole. It is thus proposed that the acetyl

transferase activity of MSHD involves its C-terminal

GNAT domain (Fig. 4). However, in the case of the tan-

dem GNAT protein N-myristoyl transferase (NMT)10 the

N-terminal domain binds to myristoyl CoA and has an

oxyanion hole and hence is assigned as the catalytic

domain; in NMT the C-terminal domain of NMT does

not bind a CoA derivative.

Comparison of the proposed acyl-CoA binding sites in

CBG and other GNAT proteins reveals significant differ-

ences in the thioester binding sites (Fig. 4), both in the

domains that are proposed to be directly involved in cat-

alytic acyl-transfer [Fig. 4(b–d) and (f–k)] and those that

are not [Fig. 4(a,e)]. In the case of the catalytic domains,

there is considerable variation in the proposed residues

proposed to form oxyanion holes which can include

backbone amides and tyrosine phenol groups (Fig. 4).

The available structures imply that in some cases only

one residue is involved in forming the oxyanion hole,

and in the cases where there are two proposed oxyanion

hole residues, the hydrogen bonds of one of the proposed

oxyanion hole residues may appear not to be optimally

oriented for binding an oxyanion in the oxyanion hole

(Fig. 4). Although variations from the classical backbone

amide oxyanion holes occur in the serine/cysteine pro-

teases, the available structures suggest that these may be

a more widespread distribution amongst variants with

acyl-CoA transfer proteins. It is possible that apparent

lack of an essential second hydrogen binding residue in

the oxyanion holes of GNAT enzymes reflects the

enhanced reactivity of thioesters relative to amides/esters,

at least with respect to initial nucleophilic attack on the

carbonyl group at neutral pH.23

The finding that CBG is a tandem GNAT protein may

assist in functional assignments of its role in clavulanic

acid biosynthesis and, in particular, whether or not it is

involved in production of N-acetyl-clavaminic acid or

N-acetyl-glycyl-clavaminic acid. However, in addition to

acetyl CoA, various other acyl-CoA derivatives were

observed to bind to CBG, likely at the C-terminal do-

main, opening up the possibility that it may catalyze

acyl-transfer reaction with a donor substrate other than

AcCoA. However, it is also possible that these derivatives

bridge the acyl-CoA and prime substrate binding sites. It

is notable that the acyl-CoA derivative that bound most

tightly as observed by ESI-MS was succinyl-CoA, which

like intermediates in clavulanic acid biosynthesis contains

a carboxylic acid. This information may be useful in

assigning a role to CBG in CA biosynthesis.

ACKNOWLEDGMENTS

The authors gratefully acknowledge Dr. Jasmin

Mecinovic for his assistance with nondenaturing ESI-MS

analysis.

REFERENCES

1. Baggaley KH, Brown AG, Schofield CJ. Chemistry and biosynthesis

of clavulanic acid and other clavams. Nat Prod Rep 1997;14:309–

333.

2. Kershaw NJ, Caines MEC, Sleeman MC, Schofield CJ. The enzymol-

ogy of clavam and carbapenem biosynthesis. Chem Commun 2005;

4251–4263.

3. Mackenzie AK, Kershaw NJ, Hernandez H, Robinson CV, Schofield

CJ, Andersson I. Clavulanic acid dehydrogenase: structural and

biochemical analysis of the final step in the biosynthesis of the

beta-lactamase inhibitor clavulanic acid. Biochemistry 2007;46:

1523–1533.

4. Jensen SE, Paradkar AS, Mosher RH, Anders C, Beatty PH, Brumlik

MJ, Griffin A, Barton B. Five additional genes are involved in clavu-

lanic acid biosynthesis in Streptomyces clavuligerus. Antimicrob

Agents Chemother 2004;48:192–202.

5. Arulanantham HA, Kershaw NJ, Hewitson KS, Hughes CE, Thirket-

tle JE, Schofield CJ. ORF17 from the clavulanic acid biosynthesis

gene cluster catalyzes the ATP-dependent formation of N-glycyl-

clavaminic acid. J Biol Chem 2006;281:279–287.

6. Vetting MW, De Carvalho LPS, Yu M, Hegde SS, Magnet S, Roderick

SL, Blanchard JS. Structure and functions of the GNAT superfamily of

acetyltransferases. Arch Biochem Biophys 2005;433:212–226.

7. Dyda F, Klein DC, Hickman AB. GCN5-related N-acetyl transfer-

ases: a structural overview. Annu Rev Biophys Biomol Struct 2000;

29:81–103.

8. He H, Ding Y, Bartlam M, Sun F, Le Y, Qin X, Tang H, Zhang R,

Joachimiak A, Liu J, Zhao N, Rao Z. Crystal structure of tabtoxin

resistance protein complexed with acetyl coenzyme A reveals the

mechanism for b-lactam acetylation. J Mol Biol 2003;325:1019–

1030.

9. Vetting MW, Errey JC, Blanchard JS. Rv0802c from Mycobacterium

tuberculosis: the first structure of a succinyltransferase with the

GNAT fold. Acta Crystallogr F Biol Crystallogr 2008;64:978–985.

10. Bhatnagar RS, Futterer K, Farazi TA, Korolev S, Murray CL, Jack-

son-Machelski E, Gokel GW, Gordon JI, Waksman G. Structure of

N-myristoyltransferase with bound myristoylCoA and peptide sub-

strate analogs. Nat Struct Mol Biol 1998;5:1091–1097.

11. Walter TS, Diprose JM, Mayo CJ, Siebold C, Pickford MG, Carter

L, Sutton GC, Berrow NS, Brown J, Berry IM, Stewart-Jones GBE,

Grimes JM, Stammers DK, Esnouf RM, Jones EY, Owens RJ, Stuart

DI, Harlos K. A procedure for setting up high-throughput nanolitre

crystallization experiments. Crystallization workflow for initial

screening, automated storage, imaging and optimization. Acta Crys-

tallogr D Biol Crystallogr 2005;61:651–657.

12. Otwinowski Z, Minor W. Processing of x-ray diffraction data col-

lected in oscillation mode. Methods Enzymol 1997;276:307–326.

13. Terwilliger T. Maximum-likelihood density modification. Acta Crys-

tallogr D Biol Crystallogr 2000;56:965–972.

14. Jones TA, Zou J-Y, Cowan SW, Kjeldgaard M. Improved methods

for the building of protein models in electron density maps and the

location of errors in these models. Acta Crystallogr A Foundation

Crystallogr 1991;47:110–119.

A. Iqbal et al.

1406 PROTEINS

15. Brunger AT, Adams PD, Clore GM, Delano WL, Gros P,

Grosse-Kunstleve RW, Jiang J-S, Kuszewski J, Nilges M, Pannu

NS, Read RJ, Rice LM, Simonson T, Warren GL. Crystallogra-

phy and NMR system: a new software suite for macromolecular

structure determination. Acta Crystallogr D Biol Crystallogr

1998;54:905–921.

16. Kissinger CR, Gehlhaar DK, Fogel DB. Rapid automated molecular

replacement by evolutionary search. Acta Crystallogr D Biol Crys-

tallogr 1999;55:484–491.

17. Wolf E, De Angelis J, Khalil EM, Cole PA, Burley SK. X-ray crystal-

lographic studies of serotonin N-acetyltransferase catalysis and inhi-

bition. J Mol Biol 2002;317:215–224.

18. Scheibner KA, De Angelis J, Burley SK, Cole PA. Investigation of

the roles of catalytic residues in serotonin N-acetyltransferase. J Biol

Chem 2002;277:18118–18126.

19. Vetting MW, Roderick SL, Yu M, Blanchard JS. Crystal structure of

mycothiol synthase (Rv0819) from Mycobacterium tuberculosis

shows structural homology to the GNAT family of N-acetyltransfer-

ases. Protein Sci 2003;12:1954–1959.

20. Bornemann C, Jardine MA, Spies HS, Steenkamp DJ. Biosynthesis

of mycothiol: elucidation of the sequence of steps in Mycobacterium

smegmatis. Biochem J 1997;325:623–629.

21. Newton GL, Ta P, Fahey RC. A mycothiol synthase mutant of Myco-

bacterium smegmatis produces novel thiols and has an altered thiol

redox status. J Bacteriol 2005;187:7309–7316.

22. Koledin T, Newton G, Fahey RC. Identification of the mycothiol

synthase gene (mshD) encoding the acetyltransferase producing

mycothiol in actinomycetes. Arch Microbiol 2002;178:331–337.

23. Douglas KT. Elimination-addition pathways for thiol esters. Acc

Chem Res 2002;19:186–192.

Analyzes of a Tandem GNAT Protein

PROTEINS 1407