proteins STRUCTURE O FUNCTION O BIOINFORMATICS

An intersubunit disulfide bridge stabilizes the tetrameric nucleoside diphosphate kinase of Aquifex aeolicus Fanny Boissier,1,2 Florian Georgescauld,1,2 Lucile Moynie´,1,2 Jean-William Dupuy,3 Claude Sarger,1,2 Mircea Podar,4 Ioan Lascu,1,2 Marie-France Giraud,1,2 and Alain Dautant1,2* 1 Universite´ de Bordeaux, Institut de Biochimie et de Ge´ne´tique Cellulaires, UMR 5095, F-33077 Bordeaux, France 2 CNRS, Institut de Biochimie et de Ge´ne´tique Cellulaires, UMR 5095, F-33077 Bordeaux, France 3 Universite´ de Bordeaux, Centre Ge´nomique Fonctionnelle Bordeaux, Plateforme Prote´ome, F-33076 Bordeaux, France 4 Oak Ridge National Laboratory, Biosciences Division, Oak Ridge, Tennessee 37831

ABSTRACT

INTRODUCTION

The nucleoside diphosphate kinase (Ndk) catalyzes the reversible transfer of the c-phosphate from nucleoside triphosphate to nucleoside diphosphate. Ndks form hexamers or two types of tetramers made of the same building block, namely, the common dimer. The secondary interfaces of the Type I tetramer found in Myxococcus xanthus Ndk and of the Type II found in Escherichia coli Ndk involve the opposite sides of subunits. Up to now, the few available structures of Ndk from thermophiles were hexameric. Here, we determined the X-ray structures of four crystal forms of the Ndk from the hyperthermophilic bacterium Aquifex aeolicus (AaNdk). Aa-Ndk displays numerous features of thermostable proteins and is made of the common dimer but it is a tetramer of Type I. Indeed, the insertion of three residues in a surface-exposed spiral loop, named the Kpn-loop, leads to the formation of a two-turn a-helix that prevents both hexamer and Type II tetramer assembly. Moreover, the side chain of the cysteine at position 133, which is not present in other Ndk sequences, adopts two alternate conformations. Through the secondary interface, each one forms a disulfide bridge with the equivalent Cys133 from the neighboring subunit. This disulfide bridge was progressively broken during X-ray data collection by radiation damage. Such crosslinks counterbalance the weakness of the common-dimer interface. A 40% decrease of the kinase activity at 608C after reduction and alkylation of the protein corroborates the structural relevance of the disulfide bridge on the tetramer assembly and enzymatic function.

The nucleoside diphosphate kinase (Ndk) catalyzes the transfer of the g-phosphate from nucleoside triphosphate to nucleoside diphosphate via a ping-pong mechanism. Ndks are known to form hexamer in eukaryotes and some bacteria,1 Type I tetramer in Myxococcus xanthus Ndk (Mx-Ndk),2 and Type II tetramer in Escherichia coli Ndk (Ec-Ndk).3 The structure of Dictyostelium discoideum Ndk (Dd-Ndk, PDB id: 1npk) is considered as the reference for the hexameric form.4 Nevertheless, the oligomerization/ function relationship is up to now not clearly established. Whatever the quaternary architecture, Ndks share a common-dimer unit. Such building blocks can assemble through secondary interfaces to form oligomers. By three, they form a highly stable hexamer with 32 symmetry burying about 1500 A˚2 per monomer. By two, they form less stable tetramers with 222 symmetry burying less than 500 A˚2 per monomer. In Type I and Type II tetramers, the buried secondary interfaces are located on opposite sides of the protein.3 Ndks exhibit a ferredoxin fold composed of a fourstrand anti-parallel b-sheet1 covered on one face by three a-helices aA, a2, and a4 and on the opposite face by the helices a0, a1, and a3 and the Kpn-loop. The Kpn-loop adopts a spiral shape and connects the helix a3 to the strand b4 that harbors the catalytic histidine (Supporting Information Fig. S1). In Dd-Ndk, Arg109 (Arg107 in Aa-Ndk) is a ligand of the b-phosphate group of ADP and essential for catalytic mechanism.5

Proteins 2012; 00:000–000. C 2012 Wiley Periodicals, Inc. V

Key words: crystal structure; oligomerization; hyperthermophile; radiation damage.

C 2012 WILEY PERIODICALS, INC. V

Additional Supporting Information may be found in the online version of this article. Abbreviations: Aa, Aquifex aeolicus; AS, ammonium sulfate; asa, accessible surface area; Ba, Bacillus anthracis; bsa, buried surface area; Dd, Dictyostelium discoideum; Ec, Escherichia coli; LC-MS, liquid chromatography-mass spectrometry; Kpn-loop, a 20-residue segment which connects the third a-helix to the fourth b-strand, named after the killer of prune (kpn) mutation in Drosophila; Mx, Myxococcus xanthus; Ndk, nucleoside diphosphate kinase; Pa, Pyrobaculum aerophilum; Ph, Pyrococcus horikoshii; rmsd, root mean square deviation; Tt, Thermus thermophilus. Grant sponsors: Re´gion Aquitaine, ESRF, SOLEIL *Correspondence to: Dr. Alain Dautant, IBGC- CNRS, 1, rue Camille Saint-Sae¨ns, 33077 Bordeaux cedex, France. E-mail: [email protected]. Received 10 November 2011; Revised 23 January 2012; Accepted 6 February 2012 Published online 22 February 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/prot.24062

PROTEINS

1

F. Boissier et al.

Ndk of the halophilic bacterium Halomonas sp. 593 exhibits a dimeric form.6,7 Ndks from mesophilic bacteria are active as dimers and tetramers (Ec-Ndk8) or, surprisingly, as hexamers (Ba-Ndk from Bacillus anthracis (Ba)).9 Hyperthermophilic organisms were good candidates to study the stability of Ndk interfaces. In the thermophile bacterium Thermus thermophilus (Tt)10 and in hyperthermophilic archaea Pyrococcus horikoshii (Ph)11 and Pyrobaculum aerophilum (Pa),12 Ndks were hexameric. After multiple sequence alignments, analysis of known oligomerization interfaces predicted that most of the Ndks from thermophilic organisms could be hexameric, while Aquifex aeolicus Ndk (Aa-Ndk) may be tetrameric or even dimeric. To understand its thermal stability and to further determine its oligomeric state, we solved the crystal structure of Aa-Ndk. It assembled as a Type I tetramer with a Cys133

Cys133 disulfide bridge through the secondary interface. The bridge was also observed by liquid chromatography-mass spectrometry (LC-MS) and when reduced, the significant decrease of kinase activity indicated that oligomerization plays an important role in the stability of the enzyme. MATERIALS AND METHODS Expression, purification, and activity assays

The construct used for protein expression and purification was obtained by cloning the ndk gene from Aa into the NdeI and XhoI sites of the expression vector pET-21a (Novagen). The overexpression of the recombinant AaNdk fusion was carried out in E. coli BL21(DE3) C41. Cells were grown in 2YT medium containing 100 lg mL21 ampicillin and overexpression was induced with 1 mM isopropyl-1-thio-b-D-galactopyranoside when cell density reached an A600nm of 0.5–0.9. Cell cultures were incubated overnight at 378C and harvested by centrifugation at 3000g for 15 min at 48C. The cell pellet was resuspended in buffer A (50 mM Tris/HCl, pH 7.4). Bacteria were lysed by sonication and centrifuged at 10,000g for 1 h at 48C. The supernatant was applied to a Q-Sepharose column (Pharmacia) equilibrated with buffer A and eluted with a linear gradient of 0–1M NaCl. The active fraction was precipitated by 80% saturated ammonium sulfate (AS). The precipitate was dissolved in a minimum volume of 60% saturated AS buffer A and applied to a Sepharose 6B column (Pharmacia). Enzymes were eluted with a decreasing linear gradient of AS from 60% to 20% saturation (salting-out chromatography). Active fractions were dialyzed against 100% saturated AS buffer A and the suspension was stored at 48C for future use. For crystallization, protein solutions were desalted on a HiTrap2 column (Sephadex2 G-25 Superfine) equilibrated with buffer A. The activity of Aa-Ndk was measured spectrophotometrically by a coupled assay.13 Reduction was performed at 608C for 1 h in the presence of 10 mM dithio-

2

PROTEINS

threitol (DTT) in 0.1M ammonium bicarbonate buffer. Then alkylation was carried out by adding 200 mM of iodoacetamide followed by incubation in the dark for 2 h. Molecular mass determination of protein using LC-MS

Protein samples were analyzed by online capillary high performance liquid chromatography (HPLC) (LC Packings, Amsterdam, The Netherlands) coupled to a nanospray LCQ Deca XP Ion Trap mass spectrometer (Thermo Finnigan, San Jose, CA). About 100 ng of purified protein were separated onto a 75 lm i.d. 3 5 cm C4 PepMap3002 column (LC Packings). The mass spectrometer operated in positive ion mode at 2 kV needle voltage. Data were acquired in a MS scan survey over the range m/z 400–2000 and deconvoluted using Bioworks 3.3.1 (Thermo Finnigan). Crystallization

Protein samples (e280 5 4470 mol L21 cm21) were concentrated to 10 mg mL21 in the appropriate buffer (50 mM NaCl, 50 mM Tris/HCl pH 7.4). Drops were prepared by vapor diffusion at 208C by mixing 200 nL of protein and reservoir solutions with a Honeybee 961 robot (Cartesian Technology). Initial screening was performed using commercial kits (Index and Crystal Screen from Hampton Research and Structure Screen from Molecular Dimensions). Crystals grew in four different conditions (Form 1: 25% polyethylene glycol (PEG) 3350, 0.2M monohydrated LiSO4, 0.1M Bis-Tris pH 5.5; Form 2: 25% PEG 3350, 0.2M AS, 0.1M Tris pH 8.5; Form 3: 2M magnesium chloride, 0.1M bicine pH 9; Form 4: 25% PEG 3350, 0.2M NaCl, 0.1M Bis-Tris pH 5.5). X-ray diffraction quality crystals of Aa-Ndk were obtained in 3–4 days. Crystals were cryo-protected with the mother liquor containing 20% (v/v) glycerol and flash frozen in liquid nitrogen. Data collection

Diffraction data were collected for crystal Forms 1, 3, and 4 on PROXIMA-1 beamline at SOLEIL synchrotron (St Aubin, France) and for crystal Form 2 on ID14-1 beamline at the European Synchrotron Radiation Facility (Grenoble, France). The crystal Forms 1 and 2 of AaNdk belonged to the orthorhombic space groups C222 (a 5 43.12 A˚; b 5 101.03 A˚; c 5 62.03 A˚) and P21212 (a 5 52.19 A˚; b 5 96.04 A˚; c 5 62.14 A˚) with 1 and 2 molecules per asymmetric unit and a solvent content of 42% and 49.7%, respectively. The crystal Forms 3 and 4 of Aa-Ndk belonged to the hexagonal space groups P61 (a 5 b 5 116.12 A˚; c 5 246.9 A˚) and P6122 (a 5 b 5 200.98 A˚; c 5 246.74 A˚) with 8 and 12 molecules per asymmetric unit, respectively. The crystal lattices of the hexagonal Forms 1 and 3 were related with a ratio of H3 between the two a cell parameters and almost equal c parameters. Thus the packing was similar with a solvent

A. aeolicus Ndk Crystal Structure

Table I X-Ray Data and Refinement Statistics Crystal form PDB id Space group Unit-cell parameters (
)

Molecules per a.u. VM; % solvent X-ray source Wavelength (
) Resolution (
)a High resolution bin (
)a Frames number No of unique reflectionsa Completeness (%)a Multiplicitya I/r(I)a Rmergea Bfact from Wilson plot (
2) Rcrysta Rfree (5% of reflections)a Root-mean-square deviations Bond lengths (
) Bond angles (8) No. of protein atoms No. of solvent molecules No. of heterogen groups Average Bfactor (
2) Protein Waters Ramachandran plot (%) Most favored Additionally allowed Disallowed

Form 1 (pH 5.5)

Form 2 (pH 8.5)

Form 3 (pH 9.0)

3zto C222 a 5 43.13 b 5 101.06 c 5 62.04 1 (chain A) 2.12; 42.0 SOLEIL PROXIMA-1 0.980 15.5–1.47 1.55-1.47

3ztp P21212 a 5 52.18 b 5 95.12 c 5 61.98 2 (chains A and C) 2.44; 49.7 ESRF ID14-1 0.933 23.3–1.37 1.43-1.37

3ztq P61 a 5 b 5 116.36 c 5 247.32

Form 4 (pH 5.5)

8 (chains A to H) 3.77; 67.4 SOLEIL PROXIMA-1 0.980 38.0–2.10 2.21-2.10

23,200 (3313) 99.0 (98.1) 3.0 (2.7) 9.4 (2.4) 0.067 (0.392) 13.08 0.150 (0.264) 0.180 (0.271)

65,505 (8336) 97.8 (86.8) 3.3 (2.1) 7.1 (4.1) 0.060 (0.311) 11.45 0.141 (0.248) 0.159 (0.271)

107,031 (15,851) 97.2 (98.9) 2.2 (2.2) 6.3 (3.4) 0.090 (0.232) 17.58 0.160 (0.219) 0.192 (0.265)

0.008 1.21 1170 185 2 (SO4) in A

0.008 1.22 2457 350 1(SO4) in A & C, 1 glycerol in A

0.008 1.00 9030 1282 none

0.008 1.00 13446 1061 none

0.007 0.97 13422 1092 none

14.3 30.2

17.3 32.4

22.7 34.3

23.8 29.1

26.2 31.9

96.15 2.31 1.54

96.58 1.71 1.71

96.28 2.23 1.49

96.65 1.89 1.46

96.64 1.92 1.44

3ztr

3zts P6122 a 5 b 5 200.71 c 5 246.73

12 (chains A to L) 3.75; 67.2 SOLEIL PROXIMA-1 0.980 52.4–2.30 2.42–2.30 1–65 217–300 117,919 (17,316) 110,849 (16,494) 92.0 (93.4) 86.4 (89.0) 4.2 (4.2) 5.6 (5.4) 8.0 (3.6) 8.5 (3.4) 0.136 (0.353) 0.152 (0.484) 18.39 20.67 0.193 (0.27) 0.190 (0.25) 0.219 (0.37) 0.220 (0.33)

a

Values in parentheses are for the highest resolution shell.

channel of 65 A˚ in diameter along the helicoidal sixfold axis and thus a high solvent content of 67%. Accordingly the two low solvent content crystal forms diffracted to high resolution, 1.47 and 1.37 A˚, whereas the high solvent content crystal forms diffracted to lower resolution, 2.1 and 2.3 A˚. For the crystal Form 4, 300 frames were collected with a total collection time of 5 min. In order to evaluate the radiation damage, two data sets which included the first and the final stages of the collect with about 90% of completeness were separately merged. Xray diffraction data were processed using MOSFLM14 and scaled with SCALA.15 Data collection details and refinement statistics were summarized in Table I. Figures were made using PYMOL16 and surface areas were calculated using PISA.17 Structure solution and refinement

The crystal Form 1 structure of Aa-Ndk was solved by molecular replacement with the PHASER program18 using the coordinates of one subunit of Mx-Ndk2 (47% identity of sequence, PDB id: 1nlk) as search model.

There was a single molecule labeled A present in the asymmetric subunit. The side chains were constructed manually using COOT.19 The three subsequent structures were solved by molecular replacement with PHASER using the refined model of the Aa-Ndk crystal Form 1 (PDB id: 3zto). Restrained refinements were performed with the PHENIX.REFINE program20 using individual atomic displacement parameters. In the two high resolution structures (Forms 1 and 2), hydrogen atoms were positioned geometrically and refined using a riding model. In the Form 2, the asymmetric unit contained two chains labeled A and C to comply with chain labeling of the Forms 3 and 4. In the Form 2, non-crystallographic symmetry (NCS) restraints were applied as three sets: residues 2–41, 42–69, or 70–142 during the first refinement steps then progressively weakened and removed. For the Form 3, NCS restraints were applied as two sets: the Chains A/B/E/F or C/D/G/H. For the Form 4, NCS restraints were applied as three sets: the main chain of all chains, the side chains of Chains A/B/E/F/I/J, or the side chains of Chains C/D/G/H/K/L. For the four structures, PROTEINS

3

F. Boissier et al.

Figure 1 View along the twofold axis of the tetramer assembly of the Form 1 of Aa-Ndk tetramer. The secondary structure elements are labeled. The Kpn-loops are in light colors. Catalytic residues (H120 and E131), the intersubunit Cys133

Cys133 disulfide bridges (yellow star), the sulfate ions (S1 and S2), and its ligand (R107) are drawn in stick representation. For clarity, the labels are drawn on only one chain. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

anisotropic displacement parameters were refined via a two groups translation-liberation-screw (TLS) parameterization identified using the TLSMD server.21 For each chain, one TLS group included the residues 52–57, the other one the residues 2–51 and 58–142. Main and side chains were unambiguously defined in the electron density maps except for the loop 52–57 of some chains. Solvent molecules were automatically added using COOT. The final stereo-chemistry was checked using the MOLPROBITY program.22 A particular attention was brought to the refinement of alternate conformations and individual occupancies. RESULTS Overall structure of the Aa-Ndk subunit

The X-ray structure of the crystal Form 2 was refined at 1.37 A˚ resolution and was the highest resolution of Aa-Ndk. Whatever the crystal form, Aa-Ndk subunits are almost identical and their Ca atoms superimpose with a root mean square deviation (rmsd) less than 0.4 A˚.

4

PROTEINS

Moreover, the overall structure of the Aa-Ndk monomer is similar to other Ndk structures (rmsd of 0.70 A˚ with the Mx-Ndk monomer) (Fig. 1). Three residues (98–100) are specifically inserted in the Kpn-loop sequences of AaNdk (Fig. 2 and Supporting Information Fig. S1) and then the one-turn 310-helix and the polyproline II stretch found in other Ndks become the two-turn a-helix aK. The Kpn-loop exhibits successively a first b-turn (92–94), the a-helix aK (95–101), a second b-turn (102–105), a short a-helix a30 (106–108), and a turn (109–110). One behind the other, the a-helices a3 and a30 form a pseudo continuous helix with a bend of about 338 between them with two intra-main chain hydrogen bonds (Ile89-Ile106 and Ile90-Arg107). The bend angle is intermediate between that found in Ec-Ndk (368) and Mx-Ndk (278).2,3 The helix aK is almost perpendicular to the pseudo helix a3–a30 preventing the formation of a Type II tetramer. The location of the essential Arg107 is conserved because it is tightly hydrogen bonded between the Kpn-loop (Gly91) and the strand b4 (Asn117, Ile119). The two Ramachandran plot outliers Asp113 and Ala118 have well defined electron density. Asp113 is in an extended conformation stabilized by three hydrogen bonds (N:Asp113 with O:Lys116; Asp113 carboxylate with N:Gly115 and N:Lys116) and Ala118 forms a g-turn stabilized by two hydrogen bonds (O:Ala118 with N:Lys11; Asp13 carboxylate with N:Ala118). Among all chains in the four structures, the largest deviations found in the crystal Form 3 in the loop (52– 59) linking the helices aA and a2 of the Chain C are due to crystal packing. This loop highly solvent-exposed is not well defined in the electron density maps, like frequently observed in nucleotide-free Ndk structures. The flexibility of this region is required for nucleotide binding.24 At the opposite side of the binding pocket, like in the tetramer Mx-Ndk,2 the Kpn-loop is also solvent accessible (Fig. 1 and Supporting Information Fig. S1) whereas, in hexameric Ndks and tetrameric Ec-Ndk, it is buried at the secondary interface. Whatever the case, the Kpn-loop participates to the binding site. Sulfate ions binding in the catalytic site

In the active site of the Form 1 crystallized with lithium sulfate as precipitant, two sulfate ions are clearly visible in the difference electron density maps (Fig. 3). The two sulfate ions are 6 A˚ apart and bound through one water molecule. The most buried site (Site 1) is fully occupied by sulfate S1 that is directly bound to Nd1His120, Nf-Lys11, and OH-Tyr51 and through water molecules to Nd1-His54, Nd2-Asn117, and N-Ala121. The less buried site (Site 2) is partially occupied (0.63) by sulfate S2 that is directly bound to the Arg107 guanidinium and Thr93 hydroxyl groups and through water molecules to Ala121 backbone carbonyl oxygen and amide nitrogen and Arg87 guanidinium.

A. aeolicus Ndk Crystal Structure

Figure 2 Amino acids sequence alignment of Aa-Ndk with some Ndks of known 3D structure. The alignment includes Ndks from E. coli (UniProt: P0A763; PDB id: 2hur), Halomonas sp. no. 593 (UniProt: Q83WH5; PDB id: 3vgs), M. xanthus (UniProt: P15266; PDB id: 2nck), A. aeolicus (UniProt: O67528; PDB id: 3zto), T. thermophilus (UniProt: Q5SLV5; PDB id: 1wkl), Pyrococcus horikoshii OT3 (UniProt: O58429; PDB id: 2cwk), Pyrobaculum aerophilum (UniProt: Q8ZWY4; PDB id: 1xqi), and D. discoideum (UniProt: P22887; PDB id: 1npk). The alignment was carried out using the CLUSTALW server and the figure was generated using the ESPRIPT server.23 The residues involved in the various interfaces are shaded in different colors: cyan for the common-dimer interface, dark and light violet for the two trimer interfaces, green for Type I tetramer interface (Mxand Aa-Ndk), and brown for Type II tetramer interface (Ec-Ndk). The putative protein ligands of a-, b-, and g-phosphate of ATP are indicated by green, red, and blue arrows, respectively. The cyan star marks Ile28 in Aa-NDK instead of usual Glu28.

Superimposition of Aa-Ndk active site with the one of Dd-Ndk in complex with ADP and AlF325 shows that the less buried sulfate ions (S2) is close to the b-phosphate group of ADP and the most buried sulfate ions (S1) is close to AlF3 that mimics the g-phosphate group in the transition state of protein phosphorylation (Fig. 3). Moreover, the His54 and Arg57 side chains are putative ligands of the a-phosphate group of nucleotide whereas the Phe59 aromatic ring is able to stack with the nucleobase. The Form 2 was crystallized with AS as precipitant. The Site 1 is partially occupied (0.60) by a sulfate whereas the Site 2 is occupied by a water molecule and Thr93 and Arg87 side chains adopt alternate conformations. Surprisingly, the Phe59 protrudes out of the protein with well defined density. In the Chain A, a glycerol molecule nestled between the helices aA and a4, make six hydrogen bonds with Phe50, Val53, His54, Glu131 and two water molecules. Since the aA–a2 loop is not involved in crystal contacts, the helices aA and a2 can move away from a4. Conversely, in the Chain C, the aA– a2 loop interacts with the neighboring molecule (430 A˚2 of buried surface area (bsa)) pushing the helices aA and a2 to move closer to the helix a4 with a maximum shift of 2.4 A˚ compared with backbone of Chain A. The interactions between aA and a4 are enhanced by an edge-toface aromatic interaction between Tyr52 and Tyr130.

In the hexagonal Forms 3 and 4, though there is a water molecule in each site, the location of helices aA and a2 and the active site are similar to that of the Form 1. The common-dimer unit of Aa-Ndk

Most Ndks are hexameric but tetrameric forms exist in some bacteria. Though their overall quaternary structures are different, Ndks contain a common-dimer unit. Indeed, such a dimer was found in Aa-Ndk. At the common-dimer interface, anti-parallel interactions of adjacent strands b2 and helices a1 extend the central b-sheet and the hydrophobic core, respectively (Fig. 1). Noticeably, the glutamate residue of helix a1 (at position 28 in AaNdk numbering), which makes two hydrogen bonds with main chain amino groups of strand b2 from the facing monomer in known Ndk structures, is replaced by isoleucine in Aa-Ndk (Fig. 2). Ile28 enhances the overall hydrophobicity of the protein core and of the commondimer interface. Across this interface, there is one hydrogen bond (Leu37 with Met39) and one salt bridge (Lys22 with Asp25). Overall, the Aa-Ndk common dimer is very similar to other Ndk common dimers. For instance, rmsd were lower than 0.8 A˚ after superimposition of 253 Ca atoms of Mx-Ndk dimer (90% of the total number of Ca in the dimers). The bsa of the common-dimer PROTEINS

5

F. Boissier et al.

Figure 3 2mFo 2 DFc electron density map at 1.47 A˚ resolution contoured at 1.5 r around the active site of the Aa-Ndk crystal Form 1. Hydrogen bonds involving the catalytic residues, the most (S1) and the less (S2) buried sulfate ions, and water molecules are drawn. For clarity, the Arg87 guanidinium group was omitted. The active sites residues of Aa-Ndk and Dd-Ndk:AlF3:ADP were superimposed. ADP and AlF3 molecules were shown in thin stick drawings (PDB id: 1kdn25). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

interface in Aa-Ndk (745 A˚2) is smaller than in Mx-Ndk and Ec-Ndk tetramers (1070–1200 A˚2) because the C-terminus region that contributes to this interface is shorter of at least four residues in Aa-Ndk (Fig. 2). It is in the same range as in hexamer Ndks (665–750 A˚2) (Table II). The interface in Ph-Ndk is greater (1045 A˚2) because of the presence of a C-terminal extension (Fig. 2). The small interfaces of the common dimer in hexameric Ndks are counterbalanced by oligomeric assembly, so the stability of the Aa-Ndk common dimer could be questioned. The secondary interface in Aa-Ndk is covalently linked

The common dimers assemble through secondary interfaces to form tetramers of Type I and Type II or hexamers. In the four crystal forms, common dimers of Aa-Ndk interact through a secondary interface of Type I

6

PROTEINS

(Figs. 1 and 4). This interface involves a single hydrophobic patch made up of the strand b2, the helix aA (residues 38–46), and essentially the C-terminus region (residues 129–142) with an intersubunit Cys133

Cys133 disulfide bridge. Though the bsa of the secondary interface is as small (520 A˚2) as those of Mx- and Ec-Ndk (480 and 450 A˚2), its calculated DiG of 230 kJ mol21 is more negative than those of Mx- and Ec-Ndk (210/219 kJ mol21), especially as this value does not include the contribution of the disulfide bridge (Table II). The interface is further stabilized by three main chain–main chain (Phe134 with Leu139; Phe134 with Gly138; Cys133 with Gly138) and two main chain–side chain (Gly138 and Ile141 with Lys46) intermolecular hydrogen bonds. The crosslink has no significant effect on the secondary-dimer geometry since rmsd between Aa-Ndk and Mx-Ndk was 1.28 A˚ after superimposition of 88% (248 Ca atoms) of the total number of Ca in secondary-dimers. Moreover, the bulge produced by the helix aK over the Kpn-loop definitely prevents hexamer assembly. The electron densities in the vicinity of the disulfide bridge are clearly well defined (Fig. 5). Cys133 is in the a-helix a4. Two characteristic parameters for the disulfide bond are the dihedral angle around the S

S bond (v3) and the Ca

Ca distance. In proteins, there are equal numbers of left-handed (v3 5 2908, dCa

Ca 5 6.1 A˚) and right-handed disulfides (v3 5 1908, dCa

Ca 5 5.2 A˚).26 The average v3 value of the Cys133

Cys133 disulfide bond is 2938 and dCa

Ca is 5.65 A˚, showing that it belongs to the left-handed spiral conformation with a 2.02 A˚ average S

S bond length. Each Cys133 side chain adopts two alternate conformations, g and t, with a refined occupancy close to 0.5 for the Forms 1 and 3 and in the range 0.39–0.64 for the Form 2. A crystallographic twofold (Form 1) or a pseudo twofold (Forms 2, 3, and 4) axis lies perpendicular to the plane defined by the two alternate disulfide bridges and goes between the two S

S bonds. So, in the Aa-Ndk Form 2, the side chain Cys133(g) of Chain A makes a disulfide bridge with the neighboring Cys133(t) of Chain C and symmetrically Cys133(g) of Chain A with Cys133(t) of Chain C [Fig. 5(A)]. Therefore, the disulfide bridges are in the conformations ggt and tgg. Such conformations are 2.5 kJ mol21 less stable than the ggg conformation.27 The presence of the intersubunit disulfide bond was confirmed in LC-MS under native conditions after dissolution of crystals which had not been used for diffraction experiments. The deconvoluted spectrum shows three major mass peaks at 31618.1 Da, 31651.3 Da, and 31683.1 Da and several minor peaks with increments of a multiple of 16 Da [Fig. 6(A,B)]. The theoretical masses of the monomer protein with and without N-terminal methionine were 15942.3 Da and 15811.1 Da, respectively. The covalent dimer with a disulfide bridge between two monomers, which both underwent the loss

A. aeolicus Ndk Crystal Structure

Table II

Accessible (asa) and buried (bsa) surface area interfaces (in A˚2) and some factors known to contribute to protein thermostability calculated for DdNdk, Dictyostelium discoideum (PDB id: 1npk); Ph-Ndk, Pyrococcus horikoshii OT3 (PDB id: 2cwk); Tt-Ndk, Thermus thermophilus HB8 (PDB id: 1wkj); Pa-Ndk, Pyrobaculum aerophilum (PDB id: 1xqi); Aa-Ndk, Aquifex aeolicus (PDB id: 3zto); Mx-Ndk, Myxococcus xanthus (PDB id: 2nck); Ec-Ndk, Escherichia coli (PDB id: 2hur) Dd-Ndk

Ph-Ndk

Tt-Ndk

Pa-Ndk

Aa-Ndk

8030 665

8982 1045

7330 710

9620 750

7560 745

7560 1070

7500 1200

236 720*2

285 965*2

259 478*2

243 580*2

255 520

263 450

256 480

226*2 35,470 12,720 0.34 0.44 6 1.22 572

244*2 35,330 10,630 0.57 0.40 1 1.28 616

29*2 32,290 15,890 0.47 0.68 1 1.17 594

212*2 45,725 13,820 0.40 0.48 4 1.19 802

230 24,780 7470 0.45 0.47 2 1.11 740

210 23,650 7140 0.31 0.39 5 1.18 580

219 23,490 7080 0.40 0.64 7 1.17 446

Oligomeric state Monomer asa (
2) bsa (
2) Common-dimer DiG (kJ/mol) bsa (
2) Secondary interface DiG (kJ/mol) Oligomer asa (
2) Oligomer bsa (
2) Salt-bridge per residue Arg / (Arg1Lys) Deamidation sites RSA H-bond

Hexamer

Mx-Ndk Tetramer I

Ec-Ndk Tetramer II

Ndk boxes from thermophile organisms are gray. RSA 5 asa/15.0 N0.866 where N is the total number of protein atoms. DiG indicates the solvation free energy gain upon formation of the interface, in kJ/mol. Negative DiG corresponds to hydrophobic interfaces or positive protein affinity. The value does not include the contribution of the disulfide bridge of about 16.7 kJ/mol.

Figure 4 Views of the crystal packing that show the presence of similar tetramers in the four crystal forms of Aa-Ndk. (A) C222, (B) P21212, (C) P61, and (D) P6122. The equivalent chains of each tetramer are similarly colored with the Kpn-loops in light colors. (cdi, common-dimer interface; si, secondary interface). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] PROTEINS

7

F. Boissier et al.

Figure 5 Electron density map around the disulfide bridge contoured at 1.5 r at 1.37 A˚ resolution in P21212 (A) and at 2.3 A˚ resolution in P6122 using the starting (B) and the ending data (C) (Table I). (A) The Chain A and the (2x,y, 2z) symmetry related A# are labeled and differently colored. The ˚ twofold axis is horizontal. (A and B) Half of the Cys133 in conformation g of one subunit forms a disulfide bridge (dS

S 5 2.02 A) with half of the Cys133 in conformation t of the adjacent subunit and symmetrically. (C) The two alternate intersubunit Cys133

Cys133 disulfide bridges were ˚ clearly broken (dS

S 5 3.2 A) during data collection. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

of their N-terminal methionine, will have a theoretical mass of 31,620.3 Da. Thus, the 31,618.1 Da peak corresponded to such a covalent-dimer while the other peaks were likely identical dimers but with oxidized methionine residues. After reduction and alkylation of samples, two mass peaks of 15,807.9 Da and 15,840.8 Da were observed [Fig. 6(C,D)]. The former corresponded to monomer proteins without N-terminal methionine, the latter to equivalent monomers with two oxidized methionine residues. There were clearly no peaks corresponding to dimers. Note that peaks from alkylated cysteine proteins were not detected. Aa is a hyperthermophilic bacterium that grows in extreme conditions (85–958C). Accordingly, kinase activity was about 400 lmol min21 mL21 from 608C up to 858C then decreased to 285 lmol min21 mL21 at 958C. Moreover, reduction/alkylation of the enzyme preparation at 608C results in a 40% decrease of the initial activity. Together these data show that under native conditions, the disulfide bridge stabilizes the tetramer assembly. It is well known that disulfide bond breakage occurs during data collection by X-ray irradiation or mediated by free radicals generated by the X-rays within the protein or the solvent.28 In the hexagonal crystal Form 4, gradual Fourier maps showing the time course of structural changes in the Cys133

Cys133 disulfide bond were inspected. Comparison of the maps calculated using the beginning [Fig. 5(B)] and the ending [Fig. 5(C)] data set clearly shows that the disulfide bridge was broken.

8

PROTEINS

Refined occupancies of sulfur atoms different from 0.5 mean that the disulfide bridges were broken. At the end, the Cys133 side chains only adopt the most stable conformer (g). Furthermore, we suspect that similarly X-rays induce breakage of the disulfide bridge in the three other crystal forms. Aa-Ndk forms a tetramer similar to Mx-Ndk

In the four crystal forms of Aa-Ndk, Type I tetramers were found (Figs. 1 and 4). It is the second example of Type I tetramers. The small interface (330 A˚2) of Type II only found in the crystals with orthorhombic space groups (Forms 1 and 2) is thus a crystal contact [Fig. 4(A,B)]. The Aa-Ndk tetramers are each similar to the Mx-Ndk tetramer. The rmsds were lower than 1.8 A˚ after superimposition of 94% (528 Ca atoms) of the total number of Ca in the tetramers. DISCUSSION Thermostability of the Aa-Ndk monomer

Protein thermostability is related to a set of structural features.29,30 Compared to Ndks from mesophile organisms (Table II), Aa-Ndk has a small number of asparagine residues and a higher number of hydrogen bonds. Hydrogen bonding increases packing density in the interior of the protein. Although a high number of ion pairs

A. aeolicus Ndk Crystal Structure

ity is related to oligomerization.34,35 For example, intersubunit disulfide interactions play a critical role in thermostability of the glucose-6-phosphate dehydrogenase from A. aeolicus36 and in a few thermophilic heterooligomer complexes.32 The intersubunit Cys187

Cys187 disulfide bridge introduced in the malate dehydrogenase at the dimer–dimer interface made the enzyme more resistant to thermal denaturation.37 Similarly, it could be expected that the intersubunit Cys133

Cys133 bond in Aa-Ndk prevents the dissociation of the tetramer. The disulfide bridge appears to be a specific marker of thermostability.38 NDK of Aa have been produced as recombinant proteins in E. coli. Since the reducing environment of E. coli cytoplasm probably does not allow the formation of disulfide bridges, it is very likely that the disulfide bridge Cys133

Cys133 was formed during purification and crystallization steps. As evidenced by a phylogenetic analysis, a specific protein disulfide oxidoreductase could be efficient in forming and maintaining intracellular disulfide bonds in Aa.38 Among all the known Ndk sequences, none have a cysteine at the equivalent position

Figure 6 Deconvoluted LC-MS spectra of protein from dissolved crystals. Representative spectrum (A) and zoom of the peaks (B) corresponding to covalently bonded dimer. Representative spectrum (C) and zoom of the peaks (D) after reduction/alkylation of the disulfide bridge corresponding to monomer.

is known to stabilize the native structure at high temperature resulting from reduction in desolvation penalty,31 the ratio Arg/(Arg1Lys) is in the range of other Ndks and does not appear as a determining factor. Finally, relative surface area shows smaller values according to what is observed in other thermophile enzymes. All these values are in the same ranges as those found for Ndk structures from thermophiles. Thermophilic organisms have utilized a variety of molecular mechanisms to increase the thermostability of proteins or complexes, among which intrasubunit and intersubunit disulfide bridges hold a potentially important role.32,33 Figure 7 The covalent secondary-dimer stabilizes the Aa-Ndk tetramer

Genome analysis of hyperthermophilic archaea and bacteria suggests that disulfide bonds are likely a result of selective pressure for thermostable proteins and likely represent a strategy for adaptation to high temperature.29,32for review For some enzymes, a high thermal stabil-

Diversity of the assembly mode in the Ndk family, from the monomer to the common-unit dimer, tetramers, and hexamer. Equilibrium between the common dimer and Type II tetramer observed with Ec-Ndk is shown.8 Putative pathways via dimeric Aa-Ndk and trimeric Ba-Ndk9 lacking experimental data are drawn with dotted lines. The interfaces are colored in cyan for the common-dimer interface, blue and violet for the trimer interfaces, green for Type I tetramer interface, and brown for Type II tetramer interface (as defined on Fig. 2). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] PROTEINS

9

F. Boissier et al.

133. Even in the most homologous Ndks from thermophiles, it was replaced by an alanine (Fig. 2). The protein was 40% less active after reduction and alkylation showing the importance of the disulfide bridge on tetramer assembly and protein activity. The tetrameric structure is necessary for full enzymatic activity. Oligomerization artifacts were suggested to explain phosphotransfer from catalytic histidine to serine at position 12 in S. typhimurium Ndk assuming a Type II tetramer.39 Whatever the tetramer type, a direct intermolecular event between the two active site residues at the bottom of the binding pocket seems unlikely. Halomonas sp. no. 593 Ndk is a dimer6,7 because Glu134 (Ser137 in Aquifex) results in steric hindrance and charge repulsion preventing the tetramerization (Fig. 7). However, the mutation to a small and hydrophobic residue, E134A but not E134G, forms a hydrophobic cluster that allows the formation of a Type I tetrameric structure.6,40 Ala130 in Halomonas-Ndk, which is homologous to Cys133 in Aa-Ndk, was suspected by molecular modeling to be another key position of the possible tetramerization. Similarly, the formation of the Cys133

Cys133 bridge in Aa-NDK contributes to the stabilization of dimer unit association to form tetramer inside Aa cells. CONCLUSION Ndk family has a very similar monomer structure and also shares a common-dimer unit. However their quaternary structures exhibit an amazing diversity. The new example of Type I tetramer structure which has been solved here, differs from other Ndk structures by the presence of an inter subunit disulfide bridge. Current structural data suggest the following schema for Ndk assembly (Fig. 7). In the case of hexamer, the formation occurs generally through trimer of dimers1 and in one case of dimer of trimers.9 Given the larger surface area (bsa > 1000 A˚2) of the common-dimer interface of Mx- and Ec-Ndks than in Dd-Ndk (bsa 5 665 A˚2) (Table II), the formation of common dimers should be favored. The secondary interface in Aa-Ndk, Mx-Ndk, and Ec-Ndk has similar area and, in the former, is stabilized by a disulfide bridge that contributes to the stability of the tetramer. In any case, Ndk needs to form a common dimer to achieve its maximum activity. The hexameric and tetrameric assemblies do not have significantly different catalytic rates (kcat 300– 400 sec21)41 but different thermodynamic stability. Hence, Ndks have found several ways for reaching its stability, by varying its oligomerization degree, by adjusting the length of its C-terminal extension, or by using a disulfide bridge. ACKNOWLEDGMENTS Authors acknowledge Dr. Didier Thoraval who cloned the Aa-Ndk gene. AD thank Pr. Masao Tokunaga for providing unpublished results on Ndks from halophile organ-

10

PROTEINS

isms. The authors wish to thank the beam line staffs of SOLEIL (Saint-Aubin, France) and ESRF (Grenoble, France) for their assistance in X-ray diffraction data collection. REFERENCES 1. Janin J, Dumas C, More´ra S, Xu Y, Meyer P, Chiadmi M, Cherfils J. Three-dimensional structure of nucleoside diphosphate kinase. J Bioenerg Biomembr 2000;32:215–225. 2. Williams RL, Oren DA, Mun˜oz-Dorado J, Inouye S, Inouye M, Arnold E. Crystal structure of Myxococcus xanthus nucleoside diphosphate kinase and its interaction with a nucleotide substrate at 2.0 A resolution. J Mol Biol 1993;234:1230–1247. 3. Moynie´ L, Giraud M-F, Georgescauld F, Lascu I, Dautant A. The structure of the Escherichia coli nucleoside diphosphate kinase reveals a new quaternary architecture for this enzyme family. Proteins 2007;67:755–765. 4. Tepper AD, Dammann H, Bominaar AA, Veron M. Investigation of the active site and the conformational stability of nucleoside diphosphate kinase by site-directed mutagenesis. J Biol Chem 1994;269:32175–32180. 5. More´ra S, LeBras G, Lascu I, Lacombe ML, Ve´ron M, Janin J. Refined X-ray structure of Dictyostelium discoideum nucleoside diphosphate kinase at 1.8 A˚ resolution. J Mol Biol 1994;243:873–890. 6. Tokunaga H, Ishibashi M, Arisaka F, Arai S, Kuroki ,R, Arakawa T, Tokunaga M. Residue 134 determines the dimer-tetramer assembly of nucleoside diphosphate kinase from moderately halophilic bacteria. FEBS Lett 2008;582:1049–1054. 7. Arai S, Yonezawa Y, Okazaki N, Matsumoto F, Tamada T, Tokunaga H, Ishibashi M, Blaber M, Tokunaga M, Kuroki R. A structural mechanism for dimeric to tetrameric oligomer conversion in halomonas sp. nucleoside diphosphate kinase. Protein Sci 2012; doi: 10.1002/pro.2032. 8. Shen R, Wheeler LJ, Mathews CK. Molecular interactions involving Escherichia coli nucleoside diphosphate kinase. J Bioenerg Biomembr 2006;38:255–259. 9. Misra G, Aggarwal A, Dube D, Zaman MS, Singh Y, Ramachandran R. Crystal structure of the Bacillus anthracis nucleoside diphosphate kinase and its characterization reveals an enzyme adapted to perform under stress conditions. Proteins 2009;76:496–506. 10. Takeishi S, Nakagawa N, Maoka N, Kihara M, Moriguchi M, Masui R, Kuramitsi S. Crystallization and preliminary X-ray diffraction studies of nucleoside diphosphate kinase from Thermus thermophilus HB8. Acta Crystallogr D 2003;59:1843–1845. 11. Kato-Murayama M, Murayama K, Shirouzu M, Yokoyama S. Crystal structure of nucleotide diphosphate kinase from Pyrococcus horikoshii, doi:10.2210/pdb2cwk/pdb. 12. Pe´delacq J, Waldo GS, Cabantous S, Liong EC, Terwilliger TC. Structural and functional features of an NDP kinase from the hyperthermophile crenarchaeon Pyrobaculum aerophilum. Protein Sci 2005;14:2562–2573. 13. Lascu I, Kezdi M, Goia I, Jebeleanu G, Baˆrzu O, Pansini A, Papa S, Mantsch HH. Enzymatic properties of 8-bromoadenine nucleotides. Biochemistry 1979;18:4818–4826. 14. Leslie AGW. Recent changes to the MOSFLM package for processing film and image plate data. Joint CCP4 1 ESF-EAMCB Newsletter on Protein Crystallography 1992;26. 15. CCP4. Collaborative Computational Project, Number 4. Acta Crystallogr D 1994;50:760–763. 16. DeLano WL. The PyMOL molecular graphics system. CA: DeLano Scientific; 2002. 17. Krissinel E. Crystal contacts as nature’s docking solutions. J Comput Chem 2009;31:133–143. 18. McCoy AJ, Grosse-Kunstleve RW, Adams PD, Winn MD, Storoni LC, Read RJ. Phaser crystallographic software. J Appl Crystallogr 2007;40:658–674.

A. aeolicus Ndk Crystal Structure

19. Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Cryst D 2004;60:2126–2132. 20. Adams PD, Gopal K, Grosse-Kunstleve RW, Hung LW, Ioerger TR, McCoy AJ, Moriarty NW, Pai RK, Read RJ, Romo TD, Sacchettini JC, Sauter NK, Storoni LC, Terwilliger TC. Recent developments in the PHENIX software for automated crystallographic structure determination. J Synchrotron Radiat 2004;11:53–55. 21. Painter J, Merritt EA. TLSMD web server for the generation of multi-group TLS models. J Appl Crystallogr 2006;39:109–111. 22. Chen VB, Arendall WB, Headd JJ, Keedy DA, Immormino RM, Kapral GJ, Murray LW, Richardson JS, Richardson DC. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Cryst D 2010;66:12–21. 23. Gouet P, Courcelle E, Stuart DI, Metoz F. ESPript: multiple sequence alignments in PostScript. Bioinformatics 1999;15:305–308. 24. Chen Y, Gallois-Montbrun S, Schneider B, Ve´ron M, More´ra S, Deville-Bonne D, Janin J. Nucleotide binding to nucleoside diphosphate kinases: X-ray structure of human NDPK-A in complex with ADP and comparison to protein kinases. J Mol Biol 2003;332:915–926. 25. Xu YW, More´ra S, Janin J, Cherfils J. AlF3 mimics the transition state of protein phosphorylation in the crystal structure of nucleoside diphosphate kinase and MgADP. Proc Natl Acad Sci USA 1997;94:3579–3583. 26. Richardson JS. The anatomy and taxonomy of protein structure. Adv Protein Chem 1981;34:167–339. 27. Haworth NL, Gready JE, George RA, Wouters MA. Evaluating the stability of disulfide bridges in proteins: a torsional potential energy surface for diethyl disulfide. Mol Simul 2007;33:475–485. 28. Weik M, Ravelli RB, Kryger G, McSweeney S, Raves ML, Harel M, Gros P, Silman I, Kroon J, Sussman JL. Specific chemical and structural damage to proteins produced by synchrotron radiation. Proc Natl Acad Sci USA 1999;97:623–628. 29. Karshikoff A, Ladenstein R. Ion pairs and the thermotolerance of proteins from hyperthermophiles: a ‘traffic rule’ for hot roads. Trends Biol Sci 2001;26:550–556. 30. Yano JK, Poulos TL. New understandings of thermostable and peizostable enzymes. Curr Opin Biotechnol 2003;14:360–365.

31. Ladenstein R, Ren B. Reconsideration of an early dogma, saying ‘‘there is no evidence for disulfide bonds in proteins from archaea’’. Extremophiles 2008;12:29–38. 32. Mallick P, Boutz DR, Eisenberg D, Yeates TO. Genomic evidence that the intracellular proteins of archaeal microbes contain disulfide bonds. Proc Natl Acad Sci USA 2002;99:9679–9684. 33. Boutz DR, Cascio D, Whitelegge J, Perry LJ, Yeates TO. Discovery of a thermophilic protein complex stabilized by topologically interlinked chains. J Mol Biol 2007;368:1332–1344. 34. Vieille C, Zeikus GJ. Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability. Microbiol Mol Biol Rev 2001;65:1–43. 35. Walden H, Bell GS, Russell RJM, Siebersm B, Hensel R, Taylor GL. Tiny TIM: a small tetrameric, hyperthermostable triosephosphate isomerase. J Mol Biol 2001;306:745–757. 36. Nakka M, Iyer RB, Bachas LG. Intersubunit disulfide interactions play a critical role in maintaining the thermostability of glucose-6phosphate dehydrogenase from the hyperthermophilic bacterium Aquifex aeolicus. Protein J 2006;25:17–21. 37. Bjørk A, Dalhus B, Mantzilas D, Eijsink VGH, Sireva˚g R. Stabilization of a tetrameric malate dehydrogenase by introduction of a disulfide bridge at the dimer-dimer interface. J Mol Biol 2003;334:811–821. 38. Beeby M, O’Connor BD, Ryttersgaard C, Boutz DR, Perry LJ, Yeates TO. The genomics of disulfide bonding and protein stabilization in thermophiles, PLoS Biol 2005;3:e309. 39. Dar HH, Chakraborti PK. Intermolecular phosphotransfer is crucial for efficient catalytic activity of nucleoside diphosphate kinase. Biochem J 2010;430:539–549. 40. Tokunaga H, Izutsu K-I, Arai S, Yonezawa Y, Kuroki R, Arakawa T, Tokunaga M. Dimer-tetramer assembly of nucleoside diphosphate kinase from moderately halophilic bacterium Chromohalobacter salexigens DSM3043: Both residues 134 and 136 are critical for the tetramer assembly. Enzyme Microb Technol 2010;46:129–135. 41. Gonin P, Xu Y, Milon L, Dabernat S, Morr M, Kumar R, Lacombe M-L, Janin J, Lascu I. Catalytic mechanism of nucleoside diphosphate kinase investigated using nucleotide analogues, viscosity effects, and X-ray crystallography. Biochemistry 1999;38:7265–7272.

PROTEINS

11