Crystal structure of carbonic anhydrase from Neisseria gonorrhoeae and its complex with the inhibitor acetazolamide1

Article No. mb982077

J. Mol. Biol. (1998) 283, 301±310

Crystal Structure of Carbonic Anhydrase from Neisseria gonorrhoeae and its Complex with the Inhibitor Acetazolamide Shenghua Huang1,2, Yafeng Xue3, Elisabeth Sauer-Eriksson2 Laura Chirica1, Sven Lindskog1* and Bengt-Harald Jonsson1 1

Department of Biochemistry UmeaÊ University S-90187 UmeaÊ, Sweden 2

UmeaÊ Center for Molecular Pathogenesis, UmeaÊ University S-90187 UmeaÊ, Sweden 3

Department of Biochemistry and Biophysics, GoÈteborg University, S-41390 GoÈteborg Sweden

The crystal structure of carbonic anhydrase from Neisseria gonorrhoeae has Ê by molecular replacement using been solved to a resolution of 1.78 A human carbonic anhydrase II as a template. After re®nement the R factor was 17.8% (Rfree ˆ 23.2%). There are two molecules per asymmetric unit (space group P21), but they have essentially identical structures. The fold of the N. gonorrhoeae enzyme is very similar to that of human isozyme II; 192 residues, 74 of which are identical in the two enzymes, have equivalent positions in the three-dimensional structures. This corresponds to 85% of the entire polypeptide chain of the bacterial enzyme. The only two cysteine residues in the bacterial enzyme, which has a periplasmic location in the cell, are connected by a disul®de bond. Most of the secondary structure elements present in human isozyme II are retained in N. gonorrhoeae carbonic anhydrase, but there are also differences, particularly in the few helical regions. Long deletions in the bacterial enzyme relative to human isozyme II have resulted in a considerable shortening of three surface loops. One of these deletions, corresponding to residues 128 to 139 in the human enzyme, leads to a widening of the entrance to the hydrophobic part of the active site cavity. Practically all the amino acid residues in the active site of human isozyme II are conserved in the N. gonorrhoeae enzyme and have similar structural positions. However, the imidazole ring of a histidine residue, which has been shown to function as a proton shuttle in the catalytic mechanism of the human enzyme, interacts with an extraneous entity, which has tentatively been identi®ed as a 2-mercaptoethanol molecule from the crystallization medium. When this entity is removed by soaking the crystal in a different medium, the side-chain of His66 becomes quite mobile. The structure of a complex with the sulfonamide inhibitor, acetazolamide, has also been determined. Its position in the active site is very similar to that observed in human carbonic anhydrase II. # 1998 Academic Press

*Corresponding author

Keywords: carbonic anhydrase; Neisseria gonorrhoeae; crystal structure; molecular replacement; acetazolamide

Introduction

Abbreviations used: B factor, crystallographic temperature factor; HCA, human carbonic anhydrase; NGCA, Neisseria gonorrhoeae carbonic anhydrase; MR, molecular replacement; NCS, non-crystallographic symmetry; rmsd, root-mean-square deviation; SA, simulated annealing. E-mail address of the corresponding author: [email protected] 0022±2836/98/410301±10 $30.00/0

Carbonic anhydrase, which catalyzes the interconversion between carbon dioxide and bicarbonate ion, is a ubiquitous enzyme present in animals, plants, algae and some bacteria (Lindskog, 1997). Three evolutionarily unrelated classes of the enzyme are known and have been designated a, b, and g. All of them have been found in prokaryotic organisms (Hewett-Emmett & Tashian, 1996). However, the three-dimensional structure of only one prokaryotic carbonic anhydrase has been pre# 1998 Academic Press

302

Neisseria gonorrhoeae Carbonic Anhydrase

sented, namely that of the enzyme from the archaeon Methanosarcina thermophila. This is a trimeric gcarbonic anhydrase with a prominent left-handed b-helical structure motif (Kisker et al., 1996). Several bacteria, such as Escherichia coli (Guilloton et al., 1992) and the cyanobacterium Synechococcus (Fukuzawa et al., 1992), produce carbonic anhydrases of the b type under certain conditions, but the three-dimensional structures are unknown. The a-carbonic anhydrases were thought to belong exclusively to the eukaryotic world, but it was recently demonstrated that the enzyme from Neisseria gonorrhoeae (NGCA) is of the a type and, thus, homologous to carbonic anhydrases from animal sources (Chirica et al., 1997; Hewett-Emmett & Tashian, 1996). In addition, a-carbonic anhydrases have been discovered in the cyanobacteria Anabaena and Synechococcus (Soltes-Rak et al., 1997). The carbonic anhydrase gene from N. gonorrhoeae encodes a 252-residue polypeptide, but a 26-residue signal peptide is cleaved off when the enzyme is expressed in E. coli, and most of the enzyme appears as a soluble, 226-residue protein in the periplasmic space (Chirica et al., 1997). It is a highactivity carbonic anhydrase with a CO2 hydration turnover number of 1  106 sÿ1 at pH 9 and 25 C. A comparison between the amino acid sequences of the N. gonorrhoeae enzyme (NGCA) and human

carbonic anhydrase (HCA) II, a 259-residue protein, suggests that most of the secondary structure has been conserved despite the long evolutionary distance between the two species, while a number of loops must be considerably shorter in the bacterial form than in HCA II. A more detailed structural comparison at the three-dimensional level is made here where the crystal structure of NGCA is presented.

Results and Discussion Quality of the model Table 1 gives the ®nal re®nement statistics and stereochemical parameters for the model of NGCA. The crystals have two molecules, A and B, in the asymmetric unit. Superposition of these two molecules resulted in a root-mean-square deviation Ê between corresponding Ca atoms (rmsd) of 0.194 A (residues 4 to 226). Thus, there are no signi®cant differences between the structures of molecules A and B. The N-terminal residues 1 to 3 were not observed in the electron density maps, suggesting that they are disordered in the structure. These residues are, therefore, not included in the ®nal model. The electron densities of some polar residues on the molecular surface are weak or unob-

Table 1. Data collection and re®nement statistics Detector Generator Ê) Wavelength (A Monochromator Conditions Temperature Distance (mm) Step of rotation Range of rotation ( ) No. of observations No. of unique reflections Ê) Range of resolution (A Redundancy (>4) (%) (>2) (%) Completeness (%) Rmergea Space group Cell parameters: Ê) a (A Ê) b (A Ê) c (A b ( ) Refinement: No. of reflections (2s) R factorb Rfreec Ê) rmsd: bonds (A Angles ( ) Dihedrals ( ) Impropers ( ) No. of water molecules

NGCA

NGCA(S)

NGAZ

Rigaku Raxis II Rigaku RU200 1.5418 Graphite 50 kV, 100 mA 21 C 90 2 /20 min 0 ±180 159,921 43,127 40 ±1.78 71.4 94.3 95.3 0.033 P21

DIP2030H Nonius FR591 1.5418 Double-mirror 45 kV, 95 mA 20 C 100 2 /25 min 0±98 65,932 28,517 35±1.90 16.3 83.0 83.3 0.072 P21

DIP2030H Nonius FR591 1.5418 Double-mirror 49 kV, 85 mA 100 K 100 1 /15 min 0±95 72,529 30,261 30± 1.90 13.2 75.0 91.9 0.045 P21

47.3 74.4 62.5 93.7

47.2 74.9 62.4 93.9

46.7 73.0 62.1 94.1

37,415 0.178 0.232 0.005 1.27 25.9 1.10 237

26,936 0.206 0.271 0.006 1.24 26.1 1.08 147

27,708 0.208 0.285 0.007 1.24 26.2 1.06 344

a Rmerge for replicate re¯ections, R ˆ  (Ihi ÿ hIhi)/hIhi; Ihi, intensity measured for re¯ection h in data set i; hIhi, average intensity for re¯ection h calculated from replicate data. b R-factor ˆ jjFoj ÿ jFcjj/jFoj; Fo and Fc are the observed and calculated structure factors, respectively. c Rfree based upon 10% of the data randomly culled and not used in the re®nement.

303

Neisseria gonorrhoeae Carbonic Anhydrase

served. Otherwise the electron density is well de®ned in the ®nal 2Fo ÿ Fc map. No signi®cant peaks can be seen in the difference Fourier (Fo ÿ Fc) map contoured at 3s. However, small positive densities were found around Zn in this difference map. They appear on both sides of the metal in the direction of the Zn ± His111(Nd1) bond. This can be interpreted as a result of an anisotropic motion of the metal in this direction. Note that the data were collected at 21 C; the crystallographic Ê2 Ê 2 and 15.6 A temperature factor (B factor) is 15.2 A for the zinc ions in molecules A and B, respectively. An electron density near the zinc ion has been interpreted as a mixture of an inhibiting azide ion and water molecules belonging to the uninhibited enzyme. Two additional electron densities representing extraneous entities are observed. One of them is located near the N terminus and has tentatively been identi®ed as a 2-mercaptoethanol molecule from the crystallization medium interacting with His4 and His66. The other density was identi®ed as a sulfate ion linked to Gly135 (amide N), Arg136 (side-chain nitrogen atoms), and His195 (Ne2) in both molecules. Three of the four sulfate oxygen atoms have close contacts (between Ê and 3.02 A Ê ) with these nitrogen atoms. 2.71 A Finally, the model contains 237 water molecules, and the average B factor for these molecules is Ê 2. 35.6 A The geometry of the model is good as judged from the output of PROCHECK (Laskowski et al., 1993). The overall G factor is ÿ0.3. Most (91.1%) residues lie in the favored regions in the Ramachandran plot and 8.4% in additionally allowed regions. One residue, Asn220, has an unusual backbone conformation and lies in the disallowed region in the Ramachandran plot. It has a well de®ned electron density including all side-chain atoms. Its HCA II counterpart, Lys252, has a similar conformation in the high-resolution model reported by HaÊkansson et al. (1992). The average B factor for all protein atoms (excluding hydrogen) is Ê 2, close to that estimated from a Wilson 22.94 A Ê 2). plot (20.71 A In an attempt to remove the extraneous ligands, a crystal was soaked for three days in a solution containing 40 mM Tris-SO4 (pH 7.8), 1 mM EDTA and 25% (w/v) polyethylene glycol 3350. The results of X-ray diffraction data collection and structure re®nement are listed in Table 1 under the heading NGCA(S). The rmsd between the mainchain atoms of NGCA and NGCA(S) structures is Ê . The zinc-bound water and the ``deep'' 0.194 A water (Eriksson et al., 1988) are well de®ned in the NGCA(S) structure, suggesting that the azide ion has dissociated from the active site. The sulfate ion has been replaced by water molecules. The putative 2-mercaptoethanol molecule has disappeared, and residues 1 to 6 are not observed, indicating an increased disorder of the N-terminal region. In molecule A, the main-chain atoms of His66 have Ê inwards while the electron density shifted 0.25 A from the side-chain is extended indicating that the

imidazole ring oscillates between two positions by a rotation of approximately 55 of the Ca ± Cb bond. In molecule B, the omit difference density for the imidazole ring of His66 is very poor, suggesting a high degree of mobility. Crystals of the complex with acetazolamide (2acetylamido-1,3,4-thiadiazole-5-sulfonamide) were obtained by soaking against a solution containing 1.5 mM inhibitor. Details of the data collection and re®nement are given in Table 1 under the heading NGAZ. Description of the structure Figure 1 shows a stereo diagram of the folding of NGCA. The structural similarity between NGCA and HCA II is illustrated schematically in Figure 2, where some major structural differences are also indicated. A sequence alignment of these two enzymes, based on superposition of their crystal structures, is shown in Figure 3. In accordance with a generally accepted convention, the numbering system of human isozyme I is used for HCA II, and residues 1 and 126, which are present in HCA I but absent in HCA II, are indicated by dashes in Figure 3. In the following text, residue numbers for HCA II are shown in parentheses. The alignment yielded 14 sequence segments of residues with equivalent structural positions. These segments comprise 192 residues corresponding to 85% of the entire polypeptide chain of NGCA. The two proteins have identical amino acids in 74 of these 192 sequence positions (38.5%). The identities include two cis-proline residues, Pro(30) and Pro(202) in HCA II corresponding to Pro36 and Pro180 in NGCA. The best ®t of Ca atoms between NGCA and HCA II based on the 14 equivalent sequence Ê. segments gave an overall rmsd of 1.35 A The central, ten-stranded b sheet of HCA II (Eriksson et al., 1988), as well as a number of additional, short b strands, are essentially intact in NGCA (Figure 3). However, the last two strands of the central sheet, bI and bJ, each of which comprises two residues in HCA II, are not identi®ed as such in NGCA by the program DSSP (Kabsch & Sander, 1983). In addition, the lengths and positions of the individual strands in the central sheet vary slightly between the two enzymes. Interest-

Figure 1. Stereo drawing of the a-carbon backbone of the NGCA model. The N terminus is in the upper left part of the drawing.

304

Neisseria gonorrhoeae Carbonic Anhydrase

Figure 2. Schematic drawings of the structures of NGCA (left) and HCA II (right). The molecules are shown in a slightly different orientation from that in Figure 1. The N termini are in the upper left parts of the drawings. The zinc ions are shown as ®lled circles. Arrows indicate the three loops in HCA II which are deleted in NGCA. The Figure was produced with MOLSCRIPT (Kraulis, 1991).

ingly, the conserved sequence motif LAVL in strand bF, containing the important active-site residue Val(143) in HCA II, has a different location in the three-dimensional structure of NGCA, so that another valine residue, Val123, just upstream of the LAVL motif in the NGCA sequence, has a position equivalent to that of Val(143) in HCA II. In Ê closer to NGCA, strand bC has moved about 1 A strand bD compared to its position in HCA II. This difference probably comes from the replacement of Ala(65) in HCA II with Thr67 in NGCA. In HCA II, a buried water molecule forms a hydrogenbonded link between Phe(66) and Phe(95) (Eriksson et al., 1988). In NGCA, the side-chain hydroxyl group of Thr67 has replaced this water molecule and hydrogen bonds with the amide NH of Ile68 and the carbonyl O of Phe93, thus, bringing the two b strands closer together. The helical regions are not conserved to quite the same extent as the b-structure elements. Thus, the helical region near the N terminus is slightly more extensive in NGCA than in HCA II. However, the short 310 helices at positions (125) to (128) and (131) to (134) in HCA II are absent in NGCA. While the helical turn at residues (181) to (184) is intact in NGCA, the long helix comprising residues (155) to (167) in HCAII is shorter in NGCA, and only the central a-helical portion is conserved. On the other hand, the a-helix at residues (220) to (226) in HCAII is four residues longer in NGCA and runs from Gln198 to Val208. The sequence alignment involves a number of deletions and insertions (Figure 3). The most extensive of these are deletions in NGCA of a 12-residue segment corresponding to residues (128) to (139) in HCAII and a ten-residue segment corresponding to residues (231) to (240). Another large deletion comprises residues (98) to (103) in HCAII. The structural consequence of this ``trimming'' of the NGCA

sequence is a shortening of surface loops or, in the case of the (128) to (139) deletion, the replacement of a helix-containing surface loop by a hairpin turn (see Figure 2, where these deletions are indicated by arrows). Approximately the same sequence regions seem to be deleted also in the newly discovered a-carbonic anhydrases from the cyanobacteria Anabaena and Synechococcus (Soltes-Rak et al., 1997). The structure of the region around residue (131) has previously been shown to vary between different forms of carbonic anhydrase. Thus, in the membrane-associated HCA IV, an extended loop is observed rather than a helix (Stams et al., 1996), and in murine mitochondrial isozyme V the helix Ê compared to its position in is shifted about 2 A HCA II (Boriack-Sjodin et al., 1995). Since some sulfonamide inhibitors interact with residue (131) in HCA II (Jain et al., 1994; Vidgren et al., 1990), it has been proposed that these structural differences might be exploited to construct selective inhibitors (Stams et al., 1996). Disulfide bond The only two cysteine residues in NGCA, Cys28 and Cys181, form a disul®de bond. A stereo diagram is shown in Figure 4. This bond connects the N-terminal segment with a loop containing several active-site residues as well as the cis-proline, Pro180(202) (Figure 2). These cysteine residues, with residue numbers 23 and 203 in the HCA I numbering system, are quite conserved among membrane-linked or extracellular members of the a-carbonic anhydrase family. In some cases, the presence of a disul®de bond between these residues has been demonstrated. This applies to the membrane-associated, mammalian isozyme IV (Stams et al., 1996) and the secreted, mammalian isozyme VI (Fernley et al., 1988) as well as to the

Neisseria gonorrhoeae Carbonic Anhydrase

305

Figure 3. Sequence alignment of HCA II and NGCA based on superposition of their three-dimensional structures. Residues that are identical in the two proteins and have equivalent structural positions are indicated by asterisks. Secondary structure elements were identi®ed by the program DSSP (Kabsch & Sander, 1983). The strands of the central b sheet are labelled bA to bJ.

periplasmic a-carbonic anhydrase from the unicellular green alga Chlamydomonas reinhardtii (Kamo et al., 1990). In addition, cysteine residues are found in these positions in the extracellular a-carbonic anhydrases from Anabaena and Synechococcus

(Soltes-Rak et al., 1997), in nacrein, which is a soluble matrix protein in the nacreous layer of oyster pearls (Miyamoto et al., 1996), in the carbonic anhydrase-like, extracellular domains of the transmembrane protein tyrosine phosphatases b (or z)

Figure 4. Stereo diagram showing the Cys28 ±Cys181 disul®de bond in NGCA. Thin lines represent the corresponding residues in HCA II.

306

Neisseria gonorrhoeae Carbonic Anhydrase

Figure 5. Stereo diagram showing the central part of the active site of NGCA including some of the water molecules. Thin lines represent the corresponding residues in HCA II. Note that the imidazole ring of His66(64) is in the ``out'' position, while the ``in'' position of His(64) in HCA II is also indicated. AZI, azide.

and g (Barnea et al., 1993; Krueger & Saito, 1992) and in the membrane-associated protein MN (Opavsky et al., 1996). From the location of these Cys23-Cys203 containing members of the a-carbonic anhydrase family in the phylogenetic tree derived by Hewett-Emmett & Tashian (1996), one might hypothesize that the ancestral a-carbonic anhydrase was attached to a cell membrane and/ or located outside the plasma membrane, while the intracellular, cytoplasmic forms of the enzyme evolved later. Results from protein folding studies of HCAII (Fransson et al., 1992; Jonasson et al., 1997) indicated that the conformation of the loop structure at Pro(202) is strained, a conclusion supported by the observation that the cis conformation at position (202) is retained when proline is replaced by alanine (Tweedy et al., 1993). The Cys28(23)±

Cys181(203) disul®de bond might aid in the stabilization of this loop. In the intracellular carbonic anhydrases the Pro(202) loop might be stabilized by interactions with the (122)-(139) loop which is absent in NGCA. Hydrophobic cluster The major features of the hydrophobic cluster involving, for example, Phe93 (95) and located beneath the active site in the standard view of the molecule (Figure 2, see also Figure 6 of Eriksson et al., 1988) are retained in NGCA, although Phe(176) is replaced by Leu153 and Phe(70) by Tyr72. The hydrogen-bonding requirements of the phenolic OH of Tyr72 are satis®ed by links to several groups via a buried water molecule. The few additional differences between NGCA and HCA II are of a complementary nature. These differences

Figure 6. Stereo diagram of the putative 2-mercaptoethanol molecule bound to His4 and His66(64) in NGCA. Thin lines represent the correponding residues in HCA II. Note that both the ``in'' and the ``out'' positions of His(64) in HCA II are included.

307

Neisseria gonorrhoeae Carbonic Anhydrase

Figure 7. Stereo diagram of acetazolamide bound in the active site of NGCA. Thin lines represent corresponding residues in HCA II. The zinc ion and some water molecules are represented by crosses.

are represented by two interacting triads found in the two structures, Trp(97)-Phe(226)-Val(161) in HCAII corresponding to Val95-Phe204-Trp141 in NGCA. Although not in homologous sequence positions, the aromatic rings of Phe204 and Trp(97) occupy approximately the same space in the superimposed structures as do the rings of Trp141 and Phe(226). However, the ring planes of Phe204 and Trp(97) are roughly perpendicular. The active site A stereoview of the central part of the active site is shown in Figure 5. Most residues in the activesite cavity are conserved including the zinc ligands, His92 (94), His94 (96) and His111 (119). Other conserved residues are Tyr9 (7), His66 (64), Glu98 (106), and Thr177 (199) in the hydrophilic part of the active site, and Val113 (121), Val123 (143), Leu176 (198), Val185 (207), and Trp187 (209) in the hydrophobic part. In addition, some residues that are hydrogen bonded to the zinc ligands are conserved; these include Gln90 (92), Glu109 (117), and Asn212 (244). Superposition of the structures of NGCA and HCA II based on Ca atoms of the three histidine Ê for these three ligands yielded an rmsd of 0.04 A atoms. With this superposition, shifts of atoms in the zinc coordination sphere, namely, the three imidazole nitrogen atoms and the zinc ion, are Ê and 0.26 A Ê . The shifts for atoms in between 0.16 A the catalytically important H-bonded system involving Thr177 (199) and Glu98 (106) are larger; Og1 Ê and Oe1 of Glu98 of Thr177 (199) is shifted 0.35 A Ê (106) 0.42 A. In HCA II, an H-bond network extends all the way from the zinc ligand His(119) via Glu(117), His(107), Tyr(194), and Ser(29) to Trp(209). In NGCA, His(107) is replaced by Asn99 and Tyr(194) by Phe172. Although the H-bond network is broken by Phe172, other H-bond connections of this network are intact and the side-chain positions are virtually unchanged except for a shift Ê comof the aromatic ring of Phe172 by about 1 A pared to the position of its counterpart in HCA II.

Some shifts are also observed for residues in the hydrophobic pocket, which is located near the zinc ion and assumed to harbor the CO2 substrate (Fierke et al., 1991). The Cd atoms of Leu176 (198) Ê and the Cg atoms of Val123 (143) shift 0.5-0.6 A Ê , while a small rotation (17 ) around shift 0.3-0.4 A Cb ±Cg was found for the side-chain of Trp187 (209). The deletion of the (128)-(139) segment of HCA II makes the entrance to the hydrophobic part of the active site wider and more exposed in NGCA. Two additional features contribute to this, namely, the replacements of Leu(141) and Cys(206) with the smaller residues Pro121 and Gly184. Presumably, His66 in NGCA has the same catalytic function as a proton shuttle as His(64) in HCAII (Tu et al., 1989). Indeed, preliminary kinetic results on a His66 ! Ala mutant appear to con®rm this assumption. In the absence of the extraneous ligand, tentatively identi®ed as 2-mercaptoethanol, the side-chain of His66 seems to be very mobile. However, the binding of this ligand appears to lock the imidazole ring of His66 into a position similar to the ``out'' position observed in HCA II under certain conditions (Figures 5 and 6). We are presently testing the possibility that 2-mercapÊ ) between various Table 2. A comparison of distances (A atoms in acetazolamide and active-site residues in NGCA and HCA II (Vidgren et al., 1990) Atom N1 N1 O1 O2 N2 N2 N3 N3 S2 O3 C4

Residue Zn Thr177(199) Og Zn Thr177(199) N Thr178(200) Og Leu176(198) Cd2 Thr178(200) Og Leu176(198) Cd2 Leu176(198) Cd2 Gln90(92) Ne2 Phe(131) Cz

NGCA 2.13 2.63 3.02 2.63 3.13 3.93 3.07 3.95 3.91 2.89

HCA II 1.9 3.0 3.2 2.9 3.2 3.7 3.2 3.8 4.0 2.8 3.5

The atoms N1, O1, and O2 refer to the sulfonamide group, N2, N3, and S2 are the heteroatoms in the thiadiazole ring, while C4, N4, and O3 belong to the acetylamido side-chain.

308 toethanol, or a derivative of it, is a speci®c inhibitor of the proton-transfer step in the catalytic cycle. Inhibitor binding Figure 7 shows a stereoview of the sulfonamide inhibitor, acetazolamide, bound to the active site of NGCA. The inhibitor has displaced the zinc-bound water, the ``deep'' water and another two water molecules in the active site cavity. The interaction of the sulfonamide group with the zinc ion and with Thr177(199) is practically identical with that previously observed in numerous studies of sulfonamide-carbonic anhydrase complexes (Lindskog, 1997). As illustrated in Table 2, the orientation of the inhibitor in the active site of NGCA is very similar to that observed in HCA II (Vidgren et al., 1990). These similarities include potential hydrogen bonds between the ring nitrogen atoms of the inhibitor and the hydroxyl O of Thr178(200) and a hydrogen bond between the acetylamido N and Gln90(92). In the complex with HCA II the acetyl group of the acetylamido side-chain forms a van der Waals contact with Phe(131). Of course, this interaction is not observed in the complex with NGCA, since the (128)-(139) sequence segment is absent in the bacterial enzyme.

Materials and Methods Crystallization and data collection The expression of carbonic anhydrase from N. gonorrhoeae in E. coli cells and the puri®cation of the enzyme were performed according to published procedures (Chirica et al., 1997). The crystals used for data collection were grown at room temperature by the hanging drop method from a solution containing 40 mM Tris-SO4 buffer (pH 7.8), 50 mM (NH4)2SO4, 30 mM MgSO4, 2.2 mM NaN3, 2.2 mM 2-mercaptoethanol, 9 to 10% (v/v) dimethylsulfoxide, and 16 to 18% (w/v) polyethylene glycol 3350. The enzyme concentration was 20 to 22 mg/ml. Crystals suitable for X-ray diffraction studies were obtained after about two weeks. Diffraction data were collected from a single crystal (0.2 mm  0.2 mm  0.3 mm) as summarized in Table 1. Programs Denzo and Scalepack (Otwinowski & Minor, 1996) were used to process the data. The data have an overall Rmerge of 0.034 and a completeness of 83.1%. Since less than 50% of the re¯ections were measured for Ê , the effective resolthe resolution shell of 1.65 to 1.78 A Ê with 53.5% of the re¯ections in ution is taken as 1.78 A Ê being above 2 s. Further resolution shell 1.78 to 1.86 A details are given in Table 1, which also contains data for a crystal that was soaked for three days in a solution containing 40 mM Tris-SO4 (pH 7.8), 1 mM EDTA and 25% (w/v) polyethylene glycol 3350. In addition, Table 1 shows data for a crystal (NGAZ) that was soaked for two days against a solution containing 50 mM Tris-SO4 (pH 7.8), 1 mM 2-mercaptoethanol, 25% (w/v) polyethylene glycol 3350 and 1.5 mM acetazolamide. Structure solution by molecular replacement (MR) Program O (Jones et al., 1991) was used for graphics display and model building. Programs in the CCP4 suite

Neisseria gonorrhoeae Carbonic Anhydrase (Collaborative Computational Project, 1994) were used for most of the crystallographic computing including electron density map calculation. The Matthews coef®cient (Matthews, 1968) and results from self-rotation calculations indicated the presence of two molecules in the asymmetric unit. A signi®cant peak was found in the self-rotation calculation using the program polarrfn (CCP4 suite); its peak height is 46.6% of the origin peak, and about twice as high as the next peak. This peak corresponds to a Kappa angle of 18 . The Matthews coef®cient is 2.16 (solvent content of 42.7%) assuming two molecules per asymmetric unit. The structure of HCA II (HaÊkansson et al., 1992) was used as a template for generating search models for molecular replacement. The relatively low sequence homology and the fact that there might be a dimer in the asymmetric unit make the molecular replacement calculation a challenging task. Therefore, differently trimmed search models, in addition to varying different parameters like Pattersson integration radius, resolution limits, etc. were explored in the MR trials. By inspecting the structure of HCAII with the sequence of NGCA aligned onto it, identical residues or residues with similar properties were identi®ed. The model was trimmed to take these homologous features into account and retain them as complete secondary structure elements. The correct solution was reached using a search model in which only 104 out of 259 residues in the re®ned human CA II structure were retained, including residues 28 to 34, 53 to 71, 86 to 96, 105 to 124, 139 to 146, 189 to 213 and 244 to 257. MR was performed using the program AMoRe (Navaza, 1994). Resolution Ê and 3.2 to 10 A Ê were used for limits of 4.2 to 10 A rotation/translation search and rigid body ®tting, respectively. The Pattersson integration radius was set to Ê . The top two solutions after a complete run, 22.7 A including rotation search, translation search, crystal packing examination and rigid body ®tting, have correlation coef®cients of 12.9 and 12.4 and R-factors of 0.530 and 0.535, respectively. These corresponded to the ®rst and the 43rd peaks from the rotation search. These two solutions are related by polar angles of (89.8 , 1.9 , 17.7 ) in terms of orientation, which is virtually the same as from the self-rotation calculation (90.7 , 0.0 , 18.1 ). This is the ®rst sign to indicate the validity of the solution. Search for the translation vector for the second molecule was then performed using AMoRe by ®xing the ®rst one, which increased/lowered the correlation coef®cient/R-factor to 27.2/0.499. The correctness of the solution was ®nally veri®ed based on the electron density map calculated after phase improvement procedures, including solvent ¯attening, histogram matching and density averaging, using program DM (CCP4 suite). Model building and refinement The program X-PLOR (BruÈnger, 1992) was used for molecular dynamics re®nement (simulated annealing, SA) and the program REFMAC was used for maximum Ê, likelihood re®nement. First round SA (data 5.5 to 2.5 A starting temperature 3000 K) with constraints of strict non-crystallographic symmetry (NCS) was ®nished with a R/Rfree of 0.42/0.49. Repetition of the same SA protocol after adding 20 residues into the model lowered the R/Rfree to 0.34/0.42. Extensive (re)building was then carried out to include 223 more residues in the model which is now 98% complete. More SA re®nement with-

Neisseria gonorrhoeae Carbonic Anhydrase Ê , starting temperature out NCS constraint (data 8 to 2 A 4000 K) resulted in a R/Rfree of 0.27/0.31. Re®nement of individual temperature factors normally lowered the R/ Rfree by about 2% during this stage. Subsequently, programs REFMAC and ARP (Lamzin Ê ) were used to & Wilson, 1993, 1997; data 20 to 1.65 A include bulk solvent correction and to build ordered solvent molecules into the model. All the electron density maps thereafter were calculated using sA weighted coef®cients (Read, 1986) to enhance map quality. The re®nement with NCS restraints led to the identi®cation of 230 solvent molecules in the asymmetric unit and the ®nal R/Rfree is 0.19/0.22. Modeled water molecules were justi®ed and kept according to the following criteria: the electron density in the 2Fo ÿ Fc map (contoured at 1s) should be signi®cant and spherical; its interaction with other protein atoms should have reasonable geometry Ê ). (among other things, distances no larger than 3.3 A One sulfate ion per molecule was subsequently identi®ed and modeled. The ®nal re®nement was carried out using REFMAC without NCS restraints. Coordinates of the NGCA model have been deposited in the Brookhaven Protein Data Bank (accession code 1KOP) and are available from the authors until they have been processed and released.

Acknowledgments We are very grateful to Ms Katarina Wallgren for skilful technical assistance. Financial support was obtained from the Swedish Natural Science Research Council. S.H. was partially supported by a grant to B.-H.J. from the Wenner-Gren Foundation.

References Barnea, G., Silvennoinen, O., Shaanan, B., Honegger, A. M., Canoll, P. D., D'Eustacio, P., Morse, B., Levy, J. B., Laforgia, S., Huebner, K., Musacchio, J. M., Sap, J. & Schlessinger, J. (1993). Identi®cation of a carbonic anhydrase-like domain in the extracellular region of RPTPg de®nes a new subfamily of receptor tyrosine phosphatases. Mol. Cell. Biol. 13, 1497± 1506. Boriack-Sjodin, P. A., Heck, R. W., Laipis, P. J., Silverman, D. N. & Christianson, D. W. (1995). Structure determination of murine mitochondrial carbonic anhydrase V at 2.45-A resolution: implications for catalytic proton transfer and inhibitor design. Proc. Natl Acad. Sci. USA, 92, 10949± 10953. BruÈnger, A. T. (1992). X-PLOR Manual, Version 3. 0, Yale University, New Haven, CT. Chirica, L. C., Elleby, B., Jonsson, B.-H. & Lindskog, S. (1997). The complete sequence, expression in Escherichia coli, puri®cation and some properties of carbonic anhydrase from Neisseria gonorrhoeae. Eur. J. Biochem. 244, 755± 760. Collaborative Computational Project N.4. (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, 760± 763. Eriksson, A. E., Jones, T. A. & Liljas, A. (1988). Re®ned Ê structure of human carbonic anhydrase II at 2.0 A resolution. Proteins: Struct. Funct. Genet. 4, 274± 282. Fernley, R. T., Wright, R. D. & Coghlan, J. P. (1988). Complete amino acid sequence of ovine salivary carbonic anhydrase. Biochemistry, 27, 2815± 2820.

309 Fierke, C. A., Calderone, T. L. & Krebs, J. F. (1991). Functional consequences of engineering the hydrophobic pocket of carbonic anhydrase II. Biochemistry, 30, 11054± 11063. Fransson, C., FreskgaÊrd, P.-O., Herbertsson, H., Ê . , Jonasson, P., MaÊrtensson, L.-G., Johansson, A Svensson, M., Jonsson, B.-H. & Carlsson, U. (1992). Cis-trans isomerization is rate-determining in the reactivation of denatured human carbonic anhydrase II as evidenced by proline isomerase. FEBS Letters, 296, 90± 94. Fukuzawa, H., Suzuki, E., Komukai, Y. & Miyachi, S. (1992). A gene homologous to chloroplast carbonic anhydrase (icfA) is essential to photosynthetic carbon dioxide ®xation by Synechococcus PCC7942. Proc. Natl Acad. Sci. USA, 89, 4437± 4441. Guilloton, M. B., Korte, J. J., Lamblin, A. F., Fuchs, J. A. & Anderson, P. M. (1992). Carbonic anhydrase in Escherichia coli. A product of the cyn operon. J. Biol. Chem. 267, 3731± 3734. HaÊkansson, K., Carlsson, M., Svensson, L. A. & Liljas, A. (1992). Structure of native and apo carbonic anhydrase II and structure of some of its anionligand complexes. J. Mol. Biol. 227, 1192± 1204. Hewett-Emmett, D. & Tashian, R. E. (1996). Functional diversity, conservation, and convergence in the evolution of a-, b-, and g-carbonic anhydrase gene families. Mol. Phylogenet. Evol. 5, 50 ±77. Jain, A., Whitesides, G. M., Alexander, R. S. & Christianson, D. W. (1994). Identi®cation of two hydrophobic patches in the active-site cavity of human carbonic anhydrase II by solution-phase and solid-state studies and their use in the development of tight-binding inhibitors. J. Med. Chem. 37, 2100± 2105. Jonasson, P., Aronsson, G., Carlsson, U. & Jonsson, B. H. (1997). Tertiary structure formation at speci®c tryptophan side-chains in the refolding of human carbonic anhydrase II. Biochemistry, 36, 5142± 5148. Jones, T. A., Zhou, J.-Y., Cowan, S. W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, 110± 119. Kabsch, W. & Sander, C. (1983). Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers, 22, 2577± 2637. Kamo, T., Shimogawara, K., Fukuzawa, H., Muto, S. & Miyachi, S. (1990). Subunit constitution of carbonic anhydrase from Chlamydomonas reinhardtii. Eur. J. Biochem. 192, 557± 562. Kisker, C., Schindelin, H., Alber, B. E., Ferry, J. G. & Rees, D. C. (1996). A left-handed b-helix revealed by the crystal structure of a carbonic anhydrase from the archaeon Methanosarcina thermophila. EMBO J. 15, 2323± 2330. Kraulis, P. J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, 946± 950. Krueger, N. X. & Saito, H. (1992). A human transmembrane protein-tyrosine-phosphatase, PTPz, is expressed in brain and has an N-terminal receptor domain homologous to carbonic anhydrases. Proc. Natl Acad. Sci. USA, 89, 7417± 7421. Lamzin, V. S. & Wilson, K. S. (1993). Automated re®nement of protein models. Acta Crystallog. sect. D, 49, 129± 147.

310

Neisseria gonorrhoeae Carbonic Anhydrase

Lamzin, V. S. & Wilson, K. S. (1997). Automated re®nement for protein crystallography. Methods Enzymol. 277, 269± 305. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283± 291. Lindskog, S. (1997). Structure and mechanism of carbonic anhydrase. Pharmacol. Ther. 74, 1 ± 20. Matthews, B. W. (1968). Solvent content of protein crystals. J. Mol. Biol. 33, 491± 497. Miyamoto, H., Miyashita, T., Okushima, M., Nakano, S., Morita, T. & Matsushiro, A. (1996). A carbonic anhydrase from the nacreous layer in oyster pearls. Proc. Natl Acad. Sci. USA, 93, 9657± 9660. Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Crystallog sect. A, 50, 157± 163. Opavsky, R., Pastorekova, S., Zelnik, V., Gibadulinova, A., Stanbridge, E. J., Zavada, J., Kettmann, R. & Pastorek, J. (1996). Human MN/CA9 gene, a novel member of the carbonic anhydrase family: Structure and exon to protein domain relationships. Genomics, 33, 480± 487. Otwinowski, Z. & Minor, W. (1996). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307±326.

Read, R. J. (1986). Improved Fourier coef®cients for maps using phases from partial structures with errors. Acta Crystallog. sect. A, 42, 140± 149. Soltes-Rak, E., Mulligan, M. E. & Coleman, J. R. (1997). Identi®cation and characterization of a gene encoding a vertebrate-type carbonic anhydrase in cyanobacteria. J. Bacteriol. 179, 769± 774. Stams, T., Nair, S. K., Okuyama, T., Waheed, A., Sly, W. S. & Christianson, D. W. (1996). Crystal structure of the secretory form of membrane-associated human carbonic anhydrase IV at 2. 8-A resolution. Proc. Natl Acad. Sci. USA, 93, 13589± 13594. Tu, C. K., Silverman, D. N., Forsman, C., Jonsson, B.-H. & Lindskog, S. (1989). The role of histidine 64 in the catalytic mechanism of human carbonic anhydrase II studied with a site-speci®c mutant. Biochemistry, 28, 7913± 7918. Tweedy, N. B., Nair, S. K., Paterno, S. A., Fierke, C. A. & Christianson, D. W. (1993). Structure and energetics of a non-proline cis-peptidyl linkage in a proline-202 ! alanine carbonic anhydrase II variant. Biochemistry, 32, 10944± 10949. Vidgren, J., Liljas, A. & Walker, N. P. C. (1990). Re®ned structure of the acetazolamide complex of human Ê . Int. J. Biol. Macromol. carbonic anhydrase II at 1.9 A 12, 342± 344.

Edited by R. Huber (Received 3 March 1998; received in revised form 13 July 1998; accepted 13 July 1998)