Structure of the ATP Binding Domain from the Archaeoglobus fulgidus Cu+-ATPase*

The P-type ATPases translocate cations across membranes using the energy provided by ATP hydrolysis. CopA from Archaeoglobus fulgidus is a hyperthermophilic ATPase responsible for the cellular export of Cu+ and is a member of the heavy metal P1B-type ATPase subfamily, which includes the related Wilson and Menkes diseases proteins. The Cu+-ATPases are distinct from their P-type counter-parts in ion binding sequences, membrane topology, and the presence of cytoplasmic metal binding domains, suggesting that they employ alternate forms of regulation and novel mechanisms of ion transport. To gain insight into Cu+-ATPase function, the structure of the CopA ATP binding domain (ATPBD) was determined to 2.3 Å resolution. Similar to other P-type ATPases, the ATPBD includes nucleotide binding (N-domain) and phosphorylation (P-domain) domains. The ATPBD adopts a closed conformation similar to the nucleotide-bound forms of the Ca2+-ATPase. The CopA ATPBD is much smaller and more compact, however, revealing the minimal elements required for ATP binding, hydrolysis, and enzyme phosphorylation. Structural comparisons to the AMP-PMP-bound form of the Escherichia coli K+-transporting Kdp-ATPase and to the Wilson disease protein N-domain indicate that the five conserved N-domain residues found in P1B-type ATPases, but not in the other families, most likely participate in ATP binding. By contrast, the P-domain includes several residues conserved among all P-type ATPases. Finally, the CopA ATPBD structure provides a basis for understanding the likely structural and functional effects of various mutations that lead to Wilson and Menkes diseases.

The P-type ATPases encompass a large family of integral membrane proteins that couple ATP hydrolysis with the transport of cations across cell membranes (1,2). Within this protein family, members of the P 1B subgroup specifically transport transition metal ions including Cu ϩ / Ag ϩ or Zn 2ϩ /Cd 2ϩ /Pb 2ϩ (3,4). Widely distributed in nature, the P 1Btype ATPases confer heavy metal tolerance to microorganisms (5,6) and are essential for the absorption, distribution, and bioaccumulation of metal micronutrients by multicellular eukaryotes (7,8). The impor-tance of these enzymes for metal homeostasis is underscored by the two Cu ϩ -ATPases present in humans, ATP7A (MNK) and ATP7B (WND) (7,9). Mutations in the genes encoding the ATP7A and ATP7B proteins lead to Menkes and Wilson diseases, respectively. Menkes disease is characterized by impaired transport of dietary copper across the small intestine, resulting in a copper deficiency in peripheral tissues. Improper function of WND leads to toxic copper accumulation in the liver because of reduced biliary excretion (10,11).
Although all P-type ATPases share essential structural elements, the subgroups differ in the number of predicted transmembrane helices and the arrangement of these segments with respect to the cytoplasmic ATP binding domain (ATPBD). 4 In addition, the soluble regions vary in length and composition (4,5,12). The P 1B -type ATPases comprise eight transmembrane helices, a large ATPBD, an actuator domain (A-domain), and N-or C-terminal soluble metal binding domains (MBDs). Binding sites responsible for metal coordination during transport are located within the membrane and are believed to involve a CPX sequence motif as well as other key residues that confer metal ion selectivity (3,13). The N-terminal MBDs (N-MBDs), ranging from one to six depending on the organism, receive copper ions from metallochaperones (14), participate in ATPase regulation (15), and facilitate intracellular relocalization (16). In the absence of copper, the six WND N-MBDs interact with the ATPBD (17), suggesting distinct mechanisms of regulation by substrates in P 1B -type ATPases.
The P 1B -type ATPase transport mechanism follows the classical Post-Albers E1/E2 model (7,18,19). The central element of this catalytic cycle is the coupling of metal transport to enzyme phosphorylation by ATP. Phosphorylation of the aspartic acid residue from the signature sequence DKTGT is the unifying characteristic of all P-type ATPases (1,2). The ATP binding, hydrolysis, and enzyme phosphorylation steps occur within the ATPBD, which consists of a phosphorylation domain (P-domain) and a nucleotide binding domain (N-domain). The ATPBD from WND (17), as well as those from P 2 and P 1A -ATPases (20 -22), have been isolated in soluble form. These cytoplasmic fragments are able to bind nucleotides and hydrolyze ATP, albeit at a low rate, and thus contain the key components needed for energy dependent ion transport by these ATPases.
With the exception of solution structures of N-MBDs (23,24), structural information for P 1B -type ATPases is lacking. Several structures of the sarcoplasmic reticulum (SERCA1) Ca 2ϩ -ATPase have been reported recently (25)(26)(27)(28)(29) and provide insight into the conformational changes that P 2 -type ATPases such as the Na ϩ ,K ϩ -ATPase or the H ϩ ,K ϩ -ATPase undergo during catalysis. Because of significant archi-tectural differences between heavy metal transporting and P 2 -type ATPases, homology modeling based on the SERCA1 structures is of limited utility in understanding P 1B -type ATPase function and transport mechanisms at the molecular level. We have determined the 2.3 Å resolution crystal structure of the ATPBD from CopA from the hyperthermophile Archaeoglobus fulgidus, a well characterized Cu ϩ -ATPase that contains all of the key elements present in eukaryotic Cu ϩ -ATPases (13,15,30). The CopA ATPBD structure, because of its small size and compactness, constitutes a minimal ATP binding-phosphorylation unit. As such, the structure allows for the identification of the central catalytic features independently of the secondary structural elements likely associated with trafficking, targeting, and regulatory functions.

EXPERIMENTAL PROCEDURES
Cloning of the CopA ATPBD-The CopA ATPBD (residues Lys 407 -Lys 671 ) was PCR-amplified from CopA cDNA (30) by using the primers 5Ј-GCCCTTGGTCTCTAATGAAAAATGCCGACGCT-CTCGAA-3Ј and 5Ј-CGGGAAGGTCTCTGCGCTTTTGCTCAT-GGTCTTTCTGCT-3Ј, which introduce a Bsa1 restriction site at the 5Ј-and 3Ј-ends of the PCR product. The PCR product was digested with Bsa1 and cloned into the pPR-IBA1vector (IBA), which introduces an 8-amino-acid (WSHPQFEK) streptactin tag at the C terminus of the protein. The accuracy of the insert was confirmed by DNA sequencing. BL21Star(DE3)pLysS Escherichia coli cells carrying the plasmid pSJS1240 encoding for rare tRNAs (tRNA arg AGA/AGG and tRNA ile AUA) (31) were transformed with the pATPBD construct.
Expression and Purification of the CopA ATPBD-E. coli cells containing the pATPBD plasmid were grown at 37°C in Luria-Bertani medium containing 100 mg/liter ampicillin and 30 mg/liter chloramphenicol. ATPBD expression was induced with 250 M isopropyl ␤-Dthiogalactopyranoside at an A 600 of 0.7. The cells were harvested by centrifugation at 6000 ϫ g for 5 min 3-4 h after induction. The pellet was frozen in liquid nitrogen and stored at Ϫ80°C.
For purification, the frozen cells were resuspended in 100 mM Tris, pH 7.5, 150 mM NaCl (Buffer W), to which DNase, 5 mM MgCl 2 , and 1 mM phenylmethylsulfonyl fluoride were added. The cell suspension was sonicated for 10 min (30-s pulses with a 30-s rest) and centrifuged at 163,000 ϫ g for 1 h. The supernatant was then loaded onto a 10-ml streptactin column (IBA) conditioned in buffer W, and the ATPBD was purified according to the manufacturer's protocol. SDS-PAGE of the eluted protein indicated it was greater than 95% pure. The ATPBD was exchanged into 20 mM MOPS, pH 7.0, 20 mM NaCl, and 5% glycerol by several concentration and dilution steps using an Amicon Ultra YM-10 concentrator, frozen at 30 mg/ml in liquid nitrogen, and stored at Ϫ80°C. This protocol yielded ϳ15 mg of protein/liter of medium. The protein concentration was estimated by using the theoretical extinction coefficient, ⑀ 280 ϭ 11,400 cm Ϫ1 M Ϫ1 , and a molecular mass of 29.6 kDa. A selenomethionine derivative of the CopA-ATPBD was prepared using LeMaster's medium (32) and purified as described above.
ATPase Activity-Activity assays were carried out at 70°C for 20 min in a 500-l solution containing 50 mM Tris, pH 7.5, 25 mM NaCl, 1 mM MgCl 2 , 5 mM ATP, and Ϯ 20 M ATPBD. The reaction was stopped by quickly cooling the solution to 4°C. The free phosphate concentration was measured at room temperature by using the EnzChek phosphate assay kit (Molecular Probes). The ATPBD had an activity of 4.2 nmol P i /mg/min. Background activity (ATP hydrolysis in the absence of protein) was less than 50% in all of our determinations.
Crystallization-Both wild-type and selenomethionine CopA-ATPBD were diluted to 10 mg/ml with a 20 mM MOPS, pH 7.0, 20 mM NaCl, and 5% glycerol solution and crystallized by using the hanging drop method at room temperature (22-24°C). Equal volumes of protein and a precipitant solution containing 100 mM sodium acetate, pH 4.6, 5-8% polyethylene glycol 4000, and 10 -20% glycerol were combined. Large rectangular crystals grew within 2-3 days. Prior to data collection, the ATPBD crystals were quickly exchanged into a cryosolvent comprising 50 mM sodium acetate, pH 4.6, 5% polyethylene glycol 4000, and 25% glycerol and flash frozen in liquid nitrogen.
Data Collection and Structure Determination-A SAD data set was collected at 100 K at the sector 32 beamline at the Advanced Photon Source (see Table 1). The crystals diffracted to 2.3 Å resolution, belong to the space group P3 1 , and have unit cell dimensions of a ϭ b ϭ 80.78 Å and c ϭ 105.96 Å. The data sets were integrated by using MOSFLM (33) and scaled with SCALA (34). Both SOLVE (35) and CNS (36) were used to find nine heavy atom sites corresponding to three selenium positions in each of the three ATPBD molecules in the asymmetric unit. CNS yielded the most interpretable electron density maps after density modification. The three ATPBDs comprising the asymmetric unit were built using XtalView (37), and the model was refined with CNS (Table  1). Non-crystallographic symmetry restraints were imposed throughout the refinement. Most of the residues in the three asymmetric units were observed in the electron density maps except for residues 407-410 at the N terminus, 670 -671 at the C terminus, and 637-645 in the P-domain. A Ramachandran plot calculation with PROCHECK (38) indicated that 91% of the residues had the most favored geometry with the rest in additionally allowed regions. All figures were generated with PyMOL (39).

RESULTS AND DISCUSSION
Overall Structure-The A. fulgidus CopA ATPBD spanning residues 407-671 was expressed and purified to homogeneity. This isolated domain was able to bind and hydrolyze ATP at a measurable rate (4.2 nmol P i /mg/min) when incubated at 70°C, the growth temperature of the source organism. The A. fulgidus ATPBD exhibits a kidney bean-like topology with a cleft for ATP binding at the interface between the Nand P-domains (Fig. 1a). Two 5-6-amino-acid loops (residues 429 - 434 between ␤1 and ␤2 and residues 547-552 between ␤7 and ␣6) form a hinge region connecting the N-and P-domains. Notably, the CopA ATPBD adopts a closed conformation similar to the SERCA1 structures with bound AMP-PCP or AlF 4 Ϫ and ADP, which both mimic the E1P-ADP state of the protein (27, 29) (Fig. 1b). This closed conformation is characterized by a quasiparallel orientation of helices ␣5 and ␣7, similar to helices ␣7 and ␣10 in the nucleotide-bound SERCA1 structures (Fig.  1b). By contrast, these helices adopt a perpendicular orientation in the open forms of P 2 -type ATPases (25) (Fig. 1c). Among the different P-type ATPases, and within the specific P 1 , P 1B , and P 2 subfamilies, there is little sequence conservation in these helices ( Fig. 2 and supplemental Fig. S1). Nevertheless, mutations in the corresponding MNK and WND helices are known to lead to Wilson and Menkes diseases (Fig. 2) (40). A crystal structure with a nucleotide analogue bound to the CopA ATPBD could not be obtained despite several cocrystallization and soaking efforts. The ATPBD may be locked in this closed state either by the crystallization conditions or by intrinsic interactions that confer thermostability to CopA. Hinge Region-The two short loops linking the N-and P-domains likely function as a hinge, allowing the two domains to undergo structural rearrangements concomitant with nucleotide binding and phosphoester intermediate formation (25,27,29). Several highly conserved residues are present in the hinge region. These include Thr 430 and Asp 548 , which appear to be universally conserved among all ATPases, and Gly 432 , which is conserved only among the P 1B family members. The Thr 430 and Asp 548 side chains interact via hydrogen bonding. In addition, the Gly 432 amide nitrogen interacts weakly with the Asp 548 side chain. A similar pattern is observed in SERCA1 (supplemental Fig. S2) (25)(26)(27)(28)(29). These interactions may be critical for stabilizing the conformation of the phosphorylation loop (D 424 KTGT), which is just upstream of Thr 430 in the P-domain, and are the focal point about which the hinge flexes. In support of a key role for these residues, mutations in Thr 430 and Gly 432 in WND are linked to disease states (Fig. 2) (40).
The N-domain-The N-domain consists of a six stranded antiparallel ␤-sheet sandwiched between two ␣-helices on each side of the sheet (Fig. 3a). Structures of N-domains from SERCA1 (25), the Na ϩ ,K ϩ -ATPase (41), the Kdp-ATPase (KdpB) (42) and more recently WND, 5 all exhibit this overall fold despite a lack of sequence homology ( Fig. 3 and supplemental Fig. S1). Moreover, SERCA1, the Na ϩ /K ϩ -ATPase, and WND differ significantly from CopA in that their sequences include several insertions ranging from 10 to 60 residues in  length. The WND N-domain has an unstructured 39-amino-acid loop between ␤4 and ␤5 that extends into the cytosol away from the N-domain (Fig. 3, b and c). This loop, which spans 56 amino acids in MNK, is not essential for catalysis 5 and comprises just four residues in CopA. The MNK and WND loops do not bear any sequence resemblance to each other, suggesting that these elements may specifically target the two proteins to different cellular compartments. Like WND, the Ca 2ϩ and Na ϩ ,K ϩ -ATPases have additional loops and structural elements, such as a seventh strand on the ␤-sheet ( Fig. 1 and supplemental Fig.  S1). Some of these inserts in SERCA1 participate in interactions with the A-domain (25)(26)(27)(28)(29). The additional structural elements found in the various P-type ATPases may therefore be specific for their function and help to promote domain-domain interactions as well as proper cellular localization.
The CopA N-domain is quite similar to the KdpB N-domain with bound AMP-PNP. There are no extra insertions in the KdpB sequence, and the root mean square deviation for C␣ atoms between KdpB and CopA is 1.54 Å (Fig. 3d). When KdpB is superimposed onto CopA, the adenine and ribose rings of the ATP analogue fit well within a conserved pocket (Fig. 3e). Sequence analysis of the different P 1B -type ATPase N-domains reveals that five amino acids, Glu 457 , His 462 , Gly 490 , Gly 492 , and Gly 501 , are universally conserved (Fig. 2). 5 All of these residues are predicted to interact with the docked nucleotide in CopA on the basis of the superimposed KdpB structure. Residues Glu 457 , His 462 , and Gly 492 may function to orient and bind the adenine ring through hydrogen bonding, whereas Gly 490 and Gly 501 likely interact primarily with the ribose moiety and ␣-phosphate. Mutagenesis and NMR studies on WND (44) 5 and the E. coli Zn 2ϩ -transporting P 1B -type ATPase ZntA (45) are consistent with a role for these residues in ATP binding. Interestingly, none of these residues are conserved in the other ATPase families, and each family appears to preserve its own set of residues specific for ATP binding. For example, in KdpB (42) and SERCA1 (27,29), a phenylalanine residue on the strand equivalent to ␤3 forms -stacking interactions with the adenine ring.
In the model of nucleotide-bound CopA, His 462 is not positioned properly for strong contacts with the adenine ring (Fig. 3e). Given the morphology of the adenine-binding cleft, the adenine may form better interactions with His 462 and Glu 457 by rotating 90°with respect to its position in KdpB. Further movement of the S 460 EHP loop between ␣2 and ␣3 may also facilitate interaction. The WND N-domain structure, which was determined with bound ATP, differs from CopA in that ␣2 has rotated by ϳ30°to form a different set of interactions with the ␤-sheet (Fig. 3c). As a result, the position of the SEHP loop changes and ␣3 rotates by ϳ10°. These differences may be related to interdomain regulatory interactions associated with the multiple MBDs present in eukaryotic proteins. It has been demonstrated that the first four MBDs of WND interact with the N-domain and that this interaction modulates ATP affinity (17,44). CopA, which requires only one MBD for maximum activity (15), may function in a slightly different fashion. Because the N-domain in the full-length ATPase is likely to participate in domain-domain interactions required for catalysis, movement of the SEHP loop is likely to be important for protein function.
The P-domain-The P-domain comprises a six-stranded parallel ␤-sheet sandwiched between six ␣-helices, three on each side of the sheet (Fig. 4a). The SERCA1 P-domain, which is the only other known structure of a P-domain, has a similar fold (root mean square deviation for C␣ atoms ϭ 1.7 Å) with the exception of a 50-amino-acid insert between ␣10 and ␤14 (Fig. 4, b and c). In CopA, both the invariant D 424 KTGT and T 572 GD loops are positioned in the back of the ATPBD crevice near the hinge region (Fig. 1a). These loops, which are critical for enzyme phosphorylation during catalysis, adopt similar conformations to those observed in the different SERCA1 structures (Fig. 4, c and d) (25)(26)(27)(28)(29). The loop between ␤10 and ␣9 containing the sequence A 614 XXGDGXND rests at the edge of the crevice marking the interface between the N-and P-domains. In the SERCA1 structures, the equivalent loop interacts with the A-domain and adopts different orientations. Finally, residues 637-645, for which no electron density was observed, would be located on a loop between ␤11 and ␤12 far from the phosphorylation site on the protein exterior (Figs. 1a and 4a). The analogous region in SERCA1 forms a small helix (Fig. 1b)   likely coordinate Mg 2ϩ , and Asp 574 hydrogen bonds to the nucleotide ribose moiety. The remaining conserved residues likely participate in second coordination sphere hydrogen bonding or serve a specific structural role to preserve the active site geometry. Domain Interactions-In P-type ATPases, transmembrane ion transport is coupled to ATP hydrolysis via key structural and conformational changes (1,2). These various movements include transient domaindomain interactions such as those described for the P-and N-domains of P 2 -type ATPases (25)(26)(27)(28). Electrostatic surface maps of the CopA ATPBD (Fig. 5a) do not reveal any obvious patches of positive or negative charge suggestive of a domain interaction site. Comparisons to the SERCA1 structures provide some insight into potential interaction surfaces, however. Mapping the known A-domain interaction sites in SERCA1 (28) onto the surface of the CopA ATPBD identifies specific parts of the P-domain as likely to interact with the A-domain (Fig. 5, b and c). These sites correspond to the slightly positively charged surface at the crest of the P-domain ATP-binding cleft (Fig. 5a). Notably, interactions between the A-domain and the N-domain in the SERCA1 structures (26 -29) involve regions of secondary structure not present in P 1B -ATPases. This observation does not preclude possible interactions between the A-and N-domains in CopA but could be related to the additional requirement for metaldependant interactions between the N-MBDs and the ATPBD (17,44) and between the N-MBDs and metallochaperones (46).
Structural Significance of WND and MNK Mutations-A number of mutations in the ATPBDs of WND and MNK have been identified in Wilson and Menkes diseases patients (Fig. 2) (40). The structure of the CopA ATPBD provides new insight into the probable effects of these mutations on enzyme function. For example, mutations T1031S, T1033A, G1035V, and R1038K occur in the WND hinge region and are predicted to alter P-and N-domain movement. Other mutations might affect ligand binding by direct participation in ligand-protein interactions or by proximity to interacting residues. These mutations not only include the well characterized mutations in the SEHP region like H1069Q but also the I1102T, C1104F, I1148T, and R1151H WND mutations that probably affect the positions of glycines interacting with the nucleotide. A similar case can be made for mutations that involve the phosphate binding site in the P-domain, like T1220M, D1222N, D1267A, and N1270S in WND. Finally, several mutations occur in regions not directly involved in ligand interaction or conformational changes. In WND, these mutations include L1043P, L1083P, T1232P, V1252I, and D1296N. Such mutations may affect the folding and stability of the protein.
In sum, the CopA ATPBD structure reveals the key components required for ATP binding and hydrolysis by P 1B -type ATPases. The overall fold, which resembles that observed in P 2 -type ATPases, represents the basic scaffold capable of performing key steps in the classical P-type ATPase E1/E2 mechanism. The structure additionally serves as a basis to explore domain-domain interactions specific to heavy metal transport. Finally, the CopA ATPBD structure provides a new framework for the structural analysis of the many mutations leading to Wilson and Menkes diseases. FIGURE 5. Surfaces on the P-domain predicted to interact with the A-domain. a, electrostatic surface of the CopA ATPBD. Positively charged surfaces are colored blue, and negatively charged surfaces are colored red. b, the A-domain interaction sites from SERCA1 mapped onto CopA (left). Magenta surfaces correspond to predicted interactions in the E1 and E1P-ADP states, and green surfaces correspond to additional interactions occurring upon formation of the E2 state. Residue His 462 is colored orange to denote the nucleotide binding site. A ribbon representation of the structure with the same color coding is shown on the right. The electrostatic surfaces were generated by using APBS (43).