Analysis of the Catalytic and Binding Residues of the Diadenosine Tetraphosphate Pyrophosphohydrolase from Caenorhabditis elegans by Site-directed Mutagenesis*

The contributions to substrate binding and catalysis of 13 amino acid residues of theCaenorhabditis elegans diadenosine tetraphosphate pyrophosphohydrolase (Ap4A hydrolase) predicted from the crystal structure of an enzyme-inhibitor complex have been investigated by site-directed mutagenesis. Sixteen glutathioneS-transferase-Ap4A hydrolase fusion proteins were expressed and their k cat andK m values determined after removal of the glutathione S-transferase domain. As expected for a Nudix hydrolase, the wild type k cat of 23 s−1 was reduced by 105-, 103-, and 30-fold, respectively, by replacement of the conservedP 4-phosphate-binding catalytic residues Glu56, Glu52, and Glu103 by Gln.K m values were not affected, indicating a lack of importance for substrate binding. In contrast, mutating His31 to Val or Ala and Lys83 to Met produced 10- and 16-fold increases in K m compared with the wild type value of 8.8 μm. These residues stabilize theP 1-phosphate. H31V and H31A had a normalk cat but K83M showed a 37-fold reduction ink cat. Lys36 also stabilizes theP 1-phosphate and a K36M mutant had a 10-fold reduced k cat but a relatively normalK m . Thus both Lys36 and Lys83 may play a role in catalysis. The previously suggested roles of Tyr27, His38, Lys79, and Lys81 in stabilizing theP 2 and P 3-phosphates were not confirmed by mutagenesis, indicating the absence of phosphate-specific binding contacts in this region. Also, mutating both Tyr76 and Tyr121, which clamp one substrate adenosine moiety between them in the crystal structure, to Ala only increased K m 4-fold. It is concluded that interactions with the P 1- andP 4-phosphates are minimum and sufficient requirements for substrate binding by this class of enzyme, indicating that it may have a much wider substrate range then previously believed.

Ap 4 A 1 hydrolases are enzymes that hydrolyze dinucleoside polyphosphates. Structurally and mechanistically, they fall into two groups. The symmetrically cleaving enzymes (EC 3.6.1.41), such as Escherichia coli ApaH, generate 2-ADP from Ap 4 A, whereas the asymmetrically cleaving enzymes (EC 3.6.1.17) produce AMP and ATP (1,2). The latter are members of the Nudix hydrolases, a family of structurally and catalytically similar enzymes that act upon a wide range of different nucleotide substrates. Some are highly specific whereas others appear to have a broad substrate range in vitro (3)(4)(5). The Nudix Ap 4 A hydrolases can be further subdivided into "plant" and "animal"-types, according to their primary structure (6). The plant-type includes enzymes from the Proteobacteria that have in some cases been shown to be associated with the invasion of mammalian cells, whereas the animal-type includes putative Ap 4 A hydrolases from Archaea (6 -10). Early studies of both animal and plant Ap 4 A Nudix hydrolases employing a combination of substrate analogues and labeling with heavy isotopes of oxygen revealed the mechanism of hydrolysis to involve in-line nucleophilic attack of a water molecule at the P 4 (P␣) phosphate with subsequent breakage of the P 4 -(O)P 3 bond (8,(11)(12)(13)(14). Recently, the catalytic residues of the lupin Ap 4 A hydrolase involved in this process were identified by a combination of structural analysis and site-directed mutagenesis (15)(16)(17). This study supported the catalytic mechanism previously described in detail for the prototypical Nudix hydrolase, the E. coli  Detailed structural studies of E. coli MutT first showed the importance of the highly conserved residues in the loop-helixloop Nudix motif (Fig. 1). Glu 53 , Glu 56 , Glu 57 , Glu 98 (outside the linear motif but structurally close), and the carbonyl of Gly 38 coordinate an enzyme-bound Mg 2ϩ ion. A water ligand of this ion is oriented or deprotonated for nucleophilic attack by Glu 53 , which is itself oriented by Arg 52 . A second metal ion is complexed to the substrate and neutralizes the charge on the attacked phosphate while Lys 39 activates the NMP leaving group (18 -21). The importance of Glu 57 was indicated by a 10 5 -fold reduction in k cat in a E57Q mutant (19). The contributions of the other residues to catalysis were also confirmed by site-directed mutagenesis: E53Q, E56Q, and E44Q led to 10 4.7 -, 25-, and 14-fold decreases in k cat , respectively (20), whereas K39Q and R52Q produced 8-fold and Ͼ10 3 -fold reductions, respectively (22). The principle of this catalytic mechanism appears to be well conserved among the Nudix hydrolases, including the lupin and Bartonella bacilliformis Ap 4 A hydrolases (15)(16)(17)23), human MTH1 (24), yeast Dcp2p (25), and human NUDT3 (DIPP1) (26).
Among the asymmetrically cleaving Ap 4 A hydrolases, identification of residues responsible for substrate binding as well as catalysis is of interest for two reasons. First, it will help our understanding of the evolution of substrate specificity among the Nudix hydrolases. Second, if the plant-type Ap 4 A hydrolase of invasive pathogenic bacteria is to be considered as a target for new antibacterial agents, the design of such agents will require knowledge of the subtle differences between the plant and animal types if selectivity is to be achieved. Recently we reported the crystal structure of an animal Ap 4 A hydrolase from the nematode Caenorhabditis elegans in both free form and after crystallization in the presence of the substrate analogue, AppCH 2 ppA (27,28). The structure of the resulting binary complex allowed some predictions to be made about the importance of certain residues for substrate binding and catalysis and comparisons to be drawn with the lupin enzyme. Here, we extend these studies to include the effects of 19 site-specific mutations on Ap 4 A binding and hydrolysis by the C. elegans enzyme.

EXPERIMENTAL PROCEDURES
Synthesis of C. elegans First Strand cDNA Library-Total RNA was isolated and purified from washed adult nematodes (C. elegans strain N2) using Trizol solution (Invitrogen) according to the manufacturer's instructions. Full-length C. elegans first strand cDNA was synthesized from this RNA using a first strand cDNA synthesis kit (MPI Fermentas). RNA (2 l of 2.5 g/l) was added to 10 l of RNase-free ddH 2 O. The solution was mixed gently, incubated at 70°C for 5 min, and chilled on ice for 3 min before adding to a mixture containing 4 l of 5ϫ reaction buffer (250 mM Tris-HCl, pH 8.3, at 25°C, 375 mM KCl, 15 mM MgCl 2, 5 mM dithiothreitol), 1 l of ribonuclease inhibitor (20 units/l), 2 l of oligo(dT) 18 (0.5 g/l), 2 l of dNTPs (10 mM each), and 2 l of Moloney murine leukemia virus reverse transcriptase (20 units/l, Promega). The reaction was incubated at 42°C for 1 h, and then heated to 90°C for 5 min. The library was stored at Ϫ20°C.
Cloning of C. elegans Ap 4 A Hydrolase as a Glutathione S-Transferase (GST) Fusion Protein-A cDNA corresponding to the C. elegans Y37H9A.6 Ap 4 A hydrolase gene (6) was amplified from the cDNA library by PCR using the forward and reverse primers d(CAGCGCCAG-AATTCAATGGTCGTAAAAGCCGCGGG) and d(GAAATTACTCGAGA-AAAATCGTTAAAATCCGGC), respectively. These primers provided an EcoRI restriction site at the start of amplified cDNA and a XhoI site at the end. After amplification with Taq DNA polymerase, the DNA was recovered, digested with EcoRI and XhoI, and the required restriction fragment ligated between the EcoRI and XhoI sites of the pGEX-6P-3 vector (Amersham Biosciences). The resulting construct, pGEX-Y37H9A, encoded the 137-amino acid Ap 4 A hydrolase fused to the C terminus of GST through a 6-amino acid linker.
Generation of Site-specific Mutants-Site-directed mutagenesis was performed by PCR using the QuikChange TM site-directed mutagenesis kit (Stratagene). PCR reactions contained pGEX-Y37H9A as template, Pfu Turbo DNA polymerase, and pairs of complementary oligonucleotide primers 37 to 43 nucleotides long containing the required mutations (Table I). Each reaction volume was 50 l and contained the following: 50 -100 ng of plasmid DNA, 125 ng of each mutagenic primer, 200 M dNTPs, 10 mM KCl, 6 mM (NH 4 ) 2 SO 4 , 20 mM Tris-HCl, pH 8.0, 2 mM MgCl 2 , 0.1% Triton X-100, 10 g/ml bovine serum albumin, and 2.5 units of Pfu Turbo DNA polymerase. The PCR reaction protocol consisted of 2 min at 95°C followed by 16 cycles of 95°C for 1 min, 55°C for 1 min, 68°C for 14 min, followed by a final incubation at 72°C for 15 min. Parental DNA was digested with 10 units of DpnI to degrade the methylated parental strands and the remaining plasmid DNA was used to transform E. coli XL1-Blue cells. For production of the Y76A/ Y121A double mutant, the Y76A DNA construct was used as template in a PCR containing the Y121A mutagenic primers (Table I). The identities of all mutants were verified by complete sequencing of both DNA strands.
Expression and Purification of GST-Ap 4 A Hydrolase Fusion Proteins-E. coli strain BL21(DE3) was transformed with pGEX-Y37H9A or its mutant derivatives. Cultures (250 ml) in LB medium containing 50 g/ml ampicillin were grown to an A 600 of 0.7 at 37°C. Isopropyl-1thio-␤-D-galactopyranoside was added to 1 mM and incubation continued for 2 h. Induced cells (approximately 1.6 g) were harvested by centrifugation at 10,000 ϫ g, washed, and resuspended in 10 ml of ice-cold breakage buffer: 50 mM Tris-HCl, pH 8.0, 50 mM NaCl, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 5 M transepoxysuccinyl-L-leucylamido-(4-guanidino)butane) (E-64, Sigma). Cell suspensions were sonicated and the resulting lysates cleared by centrifugation at 15,000 ϫ g and 4°C for 10 min. Supernatants were recovered and applied to columns containing 2.5 ml of glutathione-Sepharose 4B (Amersham Biosciences). Columns were washed with 25 ml of phosphate-buffered saline, followed by 25 ml of PreScission cleavage buffer (50 mM Tris-HCl, pH 7.0, 150 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 5% (v/v) glycerol). Following complete elution of the buffer, the outlets were closed and 100 units of PreScission protease in 2.5 ml of cleavage buffer added to the resin and incubated for 18 -20 h with gentle rocking at 4°C. Cleavage of the GST domain from the Ap 4 A hydrolases was complete after 20 h. Columns were remounted, the resin left to settle, and the free Ap 4 A hydrolases containing the N-terminal extension GPLGSPNS eluted.
Ap 4 A Hydrolase Assay-Ap 4 A hydrolase activity was measured using a luciferase-based bioluminescence assay as previously described (6). One ng enzyme protein was used in each case, except for K83M (10 ng), K79M (20 ng), E52Q and E103Q (60 ng), and E56Q (600 ng). This sensitive, continuous assay permits direct evaluation of initial rates. The increase in luminescence was linear for several minutes for each enzyme.
Other Methods-Protein concentrations were estimated by the Coomassie Blue binding method (29) and protein molecular masses were determined by electrospray mass spectrometry as previously described (30).

Expression and Purification of Wild Type and Mutant GST-Ap 4 A Hydrolase Fusion
Proteins-A cDNA corresponding to the C. elegans Y37H9A.6 Ap 4 A hydrolase gene was amplified from a cDNA library by PCR and inserted into the pGEX-6P-3 GST fusion vector to generate the recombinant plasmid pGEX-Y37H9A. When E. coli BL21(DE3) cells were transformed with this plasmid and induced with isopropyl-1-thio-␤-D-galactopyranoside, a major soluble 43-kDa band corresponding to the expected GST-Ap 4 A hydrolase fusion protein was detected (data not shown). The GST domain of this protein was readily removed by on-column cleavage with PreScission protease, resulting in the free Ap 4 A hydrolase with the N-terminal extension GPLGSPNS and mass 16.6 kDa.
Specific mutations were introduced into the Ap 4 A hydrolase coding region of pGEX-Y37H9A by PCR (Table I). The mutations were confirmed by DNA sequencing. A total of 16 single mutants and one double mutant involving 13 different residues was generated in this way. Their positions in the primary structure of the Ap 4 A hydrolase are shown in Fig. 2A and the locations of their ␣-carbon atoms in the three-dimensional structure of the binary complex are shown in Fig. 2B. Each was expressed as a GST fusion protein and purified after on-column cleavage and elution as described for the wild type. The predicted masses of the cleaved proteins were confirmed by mass spectrometry. With the exception of five (K36M, Y76A, Y76A/ Y121A, K79M, and W32G) mutant proteins were substantially expressed in the soluble fraction (at least 40% of the total expressed recombinant protein) and the yields of purified proteins were nearly the same as for the wild type. The first four exceptions yielded about 5% of the recombinant protein in a soluble form, whereas W32G was completely insoluble when expressed. All kinetic data were determined using enzymes purified from the soluble fractions, which were all judged to be more than 95% pure by SDS-PAGE.
Effects of Mutations on the Catalytic Properties of Ap 4 A Hydrolase-The values of K m (8.8 M) and k cat (23 s Ϫ1 ) estimated for the N-terminal extended wild type Ap 4 A hydrolase after cleavage from the GST domain were close to those previously reported for the native recombinant enzyme purified by conventional procedures (7.0 M and 27 s Ϫ1 , respectively) (6). From this we conclude that the 8-amino acid N-terminal extension does not interfere significantly with the binding of the substrate or with catalysis. The linearity of light output from the luminometric assay used was the same for all mutants as for the wild type, indicating the stability of the mutants under assay conditions. K m and k cat values were then determined for each mutant enzyme to determine the effects of each mutation on substrate binding and catalysis (Table II). For this enzyme, K m can be taken to approximate the dissociation constant of the ES complex(es), and hence as an inverse measure of affinity, based on the lack of effect of active site (k cat ) mutants on the value of K m . Previous mutational studies with Nudix hydrolases have highlighted the importance of the Glu residues within the Nudix motif for catalysis ( Fig. 1) (16,20,25,26,31,32). Not surprisingly, therefore, the E56Q mutation was found to result in a 10 5 -fold reduction in k cat and virtual abolition of detectable enzyme activity, exactly as was found for the equivalent residue (Glu 59 ) in the lupin Ap 4 A hydrolase; in contrast, the K m was unaffected, indicating that the mutation has no  effect on substrate binding. Similarly, neutralization of the charge on Glu 52 , the second of the three highly conserved Glu residues within the Nudix motif, by conversion to Gln (E52Q) reduced k cat by a factor of 10 3 but again had little effect on K m (Table II). On the basis of the 10 5 -fold reduction in k cat , we previously proposed that Glu 56 was most likely to be the catalytic base that deprotonates the attacking water molecule. However, the structural equivalents of Glu 52 in the lupin Ap 4 A hydrolase (Glu 55 ) and in the E. coli MutT protein (Glu 53 ) have been proposed as the deprotonating base (11,16,20). Glu 103 , although not in the Nudix motif, is positioned close to it in the three-dimensional structure and coordinates two of the four Mg 2ϩ ions located in the catalytic site (27). E103Q has a 30-fold lower k cat than the wild type and a similar K m . The equivalent mutations in E. coli MutT (E98Q) and the lupin Ap 4 A hydrolase (E125Q) produced 6.3-and 140-fold reductions in k cat , respectively (16,20). Whereas these values suggest that Glu 103 and its equivalents are unlikely to be the catalytic base in these enzymes, a detailed structural analysis of E. coli ADP-ribose pyrophosphatase has led to the conclusion that the equivalent Glu 162 has this role in that enzyme (33,34). Thus, although the architecture of the catalytic sites are broadly similar among the Nudix hydrolases, the mechanism of proton abstraction from the attacking water/hydroxyl appears to be subtly different in different family members.
Whereas several mutational studies of Nudix hydrolases have highlighted the importance to catalysis of particular residues in and adjacent to the catalytic motif, less attention has been focused on residues elsewhere that may be involved in substrate recognition and binding as well as catalysis. The crystal structure of the C. elegans Ap 4 A hydrolase binary complex showed that the adenine ring in the "AMP-binding pocket" distal to the P 4 -phosphate (the site of nucleophilic attack) was sandwiched between the phenolic rings of Tyr 76 and Tyr 121 and formed extensivestacking interactions with these residues. To achieve this, a 90°rotation of the phenolic ring of Tyr 121 about the 1 dihedral angle and an associated shift in Tyr 76 occurred upon substrate binding (27). Both of these residues are highly conserved in structure-based sequence alignments of animal and plant Ap 4 A hydrolases. This, coupled with further interactions between the side chain hydroxyl group of Tyr 121 and the 2Ј-OH of the attached ribose and between the ring of Tyr 76 and the ribose O 4 oxygen suggested that both Tyr residues should be important for substrate binding, therefore the effects of replacing each with Ala were investigated. As expected, both Y76A and Y121A showed an increased K m , but only by a factor of about 8 (Table II). Surprisingly, the combination of mutations in the double mutant Y76A/Y121A appeared to reduce the K m , again such that it was only 4-fold higher than the wild type. This suggests that the substrate may be able to bind effectively in a way that is independent of the Tyr residues (see "Discussion"). This alternative binding does lead to a lower catalytic rate, as evidenced by the reduced k cat values (20-fold less in the double mutant).
The crystal structure of the binary complex also showed that the P 1 -phosphate attached to the above adenosine moiety is stabilized on the enzyme via a series of hydrogen bonds/salt bridges between the phosphate oxygens and the side chain NZ nitrogens of Lys 36 and Lys 83 , the side chain hydroxyl group of Tyr 76 , and the side chain imidazole ring N ⑀2 of His 31 (27). Replacement of His 31 with Ala (H31A) or Val (H31V) had a significant and specific impact on the binding of Ap 4 A, increasing the K m 8 -12-fold while only marginally reducing k cat . Loss of the NZ nitrogen of Lys 83 (K83M) led to an even greater increase in K m (16-fold) but in this case k cat was also substantially reduced (37-fold). The K36M mutant also had a reduced k cat (10-fold) but a relatively normal K m . These results confirm the predictions of the crystal structure and indicate the importance of Lys 36 and Lys 83 , which is positioned such that it could also stabilize the P 2 -phosphate, to catalysis (27).
As there was no interpretable electron density for either the P 2 -or P 3 -phosphates in the binary complex, it was suggested that the side chains of His 38 (within the Nudix motif) plus Lys 79 , Lys 81 , and Tyr 27 (outside the Nudix motif) might be in appropriate positions to participate in P 2 -and P 3 -phosphate stabilization either by direct interaction or by metal coordination. The main chain amide of His 38 is also the only direct protein contact with P 4 via a hydrogen bond to one of the oxygen atoms (27). Potentially, His 38 (structurally equivalent to Lys 39 in the E. coli MutT protein) and/or Lys 79 and/or Lys 81 could neutralize the developing negative charge on the ATP leaving group, in much the same way as has been proposed for MutT Lys 39 (18,20). Therefore, appropriate mutants were generated to test these suggestions. Surprisingly, of the mutants analyzed (Y27A, Y27D, H38G, H38K, K79M, and K81M), only K79M showed a significant change in any kinetic constant, a substantial 140-fold reduction in k cat . However, Lys 79 , like Tyr 27 , is not well conserved among the Ap 4 A hydrolases, so this reduced activity may reflect a slight structural alteration in the protein rather than an important catalytic role. In contrast, Lys 81 is well conserved as a basic residue in animal and plant Ap 4 A hydrolases. However, its mutation to Met resulted in a slight increase in k cat to 30 s Ϫ1 , so it seems unlikely to be involved in stabilizing the leaving group. H38G and H38K also showed slight increases in k cat such that the k cat /K m ratio was 2.5-fold higher than the wild type. The equivalent residue in plant Ap 4 A hydrolases is a Gly, so, unlike Lys 39 in MutT, His 38 does not appear to be important for catalysis either. As all other residues in the region are too small to make contact, the conclusion is that there are few, if any, structurally or mechanistically important binding contacts for the P 2 -and P 3 -phosphates. This interesting point is discussed further below.
Finally, Trp 32 is a completely conserved residue among Ap 4 A hydrolases of the Nudix family and is commonly found in other family members. This residue does not form interactions with the substrate but appears to stabilize the protein fold through interactions with Leu 22 and Gln 24 in the ␤B strand and with Ile 118 in the ␣II helix (27). Consistent with this essential structural role is the fact that the W32G mutant was completely insoluble and inactive when expressed.

DISCUSSION
The results of this analysis confirm the importance of residues previously implicated by structural analysis in binding and catalysis at the P 4 -phosphate and binding of the P 1 -phosphate of Ap 4 A by the C. elegans Ap 4 A hydrolase. It also provides important information on two further aspects of substrate binding by this enzyme that have implications for substrate recognition in the Nudix hydrolase family as a whole. First, there appear to be few, if any, important binding contacts for the P 2 -and P 3 -phosphates. Our previous structural analysis was unable to provide this information. It is possible that a nucleotide-bound metal ion is more important for stabilization of the negative charge during catalysis than any individual amino acid side chain. Such an ion appears to be required by the B. bacilliformis Ap 4 A hydrolase (23) and the E. coli ADPribose pyrophosphatase (33). However, the apparent absence of specific interactions in this region is entirely consistent with our recent discovery that this and related dinucleoside polyphosphate hydrolases can bind and hydrolyze 5-phosphoribosyl 1-pyrophosphate (35). In this case the ribose ring would occupy the P 2 ,P 3 site, with binding dependent solely on interactions in the P 1 and P 4 sites, in agreement with the data above.
Second, the mutational data relating to Tyr 76 and Tyr 121 were somewhat unexpected. The 90°rotation of the phenolic ring of Tyr 121 in the binary complex compared with the apoenzyme and thestacking interactions between both rings and the adenine ring of the adenosine moiety attached to the P 1phosphate originally suggested an essential role for these residues. However, substitution of both Tyr residues by Ala did not have the expected dramatic effect on substrate binding and yielded only a 4-fold increase in K m . Again, this is consistent with the finding that 5-phosphoribosyl 1-pyrophosphate, which lacks a base altogether, is a substrate. Thus, binding of one adenine ring between the Tyr residues is not an essential requirement for catalysis, although it undoubtedly contributes to the higher specificity constant for Ap 4 A compared with 5-phosphoribosyl 1-pyrophosphate (35). Interestingly, in the NMR structure of the lupin Ap 4 A hydrolase complexed with the substrate analogue ATP-MgF x , the adenine ring is not located between the structurally equivalent residues Tyr 77 and Phe 144 , and instead Tyr 77 is suggested to be important for the structural integrity of the enzyme-substrate complex rather than for direct substrate binding (17). Although structurally very similar, plant and animal Ap 4 A hydrolases typically show only 25-30% sequence similarity outside the Nudix motif and appear to form two distinct evolutionary groups within the family (6). Thus, either the animal and plant enzymes differ substantially in the way they bind substrate, or the possibility exists that Ap 4 A can bind to both Ap 4 A hydrolases in two different ways. Conceivably, one site represents the true substrate binding site before hydrolysis whereas the other is a transitional site for the ATP leaving group. Indeed, in view of the lack of electron density for the P 2 -and P 3 -phosphates and the second adenosine moiety in the crystal structure of the C. elegans binary complex, we have suggested that the analogue AppCH 2 ppA was probably hydrolyzed during crystallization (27). Thus, the visible AMP may be the AppCH 2 p product with its mobile P 2 -and P 3 -phosphates disordered in the crystal lattice and therefore invisible, or even the AMP product that has re-bound between the Tyr residues after departure of the ATP. If binding of the adenine ring between Tyr 76 and Tyr 121 is not essential, this could also explain our observations that phosphonate analogues with isopolar halomethylene groups bridging P 1 and P 2 and P 3 and P 4 such as ApCHFppCHFpA, and P 1 ,P 4 -thiophosphates such as Ap s ppp s A can be cleaved symmetrically by Nudix Ap 4 A hydrolases (14,36). This requires attack at P 2 or P 3 . Assuming that the loop-helix-loop structural motif containing the catalytic residues cannot move significantly, symmetrical hydrolysis implies that the substrate is bound in such a way as to present P 3 rather than the usual P 4 to the attacking nucleophile. Thus, substrates are able to bind in more than one location. Taken together, the ability of some Nudix hydrolases to use non-nucleotide substrates such as 5-phosphoribosyl 1-pyrophosphate and diphosphoinositol polyphosphates and the flexibility of substrate binding noted above suggest that the substrate range and function of Nudix hydrolases may be much wider than previously believed.