Structural and catalytic similarities between nucleotide pyrophosphatases/phosphodiesterases and alkaline phosphatases.

Nucleotide pyrophosphatases/phosphodiesterases (NPPs) generate nucleoside 5'-monophosphates from a variety of nucleotides and their derivatives. Here we show by data base analysis that these enzymes are conserved from eubacteria to higher eukaryotes. We also provide evidence for the existence of two additional members of the mammalian family of ecto-NPPs. Homology searches and alignment-assisted mutagenesis revealed that the catalytic core of NPPs assumes a fold similar to that of a superfamily of phospho-/sulfo-coordinating metalloenzymes comprising alkaline phosphatases, phosphoglycerate mutases, and arysulfatases. Mutation of mouse NPP1 in some of its predicted metal-coordinating residues (D358N or H362Q) or in the catalytic site threonine (T238S) resulted in an enzyme that could still form the nucleotidylated catalytic intermediate but was hampered in the second step of catalysis. We also obtained data indicating that the ability of some mammalian NPPs to auto(de)phosphorylate is due to an intrinsic phosphatase activity, whereby the enzyme phosphorylated on Thr-238 represents the covalent intermediate of the phosphatase reaction. The results of site-directed mutagenesis suggested that the nucleotide pyrophosphatase/phosphodiesterase and the phosphatase activities of NPPs are mediated by a single catalytic site.

Nucleotide pyrophosphatases/phosphodiesterases (NPPs) generate nucleoside 5-monophosphates from a variety of nucleotides and their derivatives. Here we show by data base analysis that these enzymes are conserved from eubacteria to higher eukaryotes. We also provide evidence for the existence of two additional members of the mammalian family of ecto-NPPs. Homology searches and alignment-assisted mutagenesis revealed that the catalytic core of NPPs assumes a fold similar to that of a superfamily of phospho-/sulfo-coordinating metalloenzymes comprising alkaline phosphatases, phosphoglycerate mutases, and arysulfatases. Mutation of mouse NPP1 in some of its predicted metalcoordinating residues (D358N or H362Q) or in the catalytic site threonine (T238S) resulted in an enzyme that could still form the nucleotidylated catalytic intermediate but was hampered in the second step of catalysis. We also obtained data indicating that the ability of some mammalian NPPs to auto(de)phosphorylate is due to an intrinsic phosphatase activity, whereby the enzyme phosphorylated on Thr-238 represents the covalent intermediate of the phosphatase reaction. The results of site-directed mutagenesis suggested that the nucleotide pyrophosphatase/phosphodiesterase and the phosphatase activities of NPPs are mediated by a single catalytic site.
Mammalian nucleotide pyrophosphatases/phosphodiesterases (NPPs) 1,2 are type II transmembrane proteins with a small intracellular domain (10 -80 residues) and a large extracellular domain (ϳ830 residues) that also harbors the catalytic site (1,2). Since their catalytic site is extracellular, this family of proteins are denoted as ecto-NPPs. However, members of this family can also be processed into soluble, secreted forms by proteolysis at specific sites C-terminal to the transmembrane domain (3)(4)(5). In vitro, NPPs release nucleoside 5Ј-monophosphates from nucleotides and a variety of nucleotide derivatives. For example, NPPs efficiently hydrolyze ATP into AMP and PP i . This catalysis occurs via a covalent intermediate, i.e. a nucleotidylated threonine in the catalytic site (6). Some NPPs are also able to autophosphorylate and autodephosphorylate this catalytic site threonine. The autophosphorylation is associated with an inhibition of the nucleotide pyrophosphatase/ phosphodiesterase activity (6).
The three known members of the mammalian ecto-NPP family have a relatively broad tissue distribution and have been implicated in a variety of cellular processes. NPP1, previously known as PC-1, is likely to be involved in soft tissue calcification and bone mineralization through its ability to generate PP i , an inhibitor of calcification and mineralization (8). However, NPP1 has also been shown to oppose insulin signaling, and this effect does not require a functional catalytic site (9). The anti-insulin effects of NPP1 may be the result of its direct interaction with the insulin receptor (10). NPP2 or autotaxin was shown to stimulate cell motility (11) and to augment the invasive and metastatic potential of ras-transformed cells (12). NPP3, also known as gp130 RB13-6 (13) and B10 (5), has been reported to promote the differentiation and invasive properties of glial cells (13).
Although most of the biological effects of NPPs appear to be mediated by their NPP activity, the physiological NPP substrates remain to be identified. It is our long term goal to obtain information on the NPP substrate(s) by expressing mutants that can still form the catalytic intermediate but can no longer hydrolyze this intermediate. Characterization of this "trapped" covalent adduct should then provide data on the nature of the substrate(s). As a first step toward this goal, we report here on mutations that specifically block the second catalytic step of NPPs. These "trapping" mutants were designed on the basis of results of a structure-prediction analysis, which suggested that the catalytic domain of NPPs is structurally and catalytically related to that of well characterized enzymes such as alkaline phosphatase. Unexpectedly, these studies have also revealed that a single catalytic mechanism accounts for both the nucleotide pyrophosphatase/phosphodiesterase and the auto (de)phosphorylation activities of NPPs.

MATERIALS AND METHODS
Sequence Analysis-The organization of the genes encoding human NPP2, human NPP4, and mouse NPP5 was determined by comparing the respective cDNA sequences with the genomic sequences in the GenBank data base, using the BLAST-2 sequences program at NCBI (14). Intron-exon boundaries were verified according to Horowitz and Krainer (15) 2 The NPP nomenclature used here stems from a broad consensus at a recent international workshop on "Ecto-ATPases and Related Ectonucleotidases" (7).
Genomic Research (available via the World Wide Web). The sequence of the Clostridium acetobutylicum NPP was obtained from Genomic Therapeutics Corp. (available via the World Wide Web).
Generation and Expression of NPP Mutants-Various mutants of mouse NPP1 were expressed as HA-tagged fusion proteins in COS-1 cells (Figs. 6 and 8). The expression vectors were generated by a twostep procedure. First, the pSVL-SV40 vector was cut with XhoI and BamHI, and ligated with an adaptor encompassing the sequence of an HA tag (YPYDVPDYA) and a multiple cloning site, yielding the pMB001 vector. The mouse NPP1 coding sequence, obtained from the pSVL/PC-1 plasmid (17), was cut with XbaI and BamHI and subcloned in the pMB001 vector. Point mutations were introduced using the pMB001 vector and the QuickChange site-directed mutagenesis protocol of Stratagene. PCR amplifications were performed using Pwo proofreading polymerase (Roche Diagnostics). All mutations were verified by DNA sequencing.
COS-1 cells were grown in Dulbecco's modified Eagle's medium containing 10% (v/v) fetal bovine serum and 100 units/ml each of penicillin and streptomycin. The cells were transfected with the NPP constructs at 30 -40% confluence, using the FuGene6 reagent (Roche Diagnostics). After 48 -72 h the cells were washed twice in ice-cold phosphatebuffered saline and lysed in 20 mM Tris/HCl at pH 7.5, 0.5 mM phenylmethanesulfonyl fluoride, 0.5 mM benzamidine, 0.3 M NaCl, and 0.2% Triton X-100. After centrifugation (5 min at 5000 ϫ g), the supernatant was used for immunoprecipitation of the HA-NPP fusions with monoclonal HA tag antibodies (clone 12CA5) and protein A-TSK (Affiland). The immunoprecipitates were washed once with 0.25 M LiCl and twice with Tris-buffered saline, resuspended in 50 mM HEPES at pH 7.5, and assayed for phosphodiesterase activity and for the ability to trap the covalent intermediates of the phosphatase and the phosphodiesterase reactions.
Western Analysis-Following Tricine-SDS-PAGE (7.5%) the proteins were transferred onto polyvinylidene difluoride membranes (Bio-Rad) by electroblotting at 38 V in 50 mM Tris-base plus 50 mM boric acid at pH 8.3. Unspecific binding sites were blocked in phosphate-buffered saline containing 5% (w/v) milk powder and 0.2% (v/v) Triton X-100. Affinity-purified polyclonal antibodies against a synthetic peptide encompassing the 14 C-terminal residues of mouse NPP1 (LRLKTH-LPIFSQED) were used to visualize the HA-NPP fusion proteins. Peroxidase-labeled secondary antibodies were used for chemiluminescent detection.
Phosphodiesterase Activity and Covalent Intermediates-The nucleotide phosphodiesterase activity was measured with p-nitrophenyl thymidine 5Ј-monophosphate as substrate (6) The protein sequences were aligned using CLUSTAL X (22) and the alignments were subsequently refined manually. The GenBank accession numbers of the listed proteins are D12485 (HsNPP1), D45421 (HsNPP2), AF005632 (HsNPP3), AB020686 (HsNPP4), and AL035701 (HsNPP5). The latter represents a genomic sequence, but the exonic sequences were derived by sequencing of a commercial mouse cDNA clone (IMAGE clone 553842). The numbers indicate the distance to the beginning and the end of each protein. Conserved and identical residues are boxed. The figure also shows an NPP "consensus sequence," as derived from an alignment of all prokaryotic and eukaryotic NPPs that were obtained by data base searching (see Fig. 2). In the consensus sequence, the following symbols are used: , charged residue (D, E, K, R, N, Q); , bulky hydrophobic residue (I, L, V, M, F, Y); , small residue (G, A, S, T); , acidic residues (D, E); ⅐, no particular residue. The catalytic site threonine is underlined in the consensus sequence.

RESULTS AND DISCUSSION
A Family of NPP Proteins-A search of the nonredundant sequence data base at NCBI with NPP1-3, using the BLASTalgorithm (21), resulted in the identification of two novel mammalian NPP homologues, designated NPP4 and NPP5 (Fig. 1). An EST analysis revealed a rather broad tissue distribution of these novel putative NPPs (data not shown). Human NPP4 (453 residues) and NPP5 (477 residues) are considerably smaller than NPP1-3 (863-925 residues), which can be accounted for by the absence in NPP4 -5 of N-terminal sequences corresponding to the cytoplasmic tail, the transmembrane region, and somatomedin-B-like domains of NPP1-3. In addition, NPP4 and NPP5 are about 250 residues shorter at the C terminus. NPP4 and 5 are more closely related to each other (51% identity) than to the corresponding domain of NPP1-3 (29 -37% identity). The genes encoding human NPP4 and NPP5 (GenBank accession no. AL035701) only contain 2 introns, as compared with 24 introns for the NPP1-3-encoding genes (23,24). However, the two introns in the NPP4 and NPP5 genes coincide with introns 13 and 15 in the genes encoding NPP1-3, providing additional evidence for their common ancestry.
In addition to two novel mammalian NPPs, homology searches revealed the existence of proteins ranging from bacteria to plants that shared a significant similarity with the catalytic domain of the mammalian NPPs (Fig. 2). The conservation included the catalytic site threonine and its surrounding residues (see consensus sequence in Fig. 1). NPPs Belong to a Superfamily of Phospho-/Sulfo-coordinating Metalloenzymes-To identify more distant NPP relatives, the nonredundant protein sequence data base at NCBI was searched with the PSI-BLAST algorithm (27), using the sequence of the catalytic domain of mouse NPP1 (residues 192-568) as a query. Convergence was reached after 14 iterations, yielding proteins that could be clustered into mainly five different enzyme families. Besides the NPP family, these included the families of arylsulfatases (AS), phosphopentomutases, 2,3bisphosphoglycerate-independent phosphoglycerate mutases (iPGM), and the alkaline phosphatases (AP). The latter four families have previously also been classified in a superfamily of phospho-/sulfo-carbohydrate processing metalloenzymes, based on the conservation of metal binding motifs (28). Furthermore, the crystal structures of E. coli AP (PDB code 1ALK), human AS-A and AS-B (PDB codes 1AUK and 1FSU, respectively) and Bacillus stearothermophilus iPGM (PDB code 1EJJ) have been solved (29 -32), which has provided extensive proof of structural and catalytic similarities.
A PSI-BLAST search with the sequence of E. coli AP (SwissProt P00634) attributed a higher similarity score to NPPs than to AS (data not shown). Conversely, when the human AS-A sequence (SwissProt P15289) was used as query, NPPs emerged with a higher similarity score than did alkaline phosphatases. Thus, within the superfamily of phospho-/sulfocoordinating metalloenzymes, the NPP family is positioned intermediate between ASs and APs. Since the catalytic domains of APs, ASs, and iPGMs have been shown to assume a basically identical fold, i.e. a ␤-sheet sandwiched between ␣-helices (29 -32), it seems likely that the catalytic domain of NPPs would adopt the same fold. In agreement with this view, the fold recognition program GenTHREADER (33), which uses FIG. 4. Model of the catalytic core of NPPs. The primary structure of the catalytic domain of mouse NPP1 was threaded onto the known three-dimensional structure of the catalytic domain of alkaline phosphatase (PDB code 1ALK). Using the known (AP) and predicted (NPP) secondary structure as criteria (see Fig. 3A), the six Zn 2ϩ -coordinating residues of AP showed a superposition with identical residues in all NPPs. In addition, the catalytic site Thr of NPP1 coincided with that of the catalytic site Ser of AP. The figure shows the predicted constellation of the metal-binding residues and the catalytic site Thr of mouse NPP1. A similar active site constellation has been derived from the crystal structure of AS-A, AS-B, and iPGM (30 -32). Red, oxygen; blue, nitrogen; green, metal (Me). a threading potential to evaluate the quality of fold assignments, proposes with the highest level of confidence that the fold of the human AS-A and AS-B occurs in the catalytic domain of human NPP1.
Since a combination of various, independent prediction methods has been shown to improve the reliability of secondary structure prediction (34, 35), a consensus secondary structure was predicted for the catalytic domain of mouse NPP1, with the use of the PSI-PRED, PHD, SamT99, JPRED, and Prof prediction algorithms (36 -40). The predicted secondary structure was aligned with the known secondary structure common to the catalytic domains of APs, iPGMs, and ASs. This alignment revealed a remarkable conservation in the length and position of ␣-helices and ␤-strands between these four families (Fig.  3A). By threading the sequence of the predicted secondary structure elements of NPP1 onto the known secondary structure of the AP, iPGM, and AS backbones, we were able to produce a rough structural model of the catalytic domain of NPPs (Fig. 3B). A striking feature of this model is that the metal-binding and active site residues in NPP1 are superposed onto those of AP and iPGM. Thus, the 6 residues in APs and iPGMs that are known to coordinate the two metals in their catalytic site are conserved in the NPPs (Figs. 3A and 4) and were predicted to display a similar spatial arrangement rela-tive to the residue that forms the catalytic intermediate, i.e. a Thr in NPPs (6,42,43) and a Ser in AP and iPGM (32,44,45). Arylsulfatases have a similar catalytic core, but differ from AP and iPGMs in that they have only a single metal-binding site and that the residue that forms the catalytic intermediate is an oxidated cysteine (46).
The Reaction Mechanism of NPPs-Since we predicted the catalytic site of NPPs to have an arrangement similar to that of APs and iPGMs, this also suggested a similar catalytic mechanism. Transposing the known reaction mechanism of APs (44) and iPGMs (32) to NPPs yields the reaction scheme shown in Fig. 5. It is proposed that the Me2-activated nucleophilic hydroxyl of Thr-238 attacks the phosphate of the incoming substrate, resulting in the generation of a covalent, nucleotidylated intermediate. In the second catalytic step, a Me1activated water molecule attacks this E-NMP intermediate, regenerating Thr-238 and releasing a nucleoside 5Ј-monophosphate. An essential difference between NPPs and APs lies in the nature of their respective substrates, i.e. a phosphodiester and a phosphomonoester, respectively. In APs the substrate affinity is increased by an arginine (Arg-166 in E. coli AP), which is hydrogen-bonded to the two oxygen atoms of the phosphate moiety that are not interacting with metal ions. In the typical NPP substrates, however, one of these oxygens is FIG. 5. Proposed reaction mechanism for NPPs. The proposal is based on the known reaction mechanism of APs (44) and iPGMs (32), and the conserved catalytic site structure elements of NPPs, including the residues that coordinate Me1 and Me2 (Figs. 3 and 4). The open circles represent OH-or O-groups. R refers to various structures, including a nucleoside monophosphate (e.g. in AP 2 A), a phosphate (e.g. in ADP), or a pyrophosphate (e.g. in ATP). In the free enzyme (E), the hydroxyl group of the catalytic site threonine (Thr-238 in mouse NPP1) is stabilized by bonded with the nucleosidyl group (see Fig. 5). We suggest that binding of the latter in a nucleosidyl binding pocket serves a function similar to that of the arginine-mediated hydrogen bonding in APs.
Several lines of evidence provide support for the proposed reaction scheme of NPPs. First, it has already been established that the NPP reaction occurs in two steps and that a nucleotidylated intermediate is formed with the catalytic site Thr (6). Second, it is well known that NPPs are metalloenzymes since their activity is blocked by metal chelators such as EDTA (47,48). Third, the activity of NPP1 is blocked by imidazole (18,43) and by the histidine-acylator diethylpyrocarbonate (results not shown), in keeping with the proposed role of histidines in coordinating Me1 and Me2 (Figs. 4 and 5). We have also obtained evidence in favor of the proposed reaction mechanism by site-directed mutagenesis of NPP1. For that purpose mouse NPP1 point mutants were expressed in COS-1 cells as HAtagged fusion proteins, immunoprecipitated from the cell lysates with anti-HA antibodies, and assayed for NPP activity, using p-nitrophenyl thymidine 5Ј-monophosphate as substrate. In accordance with previous reports (9, 49, 50), mutation of the catalytic site residue, i.e. T238S or T238A, decreased the NPP activity by 95 and 100%, respectively (Fig. 6B). By contrast, mutation of the neighboring residues (K237A and F239A) merely halved the activity. Mutation of any of the 6 residues (Figs. 4 and 5) predicted to be involved in the binding of Me1 (Asp-358, His-362, and His-517) or Me2 (Asp-200, Asp-405, and His-406) essentially abolished the NPP activity (Fig. 6B). Mutation of the corresponding residues in E. coli AP and B. stearothermophilus iPGM has been reported to be inhibitory as well (41,(51)(52)(53)(54)(55)(56)(57).
It can be envisaged that the activity of NPP1 that is mutated in its metal-coordinating residues, can be (partially) restored by the addition of an excess of metals. Since the identity of the metals in the catalytic site of NPPs is not known, we have first explored which metals can restore the activity of NPP1 following metal chelation by 1 mM EDTA. We found that the activity of EDTA-inhibited NPP1 could be largely restored by the mere addition of 2-5 mM amounts of either ZnCl 2 or CaCl 2 , but only partially by 2-5 mM MgCl 2 (data not shown). In Fig. 6B it is shown that the activity of the H362Q and H517Q mutants of NPP1 was also restored to about 60% of the control value by the addition of 2 mM ZnCl 2 . The activity of the other putative metal coordination mutants (D405N, H406Q, D358Q, D200N) was also increased severalfold by the addition of Zn 2ϩ , albeit to a lesser extent than that of the H362Q and H517Q mutants. As expected, the activity of NPP1 mutated in the catalytic site residue (T238A, T238S) or its flanking residues (K237A, F239A) were not at all or only marginally affected by the addition of Zn 2ϩ . These data provide additional proof for the role of Asp-200, Asp-358, His-362, Asp-405, His-406, and His-517 in the coordination of metals and also show that the deleterious effect of these mutations is not the result of gross conformational changes.
We have also investigated whether the mutants of NPP1 are blocked in the first and/or second step of catalysis. For that purpose, we compared their ability to trap the nucleotidylated catalytic intermediate, using [␣-32 P]ATP as a substrate (Fig.  6C). As expected, the inactive T238A had lost the ability to form the nucleotidylated intermediate. In contrast, the poorly active T238S mutant accumulated the nucleotidylated intermediate, which indicates that the hydrolysis of the intermediate was hampered. A similar reasoning applies to the D358N and H362Q mutants. An increased labeling of the latter mutants could be expected since Asp-358 and His-362 were proposed to be involved in the binding of Me1, which is predicted to play an essential role in the second step of catalysis (Fig. 5). The inability of the H517Q mutant to form the covalent inter- mediate may suggest that the mutation of His-517 to a glutamine, in addition to abolishing the binding of Me1, also disturbs the coordination of Me2. Indeed, the mutation of any residue predicted to be involved in the coordination of Me2 (D200N, D405N, H406Q) abolished the formation of the nucleotidylated intermediate, in agreement with the requirement of Me2 to start the catalytic cycle (Fig. 5). The F239A mutant, which displayed about half of the wild type NPP activity (Fig.  6B), also showed a decreased ability to trap the catalytic intermediate (Fig. 6C). Phe-239 is conserved in all NPPs but not in APs, iPGMs, and ASs, indicating that the role of this residue is specific for NPP substrates. For example, Phe-239 could be involved in the coordination of the substrate. If so, the decreased activity and level of covalent intermediate of the F239Q mutant may be accounted for by a decreased binding of the substrate.
NPPs Also Exhibit a Phosphatase Activity-Having concluded that NPPs have a catalytic site that is similar to that of APs, we wondered whether the capability of NPPs to autophosphorylate and autodephosphorylate (see Introduction) could perhaps reflect an intrinsic phosphatase activity, in which autophosphorylated NPP represents the covalent phospho-intermediate. We have obtained initial data that support this view. First, autophosphorylation of NPP1 was also observed with [␤-32 P]ADP (Fig. 7), establishing that it does not represent a classical kinase reaction. Second, under identical conditions, a labeling was also observed with AP from E. coli (data not shown) and from bovine intestinal mucose (Fig. 7), showing that ATP and ADP are indeed substrates for APs. On the other hand, neither NPP1 nor AP were labeled by incubation with [ 32 P]P i , indicating that the labeling was not due to the generation of P i by hydrolysis of ATP or ADP.
We have also explored the effects of point mutations at or near the catalytic site on the ability of NPP1 to become autophosphorylated (Fig. 8). Mutations D358N and H362Q still allowed a weak labeling, whereas mutation of any of the other residues proposed to be involved in the binding of Me1 (His-517) and Me2 (Asp-200, Asp-405, and His-406) completely abolished the autophosphorylation in the presence of [␥-32 P]ATP. This suggests strongly that the same catalytic site mediates both the nucleotide pyrophosphatase/phosphodiesterase and the phosphatase activities. Moreover, as was observed for the phosphodiesterase reaction (Fig. 6C), the T238S mutant could still autophosphorylate (Fig. 8). Although the F239A mutant showed a reduced capacity for nucleotidylation (Fig. 6C), its ability to form a phosphorylated intermediate was not affected (Fig. 8). However, if Phe-239 is involved in the coordination of the nucleoside moiety, as suggested above, it can be expected that its mutation decreases the phosphodiesterase activity without affecting the phosphatase activity since both enzymic reactions require the substrate to be bound in a different way (see also "Conclusions and Perspectives"). Unexpectedly, although the D358N and H362Q mutants showed an increased labeling with [␣-32 P]ATP (Fig. 6C), the labeling with [␥-32 P]ATP was decreased (Fig. 8). These data suggest that Asp-358 and His-362, in addition to their role in coordinating Me1, may also play a role in the binding of nucleotides as substrates for the phosphatase reaction.
Conclusions and Perspectives-We have made use of the wealth of information in protein data bases to gain some initial insight into the catalytic mechanism of NPPs. Rather unexpectedly, this analysis has revealed that NPPs belong to the superfamily of phospho-/sulfo-coordinating metalloenzymes. The detailed knowledge of the structure and catalytic mechanism of some of these enzymes has helped us to identify residues in NPPs that are required for catalysis. The essential role of these residues has subsequently been confirmed by sitedirected mutagenesis. Interestingly, mutation of some of the residues that were predicted to be involved in the second catalytic step (Thr-238, Asp-358, His-362) generated enzymes that accumulated the nucleotidylated intermediate.
Our studies have also provided an entirely novel insight into the dual function of NPPs, i.e. as nucleotide pyrophosphatases/ phosphodiesterases and as auto(de)phosphorylating enzymes. We suggest that the ability of some NPPs to auto(de)dephosphorylate is actually a reflection of an intrinsic phosphatase activity, whereby the autophosphorylated enzyme represents the covalent intermediate of the phosphatase reaction. Based upon the relative production rates of ADP, AMP, PP i , and P i by NPP1 (16) or NPP2 (58), it can be deduced that their phosphatase and nucleotide pyrophosphatase/phosphodiesterase activities are of similar order of magnitude. If anything, the contribution of the phosphatase activity was underestimated in these studies since ADP, a product of the phosphatase reaction, is also a substrate of the NPP reaction. It should also be noted that the ratio of both enzymic activities is bound to be substratedependent. Thus we noted that p-nitrophenylphosphate, a classical substrate of alkaline phosphatases, is not hydrolyzed to a measurable extent by NPP1 (data not shown).
The similarities between the catalytic core of NPPs and APs (Fig. 3), combined with the results of site-directed mutagenesis (Fig. 8), strongly suggest that the phosphatase and nucleotide pyrophosphatase/phosphodiesterase reactions are catalyzed by the same catalytic site. At first glance, phosphatases and phosphodiesterases catalyze a completely different reaction. However, it can be argued that phosphatases hydrolyze phosphate esters whereas NPPs hydrolyze nucleotidyl esters (or acid anhydrides), resulting in the removal of phosphate or nucleotidyl groups, respectively. Replacing a free hydroxyl group of the substrate-bound phosphate moiety by a nucleoside transforms a phosphatase reaction scheme into a phosphodiesterase reaction scheme. For NPPs to act both as nucleotide pyrophosphatases/phosphodiesterases and as phosphatases, one only has to imply that (some) substrates can be bound in two different ways. For example, with ADP as substrate it is either the ␣or the ␤-phosphate that will become covalently bound to the catalytic site threonine in the pyrophosphatase or phosphatase reaction, respectively. Future investigations should provide a deeper insight into the binding modes of the substrates and into the relative importance of the nucleotide pyrophosphatase/ phosphodiesterase and the phosphatase reactions that are catalyzed by NPPs.