Nucleoside diphosphate kinase. Investigation of the intersubunit contacts by site-directed mutagenesis and crystallography.

NDP kinase from Dictyostelium was mutated by site-directed mutagenesis at positions indicated by structural data to be involved in the trimer interface. The mutants were substitutions at residue Pro-100 (P100S and P100G) and deletions of 1-5 residues at the C terminus. Single mutants yielded proteins that kept both activity and hexameric structure. However, they were severely affected in their stability toward temperature and urea denaturation. When the P100S mutation was combined with any of the C-terminal deletions, the enzyme lost most of its activity and dissociated into dimers. Crystallographic analysis of the P100S protein was performed at 2.6 Å resolution. The x-ray structure showed no direct alteration of intersubunits contacts at residue 100, but an induced disruption of the interaction between Asp-115 and the C terminus of another subunit. The substitution of proline 100 to serine corresponds to the Killer-of-prune mutation in Drosophila. Consequences of the mutation are discussed in view of the structural and biochemical properties observed in the mutant Dictyostelium protein.

Nucleoside diphosphate kinase (NDP kinase; EC 2.7.4.6) 1 is a ubiquitous enzyme that exchanges a phosphate between a nucleoside triphosphate and a nucleoside diphosphate (1). The mechanism of the reaction involves the formation of a phosphohistidine intermediate during the catalytic cycle. In contrast to monophosphate kinases (2), NDP kinase is not substrate-spe-cific and can use both purine and pyrimidine, ribo-or deoxyribonucleotides as substrates. The phosphate transfer is very efficient, with turnover numbers larger than 10 3 s Ϫ1 (1, 3).
All known NDP kinases are oligomers made of small polypeptides of about 150 residues (molecular mass of about 17 kDa) with a high degree of sequence similarity (4). X-ray structures are available for several NDP kinases from eukaryotes including those of Dictyostelium discoideum (5), Drosophila (6), and human (7,8), and from the prokaryote Myxococcus xanthus (9). As expected from their high sequence conservation, the polypeptide fold of all NDP kinases is very similar with a ␤␣␤␤␤␣ motif usually not found in other phosphotransferases. Eukaryotic NDP kinases for which the three-dimensional structure has been determined are all hexamers with one 3-fold axis and three 2-fold axes; i.e. the oligomer can be visualized as made of two trimers or three dimers.
The dimer interface involves strands ␤2 which form an intersubunit antiparallel ␤-sheet and helices ␣1 from two subunits (5). It is highly conserved, and the active site is almost identical in all NDP kinases (7,10,11). The C-terminal sequences form an extended segment and are more variable, although Fig. 1 shows that NDP kinases from eukaryotes all end with the sequence . . .-Tyr-Glu-COOH. The enzymes from the prokaryotes Myxococcus and Escherichia coli are 5-7 residues shorter at the C terminus and have been shown to be tetramers instead of hexamers (9,12). Fig. 2 illustrates the trimer interface in hexameric NDP kinases. The "Kpn-loop" between helix ␣3 and strand ␤4 is involved both in the active site and in the surface contacts leading to the association of the subunits into trimers. Indeed, Pro-105, close to the 3-fold axis in the Kpn-loop, interacts with other Pro-105 residues within a trimer via a water molecule (A) (13,14). Other contacts are made between the ⑀-NH 2 of Lys-35Ј and the carbonyl oxygens of Pro-100, Gly-110, and Gly-113 from an adjacent subunit (B). Apart from the contacts involving the Kpn-loop, the amino group of the C-terminal Glu-155Ј in the adjacent subunit, hydrogen bonds to Asp-115 (C).
The Kpn-loop was named after the Killer of prune (Kpn) mutation of Drosophila which leads to a lethal phenotype in the absence of the prune gene product (15). The Kpn mutation was shown to correspond to the substitution of Pro-97 by a serine in NDP kinase (13,14). This proline is totally conserved in all NDP kinases and corresponds to Pro-100 in the Dictyostelium enzyme (see Fig. 1). Using purified mutant protein from Kpn mutant flies, we previously demonstrated that the activity of the Kpn mutant protein was unstable due to dissociation of the hexamer (16). A similar loss of activity correlated to a dissociation of the hexamer was found in the mutant P105G (14,17,18).
Using NDP kinase from D. discoideum, we have investigated the role of residues participating in the trimer interface and have isolated a mutant protein which forms only dimers. The high structural homology between NDP kinases allows extrapolation to the results to other NDP kinases. We describe the high resolution structure of the Killer of prune mutation and discuss the consequence of the mutations on enzyme activity and stability.
After each mutagenesis step, several plasmids were purified and sequenced. More than 75% of the plasmids contained the desired mutation. Each plasmid used for protein expression was entirely sequenced to ascertain the absence of additional mutations.
Protein Purification-Wild-type and mutant NDP kinases were purified as described previously (20). Briefly, 500-ml cultures of E. coli XL1-Blue were grown at 37°C overnight in LB medium containing 100 g/ml ampicillin. The cells were harvested and resuspended in 50 mM Tris-HCl, pH 8.4, and broken in a French press. After centrifugation (50,000 ϫ g) for 30 min at 4°C, the supernatant was chromatographed through a DEAE-Sephacel column, pH 8.4 (Pharmacia Biotech Inc.). Wild-type and mutant NDP kinases, recovered in the flow-through fraction, were approximately 70% pure as judged by SDS-PAGE. They were completely separated from E. coli NDP kinase. The pH was adjusted to 7.4, and NDP kinases were further purified by affinity chromatography on blue Sepharose (Pharmacia), eluted by a 0 -2 M NaCl gradient. Fractions were concentrated by centrifugation on Centriprep 10 (Amicon) and dialyzed against 50 mM Tris-HCl, pH 7.4, to remove the excess salt. The purity of each protein was determined by SDS-PAGE. When the purification was performed on small amounts of proteins, the flow-through from the DEAE column was filtrated on a FPLC Superose 12 column (Pharmacia) (0.3 ml/min) equilibrated with 50 mM Tris-HCl, pH 7.4, containing 150 mM NaCl.
Assay of NDP Kinase Activity-Pure protein was used for all enzymatic assays. NDP kinase activity was measured at 20°C by a coupled assay using 0.2 mM dTDP as the phosphate acceptor in the presence of 1 mM ATP. The kinetic parameters were derived as described previously (16). All measurements were made in duplicate or triplicate. Protein concentration was determined by the method of Bradford and by measuring the optical density at 280 nm using an extinction coefficient of 0.548 for a 1 mg/ml solution.
Stability Measurements-The thermal stability of each protein was determined by measuring the residual activity after 10-min incubation at the indicated temperatures as described elsewhere (20). T m was the temperature corresponding to a 50% loss in activity. The concentration of the protein in each assay was 1.5-2 g/ml in 50 mM Tris-HCl, pH 7.4, containing 1 mg/ml bovine serum albumin. To determine the stability upon urea treatment, the protein was incubated for 24 h at the same concentration as above at 20°C in 50 mM Tris-HCl, pH 7.4, containing urea as indicated. Urea in the spectrophotometer cuvette never exceeded 40 mM and did not interfere with the assay.
Size Exclusion Chromatography-NDP kinase (200 l) was applied to a Superose 12 column (0.1-0.5 mg/ml) in a Pharmacia FPLC system and eluted with 50 mM Tris-HCl buffer, pH 7.4, with 150 mM NaCl. The flow rate was 0.3 ml/min. The eluted protein was detected by UV absorbance and the fractions were analyzed for activity and by Western blotting as described previously (20).
Analytical Ultracentrifugation-The molecular weight of the enzyme (0.5 mg/ml in 50 mM Tris-HCl, pH 7.4, 150 mM NaCl buffer) was determined by sedimentation-diffusion equilibrium in an Beckman Optimal XLA-4 analytical ultracentrifuge at 23,000 rpm at 25°C. The absorption at 280 nm was recorded after 23 h of centrifugation using the absorption scanning device and was analyzed by means of the Data-Red (Beckman) data analysis software, assuming a value of 0.73 ml/g for the protein partial specific volume and of 1.003 gm/ml for the solution density. The data were fitted either to a mixture of two noninteracting components or to a dimer-hexamer equilibrium using the "Origin-single ideal 2" or "Origin-single assoc 4" program. The molecular weight of each component, the equilibrium constant, and the absorbance at the origin were allowed to vary in the fittings.
Crystallization and Structure Determination-Crystals were grown at 20°C in hanging drops containing 50 mM Tris-HCl, 20 mM MgCl 2 , pH 7.4, 9 -11% polyethylene glycol 3350, and 6 mg/ml protein over wells containing 18 -22% (w/v) polyethylene glycol 3350 in the same buffer. The conditions are similar to those yielding hexagonal form II crystals with the wild type (10). P100S crystals are of the same space group

FIG. 1. Sequence homology between NDP kinases in the C terminus region.
The alignment of four eukaryotic NDP kinases is shown from residue 84 to the C terminus. Conserved amino acids are indicated by dots. The numbering of amino acid corresponds to the Dictyostelium enzyme. Pro-100, Tyr-154, and Glu-155 (boxed) are conserved in all of the eukaryotic enzymes. The secondary structure is also indicated.
(P6 3 22) and contain the monomer in the asymmetric unit; however, the cell differs by 0.8 Å in the a parameter, indicating poor isomorphism.
X-ray reflection data were taken at 20°C on one crystal at the D41 station of the LURE-DCI synchrotron radiation center (Orsay, France) using a MarResearch 18-cm imaging plate system. The wavelength was 1.375 Å; 150 degrees of rotation were collected at a rate of 1.5 degrees per 3 min per image. Intensities were evaluated using MOSFLM (21) and processed with programs from the CCP4 suite (22). Taking wild type NDP kinase (PDB file 1NDK, refined at 1.8 Å resolution) (13) as the initial model gave a R factor of 41% and an electron density map where the substitution at residue 100 was obvious. Refinement was conducted with the Powell and simulated annealing features of X-PLOR (23) until the R factor dropped to 19.3% for all reflections with I Ͼ 2.
As for the wild type Dictyostelium NDP kinase, residues 2-5 at the N terminus could not be located in the map and are missing from the model. Residues 6 -7, 56 -67, and C-terminal residues 145-155 have large temperature factors. Superposition of the P100S C ␣ atoms onto the wild type (file 1NDK) gave a root mean square difference of 0.33 Å for the subunit, 0.39 Å for the hexamer. These values are comparable to the estimated error in atomic positions in the mutant.

Effects of Pro-100 Substitutions and C-terminal Deletions on
Activity and Stability-Pro-100 is equivalent in Dictyostelium NDP kinase to the proline residue substituted by a serine in the Killer of prune mutant of Drosophila (Pro-97). We engineered mutants P100S and P100G in order to analyze the role of Pro-100 in intersubunit contacts within a trimer (see Fig. 2). Deletions of the C terminus were also performed to investigate its contribution in the stability of the protein. For this, a stop codon was introduced by site directed mutagenesis at positions 150, 154, and 155, leading to mutant proteins N150Stop, Y154Stop, and E155Stop respectively. Thus, E155Stop and Y154Stop lack one or both of the conserved residues at the C terminus of eukaryotic NDP kinases, while N150Stop corresponds to the polypeptide chain of Myxococcus NDP kinase (see Fig. 1).
The mutant proteins were expressed in E. coli at a high level, comparable to that of the wild-type NDP kinase as judged by SDS-PAGE. They were purified as described under "Materials and Methods." In a typical preparation 10 -20 mg of the single mutants were purified from 500 ml of bacterial culture. The specific activity of the mutant proteins deleted at the C terminus was identical to that of wild-type NDP kinase (Table I). In contrast, substitution of Pro-100 resulted in a partial loss of activity. The activity of all of these mutant NDP kinases was stable upon storage at 0°C.
In a further step in analyzing the contact between subunits within a trimer, the P100S substitution was combined with the deletions of the C terminus in the double mutants P100S,E155Stop, P100S,Y154Stop, and P100S,N150Stop. Although expression levels were similar to that of the single mutants, the purification step on blue Sepharose was less efficient resulting in a lower yield in the purification (5-10 mg/liter). Table I shows that the activity of the double mutant proteins was only 1-3% of that of wild-type NDP kinase.
The thermostability of the mutant proteins was investigated by measuring the residual activity of protein samples incubated for 10 min at various temperatures (Fig. 3). Wild-type NDP kinase was stable up to more than 50°C under these conditions, in agreement with previous observations (14,17,18). The P100S mutant protein was strongly affected in its thermostability, showing a 50% decrease of activity at 40°C. The P100G heat denaturation profile was identical to that of P100S (data not shown). The deletion of a single residue at the C terminus in the E155Stop mutant also resulted in a severe loss of stability, with a reduction of T m by more than 15°C. The Y154Stop mutant protein, was even more affected, with a denaturation profile identical to that of P100S. The further deletion of four C-terminal residues in the N150Stop mutant protein had no additional effect. The thermostability of the double mutant protein P100S,N150Stop was measured in a separate experiment because of its low residual activity compared to the wild type enzyme (see Table I). As shown in the inset of Fig. 3, the activity of P100S,N150Stop mutant protein was stable up to 40°C, a value comparable to that of the P100S or N150Stop single mutants.
The susceptibility of P100S and N150Stop mutant proteins toward urea denaturation was also strongly increased (Fig. 4). While the activity of the wild-type enzyme was unaffected upon incubation in 5 M urea, the P100S mutant protein retained less than 10% of its activity upon incubation in 1.5 M urea. The stability of the N150Stop mutant protein was also severely affected, although slightly less than the P100S mutant protein.
The stability of the low activity of the double mutant proteins in urea could not be measured using this protocol.
Oligomeric State of the Mutant NDP Kinases-The oligomeric state of the various mutant NDP kinases was determined by size-exclusion chromatography on FPLC (Fig. 5). The wildtype Dictyostelium NDP kinase eluted at a position corresponding to the elution volume of the 110-kDa hexamer (arrow A in Fig. 5). The P100S and P100G mutant proteins also eluted in a single narrow peak at the same position, as did the three single mutant proteins with deletions at the C terminus. We conclude that all single mutant proteins are assembled in hexamers under non denaturating conditions.
In contrast, the mutant proteins carrying both the P100S  3. Thermal inactivation of wild type and mutant NDP kinases. Purified proteins were incubated at the indicated temperature for 10 min and the residual activity was determined by the coupled assay. Wild type (E); E155Stop (q); Y154Stop (å); N150Stop (ࡗ); P100S (Ⅺ). The inset shows the residual activity (%) of P100S,N150Stop mutant protein after 10 min at the indicated temperature. and any one of the three C terminus deletions (P100SE,155Stop, P100S,Y154Stop, or P100S,N150Stop) eluted at a position corresponding to about 30 kDa, approximately that of a dimer (arrow B in Fig. 5). The same elution profile was obtained in successive runs in which the protein concentration injected onto the column ranged from 0.1 to 0.5 mg/ml, indicating that the protein was not in equilibrium between different oligomeric states. In order to eliminate this possibility, we have also used analytical centrifugation. Fig. 6 shows the computation of the experimental data as described under "Materials and Methods." More than 99% of the protein is present as a dimer with a molecular mass estimated at 31,800 Da, while the remainder (less than 1%) was the hexamer. These results fully agree with the gel filtration experiment described in Fig. 5.
Filtration of the P100S mutant in the presence of 2 M urea showed a peak at approximately the position of arrow B in Fig.

(data not shown). This suggests that the mutant protein was both inactivated and dissociated into dimers in the presence of 2 M urea.
Structural Consequences of the P100S Mutation-The P100S mutant was crystallized and its structure was solved by the method of molecular replacement (see "Materials and Methods") at a resolution of 2.6 Å (Table II). Fig. 7 illustrates the electron density map of the side chain of the serine residue replacing Pro-100.
The P100S mutation has no gross effect on the fold and quarternary structure of the protein, since the root mean square main chain movement is less than 0.4 Å for the whole hexamer. Very little change is seen at the mutation site itself. The main chain conformation at residue 100 is maintained, its C ␣ atom moves by only 0.28 Å and the serine side chain sterically replaces the C ␤ and C ␥ atoms of the proline. Main chain movements in the range 0.5-1 Å are observed at residues 61-63 and 145-151, but these residues have high B factors and poor density. While these movements are probably not meaningful, there are changes in the P100S subunit which are more significant and can be traced to the proline to serine substitution (Fig. 8). This introduces two new hydrogen bond donors, an NH and an OH, with a remarkable effect on the environment. Residue 100 is the first of the Kpn-loop, which ends at residues 115-118 with a ␤-turn of type I. In the wild type, the carbonyl group of Asp 115 hydrogen bonds to the NH of Arg-118. In the mutant, the 115-116 peptide bond flips and the two groups find new partners: the Ser-100 OH and NH for CO 115 , the carboxylate of Asp-115 for NH 118 . The bonds to Ser-100 induce a 0.5 Å shift of the main chain at residues 113-117. Some disorder in the ␤-turn may be due to the loss of the CO 115 -NH 118 bond. The B factors of this region are below average in the wild type. They increase in the mutant, where the Val-116 side chain appears to be mobile.
These changes occur within the Kpn-loop of a subunit, yet they affect subunit interfaces within trimers. Asp-115, the  6. Best fit of P100S-N150Stop mutant protein sedimentation equilibrium data to a dimer-hexamer model. Native enzyme was allowed to reach equilibrium as described under "Materials and Methods." The data were fitted to "Origin-single assoc 4" from Beckman allowing the molecular weight of dimer and hexamer to vary. The line through the data represents the best fit. most affected residue, is implicated in these contacts. Its carboxylate hydrogen bonds to the ⑀-amino group of Lys-85Ј and to the main chain NH of C-terminal residue Glu-155Ј of a neighboring subunit in the same trimer (13). In the mutant, the carboxylate moves back by 0.8 Å, weakening these hydrogen bonds and carrying with it the C terminus, so that Glu-155Ј C ␣ also moves by 0.7 Å.

Role of the C Terminus on Activity and Stability-
The comparison of the sequence of NDP kinase from different species shows that all eukaryotic NDP kinases have a . . .-Tyr-Glu-COOH motif at their C terminus. In contrast, the NDP kinase from M. xanthus and from E. coli have a shorter polypeptide in which this motif is missing (see Fig. 1). The data derived from the crystal structure of Dictyostelium NDP kinase as well as from Drosophila (6) or human (7) show that the C terminus participates in an important interaction between adjacent subunits by extending away from the ␣/␤ domain and reaching over to the Kpn-loop and the nucleotide binding site of another subunit (see Fig. 2, this report, and also Fig. 6 in Moréra et al. (10) and Fig. 7 in Cherfils et al. (11)). In spite of their high degree of sequence conservation, different oligomeric structures were observed for NDP kinases. If one examines only those NDP kinases for which a crystal structure has been determined, the enzyme from the prokaryote Myxococcus (9) are tetramers, while those from the eukaryotes Dictyostelium, Drosophila, and human are hexamers (6 -8, 13).
In order to analyze precisely the role of the C terminus in the stability of the hexamer, we have engineered three different mutant NDP kinases with altered C termini. E155Stop, Y154Stop, and N150Stop mutant proteins formed hexamers in recombinant bacteria showing that the lack of the C-terminal . . .-Tyr-Glu-COOH motif is not sufficient in itself to result in a tetrameric assembly of the subunits. The C terminus deletions had no measurable effect on the catalytic activity. The crystal structure of the complex of Dictyostelium NDP kinase with bound nucleotide diphosphates has shown an interaction between the terminal glutamate in one subunit with the base of the substrate bound in the active site of a neighboring subunit. This interaction is mediated by a water molecule in the complex with ADP (10) and dTDP (11), while there is a direct hydrogen bond between the terminal glutamate and with the NH 2 of the nucleobase in the case of GDP (7). The fact that the catalytic activity is not strongly affected in the mutants lacking Glu-155 indicates that this interaction does not play a major role in the binding of the substrate.
The E155Stop mutant protein had a dramatically decreased stability toward heat denaturation. Further deletions up to five of the last C-terminal residues had little additional effects, showing that the conserved Glu-155 plays a specific role in the association of the subunits as its NH interacts with Asp-115 in an adjacent subunit (see Fig. 2). The loss of this interaction in the mutant proteins is likely to be the reason for their highly increased instability.
The Dictyostelium Version of the "Kpn Mutant NDP Kinase from Drosophila"-The P100S mutation in Dictyostelium NDP kinase corresponds to the Killer-of-prune mutation (P97S) in  Drosophila. In order to study the role of Pro-100 in activity and stability, we have engineered the P100S and P100G substitutions in Dictyostelium NDP kinase. The biochemical properties of the two mutant proteins were similar, and therefore only the P100S mutant was investigated in detail. Contrary to what we observed in Drosophila (16), the mutation of Pro-100 had some effect on catalytic activity (30% of wild type activity for P100S mutant NDP kinase). However, this decrease was relatively modest if compared with the effect of mutations of the residues participating in the active site (20). In contrast, the stability was dramatically altered. In the presence of 2 M urea, activity dropped to about 1% of the native enzyme. The hexameric structure was lost and the enzyme dissociated to a dimer. Thus, the biochemical properties of the P100S mutant NDP kinase from Dictyostelium were similar to those of the NDP kinase isolated from the Kpn mutant protein from Drosophila (16).
The P100S mutation has more elaborate structural consequences than would be expected from the wild-type structure alone. There, the carbonyl group of Pro-100 hydrogen bonds across the trimer interface to the ⑀-amino group of Lys-35Ј (Fig.  2). We had surmised that the mutation would perturb this bond, making the hexamer less stable (see Fig. 1 in Dumas et al. (5)). However, the x-ray structure of the mutant showed that neither residue 100 nor residue 35 move and that the bond involving the carbonyl group is maintained. Instead, other parts of the Kpn-loop undergo conformational changes originating from the presence of a novel NH group at position 100. The perturbation is largest at Asp-115 which interacts with the C-terminal Glu-155Ј across the trimer interface. Thus, the loss of stability is more likely to result from a perturbation of this interaction, an interpretation which also explains the effect of the C-terminal deletions. The lower activity of the P100S mutant protein may also be due to the movement of the Kpn-loop which contains several amino acids involved in substrate binding (10,20).
We previously proposed that the dominant lethal phenotype of Kpn could result from the presence of dissociated forms (possibly dimers) of the enzyme in the cells (16). Alternatively, the Kpn phenotype could result from a modification of the active site. It has been hypothetized that in the absence of a prune gene product, the synthesis of a phosphorylated and toxic derivative of a yet unidentified molecule is the cause for the lethality of the Kpn mutation. In support of this hypothesis, Timmons et al. (25) isolated revertants of Kpn flies and found that all were catalytically inactive. The structure of the P100S mutant protein does not distinguish between these two possibilities. The P100S substitution induces no detectable change in any of the residues implicated directly in catalysis (25,26). The conformational change seen around Asp-115 and the increased mobility of Val-116 could allow Kpn NDP kinase to bind compounds that would not normally be substrates of the wild-type enzyme, thus explaining the gain of function character of the Kpn mutation. Alternatively, the increased tendency of the Kpn NDP kinase to dissociate may lead to the in vivo occurrence of dimers making interactions with other cell components that the hexamer cannot make.
Dimeric Mutant NDP Kinases-The combination of the P100S substitution with deletions of C-terminal residues results in an enzyme with very little activity (1-3% of wild type). Although low, this residual activity is not negligible. Indeed, taking into account the very high turnover of wild-type NDP kinase (over 1000 s Ϫ1 ), it implies that phosphate is transferred from the nucleotide substrate to the catalytic histidine (His-122) at a rate of 10 -50 s Ϫ1 . This value is well over that achieved by bacterial histidine kinases (27), indicating that the double mutant NDP kinases are properly folded when ex-pressed in recombinant bacteria. The P100S,N150Stop mutant protein is a dimer, and is not in equilibrium with other oligomeric forms as demonstrated both by its elution from gel filtration at different protein concentrations and its sedimentation at equilibrium in the analytical centrifuge.
Several lines of evidence indicate that NDP kinase may have biological function(s) unrelated to its enzymatic activity. Indeed, it was identified (4) as the protein encoded by the gene awd involved in Drosophila development (15) and to nm23, a putative metastasis suppressor gene from human (reviewed in De La Rosa et al. (28)). It has now been shown that two human NDP kinase genes, nm23-H1 and nm23-H2 encode two highly similar proteins (3). NDP kinase B encoded by nm23-H2 was shown to bind to the c-myc promoter and to activate transcription in vitro (29). This activity was maintained in a catalytically inactive mutant protein (30); NDP kinase B preferentially binds to pyrimidine-rich, single-stranded DNA in vitro (31). It is interesting that a recent analysis of the interaction between a human NDP kinase and oligonucleotides indicated that the oligomeric species binding to the oligonucleotide may be a dimer (32). In other studies, a phosphotransfer activity to protein was demonstrated for NDP kinase in the presence of low concentration of dissociating agents suggesting that a dissociated form of the oligomer might be responsible for the observed protein kinase activity (33)(34)(35). These results indicate that NDP kinase might be bifunctional. They are intriguing considering its small 17-kDa polypeptide chain which folds into a single domain. An attractive hypothesis is that different oligomeric states of the protein may have different biochemical functions. The availability of a dimeric form of NDP kinase will be a useful tool to investigate functions of NDP kinase which may be distinct from its well characterized activity of dinucleotide phosphorylation.