A Point Mutation of Human Nucleoside Diphosphate Kinase A Found in Aggressive Neuroblastoma Affects Protein Folding*

The point mutation serine 120 to glycine in the human nucleoside diphosphate kinase A has been identified in several aggressive neuroblastomas (Chang, C. L., Zhu, X. X., Thoraval, D. H., Ungar, D., Rawwas, J., Hora, N., Strahler, J. R., Hanash, S. M. & Radany, E. (1994) Nature 370, 335–336). We expressed in bacteria and purified wild-type and S120G mutant nucleoside diphosphate kinase A. The mutant enzyme had enzymatic and structural properties similar to the wild-type enzyme, whereas its stability to denaturation by heat and urea was markedly reduced. More importantly, upon renaturation of the urea-denatured mutant protein, a folding intermediate accumulated, having the characteristics of a molten globule. It had no tertiary structure, as shown by near UV circular dichroism, whereas the secondary structure was substantially recovered. The hydrophobic probe 8-anilino-1-naphthalene sulfonate bound to the intermediate species with an increase in fluorescence intensity and a blue shift. The hydrodynamic size was between that expected for a folded and an unfolded monomer. Finally, electrophoresis in a transverse urea gradient displayed no renaturation curve, and the protein showed the tendency to aggregate at the lowest urea concentrations. The existence of a molten globule folding intermediates resulting from an altered folding in the mutated protein might be related to the aggressiveness of neuroblastomas.

Nucleoside diphosphate (NDP) 1 kinases, oligomeric enzymes made of small subunits, catalyze the reversible transfer of the terminal phosphate of nucleoside triphosphates to nucleoside diphosphates (1,2). In recent years, data accumulated indicating that NDP kinases have regulatory functions. In Drosophila, NDP kinase, product of the awd gene, is essential for larvae growth and development (3,4). In humans, two highly similar enzymes, NDP kinase A and NDP kinase B, displaying 89% identity, are equally active and form mixed hexamers (5). The decreased expression level of NDP kinase A, product of the Nm23-H1 gene has been correlated with the increased metastatic potential of some human tumors but not of others (for a review see Ref. 6). The other isoform, NDP kinase B, product of the Nm23-H2 gene, acts as a transcription factor of the oncogene c-myc, in vitro and in vivo (7)(8)(9). A third homologous human gene, DR-nm23, has been discovered in humans and shown to be involved in the control of granulocyte differentiation and apoptosis of myeloid cells (10). Finally, the cDNA encoding for a forth human NDP kinase has recently been cloned (11).The last two proteins newly discovered are enzymatically active 2,3 and possess amino-terminal extensions probably involved in their targeting to specific locations.
The serine 120 to glycine point mutation in human nucleoside diphosphate kinase A has been detected in 6 of 28 advanced stage neuroblastomas, but in none of 22 limited stage tumors (12). Chang et al. (13) reported that the S120G mutant was almost as active as the wild-type enzyme, but its stability to denaturation was decreased. Serine 120 is conserved in all prokaryotic and eukaryotic NDP kinases. We show here that altered folding properties of the recombinant S120G mutant protein lead to the accumulation of a molten globule intermediate.

MATERIALS AND METHODS
Proteins-The cDNA for the wild-type NDP kinase A was cloned as described (14). Site-directed mutagenesis was carried out using the Kunkel method (15). Full-length wild-type and mutant proteins were expressed in high yields using a pET-derived expression vector, pJC20. Bacteria were grown in a rich medium (2 ϫ YT (per liter, 16 g of bacto-tryptone, 10 g of bacto-yeast extract, and 5 g of NaCl)) in the presence of 200 g/ml ampicillin. Induction was performed for 3 h with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside, at 37°C at a A 600 between 2 and 3 (additional 200 g/ml of ampicillin were added at that time). The crude extract was directly applied to a Q-Sepharose Fast Flow column. After washing with 50 mM Tris acetate (pH 8.0), 1 mM dithiothreitol, and 1 mM EDTA, the enzyme was eluted with a linear gradient of 0 to 0.5 M NaCl in the same buffer. When expression was high, essentially pure protein was obtained in just one step. Otherwise, the active fractions were subjected to an additional blue Sepharose chromatography in the same buffer as above but at pH 7.5. The enzyme was eluted with 1.5 mM ATP and then with 2 M NaCl. Both the wild-type and S120G mutant protein lacked the amino-terminal methionine as determined by automated Edman degradation. Analysis by electrospray mass spectrometry revealed, in addition to the expected peaks, minor mϩ15 and mϩ30 peaks in both proteins, probably due to oxidation of cysteines and/or methionines.
Analytical Techniques-Protein concentration was determined from the optical density at 280 nm using an extinction coefficient of 1.35 for 1 mg/ml, which was calculated from the amino acid composition according to Ref. 16. Enzymatic assay and stability studies to denaturation by heat and urea were performed as described previously (17)(18)(19). CD spectra were recorded with a Jasco J-710 A spectropolarimeter. All measurement (unless otherwise specified) were performed at 25°C in 50 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1 mM EDTA, and 1 mM dithiothreitol. Additional details can be found in the figure legends. Molar concentrations of proteins were expressed as subunit concentrations.

RESULTS
Full-length wild-type and S120G mutant NDP kinase A were expressed in bacteria in high yields and purified to apparent electrophoretic homogeneity. They elute in a position expected for a hexameric structure in size exclusion chromatography on calibrated Superose 12 column. The mutant enzyme could be fully phosphorylated (0.9 phosphate groups/enzyme subunit, using the nonisotopic method described by Lascu et al. (20)). The k cat (580 s Ϫ1 and 420 s Ϫ1 ), the apparent K m for ATP (76 M and 40 M at 100 M dTDP), and the apparent K m for dTDP (148 M and 207 M at 1 mM ATP) of the wild-type and S120G mutant enzymes were similar. The thermal stability was significantly reduced by the S120G mutation. When analyzed by differential scanning calorimetry, both proteins displayed a single calorimetric peak only having the T m of 58°C (wild-type protein) and 50°C (S120G mutant protein). The T m values correspond to the temperatures of half-inactivation temperatures, measured at similar protein concentration and heating rate. A detailed analysis of catalytic and structural properties will be reported elsewhere. Denaturation by urea was followed by intrinsic protein fluorescence. The mutant enzyme was considerably less stable than the wild-type protein (Fig. 1A). Be-cause the human NDP kinase A contains three tryptophan residues, it is not easy to correlate fluorescence changes with loss of quaternary structure and of native structure. A notable feature was the absence of a cooperative transition upon renaturation of the urea-denatured mutant protein. To detect possible folding intermediate, ANS was added to the final concentration of 56 M to each sample, and its fluorescence was measured. The hydrophobic dye ANS has no affinity for either native or completely unfolded proteins, but it strongly binds to molten globule folding intermediates (21,22). The wild-type enzyme bound very little ANS. In contrast, the S120G mutant protein bound ANS, as observed by an increase in the fluorescence intensity (Fig. 1B) and a blue shift in the spectrum (maximum moved from about 520 to 480 nm). This suggests that a molten globule intermediate accumulated both in the denaturation and the renaturation pathway of the S120G mutant NDP kinase, whereas it was undetectable with the wildtype protein.
Further support was obtained by CD experiments. CD was used to measure the extent of secondary structure (in the far UV) and tertiary structure, reflected by the asymmetric environment of the aromatic residues, (in the near UV). Diluting the urea-denatured S120G enzyme into buffer generated a species having about 60% of the secondary structure content of the native enzyme, as calculated from the signal at 220 nm, whereas no tertiary structure was detected (Fig. 2). In the control experiment with the wild-type enzyme, both secondary and tertiary structure were recovered to a similar extent (not shown).
By size-exclusion chromatography on a calibrated column of Superose 12 the Stokes radius of the intermediate species accumulated during the refolding of the S120G mutant ap-FIG. 1. Denaturation and renaturation of wild-type and S120G mutant NDP kinases. A, intrinsic protein fluorescence. Wild-type NDP kinase (circles) and S120G mutant NDP kinase (squares) were incubated overnight at room temperature at indicated urea concentrations. Filled symbols represent protein denaturation, whereas open symbols represent renaturation. Intrinsic protein fluorescence intensity was measured at 340 nm (excitation at 295 nm). The protein concentration was 1.1 M. B, binding of ANS. Native or urea-denatured NDP kinases (same symbols as in A) were incubated with urea as above. ANS at the final concentration of 56 M was then added, and its fluorescence was measured at 480 nm (excitation at 380 nm). The protein concentration was 3.5 M. The solid lines were draw for clarity and have no theoretical meaning.

FIG. 2. Refolding of the mutant S120G protein analyzed by CD.
CD spectra in the far UV (A) and near UV (B) of the native protein (thin line), urea-denatured protein (dotted line; urea 6 M), and urea-denatured protein diluted into buffer (thick line; final urea concentration 1.6 M). The far UV spectra and the near UV spectra were recorded in cuvettes having the optical path 0.1 and 1 cm, respectively. To avoid concentration-dependent differences between the samples, the same protein solution was used (18 M), despite the large difference of sensitivity in the two spectral ranges. peared between that of the folded and unfolded monomer (Fig.  3D). It was devoid of enzymatic activity (not shown). Under the same conditions, the wild-type protein refolded and assemble to hexamers (Fig. 3B), which had enzymatic activity.
Finally, the electrophoretic pattern of denaturation and renaturation of the S120G mutant of NDP kinase A subjected to electrophoresis in a transverse-urea gradient (23) showed a folding defect in the mutant protein. There is a clear transition from the native hexamer to the unfolded protein with the S120G mutant (Fig. 4A) and with the wild-type protein (not shown). A renaturation transition is absent when refolding of the S120G mutant, analyzed by this technique (Fig. 4B). At the lowest urea concentrations, the protein aggregated. Under the same conditions, the wild-type enzyme renatured mostly to folded monomer and to some hexamer (Fig. 4C). DISCUSSION The mutant S120G of the human NDP kinase A had catalytic properties similar to those of the wild-type protein, incorporated one phosphate group per subunit, and was hexameric but was less stable to denaturation by heat or urea. The most important distinctive feature was the accumulation of a folding intermediate while renaturing the urea-denatured mutant enzyme. This species had several characteristics of a molten globule (24,25): (a) absence of tertiary structure whereas the secondary structure is largely present, (b) size intermediate between the native monomer and the denatured protein, (c) binding of the hydrophobic dye ANS, and (d) the propensity to aggregate.
Despite the folding defect, the mutant protein was expressed at very high yield in bacteria in a native hexameric state. It was also found in human tumors to be active (13). It is likely that chaperone proteins assist folding/assembly in vivo. Moreover, once assembled, eukaryotic NDP kinases are stabilized by their quaternary structure (18,19). Interestingly, the natural mutant P97S in Drosophila NDP kinase (K-pn mutation) is affected in assembly but not in folding. A gel filtration experiment clearly showed that the folded monomers were the most proeminent species (17). The defect resulting from the K-pn mutation is therefore different in nature from the S120G mutant of human NDP kinase A described above. 4 The K-pn mu-tation generates a dominant, lethal, and conditional phenotype in Drosophila (26).
Human NDP kinase A is 89% identical to the other cytosolic isoform, NDP kinase B. The structure of the latter protein is known at high resolution (27,28). Serine 120 is located near the active site histidine 118, which is phosphorylated during the catalytic cycle. The O␥ is hydrogen-bonded to Glu129, which is also hydrogen-bonded to His-118. Despite the proximity of the active site, the absence of the O␥ atom does not have an important effect on the catalytic efficiency of the mutant enzyme. However, serine 120 appears to be of crucial importance for correct protein folding and stability. The total conservation of this residue in all known NDP kinases, including the bacterial enzymes, points to an essential function. We suppose that the hydrogen bond of serine 120 to glutamate 129 is necessary for the final arrangement of the protein structure after the hydrophobic collapse and secondary structure formation. Inspection of the tertiary structure suggests that this hydrogen bond anchors the carboxyl-terminal part to the protein core. By breaking just one hydrogen bond by mutation, folding intermediates were also detected in ␤-lactamase (29) and barstar (30). Alternatively, the final folding step is slowed down by local chain flexibility because two glycine residues are adjacent in the mutant protein. More experiments are under way in our laboratories to further characterize the nature of the defect.
The S120G mutation of NDP kinase A has been identified in human neuroblastomas (12). The question arises whether there exists a link between the folding defect and the tumoral phenotype of the host cells. Chang et al. (13) detected by immunoprecipitation in neuroblastoma cells a protein interacting with the mutant but not with the wild-type protein. Moreover, transfection of the highly metastatic MDA-MB-435 breast carcinoma cells with the S120G mutant of NDP kinase A cDNA failed to decrease the cell motility, whereas transfection with the wild-type cDNA has a sizable effect on cell motility. This parameter can be correlated with the metastatic potential of the tumor cells (31). Recent evidence indicates that several disease-causing mutations may exert their effect by altering protein folding (32)(33)(34). The data presented here suggest a 4 The anonymous referee made the interesting remark that a defect in subunit association is a possible cause of accumulation of the folding intermediate. The highly cooperative hexameric structure may in fact act as a "thermodynamic trap." However, the P96S (K-pn) mutation in Drosophila NDP kinase (17) and the P105G mutation in Dictyostelium NDP kinase (18) affect only the subunit association but not their folding. As a consequence, the main renaturation product were in both cases the folded monomers. possible relationship between a folding defect and aggressiveness of human neuroblastoma.