Characterization of the Amino Acids Involved in Substrate Specificity of Nonphosphorylating Glyceraldehyde-3-Phosphate Dehydrogenase from Streptococcus mutans *

In order to address the molecular basis of the specificity of aldehyde dehydrogenase for aldehyde substrates, enzymatic characterization of the glyceraldehyde 3-phosphate (G3P) binding site of non-phosphorylating glyceraldehyde-3-phosphate dehydrogenase (GAPN) fromStreptococcus mutans has been undertaken. In this work, residues Arg-124, Tyr-170, Arg-301, and Arg-459 were changed by site-directed mutagenesis and the catalytic properties of GAPN mutants investigated. Changing Tyr-170 into phenylalanine induces no major effect on k cat and K m ford-G3P in both acylation and deacylation steps. Substitutions of Arg-124 and Arg-301 by leucine and Arg-459 by isoleucine led to distinct effects on K m , onk cat, or on both. The rate-limiting step of the R124L GAPN remains deacylation. Pre-steady-state analysis and substrate isotope measurements show that hydride transfer remains rate-determining in acylation. Only the apparent affinity ford-G3P is decreased in both acylation and deacylation steps. Substitution of Arg-459 by isoleucine leads to a drastic effect on the catalytic efficiency by a factor of 105. With this R459L GAPN, the rate-limiting step is prior to hydride transfer, and theK m of d-G3P is increased by at least 2 orders of magnitude. Binding of NADP leads to a time-dependent formation of a charge transfer transition at 333 nm between the pyridinium ring of NADP and the thiolate of Cys-302, which is not observed with the holo-wild type. Accessibility of Cys-302 is shown to be strongly decreased within the holostructure. The substitution of Arg-301 by leucine leads to an even more drastic effect with a change of the rate-limiting step similar to that observed for R459I GAPN. Taking into account the three-dimensional structure of GAPN from S. mutans and the data of the present study, it is proposed that 1) Tyr-170 is not essential for the catalytic event, 2) Arg-124 is only involved in stabilizing d-G3P binding via an interaction with the C-3 phosphate, and 3) Arg-301 and Arg-459 participate not only in d-G3P binding via interaction with C-3 phosphate but also in positioning efficientlyd-G3P relative to Cys-302 within the ternary complex GAPN·NADP·d-G3P.

Nonphosphorylating glyceraldehyde-3-phosphate dehydrogenase (GAPN) 1 catalyzes the irreversible oxidation of D-glyceraldehyde 3-phosphate (D-G3P) into 3-phosphoglycerate in the presence of NADP via a two-step chemical mechanism. It belongs to the aldehyde dehydrogenase (ALDH) superfamily, which oxidizes a wide variety of aldehydes into nonactivated or SCoA-activated acidic compounds (1)(2)(3). These enzymes play parts at several levels of cellular metabolism including anabolic and catabolic pathways, detoxification processes, and embryogenesis development. Their numerous functions probably explain why most ALDHs show a wide substrate specificity, except for the CoA-dependent ALDH. In mammals, ALDH class 1 shows preference for retinal but it can also oxidize aromatic and "long-chain" aldehydes with dissimilar catalytic efficiencies, whereas ALDHs that belong to class 2 are more specific for acetaldehyde and short-chain aldehydes (4,5). Inspection of the active site of the ALDH structures (i.e. rat dimeric class 3 ALDH, bovine mitochondrial ALDH, sheep liver cytosolic ALDH, rat retinal dehydrogenase type II, cod liver betaine ALDH, human liver mitochondrial ALDH, ALDH from Vibrio harveyi, and GAPN from Thermoproteus tenax (6 -13)) gives no clear information on the binding mode of the aldehyde substrate. Therefore, characterizing the molecular and structural factors involved in substrate specificity is important for a better understanding of the efficiency of the catalytic event and of the evolution of the active sites of ALDHs.
Unlike most ALDHs, GAPN is an ALDH for which the physiological substrate is known. The fact that GAPN shows high catalytic efficiency toward D-G3P implies prerequisites with respect to the chemical mechanism. Previous enzymatic and structural studies of GAPN from Streptococcus mutans have shown that NADP binding induces a local conformational rearrangement. As a consequence, the thiol of Cys-302 becomes accessible, its pK app shifts from 8.5 to 6.2, and it is now well positioned to subsequently form a competent thiohemiacetal intermediate. On the other hand, the side chains of Glu-268 and Arg-459 also rotate and then do not interact with each other anymore. The side chain of Arg-459 is now orientated toward the substrate binding site. Glu-268 now has a pK app that shifts from probably 5.7 in the apo structure to 7.6 in the holo form and has its side chain well positioned for subsequently activating the water molecule involved in hydrolysis of the thioacyl intermediate. During this rearrangement, the distance between Glu-268 and Cys-302 passes from 7.6 to 3.6 Å. Concomitantly, the oxyanion hole, composed of at least the amide side chain of Asn-169 and the NH main chain of Cys-302, is formed. Its role is to stabilize the tetrahedral transition states formed during acylation and deacylation steps. This local conformational change has been shown to be strongly favored by binding of D-G3P to the binary complex GAPN⅐NADP (14 -16). Whereas the rate-limiting step of GAPN is deacylation, that of human liver mitochondrial ALDH has been shown to depend on the chemical structure of the aldehyde substrate (17). Thus, it appears that the nature of the substrate can modulate the formation of a competent ternary complex.
This justifies the determination of the structural factors that are implied in the catalytic efficiency of GAPNs. Several crystal structures of GAPN from S. mutans have been described so far (15,18). The apo1 structure shows a sulfate anion, called SO 4 a, which probably corresponds to the C-3 phosphate of G3P. Arg-124 and Arg-301, which are located in an ␣-helix and within a loop, respectively, interact with SO 4 a via their guanidinium groups. Arg-459, which is located within a loop, also interacts with SO 4 a but via its NH main chain. The apo2 structure shows a second sulfate anion, called SO 4 b, which probably mimics the tetrahedral transition states involved in acylation and deacylation. Guanidinium groups of Arg-301 and Arg-459 interact with SO 4 b. SO 4 b is also in interaction with the Asn-169 amide group, the Cys-302 and Thr-303 main chain NH, and the Thr-303 hydroxyl group, which in turn also interacts with SO 4 a. Inspection of the structure of the ternary complex C302S-NADP-G3P supports the potential roles of Arg-124, Arg-301, Arg-459, and Thr-303 in D-G3P oxidation process but also of the hydroxyl of Tyr-170, which forms a hydrogen bond with the oxygen that bridges the carbon chain to the phosphate in D-G3P. On the basis of sequence comparison of GAPNs to those of other ALDHs (Fig. 1) only Arg-124, Arg-301, Arg-459, and Tyr-170 seem to be specific to GAPNs. Here, Arg-124 and Arg-301 were substituted by leucine, and Arg-459 and Tyr-170 were changed into isoleucine and phenylalanine, respectively.
The catalytic properties of these GAPN mutants were determined. Altogether, the results suggest that Tyr-170 has no significant role, Arg-124 participates only in D-G3P binding via interaction with C-3 phosphate, and both Arg-301 and Arg-459 are not only involved in D-G3P binding but also participate in positioning efficiently the substrate with respect to Cys-302 within the ternary complex GAPN⅐NADP⅐D-G3P.

EXPERIMENTAL PROCEDURES
Materials-Production and purification procedures of wild-type and mutated GAPNs were carried out as previously described (14). GAPN concentration was determined spectrophotometrically as the apo form by using a molar extinction coefficient of 2.04 ϫ 10 5 M Ϫ1 ⅐cm Ϫ1 at 280 nm. Enzyme concentrations were expressed per monomer (normality, N). All other materials were reagent grade or better and were used without further purification. NADP (Roche Molecular Biochemicals) and 2,2Ј-dithiopyridyldisulfide (2PDS) were dissolved in H 2 O, and the stock concentrations were determined spectrophotometrically by using a molar extinction coefficient of 18,000 M Ϫ1 ⅐cm Ϫ1 at 259 nm and 10,100 M Ϫ1 ⅐cm Ϫ1 at 254 nm, respectively. D-G3P (Sigma) was hydrolyzed from D-G3P diethyl acetal according to the manufacturers' instructions and enzymatically titrated with GAPN.
Enzyme Kinetics and Determination of K m Values for NADP and D-G3P Constants-To determine the nature of the rate-limiting step, transient kinetics of GAPN mutants were carried out at pH 8.5 with a mixed buffer (buffer A) containing 30 mM acetic acid, 30 mM imidazole, and 120 mM Tris/HCl at constant ionic strength of 0.15 M, as already described by Marchal et al. (16). Fast kinetic measurements were carried out on a SFM 3 Biologic Instruments stopped-flow apparatus (Biologic, Grenoble, France) with a tungsten lamp and slit of 5-nm light source equipped with a 1-cm light path TC100-10 cuvette maintained at 25°C by a circulating water bath. One syringe was filled with mutant enzymes (16 N after mixing), and the other one contained NADP and D-G3P (1 mM after mixing). Under these conditions, the acylation step was shown to be rate-limiting for only the R301L and R459I mutant GAPN proteins (see "Results"). Kinetic analyses of the acylation step were carried out on a Kontron Uvikon 933 spectrophotometer by following the appearance of NADPH at 340 nm for both GAPN mutants. For the determination of kinetic parameters, initial rate measurements were carried out under the following conditions: 50 mM TES/NaOH buffer, pH 8.5, and 5 mM ␤-mercaptoethanol (ionic strength 0.05 M) FIG. 1. Amino acid sequence alignment of the regions around the conserved arginine and tyrosine residues in GAPN. The sequences of 19 GAPNs with known primary structures from the EMBL, Swiss-Prot, GenBank TM , and Pir databases have been taken into account. The sequence alignment was performed with the Bioedit software. The numbering of amino acid residues is according to Wang  using the following range: 0.05-4 mM D-G3P (with 1 mM NADP), 0.01-1 mM NADP (with 4 mM D-G3P). 0.25 M (NH 4 ) 2 SO 4 was added in the reaction mixture for the R459I GAPN. 2 Data were collected in triplicate, and kinetic constants were determined by Michaelis-Menten analysis. The acylation rate constant k ac was studied over a pH range of 5-9 with the buffer A. 0.25 M (NH 4 ) 2 SO 4 was added in the reaction mixture for the R459I GAPN protein. The data were expressed as initial rate constant per enzyme site. Initial activity measurements were carried out at subsaturating concentrations of D-G3P (see "Results"). The second-order rate constant k ac /K m , which corresponds to the efficiency of the acylation step, was calculated at each pH, dividing k ac by the concentration of D-G3P and was fitted to Equation 1 or 2 to determine the best fit pK a values, where (k ac /K m ) max represents the maximum pH-independent secondorder rate constant, where (k ac /K m ) i and (k ac /K m ) max represent the characteristic pH-independent second-order rate constants and pK a1 and pK a2 reflect the different groups involved in the acylation step (16). In the present work, an error structure of constant relative error was assumed, and weighting factors were inversely proportional to (k ac /K m ) 2 .
In the case of R124L and Y170F mutant GAPN proteins, kinetics of the acylation step were carried out on a stopped-flow apparatus at 25°C under similar conditions to those already described for the wild type (16). The pH was set by a series of low ionic strength (0.08 M) buffers over the pH range used (50 mM MES/NaOH buffer for 5.0 Ͻ pH Ͻ 6.5; 50 mM TES/NaOH buffer for 6.5 Ͻ pH Ͻ 8.2; 50 mM Tris-HCl buffer for 8.2 Ͻ pH Ͻ 9.5). The experiment was set up with one syringe filled with GAPN mutants (16 N after mixing) and the other one containing NADP and D-G3P (saturating concentration of 1 mM and subsaturating concentration of 2 mM after mixing, respectively). An average of at least five runs was recorded to perform kinetic analysis at each pH. The pH dependence of the second-order rate constant k ac /K m was analyzed as described for the kinetic studies for acylation of the R301L and R459I GAPNs.
Deuterium Isotope Effects-D-[1-2 H]G3P was prepared as described (19), and its concentration was determined enzymatically. Substrate isotope effects on R301L and R459I GAPNs were determined at 25°C in 50 mM TES buffer/NaOH buffer, pH 8.2, by direct comparison of the acylation rate constant k ac measured with D-G3P or D-[1-2 H]G3P (0.25 or 0.5 mM) in the presence of 1 mM NADP.
Isotope measurements of the pre-steady-state kinetics of the R124L mutant GAPN protein were carried out with the stopped-flow apparatus. One syringe was filled with mutant enzyme (16 N after mixing), and the other one contained NADP (1 mM and after mixing) and D-G3P or D-[1-2 H]G3P (0.25 or 0.5 mM after mixing). Isotope effects were directly calculated from the ratio of the rate constant k ac measured with D-G3P or D-[1-2 H]G3P in 50 mM TES buffer/NaOH buffer at pH 8.2 and at 25°C.
Kinetic Reactions with 2PDS-Due to the high reactivity of Cys-302 in the R459I GAPN apoenzyme, kinetic measurements were carried out with the stopped-flow apparatus. Kinetic reactions were performed at 25°C, at a constant ionic strength of 0.15 M under pseudo-first order conditions over the pH range of 5-9.5. Mutant GAPN protein and 2PDS were in buffer A containing 0.25 M (NH 4 ) 2 SO 4 . One syringe was filled with enzyme, and the other was filled with 2PDS. Reaction of 2PDS was monitored at 343 nm with 3.7 N apoenzyme and 370 M 2PDS. Release of pyridine-2-thione was quantified using ⌬⑀ 343 nm ϭ 8080 M Ϫ1 ⅐cm Ϫ1 . The second-order rate constants k 2 were calculated at each pH, dividing the observed rate constant k obs by the concentration of 2PDS. The k 2 values were fitted to Equation 3 to determine the best-fit pK app values, where kЈ represents the second order rate constant for the thiolate form, When cysteine reactivities were low, reactions were carried out on a Kontron Uvikon 933 spectrophotometer. The pH dependences of the 2PDS reactions with NADP-bound apo 3 or apo-R301L and holo-R459I GAPN mutants were performed at 25°C in buffer A over the pH range of 5.0 -9.5 under pseudo-first-order reaction conditions. The pseudofirst-order rate constants k obs1 and k obs2 were determined at each pH by fitting the absorbance A at 343 nm versus time t to Equation 4 for a double exponential profile, where a 1 and a 2 represent the burst magnitude corresponding to each thiolate species, k 1 ϭ k obs1 and k 2 ϭ k obs2 , and c represents the value of the ordinate intercept.
The second-order kinetic constants k 2 were calculated dividing the k obs values by the concentration of 2PDS and were fitted to Equation 3 to determine the best-fit pK app values.
Measurements of the absorbance of NADP-thiolate interaction within mutant GAPN R459I were recorded in a 1.0-cm path length cuvette in a Uvikon 933 spectrophotometer at 25°C. The R459I mutant samples were diluted in buffer A with 0.25 M (NH 4 ) 2 SO 4 at a final concentration of 20 N and were incubated 5-30 min in the presence of 1 mM NADP. Data acquisition was in steps of 1 nm from 395 to 290 nm over a pH range of 5.5-9.0. The protein solution was scanned relative to the buffer solution containing 1 mM NADP. The differences were then converted to molar absorption coefficients at 333 nm, and data were fitted to a model derived from the Henderson-Hasselbach equation for one pK app , where ⑀ SH and ⑀ S Ϫ are the molar absorption coefficients of the thiol and thiolate species in the binary complex of enzyme-NADP. Time courses of formation of NADP-thiolate interaction in R459I GAPN were followed by the appearance of the absorption band at 333 nm, at a constant temperature of 25°C, over the pH range of 5.5-9.5. At each pH, the enzyme (20 N final concentration) was dissolved in the buffer A with 0.25 M (NH 4 ) 2 SO 4 . 1 mM NADP was added, and the absorbance was recorded at 333 nm. Pseudo-first-order rate constants k obs were determined at each pH by fitting the absorbance A at 333 nm versus time t to an equation for a single exponential profile and were fitted to Equation 6 to determine the best-fit pK app values, where kЈ and kЉ represent two different rate constants for the thiolate form.

RESULTS
Kinetic Properties of R124L, Y170F, R301L, and R459I GAPNs-In a previous study, it was shown that acylation and deacylation steps of wild-type GAPN could be kinetically resolved and that deacylation was rate-limiting under steadystate conditions (16). It was also demonstrated that the ratelimiting step in deacylation was associated with hydrolysis, whereas hydride transfer was rate-determining in acylation. k ac and K m values for D-G3P in acylation were also shown to be about 10-fold higher compared with k cat and K m values determined under steady-state conditions at 25°C. 4 Therefore, before interpreting the steady-state kinetic data of the various mutants, it was necessary to determine whether the rate-lim- 2 Usually, wild-type GAPN remains stable at least 3 days after dialysis against a 50 mM phosphate buffer, pH 6.4, at 4°C. Under similar conditions, R459I mutant loses its activity. Only the addition of ammonium sulfate (NH 4 ) 2 SO 4 (0.25 M final concentration) in the phosphate buffer permits storage of the mutant enzyme without loss of activity. For the kinetic assays, (NH 4 ) 2 SO 4 was systematically added at a concentration of 0.25 M. 3 The term "NADP-bound apo" indicates that the local conformational rearrangement of the active site has not been induced by the binding of the cofactor on the apoenzyme. This also applies to the term "apo-like form" described previously (14). 4 To achieve adequate resolution for kinetic analyses of the acylation step for the wild-type enzyme, the acylation rate constants k ac were determined at 10°C and at 0.2 mM D-G3P (16). Kinetics were carried out at 25°C for the R124L, R301L, and R459I mutants and at 10°C for the Y170F mutant. Therefore, to compare the k ac /K m values of the mutants and the wild type, k ac /K m values of wild type and Y170F mutant have to be multiplied by 3.75, assuming a 2.5-fold increase of the rate constant k ac per 10°C (24).
iting step was the same as for wild type. Accordingly, transient kinetics were carried out in the presence of 16 N enzyme, 1 mM NADP, and 1 mM D-G3P at pH 8.5 and at 25°C. Under these experimental conditions, an enzyme-dependent burst of NADPH production was observed for the R124L and Y170F GAPNs. In contrast, no burst was observed for R301L and R459I mutant GAPN proteins. Moreover, for the latter two mutant GAPNs, the measured rate constants were similar to the rate values determined under steady-state conditions (data not shown). Altogether, these results support a rate-limiting step for mutant GAPNs R301L and R459I associated with the acylation process in contrast to wild type and R124L and Y170F GAPNs for which deacylation is rate-limiting.
R124L and Y170F GAPNs-As shown in Table I, substituting Arg-124 with leucine resulted in a strong increase of the K m value of mutant for D-G3P. Under steady-state conditions, no saturating K m effect was observed up to a concentration of 4 mM of D-G3P. Therefore, only an estimation of the K m value could be done (i.e. higher than 15 mM, which is at least 300-fold higher compared with the wild type). Because of the high K m value for D-G3P, the catalytic constant k cat could not be determined. At a subsaturating concentration of 4 mM of D-G3P, the observed rate constant k was 22-fold lower compared with the k cat of wild type. Therefore, at saturating concentration of D-G3P, it is probable that k cat for the R124L mutant is in the range of that determined for wild type. Thus, mutation at Arg-124 seems to have mainly a strong destabilizing K m effect on D-G3P binding within the deacylating subsite.
Under pre-steady-state conditions, at pH 8.5, the acylation rate was decreased by a factor of 2.1 when D-[1-2 H]G3P was used as substrate (Table II). This suggests that hydride transfer was rate-determining within the acylating process, as already shown for wild type. A linear dependence of the rate of acylation versus D-G3P concentration was observed up to 2 mM, whatever the pH tested. Only partial saturation kinetics with respect to substrate could be detected at a concentration of 9 mM D-G3P at pH 8.5 (curves not shown). Therefore, the K m value for D-G3P could be again only estimated by extrapolation: 8 mM (at pH 8.5). The second-order rate constant, k ac /K m , was determined over a pH range of 5-9. As observed for wild type, the pH-k ac /K m curve exhibited an increasing double sigmoidal profile with two pK a values of 6.1 and 7.3 (curve not shown) with a k ac /K m value of 2.1 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 at pH 8.5. Taking into account what was suggested for wild type (16), pK a values of 6.1 and 7.3 would be those of Cys-302 and Glu-268, respectively. Compared with wild type, k ac /K m and K m values for D-G3P are 17-fold lower 4 and 16-fold higher respectively (Table  II). Therefore, mutating Arg-124 has also mainly a K m effect in acylation.
The situation is different for the Y170F mutant GAPN protein. Under steady-state conditions, the K m value for D-G3P was only increased by a factor of 4, whereas the k cat constant was even higher compared with wild type (Table I). The K m and k ac values for D-G3P were also determined under pre-steadystate conditions at pH 8.5 and at 10°C. They were shown to be 0.67 mM and 260 s Ϫ1 , respectively. Compared with k ac /K m and K m values for D-G3P, these values are similar to those determined for the wild type 4 (16).
R301L and R459I GAPNs-For the R301L mutant, no saturating K m effect for D-G3P was observed in the acylation step up to a concentration of 2 mM. Only an estimation of the K m value could be obtained (i.e. higher than 15 mM). Also, the substitution of Arg-459 by isoleucine induced a large increase of the K m value for G3P up to 9.2 mM (Table I). These values are at least 20-fold higher than that of wild type.
No substrate kinetic isotope effect was observed with D-[1-2 H]G3P for either GAPN mutants (Table II). Thus, the ratedetermining step for both mutants is associated with a step prior to hydride transfer (Table II). The second-order rate constant k ac /K m was decreased by a factor around 10 5 relative to wild type. Taking into account the K m values, this indicates that the rate of the new limiting step is at least 2500-fold less efficient than with the wild type.
The low catalytic efficiency of the acylation step for both mutants is probably the consequence of the formation of an inefficient ternary complex GAPN⅐NADP⅐D-G3P due to either an unsuitable positioning of the aldehyde of D-G3P with respect to Cys-302 or to a low chemical reactivity of Cys-302 or both. A low chemical reactivity could be due to either a high pK a or a low accessibility of the thiolate of Cys-302 or both. To explore these different possibilities, two approaches were used. In the first one, the rate of acylation was studied over a pH range of 5-9. k ac values for R459I mutant GAPN protein were determined at subsaturating concentration of 1 mM of D-G3P under conditions where D-G3P does not behave as a competitive inhibitor. As shown in Fig. 2, the pH dependence of k ac /K m exhibited a double-sigmoidal profile with pK a of 6.7 and 7.6. Compared with wild type, pK a of 6.7 would correspond to that of Cys-302, whereas the one at 7.6 would be that of Glu-268. Another means to determine pK a of cysteine is to use a specific kinetic probe like 2PDS. This has the advantage of also probing thiolate accessibility. This allowed us to show that (i) Cys-302 and Cys-382 in apo wild type have similar pK app around 8.5-9.0 and exhibit a low reactivity with k 2 values of 225 M Ϫ1 ⅐s Ϫ1 and 475 M Ϫ1 ⅐s Ϫ1 , respectively, which indicates a low accessibility of both cysteines; (ii) Cys-302 is accessible within the holo form as shown by a k 2 value higher than 2 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 and has a pK app that shifts from 8.5 to 6.2; and (iii) Cys-382, which is located at the surface of the tridimensional structure, has a pK app value within the holo form that remains the same as in the apo form and displays a low accessibility as shown by a k 2 value of 13 M Ϫ1 ⅐s Ϫ1 (14). For the R459I mutant GAPN apoenzyme, two cysteines were reactive with 2PDS but with distinct k 2 constants (data not shown). The pH-k 2 curve for each thiolate form exhibited a monosigmoidal profile with a pK app value of 8.5 and higher than 9.2 and k 2 values of 18 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 and 6 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 at pH 9.5, respectively (see Fig. 3A, solid symbols). From the comparison of the results obtained for the titrations of Cys-302 and Cys-382 by 2PDS in the C382A and C302A GAPN mutants, respectively (14), the pK app of 8.5 and of 9.2 could be assigned to those of Cys-302 and Cys-382, respectively. The fact that the Cys-302 has a pK app value similar to that of Cys-302 in wild-type apoenzyme but with a high k 2 value of 18 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 indicates that Cys-302 is accessible within the active site of R459I GAPN. This result remains to be explained but is reminiscent a similar situation encountered with other active site mutants such as the N169A mutant GAPN protein. 5 Clearly, the energetic barrier between both Cys-302 conformers should be small, and therefore the proportion of both populations is likely to be controlled by a kinetic effect. The high accessibility of Cys-302 in R459I GAPN is probably not due to the presence of 0.25 M (NH 4 ) 2 SO 4 in the buffer. Indeed, the accessibility of Cys-382 is also strongly increased, whereas under similar conditions of salts, the accessibility of Cys-382 in apo-wild type remains low (results not shown). In the presence of saturating concentration of NADP, under conditions where the conformational transition upon cofactor binding has occurred (see below), the pH dependence curve exhibited a monosigmoidal profile with a pK app of 6.7 and a low k 2 value of 97 M Ϫ1 ⅐s Ϫ1 (Fig. 3A, open symbols). The fact that the observed pK app of Cys-302 is 2 pH units lower than that of Cys-382 explains why only Cys-302 was titrated under the pH range tested. As indicated above, K m value for D-G3P of R301L mutant is higher than 15 mM, and the k ac /K m value is drastically decreased. As a consequence, at the concentrations of D-G3P used, the absorption signal at 340 nm is low, particularly at pH lower than 7. Therefore, it was not possible to trace a pH-k cat /K m curve profile with sufficient confidence to attain pK a values. On the other hand, the kinetics with 2PDS could be attained. Kinetics were carried out under similar conditions to those described with the R459I mutant GAPN protein except that 0.25 M of (NH 4 ) 2 SO 4 was omitted in the reaction buffer. Two moles of pyridine-2-thione per subunit were formed with both apo form and apo form on which NADP was bound. The pH-k 2 curves obtained for both forms fitted better to a monosigmoidal profile with pK app values of 8.5 and k 2 values of 450 M Ϫ1 ⅐s Ϫ1 and 91 M Ϫ1 ⅐s Ϫ1 , respectively (see Fig. 3B, solid and dotted lines). These data indicate that 1) both Cys-302 and Cys-382 are little accessible and have similar pK app values and 2) NADP binding to R301L GAPN is not sufficient to induce a conformational rearrangement of the active site in contrast to that observed for the wild type (14). Fig. 4A, the binding of NADP to apo-R459I mutant led to formation of an absorption band at 333 nm, called Ab 333 . This is a situation similar to that observed with phosphorylating NAD-dependent glyceraldehyde-3-phosphate dehydrogenase. In this latter enzyme, it was shown that the transition was due to a charge transfer between the essential Cys-149 and the nicotinamide ring of the cofactor acting as an electron donor and electron acceptor, respectively (20,21). Therefore, it was reasonable to hypothesize that Ab 333 was due to a charge transfer between Cys-302 and the nicotinamide of NADP. The fact that Ab 333 disappeared in the presence of 10 mM iodoacetamide (Fig. 4A), which is a probe of cysteine residue, strengthens this assumption. At pH 8.5, Ab 333 showed a maximal absorbance with a molar extinction coefficient ⑀ M of 6500 M Ϫ1 ⅐cm Ϫ1 . A study of the pH dependence of the intensity of Ab 333 absorbance also revealed an inflection point at pH 6.6 (Fig. 4A). Therefore, this pH corresponds to the pK app 5 S. Marchal, unpublished results.  37 N). Second-order rate constants k ac /K m were calculated by dividing the rate constant determined at each pH (pH 5-9) by D-G3P concentration. Data (q) were analyzed by nonlinear regression analysis according to Equation 2 for a best-fit pK a value (solid line). Two active protonic species were characterized with pK app values of 6.7 and 7.6, respectively. The dashed lines represent the contribution of each species, provided by the terms of Equation 2.  4). b The K m value for D-G3P was estimated from double reciprocal plots of the acylation rate constant k ac versus D-G3P concentrations (0.5-9 mM). c From Table I. d ND, not determined. e For the R301L and R459I mutants, the second-order rate constants k ac /K m were calculated by dividing the observed rate constant by the concentration of 4 mM and 1 mM D-G3P, respectively. The values indicated are those obtained at optimum pH of 8.5. The substrate isotope effects were calculated as described under "Experimental Procedures." of Cys-302. This also correlates well with the pK app value found by titration with 2PDS. Moreover, the appearance of Ab 333 was shown to be time- (Fig. 4B) and enzyme concentration-dependent (curve not shown). This suggests a time-dependent, local conformational change of the active site of R459I mutant occurring upon NADP binding. In a previous study done with the wild type, a way to follow the rate of the conformational rearrangement was to measure the rate of the quenching of the Trp-397 fluorescence intensity upon cofactor binding (14). 5 A similar behavior was observed for R459I GAPN. Moreover, the k obs value determined from quenching fluorescence was similar to that determined with Ab 333 probe under similar conditions (data not shown). Therefore, Ab 333 could be also used as a kinetic probe to follow the conformational rearrangement of R459I mutant GAPN protein upon NADP binding. pH rate dependence of Ab 333 appearance was measured over a pH range of 5.5-9.2. As shown in Fig. 4B, the pH-k obs curve exhibited an increasing double sigmoidal profile with two pK app values of 6.6 and 7.6. This profile and the pK app values are similar to those obtained from the study of the pH dependence of the acylation rate. Therefore, once again, the pK app values of 6.6 and 7.6 should correspond to Cys-302 and Glu-268, respectively. At pH 8.5, the rate of the rearrangement was optimal, with a k obs value of 0.019 s Ϫ1 . This is at least 40-fold higher than in the case of wild type, where the optimal pH was shown to be pH 5 (14). Therefore, removing the guanidinium group of Arg-459, which interacts with the side chain of Glu-268 in the apo structure facilitates the conformational change at least within the binary complex GAPN⅐NADP. DISCUSSION GAPN belongs to the ALDH family. It is one of the ALDHs whose properties have been the most studied at both the structural and enzymatic levels. The fact that all of these studies were carried out with the physiological substrate D-G3P is also a guarantee of the generality of the interpretations of the results. Taking into account all of the enzymatic and structural data available so far, a scenario of the catalytic mechanism of GAPN from S. mutans can be described as presented in the Introduction. As shown, the binding of NADP to apo-GAPN induces a local conformational change of the active site with at least a reorientation of side chains of Cys-302, Glu-268, and Arg-459. The rate of the conformational change upon cofactor binding has a maximal k obs value of 4.7 ϫ 10 Ϫ4 s Ϫ1 at acidic pH, which is not compatible with the k ac and k cat values (14). The fact that adding D-G3P increases the rate of the reorganization by a factor of at least 10 5 -fold (14) 6 indicates that D-G3P binding to the binary complex GAPN⅐NADP strongly favors the active site reorganization. Therefore, characterizing the structural factors involved in D-G3P binding could help to a better understanding of the catalysis and of the evolution of the substrate binding sites of ALDHs. As presented in the Introduction, four residues seem to be conserved in the GAPN family (i.e. Arg-124, Tyr-170, Arg-301, and Arg-459) and could be involved in the recognition of D-G3P and/or in catalysis.

Characterization of the Charge Transfer Transition in Holo-R459I Mutant GAPN-As shown in
Substituting Arg-124 with leucine does not change the nature of 1) the rate-limiting step, which remains deacylation, and 2) the rate-determining step in acylation, which remains 6 S. Marchal and G. Branlant, unpublished results.

FIG. 3. pH dependence of the second-order rate constant (k 2 ) for the reaction of 2PDS on R459I (A) and R301L (B) mutants.
Kinetics were performed at 25°C over a pH range of 5-9.5, as indicated under "Experimental Procedures." The concentrations of mutants and 2 PDS were 3.71 N (in sites) and 111 M, respectively. Rate constants, k obs , were determined using nonlinear regression analysis, and second order rate constants k 2 were calculated by dividing values of k obs by the concentration of 2PDS. For both mutants, constants k 2 fitted to Equation 3. A, pK app values of 8.5 and higher than 9.2 were determined for the R459I mutant apoenzyme (q, f), respectively, and a pK app of 6.7 for the holoenzyme (E). B, pK app values of 8.5 were deduced from the best fit for the R301L mutant apoenzyme (q, f, solid lines) and NADPbound apoenzyme 3 (E, Ⅺ, dotted lines).

FIG. 4. Properties of the charge transfer transition in holo-R459I mutant.
A, pH dependence of Ab333 absorbance. Inset, effect of the addition of iodoacetamide (in excess of 50-fold relative to the subunit enzyme concentration) on the Ab333 absorbance. B, pH dependence of the rate of the Ab333 appearance. The dashed lines represent k values of each protonic state provided by the individual terms of Equation 6. The first pK app value of 6.6 corresponds to that of Cys-302, and the second one of 7.6 corresponds to that of Glu-268. See "Experimental Procedures" and "Results" for more details. hydride transfer. Moreover, no significant effect is observed on acylation and deacylation rates. Therefore, these results exclude a role of the guanidinium group of Arg-124 in stabilizing the transition states associated with hydride transfer and involved in hydrolysis. Only an increase in K m is observed in both acylation and deacylation steps but with a more pronounced effect in the latter step. Therefore, the only role of Arg-124 is to stabilize the binding of G3P via an interaction between the guanidinium group and the phosphate at C-3. This is in accord with the structure of 1) the ternary complex C302S⅐NADP⅐D-G3P in which the guanidinium group of Arg-124 interacts with one of the oxygens of the phosphate and 2) the apo2 form in which the guanidinium group interacts with one of the oxygens of the anion SO 4 a that is postulated to mimic the phosphate of D-G3P. The fact that the pK app of Cys-302 within the acylation complex remains similar to that observed in wild type indicates no role of Arg-124 in activating Cys-302. Again, this is in accord with the x-ray structure of apo2 GAPN, which shows a distance between both residues higher than 12 Å.
The situation is different for the Y170F mutant GAPN protein. No major effect is observed in the K m values for D-G3P and in the rates of both acylation and deacylation steps. Therefore, although the hydroxyl group of Tyr-170 within the ternary complex C302S⅐NADP⅐D-G3P has been shown to be at a hydrogen bonding distance of the oxygen that connects the C-3 carbon to the phosphate in D-G3P, the interaction seems not to be essential for D-G3P binding. The fact that, in some archeal GAPNs, Tyr-170 is replaced by a phenylalanine supports the present results (Fig. 1).
The behavior of the R459I mutant GAPN protein is very different. With this, the acylation process becomes rate-limiting with a limiting step prior to hydride transfer. The efficiency of the acylation step is strongly decreased by at least a factor of 10 5 , which includes both a K m and a k ac contributions. From different approaches, i.e. using kinetic probes like 2PDS and the Ab 333 or tracing the pH rate profile of acylation and of Ab 333 appearance, it has been possible to determine the pK app of Cys-302. Whatever the approaches, a pK app of 6.6 -6.7 is found within either the holo form or the ternary complex. Therefore, the guanidinium group of Arg-459, which is situated 5.7 Å from Cys-302 in the apo2 structure appears to participate directly or indirectly in lowering the pK app of Cys-302 by 0.5 pH units. However, the 0.5-unit increase in pK app could not explain the drastic decrease in k ac . In fact, as shown by using 2PDS probe, Cys-302 is at least 2 ϫ 10 3 -fold less reactive within the R459I holo form than within the holo wild type. Thus, Cys-302 has little accessibility in the R459I holo form. Therefore, it is probable that the location of the side chain of Cys-302 is not optimal and then necessitates a local reorganization for forming a competent ternary complex with D-G3P that could be rate-limiting. Several data favor this assumption. First, the Ab 333 is not observed in holo wild type. This supports a relative positioning of Cys-302 and of the nicotinamide ring of NADP in R459I mutant different from that of wild type. Second, pH dependence of Ab 333 of R459I mutant depends on Cys-302 pK app but not on Glu-268 pK app . This is an indication of a relative positioning of Cys-302 and Glu-268 in the R459I holo form different from that observed in competent holo wild type in which the distance between the side chains of Cys-302 and Glu-268 is around 3.6 Å. As seen from the inspection of the apo2 structure of GAPN, Arg-459 main chain NH interacts with the SO 4 a, which is representative of the phosphate of D-G3P, whereas guanidinium side chain interacts with SO 4 b, which probably mimics the sp 3 transition states. Arg-459 interacts also with Gln-455, which is a residue conserved in almost all GAPNs. Therefore, substituting Arg-459 by isoleucine can also modify the conformation of the loop on which Arg-459 is located. Together, this is in accord with the data of the present study, which shows that substituting Arg-459 by isoleucine has not only a destabilizing K m effect for D-G3P but also a strong kinetic effect, which reflects a Michaelis ternary complex that seems to be inefficient for forming the thiohemiacetal intermediate ternary complex.
The role of Arg-301 is more difficult to comprehend in detail due the very low catalytic efficiency of R301L mutant GAPN protein. As with the R459I mutant, the rate-limiting step is associated with acylation, more precisely prior to hydride transfer, and a high K m value for D-G3P and a low k ac value are observed. Taking into account the apo2 structure of GAPN, which shows an interaction of Arg-301 guanidinium side chain with both the SO 4 a and SO 4 b, it is probable that substituting Arg-301 by leucine should destabilize D-G3P binding but at the same time perturb the relative positioning of D-G3P and of Cys-302 within the ternary complex. Consequently, this would prevent an efficient attack of the thiolate group toward the aldehydic group.
GAPN from S. mutans reduces efficiently D and L isomers of G3P (16). Therefore, Arg-301 and Arg-459 are probably not involved in G3P stereospecificity. This suggests that both arginines do not interact selectively with the C-2 moiety of D-G3P. This is also supported by the observation that both residues are conserved in GAPNs from plants which are described to reduce only D isomer (22). Thus, as supported by the present study, one major role of Arg-301 and Arg-459 would be to position efficiently the D-G3P with respect to Cys-302 within the Michaelis ternary complex. This would favor the formation of the thiohemiacetal intermediate and consequently the efficiency of the hydride transfer. During this part of the catalytic event, the geometry at the C-1 carbon changes from sp 2 to sp 3 and reveals the role of Asn-169. Now, the negative charge of the oxygen of the thiohemiacetal intermediate is stabilized by the oxyanion site, which is composed at least of the amide side chain of Asn-169. This is in accord with previous studies, which showed that mutating Asn-169 results in a drastic decrease of the hydride transfer rate but has no effect on K m value for D-G3P (15).