Aspartate 142 Is Involved in Both Hydrolase and Dehydrogenase Catalytic Centers of 10-Formyltetrahydrofolate Dehydrogenase*

The enzyme 10-formyltetrahydrofolate dehydrogenase (FDH) catalyzes conversion of 10-formyltetrahydrofolate to tetrahydrofolate in either a dehydrogenase or hydrolase reaction. The hydrolase reaction occurs in a 310-residue amino-terminal domain of FDH (Nt-FDH), whereas the dehydrogenase reaction requires the full-length enzyme. Nt-FDH shares some sequence identity with several 10-formyltetrahydrofolate-utilizing enzymes. All these enzymes have a strictly conserved aspartate, which is Asp142 in the case of Nt-FDH. Replacement of the aspartate with alanine, asparagine, glutamate, or glutamine in Nt-FDH resulted in complete loss of hydrolase activity. All the mutants, however, were able to bind folate, although with lower affinity than wild-type Nt-FDH. Six other aspartate residues located near the conserved Asp142 were substituted with an alanine, and these substitutions did not result in any significant changes in the hydrolase activity. The expressed D142A mutant of the full-length enzyme completely lost both hydrolase and dehydrogenase activities. This study shows that Asp142 is an essential residue in the enzyme mechanism for both the hydrolase and dehydrogenase reactions of FDH, suggesting that either the two catalytic centers of FDH are overlapped or the dehydrogenase reaction occurs within the hydrolase catalytic center.

The liver cytosolic enzyme 10-formyltetrahydrofolate dehydrogenase (FDH) 1 converts 10-formyltetrahydrofolate (10formyl-THF) to tetrahydrofolate (THF) in an NADP ϩ -dependent dehydrogenase reaction or an NADP ϩ -independent hydrolase reaction ( Fig. 1) (1-3). 10-Formyl-THF represents one of the major forms of folate in the cell (4) and plays a significant biological role, being involved in two reactions of the de novo purine biosynthesis pathway (5). It is also involved in protein biosynthesis in bacteria, chloroplasts, and mitochondria through formylation of the specialized methionyl-tRNA that is necessary for initiation of translation (6). The dehydrogenase reaction performed by FDH is important for recycling excess 10-formyl-THF that is not needed for purine biosynthesis and restoration of the THF pool (5). The reaction is probably important also for metabolism of formate by clearing it as CO 2 and thus protecting the cell from formate intoxication (7). It is not clear whether the hydrolase reaction has any physiological significance and whether the reaction occurs in vivo. Study of the hydrolase reaction in vitro, however, is important as a means of understanding the FDH mechanism. The enzyme has a very complicated structure. It exists as a homotetramer of 902-residue subunits. The FDH monomer is composed of two functional domains (amino-and carboxyl-terminal) connected by an ϳ100-residue linker (8 -10). The carboxyl-terminal domain of FDH (residues 420 -902) has ϳ50% identity to class 1 and 2 aldehyde dehydrogenases (8,9), suggesting that this part of the FDH molecule is derived from an aldehyde dehydrogenase-related gene. Earlier, we showed that the hydrolase catalytic center is located within a 310-residue amino-terminal domain of the enzyme (10), whereas the dehydrogenase reaction requires the full-length enzyme (9,10). We have also shown that the 310-residue amino-terminal domain of FDH (N t -FDH) can be expressed separately as a functional protein.
It bears the folate-binding site and carries out the 10-formyl-THF hydrolase reaction converting 10-formyl-THF to THF and formate (10). Thus, the enzyme is an interesting example of a natural fusion of two unrelated proteins in which the aminoterminal domain bears the substrate-binding site while the machinery of the aldehyde dehydrogenase homologous carboxyl-terminal domain is used as the catalytic center in the dehydrogenase reaction. N t -FDH shares some sequence identity with other enzymes that also use 10-formyl-THF as a substrate: glycinamide-ribonucleotide formyltransferase (GART) (11), L-methionyl-tRNA formyltransferase (FMT) (6), and 10-formyl-THF hydrolase (12). Although the overall identity between proteins in this group is low (16 -30%), there is a region of ϳ70 residues in the middle of the peptide sequences where identity increases to 30 -50% (Fig. 2). This region was predicted to be a folatebinding site (8). Four out of the 14 conserved residues in this region are glycines, and two are prolines, apparently reflecting a similar overall fold of the region. Resolved crystal structures of two of the enzymes, GART (13,14) and FMT (15), revealed that despite relatively low overall sequence identity, this part of the two protein molecules, which is adjacent to the bound folate substrate, has a very similar structure (Fig. 3) (13)(14)(15). Besides the glycine and proline residues, there is a strictly conserved aspartate in this area that might be a candidate for folate binding and/or catalysis. Resolved crystal structures of GART and FMT showed that this aspartate is adjacent to the substrate bound in the active center (13,14) and imply that it might function within the catalytic center. In the GART crystal structure, the conserved aspartate is located within the flexible loop, and it is likely that binding of the cofactor is required for stabilization of this loop (16). However, the relative orientation of this loop can vary considerably in all the different GART structures (16). It is well known that protein flexibility is often directly linked to protein function (16). Earlier studies on GART showed that mutation of the conserved aspartate to asparagine results in non-active enzyme (11). Later it was concluded that the aspartate is not absolutely required for catalysis, although its replacement results in an enzyme with substantially reduced activity (17). From the crystal structure of FMT, the conclusion has also been made that, similar to GART, the conserved aspartate is located within the enzyme active center and probably participates in the catalytic mechanism (15). In this work, we studied the role of the conserved aspartate in FDH function. The dehydrogenase reaction is carried out by full-length FDH; the hydrolase reaction is carried out by either fulllength FDH or its 310-residue amino-terminal domain.
FIG. 2. Diagrammatic scheme of the FDH domain structure. Residues 1-310 compose the amino-terminal domain (N t -FDH); residues 311-419, the intermediate domain; and residues 420 -902, the aldehyde dehydrogenase homologous carboxyl-terminal domain. Enzymes utilizing 10-formyl-THF as a substrate and sharing sequence identity with N t -FDH are as follows: 10-formyl-THF hydrolase (FH), GART, and FMT. Alignment of sequence fragments with the greatest identity is shown. The alignment is between FDH and three other proteins. Identical residues are indicated by vertical lines, and conservative changes are shown by plus signs. Asterisks show positions of aspartate residues in the N t -FDH sequence.
FIG. 3. Comparison of the crystal structures of GART and FMT. The structures have been taken from the Protein Data Bank (codes 3GAR for GART and 1FMT for FMT) and were generated using Rasmol (Glaxo Wellcome). The structure of the first 211 residues is shown for FMT. The conserved region of the molecules (see Fig. 2) is in red; the conserved Asp is in yellow.

EXPERIMENTAL PROCEDURES
Materials-10-Formyl-5,8-dideazafolate (10-formyl-DDF) was obtained from Dr. John B. Hynes (Department of Pharmaceutical Chemistry, Medical University of South Carolina). All media were obtained from Difco or Life Technologies, Inc. SDS-PAGE standards were purchased from Amersham Pharmacia Biotech. Other chemicals were obtained from Sigma.
Site-directed Mutagenesis-We used the QuikChange site-directed mutagenesis kit (Stratagene) to introduce point mutations into N t -FDH cDNA cloned to the pRSET vector (18) or into full-length FDH cDNA cloned to the pVL1393 vector (18). After introduction of the mutation, the vector was transformed into Escherichia coli XL-1 Blue cells using a standard protocol, and the cells were grown on LB/ampicillin plates to allow mutant selection. Plasmids were isolated by using the Quantum Prep kit (Bio-Rad) and sequenced to identify clones carrying the mutation. The entire coding region of the clones carrying the introduced mutations was sequenced to ensure the absence of random mutations. Sequencing was carried out in the Vanderbilt Medical Center Core facility.
Expression of N t -FDH Mutants-Expression of N t -FDH mutants was done as we earlier described (18). The pRSET vector carrying the mutation was transformed into E. coli BL21(DE3) cells (Novagen) according to the manufacturer's protocol, and the cells were grown in 4 ml of NZCYM medium (Life Technologies, Inc.) containing ampicillin (50 mg/ml) overnight at 37°C with shaking. NZCYM medium (100 ml) containing ampicillin was inoculated with the overnight culture and incubated at 37°C with shaking until A 600 ϭ 0.6 (ϳ6 h), followed by induction with isopropyl-␤-D-thiogalactopyranoside (1 mM final concentration). Three hours after induction, the cells were harvested by centrifugation (5000 ϫ g, 10 min), resuspended in 2 ml of buffer (50 mM Tris-HCl, pH 8.0, 2 mM EDTA, and 0.1% Triton X-100), and incubated for 30 min with 0.2 mg/ml lysozyme at 37°C. The suspension was chilled on ice and sonicated (three times for 30 s). The cellular debris was removed by centrifugation (13,000 ϫ g, 15 min, 5°C), and the supernatant was examined for the presence of N t -FDH by SDS-PAGE followed by immunoblot analysis and for 10-formyl-DDF hydrolase activity.
Purification of N t -FDH Mutants-The mutants were purified using a procedure that we earlier developed for wild-type N t -FDH (18). All buffers used in purification steps contained 1 mM NaN 3 . The supernatant of the cell lysate (2 ml), obtained as described above, was dialyzed overnight against 20 mM Tris-HCl, pH 7.4, and was then passed through a 0.2-m filter (Nalgene) to remove insoluble contaminants. Purification was done on a DEAE MemSep 1000 HP ion-exchange membrane chromatography cartridge (Millipore Corp.) with a linear NaCl gradient (0 -0.4 M in 20 mM Tris-HCl, pH 7.4) using a ConSep LC100 chromatography system (Millipore Corp.). Lysate samples of 1 ml each containing ϳ4 mg of total protein were used in the chromatography run. The chromatography was done at a flow rate of 4.0 ml/min, and fractions of 0.8 ml were collected. The fractions were analyzed for the presence of N t -FDH by SDS-PAGE and immunoblot analysis. The fractions containing N t -FDH were combined; dialyzed against 20 mM Tris-HCl, pH 7.4; concentrated using a Centricon 10 (Amicon-Millipore); and stored at 4°C in the presence of 1 mM NaN 3 and 10 mM 2-mercaptoethanol.
Expression of the D142A Mutant of FDH-The expression of the full-length FDH mutant was done in insect cells using a baculovirus expression system as we previously described (19). The pVL1393 vector carrying mutant FDH cDNA and linearized AcNPV viral DNA were cotransfected into Sf9 cells using the BaculoGold transfection kit (Pharmingen, San Diego, CA) according to the manufacturer's protocol. Recombinant viral stock was amplified in Sf9 cells to produce high-titer virus stock. High Five cells were seeded as a monolayer in 25-cm 2 flasks (ϳ2 ϫ 10 6 cells/flask) and grown overnight. Each flask was infected with 0.1 ml of high-titer virus stock. Culture medium and cells were collected separately 2-4 days post-infection, and the recombinant protein production was detected by SDS-PAGE (20) followed by Western immunoblot analysis (21) with rabbit polyclonal antiserum raised against pure rat liver FDH and goat anti-rabbit IgG conjugated to alkaline phosphatase (Bio-Rad) (22). Protein concentration was determined using the Bradford protein assay (23). For large-scale recombinant protein production, High Five cells were seeded in 225-cm 2 flasks (ϳ20 ϫ 10 6 cells/flask); and after growing overnight, each flask was infected with 1.0 ml of high-titer virus stock. The culture medium at 96 h post-infection was collected, centrifuged to remove detached cells, and used to isolate recombinant protein.
Purification of the D142A Mutant of FDH-All buffers used in the purification steps contained 10 mM 2-mercaptoethanol and 1 mM NaN 3 . The FDH mutant was purified from the cell-free culture medium by affinity chromatography on a column of 5-formyltetrahydrofolate-Sepharose essentially as we earlier described (19). A column (1.5 ϫ 10 cm) was packed with ϳ8.0 ml of settled gel and equilibrated with 10 mM Tris-HCl, pH 7.4, containing 10 mM 2-mercaptoethanol and 1 mM NaN 3 (buffer A). Medium (200 ml) containing 2-mercaptoethanol (10 mM) and NaN 3 (1 mM) was applied to the affinity column. The column was then washed with buffer A containing 1 M KCl (100 ml). The enzyme was eluted from the column with buffer A containing 1 M KCl and 20 mM folic acid. The eluate was passed through a column of Bio-Gel P6-DG (Bio-Rad) equilibrated with buffer A and concentrated to ϳ5 ml using an Amicon filtration cell. Additional purification was done on a Mono Q column in 20 mM Tris-HCl, pH 7.4, with a linear KCl gradient (0 -0.5 M in buffer A) using a fast protein liquid chromatography system (Amersham Pharmacia Biotech).
Analysis of the Mutant Proteins-SDS-PAGE was done according to the method of Laemmli (20) on an 8.0% gel. Western immunoblot analysis was performed as described by Burnette (21) with rabbit polyclonal antiserum against pure rat liver FDH and goat anti-rabbit IgG conjugated to alkaline phosphatase (22). Protein concentration was determined using the Bradford protein assay (23). Amino-terminal sequence analysis of wild-type and mutant N t -FDH was carried out in the Vanderbilt Medical Center Core facility using automated Edman degradation with gas-phase analysis on an LF-3000 apparatus (Beckman) or on an Applied Biosystems Model 470A protein sequencer (Perkin-Elmer).
Assay of Enzyme Activity-All assays were performed at 30°C in a Perkin-Elmer Lambda 4B double-beam spectrophotometer. For measurement of hydrolase activity, the reaction mixture contained 0.05 M Tris-HCl, pH 7.8, 100 mM 2-mercaptoethanol, and 5 mM substrate 10-formyl-DDF. 10-Formyl-DDF is an alternative, stable substrate for the enzyme (24). The reaction was started by the addition of enzyme (1-20 g) in a final volume of 1.0 ml and read against a blank cuvette containing all components except the enzyme. The appearance of the product, 5,8-dideazafolate (DDF), was measured at 295 nm using a molar extinction coefficient of 18.9 ϫ 10 3 (25). The addition of NADP ϩ to the reaction mixture provided a measure of both dehydrogenase and hydrolase activities, i.e. total activity of the enzyme. Hydrolase activity measured in the absence of NADP ϩ was subtracted from the total activity to give the dehydrogenase activity. Dehydrogenase activity was also measured independently using the increase in absorbance at 340 nm due to production of NADPH and a molar extinction coefficient of 6.2 ϫ 10 3 . Aldehyde dehydrogenase activity was assayed using propanal as we earlier described (9). The reaction mixture contained 50 mM CAPS, pH 9.4, 5 mM propanal, 1 mM NADP ϩ , and enzyme in a total volume of 1 ml. Activity was estimated from the increase in absorbance at 340 nm.
Fluorescence Studies-All fluorescence experiments were done on a Perkin-Elmer Model 650-40 fluorescence spectrophotometer. Emission fluorescence spectra of wild-type and mutant N t -FDH were recorded by scanning from 300 to 460 nm with fluorescence excitation at either 280 or 295 nm. Protein samples (0.1 mg/ml) were in 20 mM Tris-HCl, pH 7.5. Binding of 10-formyl-DDF and DDF to the N t -FDH mutants was detected by measuring the quenching of protein tryptophan fluorescence. Protein samples (ϳ10.0 nM) were in 50 mM Tris-HCl, pH 7.8. The concentrations of ligands were varied from 2 to 2000 nM for 10-formyl-DDF and from 50 nM to 40 M for DDF. The experiments were done at 25°C. Fluorescence excitation was at 291 nm, and emission was monitored at 340 nm. The data were corrected for dilution effect, for 10formyl-DDF or DDF intrinsic fluorescence, and for absorptive screening caused by DDF. Intrinsic fluorescence of 10-formyl-DDF or DDF was measured experimentally. In case of DDF, the observed fluorescence signal was multiplied by the correction factor (C) to obtained the corrected data. The correction factor was calculated according to the equation C ϭ 10 ⌬A/2 , where ⌬A is the absorbance of DDF at the excitation wavelength (26). K d values for the ligands were calculated from data on fluorescence quenching in the presence of ligands that were plotted in a linear form. The value (1 Ϫ F/F o ) Ϫ1 was plotted against the inverse of ligand concentration (27). This is a modified form of the classical Stern-Volmer plot that relates the decrease in fluorescence to the concentration of a collisional quencher (28). F is intrinsic fluorescence observed at a quencher concentration; F o is fluorescence in the absence of quencher. The slope of the line (least-squares fit) gave a K d . Variation of the measured values was ϳ5%.

RESULTS
Site-directed Mutagenesis of Aspartate 142 in N t -FDH Function-We mutated aspartate 142 to alanine and expressed the mutants in E. coli. Earlier, we had expressed N t -FDH in E. coli (18) and showed that it is a functional protein identical to the protein expressed in insect cells (10). The mutant was expressed as a soluble protein, and it showed a high level of expression similar to that of wild-type N t -FDH (Fig. 4A). Analysis of 10-formyl-DDF hydrolase activity in the soluble fraction of the cell lysate in which the mutant was the major protein component did not reveal any activity. Mutant N t -FDH was purified (Fig. 4B) from the cell lysate in one step by ion-exchange chromatography as we previously described (18). Mutant and wild-type N t -FDH demonstrated identical chromatographic behavior when purified on an ion-exchange column (data not shown). Analysis of the purified mutant protein showed that it did not produce detectable hydrolase activity. Based on the limit of the assay, we estimated that the mutant had Ͻ0.02% of the activity of wild-type N t -FDH. Thus, we assumed that the mutant was practically inactive.
We additionally expressed three N t -FDH mutants in which aspartate 142 was changed to asparagine, glutamate, or glutamine. All of the expressed mutant proteins were soluble and showed levels of expression similar to those of wild-type N t -FDH and other mutants (Fig. 4A). No 10-formyl-DDF hydrolase activity was detected in the soluble fraction of cell lysates for all of the expressed mutants. The proteins were purified ( Fig. 4B) using ion-exchange chromatography (18). All the mutants demonstrated identical chromatographic behavior during the purification similar to that of the D142A mutant and wildtype N t -FDH (data not shown). Assay of the hydrolase activities of the purified mutants showed that they all completely lost activity. Based on the detectable level of the activity, we concluded that, similar to the results obtained with the D142A mutant, the hydrolase activities of the mutants were Ͻ0.02% of that of the wild-type protein. To study whether recombinant proteins were subjected to proteolytic degradation, we sequenced the first five residues from the amino-terminal end. Analysis of the amino-terminal sequences of wild-type N t -FDH and the mutants showed that the sequences were identical and corresponded to the original sequence of N t -FDH. We also recorded emission spectra of protein fluorescence to learn whether the mutations affected the protein tertiary structure. The spectra were generated at an excitation wavelength of either 280 or 295 nm. With excitation at 295 nm, the spectrum reflects fluorescence of tryptophans, whereas at 280 nm excitation, there is significant contribution of tyrosine residues to the fluorescence. N t -FDH has six tryptophan and five tyrosine residues, with one of the tryptophans located just six residues upstream of the mutated aspartate. We observed that emission fluorescence spectra were identical for wild-type N t -FDH and all Asp 142 mutants for each of the excitation wavelengths with the maximum at 337 nm (data not shown).
Site-directed Mutagenesis of Aspartates in the Putative Folate-binding Region of N t -FDH-We also mutated to alanine all other aspartate residues in the region between residues 106 and 158. In addition to D142A, the following mutants were created and expressed in E. coli: D127A, D138A, D139A, D145A, D157A, and D158A. All mutants were expressed as soluble proteins, and they revealed high and similar levels of expression. Analysis of the 10-formyl-DDF hydrolase activity in the soluble fraction of the cell revealed that all the mutants possessed activity. We purified the mutants using ion-exchange chromatography (18) (Fig. 5) and measured their specific activity. All six mutants displayed activity similar to that of wild-type N t-FDH (Table I).
Ligand Binding Study-We studied binding of 10-formyl-DDF and DDF to the purified mutant proteins using titration of tryptophan fluorescence (27,28). We previously used this approach to study binding of 10-formyl-DDF and DDF to fulllength FDH and N t -FDH (10). The fact that the hydrolase reaction does not proceed in the absence of 2-mercaptoethanol allowed us to specifically measure binding of 10-formyl-DDF since no reaction can take place (10, 24). To study binding parameters under the same conditions, we also did not use 2-mercaptoethanol in titrations with DDF. We observed significant and similar quenching of tryptophan fluorescence using both 10-formyl-DDF (38% maximal quenching) and DDF (42% maximal quenching). Such a significant level of quenching might be explained, in our opinion, by the fact that two out of a total of six tryptophans of N t -FDH are located within the conserved region, which was suggested to be a putative folatebinding site (8). Two mutants, D142E and D142Q, showed a lower level of maximal quenching with 10-formyl-DDF (26%) compared with the wild-type protein and D142N and D142A mutants (36 -38%) (Fig. 6A). We explain this phenomenon by the fact that the longer side chain of glutamate or glutamine apparently creates steric hindrance for the formyl group of the substrate within the binding pocket, preventing it from assuming the correct orientation in which maximal quenching could be achieved. Limitation of the flexibility of the Asp 142 -bearing loop after introducing a residue with a longer side chain could be the factor generating the steric hindrance. In contrast with 10-formyl-DDF, in the presence of DDF (which does not have a formyl group), fluorescence was decreased in a similar manner for all mutants as well as for wild-type N t -FDH, with maximal quenching between 39 and 42% in each titration (Fig. 6B). The fluorescence experiments showed that mutation of the aspartate resulted in a significant decrease in apparent binding affinity of N t -FDH for the substrate of the hydrolase reaction, 10-formyl-DDF, as well as for the product of the reaction, DDF (Table II). The highest decrease was observed for the D142A mutant, and the lowest changes were observed for the D142N mutant. Two other mutants, D142E and D142Q, displayed similar affinity, with apparent K d values intermediate between those of the D142A and D142N mutants (Table II). All mutants showed a proportional decrease in affinity for both 10-formyl-DDF and DDF, with the difference in ⌬G for binding of 10formyl-DDF and DDF within 2.0 -2.2 kcal/mol, similar to that for the wild-type enzyme. This suggests that Asp 142 does not contribute to the binding of the formyl group of the substrate.
Site-directed Mutagenesis of Aspartate 142 in FDH-To study the role of the conserved aspartate in FDH function, we changed the residue in the full-length enzyme to alanine. The FDH mutant was expressed in insect cells using the baculovirus expression system as we previously described (9,10). We used this expression system instead of expression in E. coli because earlier we showed (18) that, in contrast to N t -FDH, the full-length enzyme is expressed in E. coli as an insoluble protein. The D142A mutant was expressed in insect cells in amounts similar to those of wild-type FDH and was purified (Fig. 7) using affinity chromatography followed by chromatography on a Mono Q column. In contrast to the wild-type enzyme, the mutant showed a slightly different elution pattern from the affinity column, i.e. it started eluting at 1.0 M NaCl, whereas the wild-type enzyme was not eluted by 1.0 M NaCl, but required 20 mM folate to elute it (data not shown). This probably reflects changes in the affinity of folate binding due to the mutation. Nevertheless, the mutant still interacted with the affinity column strongly enough to allow the use of this procedure for purification. During ion-exchange chromatography on a Mono Q column, the mutant was eluted at ϳ0.3 M KCl, similar to wild-type FDH. All three activities, i.e. hydrolase, dehydrogenase, and aldehyde dehydrogenase, inherent to wildtype FDH were assayed for the mutant. We observed that the mutant protein displayed aldehyde dehydrogenase activity similar to that of wild-type FDH (Fig. 8). As expected, similar to the N t -FDH mutants, the D142A FDH mutant completely lost 10-formyl-DDF hydrolase activity (Fig. 8). Assay of 10formyl-DDF dehydrogenase activity revealed that, together with the hydrolase activity, the FDH mutant also lost the dehydrogenase activity (Fig. 8). Based on the detection limit of the assay used, we estimated that the mutant has Ͻ0.02% of the hydrolase activity and Ͻ0.01% of the dehydrogenase activity of wild-type N t -FDH. The difference in the figures is ex-  plained by the fact that wild-type FDH has ϳ2-fold higher dehydrogenase activity than hydrolase activity. Thus, we concluded that the mutant was inactive with respect to both hydrolase and dehydrogenase reactions. To estimate conformational integrity of the mutant, we compared its emission fluorescence spectra with the spectra of wild-type FDH at excitation wavelengths of 280 and 295 nm. At both excitation wavelengths, the spectra of the wild-type enzyme and mutant were similar, with an emission maximum at 335 nm (data not shown). DISCUSSION There are several enzymes in the cell that use 10-formyl-THF as a substrate, either transferring the formyl group to another compound or releasing it as formate. All these enzymes have a strictly conserved aspartate that has been predicted to be involved in substrate binding and/or catalysis of FDH (8). Studies on GART showed that this aspartate is an important residue in the enzyme catalytic mechanism (11,17). Yet, the simple extrapolation of the role of this residue in catalysis to other enzymes is not valid. For example, another 10-formyl-THF-utilizing enzyme (AICAR formyltransferase) that carries out a reaction that is very similar to GART reaction has no substantial identity to the other 10-formyl-THF-requiring enzymes. However, after introducing a number of gaps, the alignment of the conserved amino acid sequence of human GART with the sequence of human AICAR formyltransferase indicated that the sequence is at least partially conserved (30). As a result of such an alignment, an aspartate residue in the AICAR formyltransferase sequence that corresponds to the aspartate in the GART active site was identified. In contrast with GART, site-directed mutagenesis experiments revealed that this residue is not involved in AICAR formyltransferase function (30). This led to the conclusion that this aspartate is simply part of a common structural motif. The fact that 10formyl-THF synthetase/C 1 -synthase also reveals elements of the conserved sequence (8) supports this conclusion. These enzymes synthesize 10-formyl-THF from THF and formate in an ATP-dependent reaction (31), thus catalyzing a reaction that is the reverse of the FDH hydrolase reaction. All C 1synthetases also have a conserved aspartate corresponding to the conserved aspartate of 10-formyl-THF-requiring enzymes (8,32). It has been shown, however, that, similar to AICAR formyltransferase, this aspartate is not a critical catalytic residue (32).
In the case of N t -FDH, the single amino acid substitution of alanine for aspartate 142 resulted in complete loss of activity. The common homologous region of 10-formyl-THF-requiring enzymes is enriched with aspartate residues. In the case of N t -FDH, seven out of 14 total protein aspartates are located in this region within the sequence of 32 residues (Fig. 2). Besides the strictly conserved aspartate, three out of the other six aspartates in this region are common for three out of four proteins. However, none of the residues except the conserved one was crucial for hydrolase activity. To elucidate further the role of Asp 142 in the function of the enzyme, we expressed mutants in which the residue was changed to asparagine, glutamate, or glutamine, a common approach in the study of the Asp-Asn-Glu-Gln quartet. In some cases, if aspartate is involved in catalysis, asparagine or glutamate is able to at least partially compensate, and the proteins retain some enzyme activity (33)(34)(35)(36)(37)(38)(39). In the case of N t -FDH, none of the mutants revealed any hydrolase activity, suggesting that Asp 142 is an essential and irreplaceable residue in the hydrolase catalytic center of FDH. A similar result has been reported for GART; the D144N GART mutant was essentially inactive (11). The role of Asp 142 in FDH function is even more complex since the residue is also critical for dehydrogenase catalysis. This suggests that the dehydrogenase mechanism is not independent and is related to the hydrolase mechanism.
Site-directed mutagenesis experiments often raise the question of whether the mutant retains the same folded conformation as the wild-type protein or whether the mutation creates significant changes in protein conformation, in other words, whether the lack of activity is a result of the loss of a catalytically important residue or whether it is just a result of incorrect folding due to changes in amino acid sequence. In most cases, it is unlikely that a single amino acid substitution will influence folding significantly (40,41); and usually, the protein compensates for changes by small adjustments in its structure (42). In practice, there are simple signs, such as the level of  expression and solubility of the recombinant protein, that indicate whether the protein is functional or whether there are significant disturbances of its structure. In our experiments Asp 142 mutants were indistinguishable from wild-type N t -FDH except for lack of activity. They were expressed as soluble proteins at the same level as wild-type N t -FDH, were eluted from an ion-exchange column at the same concentration of salt, and were recognized with specific antibodies raised against FDH. Comparison of emission fluorescence spectra of wild-type and mutant N t -FDH indicated that the mutations did not alter the environment of tryptophan or tyrosine residues, also sug-gesting that protein conformation was not disturbed. All of the above were also true for the D142A mutant of full-length FDH. It is necessary to emphasize that mutation of three other aspartates (Asp 138 , Asp 139 , and Asp 145 ) located in close proximity to the conserved one (just three and four residues upstream and three residues downstream, respectively) did not change the enzyme activity. This suggests that amino acid alterations in this area do not noticeably affect protein conformation. This probably is connected with the fact that the region presents a flexible loop in protein structure. We make this assumption based on the crystal structures of GART and FMT (see Fig. 3). All this together supports the conclusion about the direct involvement of Asp 142 in the hydrolase and dehydrogenase reactions.
Based on the calculated difference in ⌬G for binding the substrate and product of the reaction, we earlier suggested that the formyl group of the substrate is hydrogen-bonded in the protein-binding site (10). Our present results suggest that the conserved aspartate is not the residue involved in binding of the substrate formyl group. This conclusion is based on the fact that, compared with the wild-type protein, the mutants revealed a proportional decrease in affinity for both the substrate (10-formyl-DDF) and the product (DDF) of the reaction. If interaction occurs through the formyl group, the mutants would display a decrease in affinity only for 10-formyl-DDF, but not for DDF, which does not have the formyl group. Binding energy is often used as a characteristic of molecular recognition in site-directed mutagenesis (43)(44)(45)(46)(47)(48)(49). The difference between ⌬G for binding of 10-formyl-DDF and DDF was similar for the wild-type enzyme and all mutants, suggesting no disruption of bond(s) between the enzyme and the formyl group of the substrate. At the same time, replacement of Asp 142 resulted in a significant decrease in affinity for folate, suggesting involvement of the residue in substrate binding. Mutation of the aspartate to alanine weakened the binding energy by ϳ1.7 kcal/mol, whereas the disruption of a hydrogen bond involving a charged group generally weakens binding by 3.5-4.5 kcal/mol (50). If the aspartate is not involved in hydrogen bonding of folate, the decrease in binding energy could probably reflect weakening of electrostatic interactions or distortion of substrate due to slight conformational changes within the binding pocket. However, the conformational alterations, if any, were not significant since we did not detect any changes in tryptophan fluorescence of the mutants compared with wild-type N t -FDH. In some cases, disruption of a hydrogen bond results in lower changes in ⌬G (51, 52). One possibility might be that an oxygen of the carboxyl group of Asp 142 is hydrogen-bonded to N-10 of folate through a molecule of water as was proposed for GART (14). The oxygen of the carboxamide group of asparagine can probably significantly compensate for the oxygen of the aspartate in the interaction with folate. Asparagine, however, is not able at all to substitute for the aspartate in the enzyme mechanism. Summarizing, we would like to emphasize that the binding experiments resulted in two conclusions: first, that Asp 142 influences folate binding, and second, that it is not involved in binding the formyl group of the substrate.
This study provides strong evidence that aspartate 142 contributes directly and profoundly to FDH catalysis and is an essential residue for both hydrolase and dehydrogenase activities of FDH. This suggests that it is involved in both FDH catalytic centers, leading to the conclusion that both centers are overlapped. Earlier, we showed that the dehydrogenase reaction requires full-length FDH (9,10). We also identified a residue (Cys 707 ) located in the carboxyl-terminal domain of FDH that is crucial for the dehydrogenase reaction, but that does not participate in the hydrolase reaction of FDH (53). FIG. 9. Model of the FDH hydrolase catalytic center. The loose model suggests that hydrolase and dehydrogenase catalytic centers are overlapped. Flexibility of the loop where Asp 142 (shown as a closed circle) is presumably located allows transfer of the substrate within the binding pocket from the "hydrolase" site (A) to the "dehydrogenase" site (B). The model with the tight substrate-binding pocket suggests that there is no separate dehydrogenase catalytic center and that the dehydrogenase reaction occurs within the hydrolase catalytic center (C). C t -FDH, carboxyl-terminal domain of FDH; DH, dehydrogenase. See "Discussion" for more details.
Thus, transfer of the substrate within the superimposed catalytic center from the hydrolase to the dehydrogenase site should allow its close contact with the cysteine in the carboxylterminal domain (Fig. 9B). Flexibility of the loop where the aspartate is presumably located (see structures in Fig. 3) should allow such transfer switching between hydrolase (Fig.  9A) and dehydrogenase (Fig. 9B) reactions. This is the "loose" model of the substrate-binding pocket. The switching from the hydrolase reaction to the NADP ϩ -dependent dehydrogenase reaction in this model might be mediated, for example, by binding of NADP ϩ . The other possibility might be that the substrate is not very flexible in the folate-binding site (Fig. 9C). This is the "tight" model of the substrate-binding pocket. In this model, the dehydrogenase reaction occurs within the hydrolase catalytic center when cysteine 707 of the adjacent carboxyl-terminal domain is involved in the catalysis. Both models are consistent with existence of one folate-binding site/FDH monomer (54). There is a principal difference between loose and tight models of the catalytic center. The loose model suggests that either the hydrolase or the dehydrogenase reaction can occur, and depending on the reaction, the formyl group is removed from the substrate by two different mechanisms either as formate or as CO 2 . The tight model suggests that the dehydrogenase mechanism includes the hydrolase mechanism as an essential part. We expect that further studies will allow us to make more definite conclusions about the structure of the hydrolase/dehydrogenase catalytic center and about the hydrolase/dehydrogenase catalytic mechanism.