D-Serine Dehydratase from Escherichia coli DNA SEQUENCE AND IDENTIFICATION OF CATALYTICALLY INACTIVE GLYCINE TO ASPARTIC ACID VARIANTS*

We have identified two glycyl residues whose integ-rity is essential for the catalytic competence of a model pyridoxal 5”phosphate requiring enzyme, D-serine dehydratase from Escherichia coli. This was accom-plished by isolating and sequencing the structural gene from wild type E. coli and from two mutant strains that produce inactive D-serine dehydratase. DNA sequencing indicated the presence of a single glycine to aspartic acid replacement in each variant. The amino acid replacements lie in a glycine-rich region of D- serine dehydratase well removed from pyridoxal 5‘-phosphate-binding lysine 118 in the primary structure of the enzyme. The striking effect of these two glycine to aspartic acid replacements on catalytic activity, the conservation of the glycine-rich region in several pyridoxal 5‘-phosphate-dependent enzymes that catalyze a//3-eliminations, and the placement of similar glycine-rich sequences in well-characterized active site struc- tures suggest that the glycine-rich region interacts with the cofactor at the active site of the enzyme. The

study (i) establishes the complete DNA and deduced protein sequence of the wild-type enzyme and (ii) identifies two G + D replacements that virtually inactivate DSD but do not prevent the binding of cofactor. The replacements are shown to lie in an unusually glycine-rich region that resembles the glycine-rich sequence found in other enzymes that bind phosphorylated ligands.

MATERIALS AND METHODS
Chemicals, Antisera, and Enzymes-Tryptone and yeast extract for LB media (19) were obtained from Difco. DTT was ordered from Behring Diagnostics or Boehringer Mannheim. Reagents for polyacrylamide gel electrophoresis were purchased from IBI, and Tween 20 and 4-Chloro-1-napthol for Western blotting from Bio-Rad. Polymin P and goat anti-rabbit horseradish peroxidase conjugate (61-202-3) were ordered from Miles. [y-32P]ATP (>lo00 mCi/mmol) and [~x-~~S]thio-dATP (>lo00 mCi/mmol) were supplied by Amersham Corp. Other chemicals were analytical grade. Restriction enzymes and enzymes used for labeling of DNA were purchased from Bethesda Research Laboratories or New England BioLabs and were used according to directions. Previously prepared rabbit antiserum to wild-type DSD (20) was purified by sodium sulfate precipitation and DEAE-cellulose chromatography as described by Mage (21) prior to use in immunoblots.
Genetic Nomenclature, Bacterial Strains, and Plasmids-The inability to convert D-serine to pyruvate (DSD-phenotype) may result from alteration of one or more of three genetic loci known to be specific for DSD expression in E. coli K-12: (i) the structural gene of the enzyme, designated dsdA, (ii), the structural gene of a dsdA activator protein, dsdC, or (iii) the control region that lies between the two structural genes and contains the dsdA and dsdC promoters (dsdAp and dsdCp,) and transcription start sites (22)(23)(24). The symbol dsdA or dsdA+ denotes the wild-type structural gene encoding DSD. Mutations in this gene that result in an inability to grow in the presence of D-serine are designated either dsdA-or dsdA(EM1201), dsdA(AC6082), etc., to denote a specific dsdA mutation. Mutations in the regulatory loci are designated dsdC-, dsdAp-, or dsdCp-.
Phenotypically DSD-strains were obtained from wild-type parent strains using methyl-N-nitro-N-nitrosoguanidine mutagenesis, penicillin selection, and selection in D-serine-containing media (25,26). The locus of mutation was tentatively assigned as dsdC, dsdCp, dsdAp, or dsdA by complementation analysis and phage transduction using the tester strains and procedures described by Bloom and McFall (27). Six strains that appeared to be dsdA-were chosen for further screening by protein and Western blot analyses. The E. coli K-12 strains and plasmids used in these studies are described in Table I. All cultures were grown at 37 'C in LB media (19). Where appropriate, ampicillin or tetracycline was added at a concentration of 50 pg/ml or 25 pg/ml, respectively. Bacteriophage M13mp19 was purchased from Bethesda Research Laboratories.
Preparation of Cell Extracts for Western Blotting-DSD-strains and strain C600 were grown overnight in 5 ml of LB broth f D-serine (0.5 mg/ml). One ml of each overnight culture was centrifuged (14,000 X g for 5 min) and the cell pellet resuspended in 25 pl of SPE lysis buffer (20% sucrose, 20 mM EDTA, 30 mM potassium phosphate, 0.5 mg/ml lysozyme, pH 7.8) and incubated at room temperature for 15 min. Cells were lysed by gently mixing the suspension with 275 pl of water (containing lpg/ml each of the protease inhibitors leupeptin, pepstatin, and aprotinin) and incubating the mixture on ice for 15 min. An aliquot (40 pl) of Buffer A (1 M potassium phosphate, 40 p M PLP, 50 mM EDTA, 10 mM DTT, adjusted to pH 7.5 with concentrated HCl) was then added and the mixture centrifuged to remove cell debris (14,000 X g for 10 min at 4 "C). Aliquots of the supernatant were removed for assay of DSD activity and for gel electrophoresis in preparation for Western blotting.
Assays of DSD Activity-Enzyme activity assays monitored the increase in absorbance at 220 nm ( A E = 1090 M" cm") associated with the conversion of D-serine to pyruvate. The standard assay solution contained 990 p1 of 0.1 M potassium phosphate, pH 7.8, and 10 pl of crude protein extract (or 10 pl of a dilute solution of purified enzyme). Fifty pl of 0.5 M D-serine was added to start the reaction. Assays were performed at 25 "C in a EU700 GCA McPherson spectrophotometer. Screening of Cell Extracts by Western Blotting-Aliquots (2 pl) of each crude extract were electrophoresed on 9% SDS-miniPAGE and transferred to nitrocellulose (Millipore GSWP 304FO) by electroblotting using a Bio-Rad TransBlot apparatus. The subsequent steps represent a modified version of the Bio-Rad Immun-Blot protocol. The nitrocellulose was rinsed with H20, then shaken for 4 h in 0.5% Tween 20, 20 mM Tris, 0.5 M NaC1, pH 7.5 (Tw/TBS) to block nonspecific protein-binding sites. The filter was then incubated overnight with antiserum to wild-type DSD (75 p1 of rabbit antiserum in 75 ml of Tw/TBS). After four washes in Tw/TBS to remove excess antibody, the filter was incubated for 2 h in 75 ml of Tw/TBS containing 10 p1 of goat anti-rabbit horseradish peroxidase conjugate. The nitrocellulose was then washed four times with Tw/TBS and once with 0.1 M MES, pH 6.0. Following equilibration in MES, the nitrocellulose was developed to detect the presence of DSD crossreactive material with a solution prepared by combining (a) 60 mg of 4-C1-1-napthol dissolved in 30 ml of methanol (ice-cold) and (b) 60 pl of 30% H20, dissolved in 100 ml of 0.1 M MES, pH 6.0. Following development for 5-10 min, the blot was rinsed several times with water, dried at room temperature, and photographed. All steps were carried out at room temperature in glass dishes.
Construction and Screening of Genomic Libraries-Unless specified, the techniques outlined by Maniatis et al. (19) were employed for plasmid DNA isolation, electroelution of DNA fragments, and transformations. Chromosomal DNA was prepared (30) from strains EM1201 and AC6082 and digested with HindIII. Following electrophoresis on 0.8% agarose (FMC Biochemicals), the 4-5 kb region was excised, the DNA recovered by electroelution, and ligated to HindIIIcleaved, dephosphorylated pUC13. The ligation mixture was used to transform JM101. AmpR colonies were replicaplated (FMC Repliplate, FMC Biochemicals) onto Whatman No. 1 filter circles and screened for the presence of dsdA (31) with a 32P-labeled nicktranslated RsaIIAuaI dsdA fragment (0.586 kb) isolated from pAC13, a plasmid that carries the entire wild-type dsd operon (Table I and Ref. 23). Plasmid DNA prepared from clones that gave a positive hybridization signal was characterized by restriction analysis and in situ gel hybridization (32) using the RsaIIAvaI probe described earlier. I n situ gel hybridization, a simplified form of Southern blotting, was carried out as described by Kidd (32) with the following modifications. After wetting the dried agarose gels in water and removing the backing paper, the gels were rinsed for several minutes in 6 X SSC, pH 7.0 (1 X SSC is 0.15 M sodium chloride, 0.015 M sodium citrate buffer as defined in Ref. 19) then prehybridized for 1-2 h at 68 "C in 6 X SSC, pH 7.0, 5 X Denhardt's solution (19), 0.1% SDS, 0.5 mg/ml sodium pyrophosphate, 0.1 mg/ml sheared, denatured salmon sperm DNA. Gels were hybridized with the 32P-labeled probe for 2-4h at 68 "C in 6 X SSC, 1 X Denhardt's solution, 0.1% SDS, 0.1 mg/ml salmon sperm DNA before washing and exposure to x-ray film (32). These modifications markedly reduced 32P background due to nonspecific binding of probe to rehydrated agarose gels. HindIII library plasmids pEM1201.1 and pAC6082.35 were chosen for further characterization of the insert DNA by dideoxy sequencing and expression of the structural gene.
Subcloning of DNA for Dideoxy Sequencing-The sequencing strategy is outlined in Fig. 1. Briefly, a 1.5-kb RsaIIHincII fragment containing all of dsdA was isolated from pEM1201.1, pAC6082.35, and pAC13. Following digestion with SmaI or HaeIII and AluI, the DNA fragments were ligated into HincII-digested, dephosphorylated M13mp19 DNA and transfected into JMlOl (33). Single-stranded DNA was prepared from 70 colorless plaques, and dideoxy sequencing  was performed with the universal sequencing primer, modified T7 polymerase (Sequenase, U. S. Biochemical Corp.) and the sequencing protocol supplied with the enzyme. A portion of each gene was also sequenced after cloning the intact 1.5-kb dsdA fragment into HincIIdigested, dephosphorylated M13mp19. Following transformation of JM101, clones containing the insert in opposite orientations were isolated and used to prepare a series of overlapping deletion clones as described by Dale et al. (34). Single-stranded DNA from these clones was isolated and sequenced as described earlier. Sequence data obtained by both cloning strategies was used to assemble the complete sequence of each gene in both directions.
Subcloning of DNA into Expression Vector ptacl2-The 1.5-kb RsaI/HincII fragment containing either the wild-type or mutant structural gene was ligated into PuuII-digested, dephosphorylated ptacl2. Following transformation of E. coli JM101, plasmid DNA was prepared from ampR colonies and screened for the presence and orientation of the insert by restriction analysis. To verify that the recombinant plasmids could express DSD under the direction of the tac promoter, JMlOl containing either PTC-1, PTC-35, or wild-type expression plasmid PTC-21 was grown in 5 ml of LB,, to an absorbance at 660 nm (Am) of 0.4. Isopropyl @-D-thiogalactoside was added to a 1 mM final concentration, and cultures were allowed to grow for an additional 8 h before harvesting. Crude extracts were prepared as described earlier, and DSD activity assays and Western blotting were used to verify the expression of wild-type or mutant DSD.
Expression and Purification of DSD Proteins-DSD expression strains were grown for 5-6 h in 10 ml of LB.,, and used to inoculate flasks containing 1 liter of LB.,,.
When the Asso reached 0.4, 20 ml of 0.5 M lactose was added to each flask to induce expression of DSD. Cells were harvested after an additional 7-8 h of growth, and the cell paste was stored at -20 "C. A typical yield of cell paste from 15 liters of medium was 40 g which in turn yielded 300-400 mg of purified wild-type or mutant protein.
Purification steps were performed at 4 "C unless otherwise specified. Protein concentration was determined from the absorbance of holoenzyme at 280 nm in 0.1 M potassium phosphate buffer, pH 7.8, using E 2 = 10.5 (1). All buffers except the SPE lysis buffer (described earlier) contained 1 pg/ml each of the protease inhibitors pepstatin, leupeptin, and aprotinin. Cell paste (30-40 g) was resuspended in 40 ml of SPE lysis buffer. After incubation with occasional stirring at room temperature for 20 min, cells were lysed by addition of 400 ml of water containing the protease inhibitors and 80 p~ PLP. After stirring the suspension for 5-10 min, the mixture was incubated on ice for 20 min and 50 ml of Buffer A' (Buffer A containing 800 p~ PLP) added. Cell debris was removed by centrifugation for 30 min at 18,000 X g. The pH of the supernatant was adjusted to pH 7.3 with 2.5 M KOH and nucleic acids precipitated with 1% Polymin P as previously described (35). The resulting supernatant was adjusted to pH 7.3 and dry ammonium sulfate added slowly, with stirring, to 70% saturation (the pH was maintained constant at pH 7.3 with 2.5 M KOH). The solution was stirred for an additional 40 min, centrifuged at 18,000 X g for 2 h, and the resulting pellet was dissolved in a minimum volume (50-70 ml) of Buffer B (10 mM potassium phosphate, 1 mM EDTA, 1 mM DTT, 80 pM PLP, pH 7.2) and dialyzed overnight against three changes (1 liter each) of Buffer B in preparation for DEAE-cellulose chromatography.
After a brief centrifugation (18,000 X g for 5 min at 4 "C) to clarify the sample, it was loaded onto a 2.5 X 20-cm column of DEAEcellulose (Whatman DE52) previously equilibrated in Buffer B. Proteins were eluted from the column with a 400-ml gradient of 0 + 200 mM KC1 in Buffer B. Fractions (5 ml) were collected at a flow rate of approximately 70 ml/h, and 0.55 ml of Buffer A' was added to each. Fractions were screened for the presence of DSD by the activity assay in the case of wild-type DSD or by dot blotting in the case of the inactive DSD variants. For dot blotting, 5 pl of each column fraction was spotted onto a sheet of nitrocellulose and allowed to dry. The nitrocellulose was then incubated with DSD antiserum and developed to reveal the presence of cross-reactive material as described for Western blotting of crude extracts. Peak fractions were pooled, and the protein was precipitated with 70% ammonium sulfate. The pellet was dissolved in a minimum volume (10-15 ml) of Buffer B, dialyzed overnight against Buffer B to remove ammonium sulfate, then dialyzed an additional 3-4 h against Buffer C (1 mM potassium phosphate, 1 mM DTT, 80 ~L M PLP, pH 7.0) in preparation for chromatography on hydroxyapatite. No PLP was added to these dialysis buffers for the purification of wild-type DSD.
After a brief centrifugation to remove precipitated material, the sample was applied to a 2.5 X 19-cm column of hydroxyapatite (Bio-Rad HTP) previously equilibrated 1 mM potassium phosphate, 1 mM DTT, pH 7.0. The column was washed with 3 column volumes of Buffer C (mutant DSD proteins) or Buffer C minus PLP (wild-type DSD). DSD was eluted with 3 column volumes of 10 mM potassium phosphate, 1 mM DTT, 80 p~ PLP, pH 7.8 (mutant and wild-type DSD). One-tenth volume of Buffer A' was added to each fraction, and those containing DSD were identified, pooled, and precipitated with 70% ammonium sulfate as described for DEAE-cellulose chromatography. The pellet was resuspended in a minimum volume of Buffer D (100 mM potassium phosphate, 1 mM DTT, 1 mM EDTA, 80 pM PLP, pH 7.8), dialyzed to remove ammonium sulfate, and stored at -70 "C.
Determination of the PLP Content of DSD-DSD (225 PM) was dialyzed for 18 h against three changes of Buffer D containing 250 pM PLP (stoichiometric with enzyme concentration plus 10%). Free and protein-bound PLP were then separated by the centrifuge column procedure of Penefsky (37). Sephadex G-50 resin used for column matrices were pre-equilibrated in Buffer D without PLP. DSD samples were loaded onto the 1-ml syringe columns in a total volume of <lo0 pl and allowed to filter into the resin for 2-3 s prior to centrifugation at 5,000 X g for 4 min at room temperature. Following two successive column centrifugations to remove unbound PLP, the sample was transferred to a 1.5-ml Eppendorf tube and centrifuged briefly (3 min, 14,000 X g) to remove any traces of Sephadex. The ratio of apo/holoDSD in the supernatant was determined spectrophotometrically.
Absorbance spectra were recorded at 25 "C on a Cary 219 spectrophotometer. Enzyme solution from the previous step (50-100 rl) was added to 900 p1 of quenching buffer (6 M guanidine hydrochloride, 0.5 M 2-hydroxyethylamine, 0.1 M KPi, adjusted to pH 7.8 with HC1) and the spectrum recorded against a reference solution of 50-100 pl of buffer D (without PLP) in 900 p1 of quenching buffer. The high concentration of guanidine hydrochloride ensured that enzyme-bound PLP was released to form a Schiff base with 2-hydroxyethylamine. Concentrations of PLP ([PLP]) and protein ([DSD]) in the assay curvette were determined from the absorbance at 418 nm (A418) and 280 nm (AzM) using Equations 1 and 2: where the molar absorptivities of the PLP:2-hydroxyethylamine Schiff base (E% = 5225 M" cm", = 4426 M" cm") at the subscripted wavelengths were determined from the spectra of solu- The molar absorptivity of denatured apoDSD, EZD (39, (1) using a value of E;% = 10.5 for native wildtype DSD. The molar absorptivity for the denatured apoDSD variants (at 280 nm in quenching buffer) was assumed to be equivalent to that of denatured wild-type apoenzyme.

RESULTS
Several strains of E. coli that could not grow in the presence of minimal medium containing D-serine were produced by NMNG mutagenesis (25)(26)(27). Presumably these strains produced insufficient active DSD to destroy D-serine, a growth inhibitor of E. coli. (38). SDS-PAGE and Western blotting allowed identification of strains likely to be expressing full length copies of inactive or partially active DSD. The structural gene for wild-type DSD and the dsdA genes from the two mutant strains were isolated and further analyzed.
Previously prepared PAC 13 containing a 7.4-kb BstEII fragment of E. coli DNA was used as source of wild-type dsdA (Table I and Ref. 23). The dsdA gene from strains EM1201 and AC6082 was isolated from genomic libraries prepared from Hind111 digests of chromosomal DNA. The DNA frag-ments coding for the wild-type and mutant proteins were subcloned and sequenced as indicated in Fig. 1 and under "Materials and Methods." The complete nucleotide and amino acid sequence of wild-type DSD are presented in Fig. 2. The 442-residue wild-type enzyme as deduced from the DNA sequence has an M, of 47,920 and the composition:
The sequence of dsdA from EM1201 and AC6082 disclosed the presence of a single glycine to aspartic acid replacement in DSD at position 279 and position 281, respectively (boxed area of Fig. 2). Interestingly, G-279 and G-281 reside in a glycine-rich region of DSD wherein glycine comprises 6 of 8 residues (279-286). The predicted secondary structure of DSD (Fig. 3) suggests that the G + D replacements reside in an

l a ASP A s n L a u T h r A l a A l a
Asp G l y L a u la v a l G l y A r g la sar G l y Pha va1 G l y A r g * l a M a t G l u A r g Lau L a u ASD G l y  I I I I I J I I I~I I I I I~I I I I I I I I I~I I I I I I I I I~I I I I I I I I I~I I I I I   I I~I I I I I I I I I~I I I I I I I l I~I I I I I  extended turn between the carboxyl terminus of a &sheet and the amino terminus of an a-helix. Wild-type and mutant proteins were produced in transformants of JMlOl that contain the DSD structural gene linked to the powerful tac promoter of plasmid ptacl2 (Table  I). DSD accounts for 7-8% of the total bacterial protein in these expression strains, or about 10 times the amount of DSD produced by the single copy of genomic dsdA in strain EM1600 (dsdApd+), a commonly used source of DSD (1,42). The E. coli JMlOl host strain expresses little or no DSD unless induced with D-serine. In control experiments, isopropyl 8-D-thiogalactoside-induced cultures of JMlOl produced <0.1% of the DSD expressed by induced transformant strains. Enzyme was purified from crude bacterial extracts using a modified version of previously described procedures (1,35,42). Because of the possibility that the DSD variants might bind pyridoxal 5'-phosphate less tightly than wild-type DSD, they were purified in the presence of excess cofactor. Wildtype DSD, DSD(G279D) and DSD(G281D) were isolated in greater than 95% purity as judged by 9% SDS-PAGE (Fig. 4). The purification protocol yielded wild-type enzyme with a specific activity of 140-150 units/mg, in good agreement with that previously reported for DSD (4). Purified DSD(G279D) and DSD(G281D) displayed specific activities less than 0.3% that of wild-type DSD. The lack of activity of the variant enzymes could not be attributed to a failure to bind PLP. Determination of the cofactor content yielded values of 1.18,0.97 and 0.99 mol bound cofactor/mol protein for DSD, DSD(G279D), and DSD(G281D), respectively. Furthermore, the activity of the variant proteins and wild-type DSD was unaffected by varying the concentration of PLP in the assay medium from 1.5-to 100-fold molar excess over enzyme.
Five cycles of Edman degradation indicated the sequence MENAK, and digestion with carboxypeptidase B liberated only MNQYLAKGR for all three proteins. These observations support the conclusion that the wild-type and variant proteins were isolated with their polypeptide chains intact. Thus, the inactivity of the variant proteins could not be attributed to proteolysis. DISCUSSION The DNA sequence of dsdA (Fig. 2) agrees with that previously reported (18) for nucleotides 1-585 with the exception of a nucleotide in the codon for A-192. GCT was observed rather than GCC. The deduced amino acid sequence agrees with that determined by direct peptide sequence analysis (17)  (A-192) were also observed in the dsdA sequence of EM1201, AC6082, and EM1502, a dsdC+Apd+ constitutive strain.
It came as a surprise that the loss of catalytic activity seen in DSD(G279D) and DSD(G281D) was not due in either case to loss of a functional side chain, since a minimum of six functional groups have been implicated in catalysis or alignment of cofactor or substrate (1,2,10,11,14,42,44,45). Introduction of an aspartyl side chain at position 279 or 281 may have perturbed the orientation of one or more of these groups, although gross perturbations of protein conformation are unlikely. The G + D replacements reside in what is predicted (39,41) to be an extended @-turn in the most flexible region of the molecule (Fig. 3). Aspartic acid is frequently found in @-turns (46) and its presence at position 279 or 281 does not alter the predicted secondary structure (39) or flex-ibility (41) of the glycine-rich region.' Also consistent with the maintenance of DSD structural integrity, G + D variants were able to bind PLP stoichiometrically.
Interestingly, the glycine-rich region is highly conserved in several PLP-requiring c~/p-eliminases,~ despite an overall sequence homology of less than 28% between DSD and each of the enzymes listed in Table I1 (the most homologous proteins were E. coli biodegradative threonine dehydratase (27%) and tryptophan synthetase P-subunit (22%)). Alignment of the glycine-rich regions of several PLP enzymes using the criteria indicated in the legend to Table I1 disclosed several interesting similarities. All of the threonine dehydratases (Tdc, ILV-1, and ILV-A in Table 11) contained glycine-rich sequences that were 36-41% homologous to a 22-amino acid segment encompassing the glycine-rich region of DSD, with the homology increasing to 50% when the comparison is limited to the 8residue segment comprising G-279 to G-281. Moreover, in the aligned sequences of DSD, the threonine dehydratases, and tryptophan synthetase p-subunit, glycine was invariant at positions corresponding to (3-279, G-281, and G-286 of DSD. It is thus tempting to suggest a functional role for this conserved glycine-rich region. In this regard, we should note Conserved regions (including the glycine-rich region) in PLPrequiring alp-eliminases have been noted previously (17,49).

TABLE I11
Glycine-rich sequences in DSD and nucleotide-binding proteins
bAK, phospholipase A*, p21, F,-ATPase B, myosin, PK-CAMP, v-src, nitrogenase alignment is from Ref. 62. PK-C, EGFR, PK-CAMP, v-src alignment is from Ref. 73 (81). It is not clear whether the failure to observe activity is due to unproductive binding of NAD or a consequence of low cofactor binding under the assay conditions employed (no added NAD). e X = all amino acids but proline.

Inactive Variants of D-Serine
Dehydratase that glycine-rich sequences are found in many unrelated proteins including IgG light chain, keratin, collagen, ribosomal proteins, enolase, and heat shock proteins (62). Thus, the presence of glycine-rich sequences in distantly related PLP enzymes (63) does not in itself imply functional significance. The conservation of glycine at three or more homologous positions in PLP-requiring a/@-eliminases, however, together with the striking effect of glycine replacement in DSD at two of these conserved positions strongly suggests that the glycine-rich region plays a structural and/or functional role in this group of enzymes. In many of the nucleotide-binding proteins, a highly conserved glycine-rich sequence is an important structural determinant of the nucleotide-binding site, as indicated by x-ray diffraction analyses (64-70) and the effect of G to X replacements on activity (Table 111). The flexible glycine-rich loop connects the carboxyl edge of a 0-sheet with the NHz-terminal end of an a-helix and is usually juxtaposed with a phosphoryl group of the bound nucleotide cofactor or substrate (65, 67-72). Although insufficient information is available to generalize about the presence of glycine-rich regions at PLP-binding sites, a precedent for PLP attachment to a nucleotide-like binding fold clearly exists in the case of glycogen phosphorylase. The PLP-binding site of glycogen phosphorylase is well removed from any subunit interface (82) in contrast to that for aspartate aminotransferase (see below). Crystallographic studies of phosphorylase (82)(83)(84) show that residues 133-139 in the nucleotide-binding domain form a glycine-rich loop that separates the phosphates of enzyme-bound PLP and glucose-1-phosphate. PLP is intimately associated with this loop, G-134 being in van der Waals contact with C6 and N1 of PLP and G-135 interacting with C5' (83). Additionally, G-135(NH) is linked via two hydrogen-bonded water molecules to 0 2 2 of the highly solvated phosphate group of PLP (84). The glycogen phosphorylase sequence from positions 130 to 137 (GLGNGGLG) is similar to the DSD sequence between residues 279 and 286 (GVGGGPGG). If the glycine-rich region of DSD were similarly juxtaposed with the bound PLP, G to D replacement at position 279 or 281 might inactivate DSD by sterically or electrostatically altering the orientation of the enzyme-bound cofactor.
The x-ray crystal structures of cytoplasmic and mitochondrial aspartate aminotransferases (85)(86)(87)(88)(89) indicate that these PLP enzymes lack glycine-rich sequences at their active centers, illustrating that a glycine-rich loop is not prerequisite for formation of a PLP-binding site. In this regard, it is important to note that most PLP enzymes are multimeric. The two PLP-binding sites of the isologous aspartate aminotransferase dimer reside at the subunit interface, where residues from both subunits contribute to cofactor binding (85)(86)(87)(88)(89). Such a structural arrangement is necessarily different from the PLP-binding site of DSD, a monomer, and may also contrast with the PLP-binding sites of multimeric PLP enzymes in which the bound cofactor is well removed from any subunit interface.
In summary, we have identified two point mutants of dsdA from E. coli that encode virtually inactive enzymes. The mutations do not alter residues with functional side chains capable of participating in catalysis, but instead affect two conserved glycyl residues that lie in an unusually glycine-rich region of the protein. This glycine-rich sequence, like the conserved glycine-rich sequences in nucleotide-binding proteins and glycogen phosphorylase, may constitute part of the cofactor binding site of DSD.