The Glycine-rich Region of Escherichia coli D-Serine Dehydratase ALTERED INTERACTIONS WITH PYRIDOXAL 5”PHOSPHATE PRODUCED BY SUBSTITUTION OF ASPARTIC ACID FOR GLYCINE*

Replacement of glycine by aspartic acid at either of two sites in a conserved, glycine-rich region inacti- vates the pyridoxal 5’-phosphate-dependent enzyme D-serine dehydratase (DSD) from Escherichia coli. To investigate why aspartic acid at position 279 or 281 causes a loss of activity, we measured the affinity of the G+D variants for pyridoxal 5’-phosphate and a cofactor:substrate analog complex and compared the UV, CD, and fluorescence properties of wild-type D- serine dehydratase and the inactive variants.

UV, CD, and fluorescence properties of wild-type Dserine dehydratase and the inactive variants.
The two G+D variants DSD(G279D) and DSD(G281D) displayed marked differences from wildtype D-serine dehydratase and from each other with respect to their affinity for pyridoxal 5'-phosphate and for a pyridoxal 5'-phosphate:glycine Schiff base. Compared to the wild-type enzyme, the cofactor affinity of DSD(G279D) and DSD(G281D) was decreased 225and 50-fold, respectively, and the ability to retain a cofactor:glycine complex was decreased 765and 1970-fold. The spectral properties of the inactive variants suggest that they form a Schiff base linkage with pyridoxal 5'-phosphate but do not hold the cofactor in a catalytically competent orientation. Moreover, the amount of cofactor aldamine in equilibrium with cofactor Schiff base is increased in DSD(G279D) and DSD(G281D) relative to that in wild-type DSD. Collectively, our findings indicate that introduction of a carboxymethyl side chain at G-279 or G-281 directly or indirectly disrupts catalytically essential proteincofactor and protein-substrate interactions and thereby prevents processing of the enzyme bound cofactor:substrate complex. The conserved glycine-rich region is thus either an integral part of the D-Serine dehydratase active site or conformationally linked to it.
The pyridoxal 5"phosphate (PLP)' dependent enzyme D-* This work was supported by Grant HL32006 from the National Institutes of Health. Part of this work is taken from a Ph.D. thesis to be submitted by Michelle Marceau to the Graduate School of The University of Michigan The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
5 To whom correspondence should be addressed.
The abbreviations used are: PLP, pyridoxal 5'-phosphate; DSD, D-serine dehydratase; DTT, dithiothreitol; EA, 2-hydroxyethylamine; GLY, glycine; DSD:PLP, PLP:EA, PLP:GLY, PLP:DTT denote the Schiff base formed between PLP and K-118 of DSD, EA, GLY, and DTT, respectively; DSD-PLP:GLY, the noncovalent complex between DSD and PLP:GLY. The single letter abbreviations for amino acids appear in bold type. The terms "aldimine" and "aldamine" denote PLP derivatives wherein the C4' carbon atom of PLP is sp2 and sp3 hybridized, respectively. serine dehydratase from Escherichia coli loses the ability to convert D-serine (a growth inhibitor) to pyruvate and ammonia upon substitution of aspartic acid for glycine at either position 279 or 281. G-279 and G-281 lie in an unusually glycine-rich region that is highly conserved in several PLP enzymes that catalyze a,B-eliminations and well-removed from the active site lysyl residue (K-118) in the primary structure of DSD (1). These observations, together with x-ray crystallographic studies that indicate the presence of glycinerich sequences at the PLP-binding site of glycogen phosphorylase (2)(3)(4) and the phosphoryl-binding sites of many nucleotide-binding proteins (5)(6)(7)(8)(9)(10)(11)(12)(13), prompted us to suggest that the glycine-rich region of DSD may constitute part of the PLPbinding site ( 1).
The earliest chemical transformation in the catalytic pathway for DSD is a transimination reaction wherein D-Serine displaces the t-amino group of K-118 from the cofactor Schiff base linkage. Glycine and certain other amino acids behave as inhibitory substrate analogs by transiminating the Schiff base linkage between PLP and K-118 to form a PLP:amino acid Schiff base that is not further processed by DSD. Studies of the interactions between substrate analogs and DSD have thus provided useful information about the initial steps in DSD catalysis (14). Transimination requires proper orientation of the Schiff base linkage between PLP and K-118 for nucleophilic attack by an amino acid substrate or substrate analog and reorientation of the DSD-PLP-amino acid gem diamine intermediate to ensure elimination of K-118 rather than the substrate (14, 15). Moreover, the product of transimination (the PLP-substrate Schiff base) must remain bound to the active site to permit further processing of the substrate. To gain further information regarding the manner in which the G + D replacements prevent catalysis, we compared wildtype DSD, DSD(G279D) and DSD(G281D) with respect to their interactions with cofactor and their ability to form an enzyme-bound Schiff base between PLP and glycine. Our findings indicate that the presence of an aspartyl residue at position 279 or 281 alters covalent and noncovalent interactions between cofactor and protein in a manner that impairs PLP binding and prevents productive interaction with amino acid substrates.

Chemicals
DTT was ordered from Behring Diagnostics. 2-Hydroxyethylamine and guanidine hydrochloride were purchased from Aldrich. D-Serine and glycine were ordered from Sigma. All other reagents were analytical grade. Buffers for UV, CD, and fluorescence measurements were filtered through 0.22-pm Millex-GS filter units (Millipore Corporation, Bedford, MA) prior to use.

General Considerations
Purification of DSD and DSD variants, assays of enzyme activity, separation of free and enzyme-bound PLP by the spun column method of Penefsky (16), and determination of the total protein and PLP content of enzyme samples in 6 M guanidine hydrochloride, 0.5 M EA, 0.1 M KH2P04, pH 7.8 (quenching buffer), have been described (1). Spehadex G-50 (Pharmacia LKB Biotechnology Inc.) used in the preparation of 1-ml centrifuge columns was pre-equilibrated in Buffer A (0.1 M KH2P04, 10 mM EDTA, 5 mM DTT, pH 7.8). Absorption spectra (520-270 nm) were recorded on a Cary 219 spectrophotometer at 25 "C.
Determination of Kp and Kpc, the Equilibrium Constants for Dissociation of PLP and PLP:GLY from DSD Preparation of Incubation Mixtures-Since DSD binds PLP very tightly, direct evaluation of the equilibrium constant (Kp) for dissociation of PLP from DSD is difficult. Kp was thus determined by perturbing the DSD:PLP equilibrium with EA, an amine that competes with DSD for cofactor by forming a Schiff base with PLP in solution. Samples of wild-type DSD, DSD(G279D) and DSD(G281D) (100-600 p~ in Buffer A) were first characterized with respect to PLP content ([PLP],) and enzyme concentrations ([Et]) as previously described (1). Incubation mixtures for Kp determinations were prepared by adding 11 pl of a solution of freshly prepared 0.1-0.5 M EA in Buffer A, pH 7.8, to 100 pl of DSD(G279D) or DSD(G281D), or by adding 11-25 pl of 5 M EA in Buffer A, pH 7.8, to 100 pl of wildtype DSD. Incubation mixtures for Kpc determinations were prepared by adding 0.1-0.5 M glycine (as a solution of 2 M glycine in Buffer A, pH 7.8) to 100 p1 of enzyme along with sufficient EA to partially resolve the enzyme (empirically determined as 1 M EA in Buffer A for wild-type DSD and 0.01-0.05 M EA for DSD variants).
Separation of Free and Enzyme-bound PLP in Equilibrium Mixtures-Following incubation at room temperature for 2-3 h (empirically determined as the time required to reach a constant value of [PLP]b/Et, as described under Calculations of Kp and Kpc, free and enzyme-bound PLP were determined by the spun column method (16) or by ultrafiltration as follows. An aliquot of each incubation mixture (75-100 pl) was loaded onto a 1-ml centrifuge column and excess PLP, EA, and glycine (when present in the incubation) were removed by centrifugation for 4 min at 4,000 X g. The spectrum of the column eluent was recorded from 520-270 nm in quenching buffer, and Kp or K~G was calculated as described below. For determination of free and enzyme-bound PLP by ultrafiltration, 500 pl of each equilibrium mixture was loaded on top of a dry Centricon-10 microconcentrator (Amicon) and centrifuged at 5,000 X g at room temperature for 5 min, or just until 50 p1 of filtrate could be collected. The concentration of PLP in the filtrate and retentate and the total enzyme concentration in the retentate ([Et]) were used to determine Kp or Kpc as described under Calculations of Kp and KPG. Equilibrium dialysis of incubation mixtures (prepared as described earlier) was also used to evaluate the concentration of free and enzyme-bound PLP or enzyme-bound PLP:GLY. Dialysis was performed at room temperature for 18 h (tth for equilibration across the dialysis membrane was approximately 1.5 h) using a plexiglass chamber equipped with 0.3-ml microdialysis cells (TechniLabs, Piscatawney, NJ). Kp or KPG was determined as described in the next section.  (14)). These PLP solution complexes may all theoretically bind to the active site. PLP that is not bound to DSD at equilibrium (either specifically or nonspecifically) is the sum of free, uncomplexed PLP ([PLPIr) and the various PLP solution complexes: where X, represents the concentration of the PLP-binding species in the incubation mixture and K,i represents the equilibrium constant for dissociation of the PLP complex in solution (PLP:Xi + PLP + X,). In our experiments: where [GI, [Dl, and [E] represent the concentrations of glycine, DTT and EA, respectively. The values of KZi were determined spectrophotometrically by previously described methods (17). For analysis of the resolution of G D variants, KG (7.0 mM) and KE (2.4 mM) were determined in 0.1 M K2HP04, pH 7.8, at room temperature. For analysis of the resolution of wild-type DSD, KG (6.0 mM) and KE (3.0 mM) were determined in 0.1 M K2HP04, 1.0 M KCl, pH 7.8. One M KC1 was added to the buffer in the case of the wild type to approximate the higher ionic strength (i.e. higher concentration of 2-hydroxyethylammonium ion) of solutions used to resolve wild-type DSD. A value of 12.7 mM for KO (determined in 0.1 M KzHPO,, pH 7.8) was used in all analyses? The observed value of r ([PLP]b/Et) was corrected for nonspecific binding of PLP to e-lysyl groups other than K-118 by assuming all 19 nonactive site lysyl residues of DSD were available for Schiff base formation (although this may not be the actual case, the assumption allows an upper limit determination of Kp). PLP specifically bound to the active site (r' X [E],) is equivalent to the experimentally determined value of total enzyme-bound PLP (r X [Elt) minus the concentration of PLP bound to e-lysyl groups other than K-118 where KNs is the average equilibrium constant for dissociation of a PLP Schiff base of a nonactive site DSD lysyl residue. A value of 4.4 mM was determined for KNS of the model Schiff base formed between PLP and c-aminocaproic acid (in 0.1 M K2HP04, pH 7.8) and used in equation 4 (17). 3 If the complexes that PLP forms with DTT, EA, and glycine can bind to the active site of DSD (in addition to uncomplexed PLP), then the total concentration of PLP bound to the active site represents the sum of these species:  (7) where (1r') is the fraction of [E], present as apoenzyme.
In experiments where glycine was absent, we observed that the first term in Equation 7 was independent of both EA and DTT concentration (Table I). This finding strongly suggests that under our experimental conditions, the solution complexes PLP:DTT and PLP:EA do not bind to DSD (ie., the second term of Equation 7 is negligible). Equation 7 was thus simplified for determinations of Kp (in the absence of glycine): The values of KP listed in Table I   . Spectra were also corrected for the contribution to the absorbance of free PLP (resulting from dissociation in the assay cuvet) and PLP nonspecifically bound to lysyl residues other than K-118. The millimolar absorptivity of active site bound PLP (6. ) was calculated from the observed absorbance (A) using the relationship where the millimolar absorptivities of free PLP (CPLP) and nonspecifically bound PLP (ENS) were obtained from spectra PLP and PLP complexed to t-aminocaproic acid. The fraction of PLP present as free PLP (fpLp), nonspecifically bound PLP (fNS), or active site bound PLP (fa) was calculated by iterative solution of Equation 10 for [PLPIf and substitution of the solution in Equations 11 and 12 along with the average values of KP listed in Table I The derivation of Equation 10 assumed that (i) equilibrium was established between cofactor and protein in the cuvet, (ii) the 19 nonactive site lysyl residues of DSD were equally accessible to PLP, and (iii) the small effect of DTT on [PLP]b and the contribution of PLPDTT to the spectra could be neglected at the low concentrations of free PLP and DTT in the cuvet.
The UV spectra of PLP:t-aminocaproic acid and PLP:EA were determined for solutions containing PLP in 0.5 M amine at pH 7.8 in the same buffer used for spectral analysis of protein samples. Schiff base formation was essentially complete under these conditions. Fluorescence Spectra Fluorescence emission and excitation spectra were recorded at 25 "C on a spectrofluorimeter built by Dr. David Ballou and Gordon Ford of the Department of Biological Chemistry. Emission spectra (400-650 nm) were obtained using an excitation wavelength of 400 nm. Excitation spectra (250-500 nm) were obtained using 490 nm as the emission wavelength. DSD samples consisted of 100 pl of a 48-p~ solution of wild-type DSD, DSD(G279D) or DSD(G281D) in 0.1 M KH2P04, 80 p M PLP, pH 7.8, diluted into 900 pl of the same buffer (final enzyme concentration 4.8 p~) .
Excess PLP was included in the buffer to minimize cofactor loss from DSD variants. DSD samples for fluorescence spectra in the presence of 0.5 M glycine consisted of 50 pl of enzyme solution (final enzyme concentration 2.4 p~) , 450 pl of buffer, and 500 pl of 1.0 M glycine (prepared in buffer and adjusted to pH 7.8 with potassium phosphate). The contribution of the sample buffer to fluorescence was determined from the spectra of 80 p~ PLP or 80 p M PLP, 0.5 M glycine in 0.1 M KHzPO,, pH 7.8.
The equilibrium constant (Kw') for dissociation of glycine from DSD was determined from the enhancement of DSD fluorescence (400 nm excitation, 470 nm emission) by glycine. The dependence of DSD fluorescence on glycine concentration was analyzed in terms of Equation 13.
where Fo, F, and FM are the measured values of fluorescence in the absence of glycine, a concentration of glycine [GLY], and at saturating glycine, respectively. The value of Kpc' was estimated from the ratio of the slope to the intercept of a linear plot (r = 1.00) of 1/(F -

CD Spectra
Wild-type DSD, DSD(G279D), and DSD(G281D) were dialyzed for 18 h at 4 "C against two changes (1 liter each) of 20 mM KH2P04, 150 mM KCl, 1 mM EDTA, 1 mM DTT, 250 p~ PLP, pH 7.8, and then for 3.5 h against buffer lacking EDTA and DTT. After removing an aliquot of the sample for determination of ([E,]) and [PLPIb, the DSD concentration was adjusted to 75 p M for 500-300 nm spectra or to 5.0 ~L M for 305-205 nm spectra. CD spectra were recorded on a Jasco model 40C spectropolarimeter at 17 "C using a cuvette of path length, 1 (1 cm for 500-300 nm or 0.2 cm for 305-205 nm spectra). The observed rotation in degrees (&,bs) was converted either to molecular circular dichroism (500-300 nm spectra) or to millimolar ellipticity (305-205 nm spectra). Molecular circular dichroism ( B E ) in M" cm" was calculated from the expression A E = fI0~/33c1, where c is the molar concentration of protein (75 X M). Millimolar ellipticity in degrees-cm2/dmmol ([e]) was calculated from the expression [SI = &t&OcMl, where CM is the molar concentration of peptide bonds (5 X 1O"j X 441).

KP and Kpc
Determinations-Equilibrium constants for dissociation of PLP ( K P ) and PLP:GLY (Kpc) from wild-type DSD, DSD(G279D), and DSD(G281D) are presented in Tables I and 11. The values of Kp and Kpc were determined by perturbing the equilibrium between DSD and PLP with EA. This amine forms the Schiff base PLP:EA which does not bind to DSD. In contrast, PLP:GLY binds to the enzyme. These conclusions follow from the observation that the experimentally measured first term of Equation 7 (as explained under "Materials and Methods") was independent of the concentration of EA but dependent on the glycine concentration.
The values of Kp and Kpo determined by the centrifuge column procedure, ultrafiltration, and equilibrium dialysis corresponded within experimental error. The spun column method proved to be the most convenient and was used to determine most of the values listed in Tables I and 11. The close agreement of values obtained by the three different methods suggests that centrifugation of incubation mixtures through the spun columns did not perturb the equilibrium established during incubation. The value of the equilibrium constant for dissociation of PLP from wild-type DSD in Table I (120 nM) is higher than the values of 7-35 nM reported in earlier studies (18,19). The increased ionic strength and medium effects from the buffer used to resolve wild-type DSD (due to the presence of 0.5-1.0 M 2-hydroxyethylammonium chloride) may be responsible in part for the increased dissociation constants. Examination of the data in Table I reveals that the inactive variants DSD(G279D) and DSD(G281D) bind PLP 225-and 50-fold less tightly than does wild-type DSD. It is important to note that the binding of PLP to the weaker binding variant is still -160-fold tighter than the binding of PLP to free lysine or taminocaproic acid in solution (17), suggesting that the variants retain some of the features of a functional PLP-binding site. Comparison of the values of KP and K , (Table 11) indicate that the affinity of wild-type DSD for PLP is reduced only 2.8-fold when the active site bound cofactor exists in a Schiff base linkage with glycine despite the fact that the covalent interaction between K-118 and PLP has been broken. In contrast, DSD(G279D) and DSD(G281D) disfavor the binding of PLP:GLY relative to free cofactor by 9.6-and 112-fold, respectively, indicating that the G + D replacements diminish the ability of DSD to retain a transaldimination complex even more than they lower the affinity of the enzyme for uncomplexed PLP. Relative to wild-type enzyme, the affinity of both G + D variants for PLP:GLY was decreased markedly, with the ratio KPG (variant):Kpc(wild type) 765 for DSD(G279D) and 1970 for DSD(G281D).

TABLE I1 PLP:GLY dissociation constants for DSD and DSD variants
The values of K~D were obtained from the concentrations of free and enzyme-bound PLP measured subsequent to perturbing the DSD:PLP equilibrium with EA and glycine, as described under "Materials and Methods." All measurements were performed at room temperature in 0.1 M KHZPO4, 5 mM DTT, pH 7.8. The lower case letters in the enzyme column refer to the experimental method (  1, curue C) reflects the predominant species in solution at neutral pH, a zwitterion (3'OH unprotonated,N1 protonated) with C4' in the aldehyde form (20, 21). About 20% of the cofactor exists in solution with the formyl group hydrated, an equilibrium reflected in the 330 nm shoulder of the spectrum (20). The absorption maximum of PLP shifts to -415 nm upon Schiff base formation with t-aminocaproic acid (Fig. 1,  curue A ) , EA (Fig. 1, curue B ) , or wild-type DSD (Fig. 1, curue D,and Refs. 22 and 23). The Schiff base formed between taminocaproic acid and PLP was included as a model for the Schiff base linkage between PLP and K-118 of DSD, since eaminocaproic acid resembles lysine in carbon chain length and pK, of the free amino group (17). The small shoulder in the 330 nm region of the PLP:EA spectrum (Fig. 1, curue B ) may represent a small amount of cyclic aldamine resulting from addition of the 2-hydroxyl group of EA to the imine bond.
The pronounced 415 nm absorbance maximum resulting from the aldimine linkage between PLP and DSD (Fig. 1, curue D and Fig. 2, curue A ) is typical of PLP-dependent holoenzymes (22,(24)(25)(26)(27)(28)(29). The small shoulder around 325 nm probably represents one or more species of enzyme bound PLP that contain sp3 hybridization at C4' rather than an sp2 aldimine linkage (see "Discussion"). The absorption spectra of DSD(G279D) and DSD(G281D) (Fig. 2, curves B and C, respectively), display considerably less Schiff base absorbance (A415) than the wild-type enzyme as well as a pronounced increase in 325 nm absorbance characteristic of sp3 hybridized PLP complexes. As noted earlier, the observed spectra (inset, Fig. 2) have been normalized for PLP concentration by plotting millimolar absorptivity (the ratio of observed absorbance to total PLP concentration) and corrected for the small amount of PLP dissociation that may have occurred in the assay cuvette, thus the decrease in 415 nm absorbance (main panel, Fig. 2) cannot reflect a simple loss of cofactor. Moreover, the spectra of DSD(G279D) and DSD(G281D) differ markedly from that of free PLP (compare Fig. 2, curves B and C with Fig. 1, curve C).
Fluorescence Spectra of DSD and DSD Variants-The fluorescence spectra of the three enzymes in the absence and presence of 0.5 M glycine are given in Fig. 3. The wild-type enzyme is weakly fluorescent with an excitation maximum at 425 nm corresponding to the absorbance maximum of the internal Schiff base and an emission maximum around 500 nm (14, 30). This intrinsic fluorescence increases markedly and shifts to slightly lower wavelengths upon transfer of the aldimine linkage to an incoming amino acid, a phenomenon attributed to a significant conformational change at the active site upon transaldimination (14). In contrast, DSD(G279D) displayed -70% less intrinsic fluorescence than wild-type DSD, and DSD (G281D)

KP
The equilibrium relationships defining this cycle require that This value of KPO' differs from the value of -6 mM determined by Federiuk et al. (14) in studies of the inhibition of DSD by glycine. The discrepancy between the two values may reflect differences in the ionic strength used for the two determinations. The experimentally determined values of KPC'/KC and KPC/ K p were 2 (i.e. 12 mM/6 mM) and 2.8 (i.e. 0.34 pM/0.12 pM), respectively. The discrepancy between these ratios may be due to experimental error' and/or medium effects from the high concentration of 2-hydroxyethylammonium ion used in the determination of K x / K P . The  CD Spectra-The CD spectra of wild-type DSD, DSD(G279D) and DSD(G281D) did not differ significantly in the far UV region (Fig. 5), consistent with our earlier suggestion that the inactivity of G + D variants is not due to gross changes in secondary structure (1). We note that amide ellipticity may be insensitive to subtle changes in secondary structure or small regional conformational changes, however (31, 32). In the visible region (Fig. 6), wild-type DSD shows the strong positive CD maximum characteristic of many PLP dependent enzymes (24,25,(27)(28)(29), with the maximum rotation at 415 nm corresponding to the absorption maximum of aldimine-bound PLP. In contrast, DSD(G279D) and DSD(G281D) show almost no measurable CD at 415 nm.
CD spectra were recorded under conditions of excess PLP (3.3 to 50 X molar excess for the visible and far UV spectra, respectively) to minimize cofactor loss from G D variants.
Excess PLP in the buffer should not have affected the CD measurements, however, since free PLP is symmetrical and thus optically inactive. Additionally, in control experiments   where the concentration of excess PLP in a sample of 100 p M wild-type DSD was increased from 1 to 250 p~ so as to increase the concentration of nonspecifically bound PLP, the CD spectrum was not affected. Balk et al. (24) have similarly found that nonspecific binding of PLP to 6-lysyl groups outside the active center did not contribute to the visible CD spectrum of tryptophan synthetase.

DISCUSSION
Our findings suggest that the presence of a carboxymethyl side chain at position 279 or 281 alters both covalent and noncovalent interactions required for proper function of the DSD active site. The nature of the covalent linkage between PLP and DSD may be inferred from the UV spectra (Figs. 1 and 2) and an extensive body of literature regarding the spectral properties of PLP and PLP derivatives (17,18,20,21,28,(34)(35)(36)(37)(38)(39).
Absorbance maxima around 330 nm have frequently been correlated with sp3 hybridization at C4' of PLP. Buell and Hansen (36) and Tobias and Kallen (34) have shown, for example, that geminal diamines and aldamines formed upon reaction of PLP with aminothiols in aqueous medium at neutral pH have absorbance maxima around 335 nm. Derivatives of enzyme-bound PLP with sp3 hybridization at C4' display similar absorption characteristics. For example, treatment of DSD or other PLP enzymes with sodium borohydride reduces the sp2-hybridized Schiff base of a secondary amine, with a simultaneous shift in the visible spectrum from -420 to -330 nm and loss of catalytic competence (22,29,40,41). An inactive derivative of cytosolic aspartate aminotransferase absorbing at 340 nm has been attributed to an aldamine derivative of PLP linked to group "X" of the protein (26).
And PLP bound at the active site of tryptophanase normally displays both 330 and 410 nm absorbance maxima, the former being attributed to an internal aldamine (28).
The 325 nm absorbance observed in DSD(G279D) and DSD(G281D), and to a lesser extent in the wild-type enzyme (Fig. 2), may represent a carbinolamine formed by the addition of water to the internal Schiff base, an aldamine wherein C4' of the cofactor is bonded to a nucleophilic group X on the protein, or a mixture of these species (Scheme I). The small 325 nm shoulder in the UV spectrum of wild-type DSD may or may not represent the same sp3 species or combination of species as the one(s) present in DSD(G279D) and DSD(G281D). The possibilities for group X on the protein include the e-amino group of lysine, a ring carbon of histidine, the " O H of serine or threonine, the "SH of cysteine, or the carboxyl group of glutamate or aspartate. Although the existence of PLP hydrates in aqueous solution is well established (20, 21), there is a lack of consensus regarding the amount of carbinolamine in equilibrium with PLP Schiff bases, Both NMR and structure-stability correlations have argued for an and against the presence of substantial equilibrium levels of carbinolamines (34,39,(42)(43)(44). The observation that the PLP:t-aminocaproic acid complex shows no shoulder in the 330 nm region (Fig. 1) suggests that little carbinolamine is present in aqueous solutions of the model Schiff base. A carbinolamine might be selectively stabilized at the active site of the enzyme, however. Thus, we cannot exclude the possibility that all or part of the increased 330 nm absorbance in the UV spectra of the G + D variants reflects a more favorable equilibrium constant for the addition of water to the azomethine bond. Since the enzyme may similarly perturb the equilibrium constant for addition of any nucleophilic amino acid side chain to the DSD:PLP imine bond, identification of the enzyme-linked nucleophile X from the propensity of X to add to model Schiff bases is difficult.
It is tempting to speculate that X is a thiolate anion, however. The rapid, reversible reaction of thiol groups with model or enzyme-bound Schiff bases has been well demonstrated (17,18,35,36,38), and Buell and Hansen (36) have shown that a wide variety of PLP complexes with aminothiols display absorbance maxima around 330 nm at neutral pH. Moreover, Dowhan and Snell (22) have shown that the presence of PLP shields two of the three titratable thiol groups of DSD from reaction with 5,5'-dith1o-bis(2-nitrobenzoic acid) and substantially reduces the reactivity of the third.6 Incubation of apoDSD with 5,5'-dithio-bis(2-nitrobenzoic acid) leads to modification of all three thiol groups and inactivation due to an inability to recombine with cofactor (22). These observations suggest that one or more thiol groups may lie close to the DSD active site. We should note one further possibility for the 330 nm absorbance maximum, namely that insertion of the intact imine linkage into an exceptionally hydrophobic microenvironment might cause the Schiff base itself to absorb around 330 nm (45). Further studies will be required to distinguish between these possibilities.
Comparison of the UV spectra of DSD, DSD(G279D), and DSD(G281D) (Fig. 2) suggests that 20-60% of the PLP bound to the active site of the G 4 D variants exists in a Schiff base linkage. The observation that the CD of both variants at 415 nm was markedly less than 20-60% that of wild-type DSD suggests that the orientation of the DSD:PLP Schiff base is different in the G + D variants. The altered orientation of bound cofactor together with the reduced affinity of DSD(G279D) and DSD(G281D) for PLP suggest that noncovalent interactions between cofactor and protein have been disrupted in the inactive variants.
The sensitivity of the PLP-binding site of DSD to subtle changes in structure is well demonstrated by studies with a wide variety of PLP analogs (19,46,47). Even minor alteration of the pyridine ring substituents severely compromises the ability of the cofactor to bind or else to bind in a catalytically competent manner (19). For example, N-methyl-PLP (where 1-H + 1-CH3) and 3-0-methyl-PLP (where 3-OH + 3-OCH3) do not bind to DSD. Replacement of the 6-H by a methyl group reduces DSD affinity 370-fold and kcat 5-fold DSD has 5 cysteinyl residues and no disulfide bridges (22).

I,,,,, 325 nm
A,,.,,, 41 5 nm k , , , 325 nm SCHEME 1. Formation of a carbinolamine or internal aldamine at the DSD active site. (19). Substitution of the 2-methyl group by an ethyl group decreases DSD affinity 3-fold and bat 4-fold (19). The PLP analog containing an isopropyl group at C2 ( i e . 2', exhibited only a 7-fold decrease in affinity for DSD but was inactive (19). The inability of enzyme bound 2',2'-dimethyl PLP to support catalysis strongly suggests that the activity of DSD is extremely sensitive to the orientation of the bound cofactor.
The limited tolerance of the active site to structural perturbations of the cofactor strongly implies the reverse is also true, i.e. that small alterations in active site structure might severely compromise the ability of normal cofactor to bind or to bind in a catalytically competent fashion. D-279 or D-281 might perturb the active site directly through unfavorable electrostatic or steric interactions with PLP or indirectly by inducing conformational changes in the vicinity of the active site. Either type of perturbation could weaken noncovalent interactions between cofactor and protein necessary for constraining PLP in a catalytically competent orientation at the active site. The virtual inability of DSD(G281D) to retain PLP:GLY (Table 11) suggests that the presence of D at position 281, in particular, perturbs noncovalent interactions important for binding of the PLP:substrate complex. The high KpG value observed for DSD(G281D) indicates that glycine behaves like an alkylamine rather than an amino acid by effectively removing PLP from the enzyme (Table I1 and Ref. 48). Since the absence of an a-COO group is the only feature distinguishing nonbinding PLP:alkylamine Schiff bases from PLP:amino acid Schiff bases which are normally retained following transaldimination (14, 48), it is tempting to speculate that D-281 may interfere with the noncovalent interaction(s) responsible for anchoring the a-carboxyl group of DSD-PLP:GLY to the active site.
In summary, G + D replacement at either position 279 or 281 of the conserved glycine-rich region substantially decreases the affinity of DSD for PLP and a PLP-amino acid complex and markedly affects the spectral properties of the holoenzyme. The simplest explanation for these findings and the observed inactivity of DSD(G279D) and DSD(G281D) is that introduction of a carboxymethyl side chain at G-279 or G-281 directly or indirectly disrupts essential protein-cofactor interactions. X-ray diffraction analysis should allow us to determine the proximity of G-279 and G-281 to the PLPbinding site of wild-type DSD and identify the catalytically important interactions that are disrupted in the G + D variants.