A Spectral Probe Near the Subunit Catalytic Site of Glutamine Synthetase from Escherichia coli REDUCED

In order to label phosphate binding sites, unadenylylated glutamine synthetase from Escherichia coli has been pyridoxylated by reacting the enzyme with pyridoxal 5'-phosphate followed by reduction of the Schiff base with NaBH4. A complete loss in Mg2+-supported activity is associated with the incorporation of 3 eq of pyridoxal-P/subunit of the dodecamer. At this extent of modification, however, the pyridoxylated enzyme exhibits substantial Mn2+-supported activity (with increased Km values for ATP and ADP). The sites of pyridoxylation appear to have equal affinities for pyridoxal-P and to be at the enzyme surface, freely accessible to solvent. At least one of the three covalently bound pyridoxamine 5'-phosphate groups is near the subunit catalytic site and acts as a spectral probe for the interactions of the manganese.enzyme with substrates. A spectral perturbation of covalently attached pyridoxamine-P groups is caused also by specific divalent cations (Mn2+, Mg2+ or Ca2+) binding at the subunit catalytic site (but not while binding to the subunit high affinity, activating Me2+ site). In addition, the feedback inhibitors, AMP, CTP, L-tryptophan, L-alanine, and carbamyl phosphate, perturb protein-bound pyridoxamine-P groups. The spectral perturbations produced by substrate and inhibitor binding are pH-dependent and different in magnitude and maximum wavelength. Adenylylation sites are not major sites of pyridoxylation.

In order to label phosphate binding sites, unadenylylated glutamine synthetase from Escherichiu coli has been pyridoxylated by reacting the enzyme with pyridoxal V-phosphate followed by reduction of the Schiff base with NaBIt. A complete loss in Mg+-supported activity is associated with the incorporation of 3 eq of pyridoxal-P/subunit of the dodecamer. At this extent of modification, however, the pyridoxylated enzyme exhibits substantial MnB+-supported activity (with increased I&, values for ATP and ADP). The sites of pyridoxylation appear to have equal affinities for pyridoxal-P and to be at the enzyme surface, freely accessible to solvent. At least one of the three covalently bound pyridoxamine 5'-phosphate groups is near the subunit catalytic site and acts as a spectral probe for the interactions of the manganeseeenzyme with substrates. A spectral perturbation of covalently attached pyridoxamine-P groups is caused also by specific divalent cations @In*+, M8+, or Ca'+) binding at the subunit catalytic site (but not while binding to the subunit high affinity, activating Me"+ site). In addition, the feedback inhibitors, AMP, CTP, ctryptophan, Lalanine, and carbamyl phosphate, perturb proteinbound pyridoxamine-P groups. The spectral perturbations produced by substrate and inhibitor binding are pH-dependent and different in magnitude and maximum wavelength. Adenylylation sites are not major sites of pyridoxylation.
Glutsmine synthetase of Escherichia coli is a dodecamer of M, = 600,000, composed of apparently identical subunits (1). Mechanisms of regulating glutamine synthetase activity have been reviewed (2,3) and include feedback inhibition and enzymatically catalyzed covalent modification by adenylylation-deadenylylation cascade systems. Each subunit of glutamine synthetase has a catalytic site (4), an adenylylation site (5), two essential divalent cation sites (4), and separate binding sites for at least some of the feedback inhibitiors (6)(7)(8). It is known that the nucleotide substrate is chelated at the subunit catalytic site by Mn2+ (4,9).
Pyridoxal V-phosphate in combination with reduction by sodium borohydride (10) has been used to label catalytically important lysine residues at or near phosphate binding sites of enzymes that do not require pyridoxal-P as a cofactor (11)(12)(13)(14)(15)(16). The reactivity of specific lysine residues is related to * 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.
$ To whom correspondence should be addressed.
abnormally low pK, values (14,17). Pyridoxal &phosphate has been used also as a photosensitizer for oxidation of neighboring histidine residues (15,(18)(19)(20). Further, a selective modification of enzyme alIosteric sites by pyridoxylation with pyridoxal-P has been achieved with several regulatory enzymes (21)(22)(23)(24)(25). With aspartate transcarbamylase from E. coli, the N-•-phosphopyridoxyllysyl group has been used as a spectral and fluorescent probe for changes in protein environment (26) caused by ligand binding (23,27). Covalently attached pyridoxamin e 5'-phosphate groups have been used in this study to monitor interactions of E. coli glutamine synthetase with divalent cations, substrates, and feedback inhibitors.

EXPERIMENTAL. PROCEDURES
Materials-Unadenylylated glutamine synthetase was isolated and stored as described previously (28). The enzyme preparation had an average of 1 eq of covalently bound 5'-adenylyl groups per mol (600,COOg) of glutamine synthetase (GS,).' The glutamine synthetase adenylyltransferase was from a previous study (29). Pyridoxal 5'phosphate (A grade) was obtained from Calbiochem Corp. Pyridoxsmine 5'-phosphate/HCl, pyridoxal, IV-a-acetyl-L-lysine, L-methionine (SR)-sulfoximine, carbamyl phosphate, sodium borohydride, amino acids, and nucleotides were obtained from Sigma Chemical Co. Di-and trivalent metal ions were removed from amino acids prior to use (28). Solutions of carbamyl phosphate, nucleotides, and other effecters were prepared at the desired pH just before use. Standardized solutions (4) of 0.486 f 0.004 M MnC12, 1.00 f 0.05 M CaClg, and 0.60 f 0.01 M MgCL kindly were supplied by Dr. John B. I&t. Deionized, ultrafiltered water was obtained bv ~assine distilled water through a MilLQ2 reagent grade water syste-& of M&pore Corp.
Pyridoxylation of Glutcwnine Syntheiase-The en&me w& dialyzed overnight at 4% against 10 llwl TEA/acetate, 10 mu MgCl, 7.2 to 7.4 containing 20 mu imidazole/HCl, 100 mM KCl, and 1 mu MnClz or 10 to 50 mM MgClz. Control samples of glutamine synthetase that were treated as above without pyridoxal-P addition retained frill enzymatic activity. Gel filtration of the dialyzed pyridoxylated enzyme through a Sephadex G-25 column (0.9 x 55 cm) showed that more than 90% of the reduced pyridoxal-P was protein-bound. All procedures were performed in the absence of direct light. Covering glassware with aluminum foil was found to be adequate protection against lightstimulated cleavage (30) of (pyridoxamine-P) GS complexes. Storage was at 4°C. Divalent cation-free enzyme was prepared and checked for EDTA removal as described previously (31). These solutions were used soon after preparation.
Protein and Protein-bound Pyridoxamine-P Concentrations-Protein concentrations of unmodified glutamine synthetase were determined from spectrophotometric measurements at 290 or 280 nm (32). Biuret measurements (33) indicated that the pyridoxylated enzyme has an unchanged absorption at 280 nm. The amount of pyridoxamine-P covalently bound per mol of enzyme subunit was calculated from absorption measurements (pH 7) at 280 and 326 nm on the basis of a subunit molecular weight of 50,000 for glutamine synthetase (l), assuming a value of 10,000 M-km-' at 326 nm for the molar absorption coefficient of protein-bound pyridoxamine-P groups 04.
Characterization of Protein-bound Pyridonamine-P Groups-AL iquots (3 mg) of the unadenylylated and pyridoxylated enzyme preparations were hydrolyzed overnight in 1 ml of 6 N HCl in evacuated tubes. N-•-Pyridoxyllysine in acid hydrolysates of the modified enzyme was identified as a fluorescent spot after ascending chromatography on Whatman No. 3MM paper using the following solvents (34): n-butyl alcohokpyridine: acetic acidwater (30:20:6:24), RF = 0.24, isoproyl alcohol:pyridine:acetic acidwater (30:20:6:24), RF = 0.49. The chromatographic reference compound, N-e-pyridoxyl-L-lysine, was synthesized from N-a-acetyl-L-lysine by the procedure of Forrey et al. (34) except that the pyridoxylation reaction with pyridoxal was run in 90% methanol after dissolving the acetylated lysine in water. A measurement of the molar absorption coefficient for covalently bound pyridoxamine-P groups to glutamine synthetase was made from 280 and 326 run absorptions at pH 7.2 and phosphate analyses (35) on acid hydrolysates, using the hydrolyzed, unmodified enzyme as a blank in the phosphate determination.
Enzyme Assays--Mg2+-or Mn*+-supported biosynthetic activities were measured at 30°C in a coupled assay system as described previously (32) except that assay solutions (with an initial absorbance at 340 mn of 2 and sufficient glutamine synthetase to give AAah/ min = 0.06 to 0.30) were mixed just prior to sampling into a thermocuvette (Gilford, model 3017-T) attached to a Beckman DU equipped with a Gilford digital readout and an automatic data lister (Gilford, model 4608). Mg*+-supported (2) and Mn*+-supported y-glutamyl transfer activities at pH 7.57 were measured as described before (4). Kinetic parameters (K, and V,,,,) were obtained from double reciprocal plots of velocity versus substrate concentration, using appropriate precautions in assays to ensure that initial velocities were measured at all substrate levels. All such plots were linear at moderate substrate concentrations with unmodified and pyridoxylated enzymes.
Difference Spectral Measurements-All spectra were recorded usiug a Gary model 15 spectrophotometer equipped with the 0 to 0.1 and 0 to 1 A slide-wires. The temperature of the sample cell for most experiments was maintained at 25 f O.l'C by using a water-jacketed silica cell (1 ml) of l-cm light path. After establishing a base-line on the 0.1 A scale with the protein solution (2 to 2.5 mg/ml) in both the sample and reference cells, ligands were added to the enzyme solution in the sample compartment while the same dilutions (<5%) of the enzyme in the reference compartment were made with appropriate buffers, Ligand or buffer additions to enzyme solutions were made with Hamilton syringes. Kinetic measurements of absorbance changes were made as described previously (36).
Ultracentrifugation-A Beckman model E analytical ultracentrifuge, equipped with a rotor temperature and control unit, phase plate, schiieren, absorption optics, and photoelectric scanner with multiplexer was used. Sedimentation velocity experiments were at 20°C with a protein concentration of 4.4 mg/ml and a speed of 40,000 rpm using a two-place An-D rotor. Sedimentation coefficients were measured from schlieren photographs and appropriately corrected (37). A 12-mm cell with a Kel F double sector centerpiece and absorption optics at 326 or 280 mn were used to look for more slowly sedimenting species. For difference sedimentation measurements, two single sector cells (with one containing a lo positive wedge window) were used. Before ultracentrifugal analyses, enzyme samples were dialyzed against several changes of 20 mM imidazole/HCl, 100 mM KCl, 1.0 mM MnClz buffer at pH 7.25.

Pyridoxylation of Unadenylylated Glutamine
Synthetase-The absorption spectrum of a protein incubated with pyridoxal 5'-phosphate provides a monitor of Schiff base formation (U-15,38).
When glutamine synthetase reacts with pyridoxal-P, a decrease in the maximum absorbance at 390 nm (free pyridoxal-P) and the appearance of a new absorption peak at 425 nm (Schiff base) are observed. In preliminary experiments (not shown), a variety of conditions (pH 6 to 9 with different buffers and either Mp or Mn" present) were tested and optimal Schiff base formation was found to occur under the conditions given under "Experimental Procedures." Nucleotide substrates of glutamine synthetase inhibited Schiff base formation with pyridoxal-P ( Table I, below). The Schiff base formed with pyridoxal-P is reversible; dialysis or gel filtration regenerates the unmodified enzyme.
All experiments reported in this paper involve the relatively stable reduced pyridoxal 5'-phosphate * enzyme complexes. Lysyl residues of the enzyme are modified in the pyridoxylation reaction; N-•-pyridoxyllysine was identified in acid hydrolysates of the pyridoxylated protein. The amount of covalently attached pyridoxamine-P in pyridoxylated glutamine synthetase samples remained unchanged after storage in the dark for as long as 6 months. The enzyme derivative was completely stable also during spectrophotometric and activity studies. However, the protein-bound pyridoxamine-P groups were photolyzed within a few minutes during excitation at 330 nm in a Hitachi Perkin-Elmer MPF-2A spectrofluorometer, which made fluorescent measurements with this derivative impossible.
In sedimentation velocity studies, the pyridoxylated enzyme had SZO,~ = 20.2 S; no more slowly sedimenting species could be detected with absorption optics. In difference sedimentation measurements, the pyridoxylated glutamine synthetase sedimented 4% faster than did the unmodified enzyme (As = 0.8 S). Modification of glutamine synthetase by pyridoxylation therefore does not cause dissociation of the dodecamer or any significant change in shape. This value is only 14% lower than the value of 10,000 M-km-' reported for N-•-phosphopyridoxyl-L-lysine (34). Also, treatment of the pyridoxylated enzyme with EDTA and 4 M urea to dissociate and unfold the protein (39) caused only a 10% increase in the absorption coefficient at 326 nm, while producing the same protein difference spectrum as observed previously for glutamine synthetase in 6 M guanidine HCl versus native dodecamer (36). These observations indicate that the protein-bound pyridoxamine-P groups are exposed to solvent (i.e. pyridoxamine-P groups are not buried in hydrophobic regions of the enzyme subunit structure).
A spectrophotometric titration curve of the enzyme-bound pyridoxamine9 groups from pH 6.6 to 9.0 was slightly broader but otherwise similar to that reported for N-e-phosphopyridoxyllysine (34). The absorption maximum of protein-bound pyridoxamine-P groups in pyridoxylated glutamine synthetase shifts from 326 nm at pH 6.6 to 7.2 to 316 nm at pH 8.5. This effect of pH is due to titrations of the pyridinium ions (34,40, 41), which have an average pK of -6.4. Some heterogeneity in the environment of the protein-bound pyridoxamine-P groups is indicated by the titration data. Fig. 2 shows the extent of pyridoxylation of unadenylylated glutamine synthetase as a function of the pyridoxal B'-phosphate concentration used in the modification reaction. A Scatchard plot of these data is given in the inset of Fig. 2. The intercept and linearity of this plot indicate that pyridoxal5'phosphate reacts with 3.2 subunit sites having about the same apparent affinity for pyridoxal-P (Kk = 700 M-l). Fig. 3 shows that there is a linear loss in the Me-supported biosynthetic activity with increasing pyridoxylation of the enzyme. A complete loss of this activity is associated with the incorporation of 3.2 eq of pyridoxal-P groups per enzyme subunit. However, the competition experiments of Table I suggest that the labeling pattern is changed by the presence of added ligands during pyridoxylation. In these experiments, there was a consistent trend for nucleotide substrates and effecters that bind to the subunit catalytic site to protect more against activity loss than against the extent of pyridoxylation.
Thus, the data of Figs. 2 and 3 might represent a random addition of pyridoxal-P to sites of about equal affinity with only one or two of the three pyridoxal-P groups being incorporated at or near the catalytic site of each subunit. Table II shows activity data for unadenylylated glutamine synthetase and for the pyridoxylated enzyme containing 2.6 eq of pyridoxamine-P/subunit. Pyridoxylation produces an equivalent loss in the Mg2'-supported biosynthetic and yglutamyl transfer reactions. Both of these reactions are specific for the unadenylylated enzyme (2,3). The residual Mg'+supported activities have K,,, values for ATP and ADP that are -3-fold higher than the corresponding values for the unmodified enzyme. We could not detect a preferential loss of higher affinity sites for MgATP by assaying at low ATP concentrations as a function of pyridoxylation (Fig. 3). The Mn2+-supported activity data of  Pyridoxyhtion of G&amine Synthetase explanation. The MI?+-supported biosynthetic activity of GSi measures only the small fraction (one-twelfth) of the adenylylated subunits in the enzyme preparation (2,3,32). The inactivity of unadenylylated subunits in this assay has been attributed by Rhee et al. (42) to the very high affinity that this enzyme form has for MnADP in the presence of Pi. In contrast, the Mn"-supported y-glutamyl transfer reaction at pH 7.57, in which MnADP and arsenate are nonconsumable substrates, measures both adenylylated and unadenylylated subunits equally (4). Further, subumts within the dodecamer independently express Mn2+-supported y-glutamyl transfer activity. The loss in this latter activity was linear with increasing equivalents of pyridoxal-P incorporated and coincided with the data shown in Fig. 3, taking into account the 42% residual Mn"-supported transfer activity of the fully modified enzyme. A pH variation from pH 6.9 to 8.1 alters the absolute velocities but not the relative activities of the native and pyridoxylated enzymes in the Mn2'-supported y-glutamyl transfer reaction. The same activity difference between tmmodified and pyridoxylated enzymes was observed also in these assays when Pi (20 mu) was substituted for arsenate.
The activity results suggest that pyridoxylation of the dodecamer occurs near the nucleotide binding site of each subunit expressing activity. In obtaining the K,,, data of Table II, double reciprocal plots of velocity uersus nucleotide-manganese concentrations were linear in Mn2+-supported y-glutamyl transfer assays. If some of the subunits within a dodecamer had not been modified, these presumably would have expressed activity with K, = lo-* M for the MnADP and double reciprocal plots of velocity uersus MnADP concentration would have been biphasic reflecting K,,, values of lo-' and lo-' M for MnADP. This was not the case.
The K, data of Table II illustrate that pyridoxylation is like adenylylation in that it causes a large decrease in the affinity of the enzyme for MnATP or for MnADP.2 Adenylylation of glutamine synthetase, however, produces an increase in the Mn2+-supported biosynthetic specific activity (3) whereas pyridoxylation of the unadenylylated enzyme causes a partial inactivation (shown by the decrease in V,, observed in the Mn2+-supported transfer assay). We conclude that pyridoxylated, unadenylylated subunits with a lower affiity for MnADP must express some Mnz+-supported biosynthetic activity and that it is therefore a coincidence that V,, in this assay remains unchanged by pyridoxylation. It is not surprising that the pyridoxylated enzyme is a poorer catalyst, but the sign&ant amount of Mn2+-supported activity is rmusual.
Me2"-induced Spectral Perturbations-The binding of specific divalent cations (Mn2+, Mgzt, or Ca") to 12 high aflinity sites of glutamine synthetase in a tightening reaction activates the enzyme and induces an ultraviolet spectral perturbation of the protein (1). Hunt et al. (4) showed that two Mn2+ sites (nI and nz) per enzyme subunit must be saturated for yglutamyl transfer activity expression.
Difference spectra are shown in Fig. 4 for absorption changes produced by adding Mn'+, Ma;"', or Ca2' to divalent cation-free, pyridoxylated glutamine synthetase. When enough Me2" was added to saturate -90% of the nl sites (using published K'A values; Refs. 1 and 4), Curves 1 and 2 of Fig. 4 were obtained after 30 min. These show the usual tryptophanyl-tyrosyl residue perturbation at 2S4 and 290 mn, attributed to a movement of these residues to a less polar environment during the tightening reaction (36,37). The kinetics of the fmt order absorbance change at 290 nm (t1,z = 110 s at 2V'C) were the same with pyridoxylated enzyme as those z Since the reactions catalyzed by glutamine synthetase involve random binding of substrates and rapid equilibria, K, = K'LI for substrates (3). Activity data for unmodified and pyridoxylated glutamine synthetase Unadenylylated enzyme (GS) and the corresponding pyridoxylated enzvme (P-PXY-GS) containing 2.6 eo of p.yridoxamine-P/subunit were as&yed & described under "Exp&me&al Procedures." Assay buffers were: 50 nm Tris/imidazole (pH 7.36) for the Me-supported biosynthetic assay at 50 mM MgC12; 50 mM 1,4-piperezinediethanesulfonic acid (Pipes) at pH 6.5 for the Mn'+-supported biosynthetic assay at 6 mM MnClz; 50 nm 3,3'-diethylglutarate (pH 7.57) for rglutamyl transfer assays with either 30 nnu MgCh or 0.4 mM MnCh. For biosynthetic assays the substrates when not varied were fixed at 30 mu L~glutamate, 5-rnM ATP, and 50 mM NH*+ in M& assays and at 100 mM r.+-elutamate. 5 mu ATP, and 100 mM NH4' in Mn2+ assays. In transfer akays, the'substrates &hen not varied were tixed at 150 mM L-glutamine, 0.4 mM ADP, 20 mM arsenate, and 40 mu NH20H. In all cases, double reciprocal plots were linear.
--_-  measured previously with unmodified glutamine synthetase (4,36). The relatively small perturbations in the absorption range of the covalently bound pyridoxamine-P groups at low concentrations of Me2+ (Curves 1 and 2, Fig. 4), presumably are due to the binding of a small amount of Me2' near the pyridoxamine-P groups at n2 sites. At lower concentrations of Me2+ than those used to obtain Curves 1 and 2 (Fig. 4), there were no absorption changes above 300 nm.
When high concentrations of Mn2", Me, or Ca2* were added to pyridoxylated glutamine synthetase (Curves 3 and 4 of Fig. 4), larger perturbations from 300 to 360 nm instantaneously occurred. This indicated that a second subunit metal ion binding site near a covalently bound pyridoxamine-P group was being titrated with Me2'. The perturbations of protein-bound pyridoxamine-P groups produced by the binding of Mn2" or Me are different, with maxima at 338 or 328 nm, respectively. Since Mn2+ can competitively displace Mg2 from the enzyme (4,43), this wavelength difference could be verified by adding MI?' to the pyridoxylated magnesium . enzyme in an experiment (not shown) and observing the perturbation maximum ehift from 328 to 338 nm. The difference in wavelength maxima produced by Mn2' versus Mg2+ (or Ca2') could be related to the electronic structures of these metal ions, in which case the aromatic ring of a covalently bound pyridoxamine-P group must be very near the bound metal ion at the subunit n2 site. Titrations of Me2+-free, pyridoxylated glutamine synthetase with Co2+ (36) were unsuccessful due to a Co'+-induced aggregation of the modified enzyme. The only groups in pyridoxamine 5'-phosphate capable of forming a stable metal ion chelate are the phenolic and adjacent aminoethylene groups (40,41). The binding of Mn2+ to pyridoxamine Y-phosphate at pH 7.2 is very weak (40); stability constanta for Me (41) and Ca2+ at neutral pH would be even lower than that for Mn2'. In the absence of tertiary structure provided by the protein, these divalent cations are not expected to form stable chelates with the secondary amine derivative N-e-phosphopyridoxyl+lysyl-GS at neutral pH.
The resulta from spectrophotometric titrations of Me2'-free, pyridoxylated glutamine synthetase with Me and Ca2+ are presented as Hill plots in Fig. 5. A stoichiometry of one Me2+ binding site per subunit for each wavelength perturbation was of data with M%+ and Ca2' at either 290 or 328 nm could be fit with slopes of unity, suggesting that each type of site in the dodecamer is equivalent or noninteracting (i.e. not cooperative). This has been found previously for Mn2' binding to the unmodified enzyme (43). The most straightforward interpretation of the results shown in Fig. 5 is that the 290 nm data measure the titration of the high affinity nI sites (Kk 1: 8000 M-") whereas the 328 nm data measure the saturation of na sites (Kk = 1000 M-l). The nl sites of pyridoxylated glutamine synthetase appear to have about a 3-fold lower affinity for Me or Ca2' than measured previously with the unmodified enzyme (4,36). It has not been possible before to measure the binding of Mg2+ or Ca2+ to n2 sites, but unadenylylated glutamine synthetase has K,,, = 1 mu for Mg?-+ (4,43), and this value reflects n2 site saturation under assay conditions.
For the experiments of Figs. 4 and 5, EDTA had been removed from the modified enzyme prior to titrations with groaps Ultraviolet difference spectral measurements at 25'C were used at an enzyme subunit concentration of 0.060 mM in a buffer of 20 to 60 mre imidamle/HCl (pH 7.3 or 8.3) and 100 mM KC1 with 1 mM MnCh, 50 mu MgCh, or no divalent cation present. Pyridoxylated glutamine aynthetaae (P-Pxy-GS) preparations contained 2.6 or 3.2 eq of covalently bound pyridoxamine-P groupa per subunit. Spectrophotometric titrationa with different liganda were performed at the pH of maximum absorbance change; AA,,, (at h,,,) ia for a saturating concentration of ligand; K'fj -CT -0.025, where CT ia the total concentration (millimolar) of ligand required to produce 50% of the maximum spectral change. When known, K'n values (with appropriate literature references) for the umnoditkd enzyme (GS) are given.  Me'* ions. When EDTA was present in difference spectra measurements similar to those shown in Curves 3 and 4 of Fig. 4, ad~tio~ perturbations of covalently bound pyridoxamine-P groups were observed. These studies were not pursued but ,the observations suggest that the metal. EDTA complex binds to the modified enzyme.

Substrate-and I~~~~~~-~d~~ed
Spectral Changes-Spectral perturbations of protein-bound pyridoxamine-P groups produced by interactions of substrates and inhibitors with the Mnac-, MgZ'-, or Me"-free forms of pyridoxylated glutamine synthetase are summarixed in Table III. Absorbance changes occurred within 15 s after the addition of each effector; no effect of time on the degree of perturbation was observed for as long as 3 h. Examples of difference spectra are shown in Fig. 6. Substrate or inhibitor concentrations required to give 50% of the maximum spectral changes were used to calculate KO values for pyridoxylated glutamine synthetase in Table III; comparable K'n values (when known) for the unmodified enzyme are included. The effect of modification on this parameter varies from a quite pronounced decrease in the apparent affinity of the modified enzyme for ATP or ADP to a negligible affinity change for the inhibitor L-tryptophan. The spectral change observed with a particular ligand can be pH-dependent and this necessitated testing perturbations with each effector at several pH values (within the range of interest for glutamine synthetase). The results from titrations at pH 7.3 and 8.3 are given in Table III, Between these pH values, spectral changes were intermediate in amplitude to those given in Table III. None of the ligands of Table III affected the spectral properties of free pyridoxamine 5'-phosphate.
Spectral perturbations of protein-bound pyridoxamine-P groups produced by the nucleotide substrates (ATP and ADP) are Me2+ ion-dependent. These are maximum with the man- 6. Substrate-induced difference spectra of pyridoxylated glu-&mine synthetase at 25°C. The modified enzyme at 0.05 mhP subunit concentration with 3 eq of covalent& bound pyridoxamine-P groups per subunit was in a buffer containing 20 nm lmidazole/HCl, 100 mM KCl, and 1.0 mu MnCb (or 50 rnM MgC& for Pi addition only) at pH 7.3 (or pi-l 8.3 for the ATP perturbation; Table III). The maximum absorbance change produced by each ligand or ligand combination is shown. Ligands added singly were: 0.6 n&x ATP, 6.5 mx4 ADP, 6. ganese . enzyme complex, greatly diminished with the magnesium.enxyme complex, and nonexistent with the Me*+-free, p~doxylated enzyme. This is consistent with the activity results of Table II, in which the pyridoxylated enzyme was shown to have Mn2+-supported but little Me-supported activity. When added to the manganese*enxyme derivative, ATP and ADP induce spectral changes that differ both in direction and pH dependence (Table III, Fig. 6). ATP produces a relatively large increase in absorption at pH 8.3 but little absorbance change at pH 7.3, whereas ADP produces a large decrease in absorption at pH 7.3 but only a small perturbation at pH 8.3. The maximum wavelength of the ATP-induced difference spectrum varied with the manganese . enzyme derivative used (2.6 or 3.2 eq of pyridoxamine-P/subunit] and was either 322 or 334 nm, respectively. This variation in maximum wavelength was observed only with ATP. Spectral changes caused by inorganic phosphate, a substrate of the reverse biosynthetic reaction (3,42), do not have the Me'+ ion-or pH-dependence observed with ATP and ADP (Table III). Titrations of the modified enzyme at low Pi concentrations (<4.5 m&Q gave similar absorbance changes in the presence of Mn" or Mg", but the complete spectral change produced by Pi presumably can be observed only with the magnesium. enzyme complex (Fig. 6) because precipitation occurs in the Mn'" system-The change in the absorption of the covalently bound pyridoxamine-P groups produced by Pi in the presence of either metal ion is positive, which may relate to a binding of Pi to the protein at the site for the yphosphate of ATP (42). Interestingly, Pi produces a small negative absorbance change with the Me'+-free, modified enzyme in which case the protein is in an inactive conformation.
The AMP-induced spectral perturbations of the proteinbound pyridoxamine-P groups in the manganese+ enzyme complex are similar in direction, amplitude, and pH dependence to those of ADP (Fig. 6). However, the binding of AMP to glutamine synthetase, unlike that of ADP, is independent of metal ion (6), as reflected in the absorption changes observed with the Me'+ ion-free enzyme. derivative (Table III). A&bough AMP is an allosteric inhibitor of this enzyme (6,49,50), AMP can substitute for ADP in a Mn2'-supported yglutamyl transfer reaction catalyzed by unadenylylated glutamine synthetase (2,42). Antagonism between AMP and ATP was indicated by titrations of the MnATP .enxyme derivative with AMP at pH 6.3.
The addition of L-glutamate or r,-glutamine to the pyridoxylated manganese .enzyme has no effect on the absorption spectrum although both substrates would be expected to bind to the protein (28, 47). A transition state analogue of the glutamine synthetase reaction (511, L-methionine ( SR)-sulfoximine, produces only a small (red-shifted) spectral perturbation (Table III). However, these compounds have marked effects on the absorption of the protein-bound pyridoxamine-P groups of the ADP*manganese.enxyme complex. Since the binding of ADP and s,-glutamine or ADP and L-glutamate to the enzyme are synergistic (7,&B), the difference spectrum for ADP + Glx in Fig. 6 is representative of a different conformation than those stabilized by either substrate alone. Spectrophotometric titrations of the ADPamanganese. enzyme.pyridoxamine-P complex with L-glutamine or L-glutamate yielded h?o values of 24 or 18 mM, respectively; these values are -lo-fold greater than those measured previously with the unmodified enzyme in the presence of Me'+ and ADP (7,28). The ATP-induced difference spectrum of Fig. 6, in contrast to that of ADP, is unchanged by the presence of 24 mM L-glutamine. There is a synergistic effect between ADP and Pi (42) and between ADP and arsenate (4) in binding to unadenylylated glutamine synthetase. Synergism between ADP and Pi is found also with the manganese. enzyme. pyridoxamine-P complex. From the effect of Pi on the ADP-induced difference spectrum in Fig. 6, K'a for Pi is -0.4 111~. This value is -lOfold that found previously (42) with the Pi. ADP . manganese. enzyme complex. Note that the spectral perturbation produced by ADP + Pi is not the same as that of ATP ( Fig. 6) even though common sites of binding are involved.
Several inhibitors of glutamine synthetase (3) cause spectral perturbations of protein-bound pyridoxamine-P groups ( Table  III). The amplitudes of the spectral changes produced by AMP, carbamyl phosphate, L-tryptophan, and CTP were large whereas that of L-alanine was very small. These effects can be mediated through interactions at a catalytic site, an allosteric site, or both. An example of the latter is carbamyl phosphate for which two distinct spectral perturbations are observed (Fig. 7). CTP probably binds at the ATP subunit site since difference spectra with CTP and ATP were identical, the modified manganese.enzyme at pH 8.3, however, has a lower affinity for CTP than for ATP (Table III). Allosteric enzyme sites are involved in the binding of AMP, L-tryptophan, and L-alanine; separate enzyme sites have been demonstrated previously for AMP and L-tryptophan (6,52), for AMP and ADP (49,50), for AMP and L-glutamate (6), and for L-alanine and L-glutamate (7,8). The data of Table III are compatible with these observations. Whereas high concentrations of Lglutamate do not perturb protein-bound pyridoxamine-P groups, L-alanine produces a small, but reproducible, spectral change. Observe also that the spectral change produced by L-tryptophan is the same with Mn2+ or Mg2+ present and, although maximum at pH 8.3, is opposite in sign to that induced by ATP binding. Modification of the enzyme did not decrease the apparent affinity of the protein for L-tryptophan, but the previously observed positive cooperativity in binding L-tryptophan to glutamine synthetase (6) was not observed with the pyridoxylated enzyme. The spectral changes observed with carbamyl phosphate are more complex than those with the other effecters of Table  III and these are illustrated in Fig. 7. Relatively low concentrations of carbamyl phosphate produce positive spectral changes when added to the modified manganese. enzyme at pH 8.3 (e.g. Curve 1 of Fig. 7). The estimated K'o value for this spectral change is 0.7 mM and is designated as that obtained by a titration of a "tight" binding site in Table III. Increasing concentrations of carbamyl phosphate (Curves 2 to 6 in Fig. 7) produce negative spectral perturbations and alter the wavelength of maximum absorption change. The K'D value estimated for this second binding site (designated "loose" in Table III) is 14 mM. The results from two types of experiments with the manganese. enzyme. pyridoxamine-P complex at pH 8.3 indicated that the tight binding of carbamyl phosphate was to the subunit catalytic site. (a) With just sufficient carbamyl phosphate to saturate the tight binding site (Curve 1, Fig. 7), the addition of ADP at 80% of the enzyme subunit concentration produced the spectral change of ATP (Fig. 6). In fact, Meister and co-workers (46) have determined that carbamyl phosphate can bind to the catalytic site of glutamine synthetase and can phosphorylate ADP to synthesize ATP. ( b) When the modified manganese. enzyme at pH 8.3 was titrated with ATP to produce the maximum spectral perturbation (Fig. 6, Table III), subsequent additions of carbamyl phosphate only caused the negative spectral change associated with the loose binding of carbamyl phosphate. Thus, the loose binding of carbamyl phosphate appears to be to allosteric inhibitor sites of the enzyme.
Adenylylation and Pyridoxylation-Since the unadenylylated enzyme was modified in these studies, two types of experiments were performed to see whether lysyl residues near adenylylation sites (1,5) are pyridoxylated. After preparing fully adenylylated glutamine synthetase (GSE) as described previously (28, 32), the adenylylated and unadenylylated enzymes were pyridoxylated in parallel reactions. Exactly 3.2 eq of pyridoxamine-P groups per subunit were covalently attached to either enzyme form. This suggests that pyridoxylation is independent of the availability of adenylylation sites. The pyridoxylated, unadenylylated, and adenylylated enzymes also had about the same Mn2+-supported activities at pH 7.57; 44 and 30 units/mg, respectively. If adenylylation caused an exposure of new pyridoxylation sites (while burying others), it must do so without much affecting the Mn2+supported transfer activity of the modified enzyme. In a separate experiment, we tried to adenylylate the pyridoxylated, unadenylylated glutamine synthetase preparation enzymatically. Using large excesses of the glutamine synthetase adenylyltransferase (29,32) in the adenylyltransferase assay system (53), the maximum incorporation of radioactively labeled [14C]AMP into the pyridoxylated protein was about 50% of that into unmodified GSi. This could be due to the pyridoxylated enzyme being a poor substrate for adenylyltransferase (which is likely) or to some pyridoxylation having occurred near adenylylation sites. Although these results are somewhat equivocal, we can conclude that lysyl residues at adenylylation sites are not major sites of pyridoxylation in unadenylylated glutamine synthetase.
Topography of Pyridoxylation Sites in Glutamine Synthetase-Whether or not pyridoxylation occurs at three unique sites of each subunit of the dodecamer must await peptide analysis. Amino acid sequence and three-dimensional structural information is not yet available for the assignment of pyridoxylation sites. Nevertheless, the results presented here indicate that at least one pyridoxylation site is near the catalytic site of each subunit of glutamine synthetase.
The uniformity of labeling near the nucleotide binding sites of the enzyme is indicated by the activity data of Fig. 3 and Table II. One pyridoxylation site appears to be near the yphosphate position of ATP. Carbamyl phosphate can bind to this site and can phosphorylate ADP to synthesize ATP. Since the maximum wavelength of the absorption change produced by ATP (Table III) was a function of the extent of pyridoxylation (2.6 or 3.2 eq of pyridoxamine-P groups/subunit), ATP binding appears to perturb two covalently bound pyridoxamine-P groups. At least one pyridoxamine-P group is very near the n2 metal ion binding site of the enzyme subunit. However, the affinity of the pyridoxylated enzyme for M$ or Ca" (Fig. 5) is too low to obtain the stoichiometry of pyridoxamine-P labeling at this site. Since the Me'+ is known to chelate the nucleotide at the catalytic site, it can be assumed nevertheless that all metal ion binding at nz subunit sites is producing the observed spectral perturbation of protein-bound pyridoxamine-P groups (Fig. 4). Interestingly, the binding of specific divalent cations (Mn", Mgz+, or Ca*+) to the nl activating site of each subunit does not perturb covalently bound pyridoxamine-P groups. Also, lysyl residues near adenylylation sites of the enzyme do not appear to be sites of pyridoxylation. Thus, the labeling of the dodecamer by pyridoxylation appears to occur at specific sites.
The spectral perturbations of protein-bound pyridoxamine-P groups caused by substrate and inhibitor binding to the enzyme at a given pH can arise from shifta in the environment that iniluence solvent accessibility and the pK of the pyridinium ion of the aromatic ring, which may involve also changes in the ionization of protein groups near pyridoxylation sites. For example, the equivalent of a 0.1 pH unit increase near the pK (at a pH of -8) of the pyridiimn ion will produce a molar absorption change of about -220 b&m-' at 326 run. The pH dependence and different wavelength maxima of absorbance changes produced by various ligands of glutamine synthetase are brought about by the different conformations stabilized. The positive absorption change produced by ATP binding to the enzyme corresponds to an increased protonation of the pyridine nitrogen which may be why the ATP perturbation is not observed at pH 7.3 but is observed at more alkaline pH values. In contrast, ADP binding to the pyridoxylated enzyme produces an absorption decrease which corresponds to a decrease in the degree of protonation of the pyridine nitrogen. This may be why the perturbation by ADP is observed at pH 7.3 and not at pH 8.3. Metal ion binding at the subunit catalytic site appears to involve a direct electronic interaction between a very near covalently bound pyridoxamine-P group and the proteinbound metal ion. The absorption change produced by Mn" and by M8+ or Ca" binding to pyridoxylated glutamine synthetase had different wavelength maxima which presumably relate to the electronic configurations of these divalent i cations. Since the Me2' chelates the y&phosphates of ATP or the a&phosphates of ADP, there must be flexibility in the conformation at the subunit catalytic site for accommodating ATP or ADP binding. An indication that the catalytic site is able to assume different conformations in the presence of various combinations of substrates is shown by the spectral perturbations produced by ADP + Glx, ADP + methionine sulfoximine, or ADP + Pi (Fig. 6). These were not equal to the mmr of the spectral perturbations produced by either ligand alone (Fig. 6, Table III). Covalently attached pyridoxamine-P groups near catalytic sites must be able to move out of the way for activity to be expressed since the pyridoxylated enzyme has substantial Mn'+-supported activity (with a lower V-). An effect of pyridoxylation is to lower the affinity of the manganese . enzyme for nucleotide substrates (Tables II and III). The inability of M%+ to support activity of the pyridoxylated enzyme may relate to the relatively low affinity that the unmodified magnesium. enzyme has for nucleotide substrates. Spectral perturbations of protein-bound pyridoxamine-P groups are observed for ATP with the manganese. enzyme but not with the magnesium.enzyme (Table III). The effect of pyridoxylation in decreasing the affinity of the enzyme for nucleotide substrates has precedent in the observations of Greenwell et al. (15) who found that the introduction of one pyridoxamine-P group per catalytic chain in aspartate transcarbamylase produced apparent inactivation but only a decrease in the affinity of the enzyme for a tightly binding, bisubstrate analogue.
The labeling of phosphate binding sites at or near allosteric inhibitor sites is difficult to assess. By definition, an allosteric inhibitor can affect the conformation of the active site of a regulatory enzyme. Desensitization of pyridoxylated glutamine synthetase to various inhibitors has not been tested. However, carbamyl phosphate clearly is bound to two types of sites, one of which is identifiable as the subunit catalytic site. Also, ~hryptophan is an interesting example of an inhibitor producing a large negative spectral perturbation of covalently attached pyridoxamine-P groups while apparently binding far enough away from pyridoxylation sites so that the enzyme has an unchanged affinity for this inhibitor. Further, the spectral change produced by tryptophan is maximum at pH 8.3 and is the same with Mn" or Me present (Table  III). Thus, the tryptophan-induced spectral change of proteinbound pyridoxsmine-P groups has a different origin than those produced by ATP or ADP binding.
The use of covalently bound pyridoxamine-P groups as spectral probes for ligand interactions with a complex regulatory protein is illustrated in the studies of this paper. The spectral and ionization characteristics of protein-bound pyridoxamine-P indicate that these groups are near the enzyme surface exposed to solvent, implying the same for the catalytic sites of glutamine synthetase.