Mechanism of Activation of the Tryptophan Synthase α2β2 Complex SOLVENT EFFECTS OF THE CO-SUBSTRATE β-MERCAPTOETHANOL

To characterize the conformational transitions that lead to activation of catalysis by the tryptophan synthase α2β2 complex, we have determined the solvent effects of a co-substrate, β-mercaptoethanol, and of a model nonsubstrate, ethanol, on the catalytic and spectroscopic properties of the enzyme. Our results show that ethanol and β-mercaptoethanol both alter the equilibrium distribution of pyridoxal 5′-phosphate intermediates formed in the reactions of L-serine at the β site in the α2β2 complex. Addition of increasing concentrations of ethanol increases the proportion of the external aldimine of L-serine and decreases the proportion of the external aldimine of aminoacrylate. Low concentrations of the co-substrate β-mercaptoethanol (Kd = ∼13 mM) decrease the proportion of the external aldimine of aminoacrylate and induce formation of the quinonoid of S-hydroxyethyl-L-cysteine. Higher concentrations of β-mercaptoethanol decrease the concentration of the quinonoid intermediate and increase the proportion of the external aldimine of L-serine. Data analysis shows that β-mercaptoethanol and ethanol both interact or bind preferentially with the conformer of the enzyme that predominates when the aldimine of L-serine is formed and shift the equilibrium in favor of this conformer. We propose that a nonpolar region of the β subunit, possibly the hydrophobic indole tunnel, becomes less exposed to solvent in the conformational transition that activates the α2β2 complex.

To fully understand the relationships between enzyme structure and function, it is important to determine how enzyme activity is regulated by protein-protein interaction and proteinligand interaction. Investigations employing spectroscopic techniques can furnish valuable insight into the regulation of catalytic intermediates and of conformational states. The chromophoric pyridoxal 5Ј-phosphate (PLP) 1 coenzyme provides an excellent spectroscopic probe for detecting catalytic intermediates and monitoring the effects of ligands on the equilibrium distribution of enzyme-substrate intermediates formed by PLPdependent enzymes. Here we utilize the PLP coenzyme of the ␤ subunit of tryptophan synthase to investigate the activation of catalysis by the tryptophan synthase ␣ 2 ␤ 2 complex.
The bacterial tryptophan synthase ␣ 2 ␤ 2 complex (EC 4.2.1.20) is a bifunctional enzyme that catalyzes the final two reactions in the biosynthesis of L-tryptophan, termed the ␣ reaction (Equation 1) and the ␤ reaction (Equation 2) (for reviews, see Refs. [1][2][3][4][5][6]. These two reactions are catalyzed by the ␣ and ␤ subunits 2 in the ␣ 2 ␤ 2 complex and, at a much lower rate, by the isolated ␣ subunit and ␤ 2 subunit, respectively. The crystal structure of the tryptophan synthase ␣ 2 ␤ 2 complex from Salmonella typhimurium (7) revealed that the active sites of the ␣ and ␤ subunits are ϳ25 Å apart and are connected by a hydrophobic tunnel. This tunnel provides a plausible structural route for transfer of indole, which is produced by cleavage of indole 3-glycerol phosphate at the active site of the ␣ subunit, to the active site of the ␤ subunit in the ␣ 2 ␤ 2 complex.
The isolated ␤ 2 subunit and the ␣ 2 ␤ 2 complex contain one PLP coenzyme at each ␤ active site and catalyze a number of PLP-dependent reactions including the ␤-replacement reactions with L-serine and indole (Equation 2) and with L-serine and ␤-mercaptoethanol (␤-ME) (Equation 3).
L-Serine ϩ ␤-ME 3 S-hydroxyethyl-L-cysteine ϩ H 2 O (Eq. 3) Equations 2 and 3 proceed through a series of PLP-substrate intermediates (Scheme I) that have characteristic absorption and fluorescence spectra. Formation of the external aldimine of aminoacrylate (E-AA) is thought to be accompanied by a conformational transition that activates the enzyme (see "Discussion"). The present work investigates the effects of ␤-mercaptoethanol, which is a co-substrate for the ␤ 2 subunit and the ␣ 2 ␤ 2 complex in Equation 3, and of ethanol, which is not a substrate, on the absorption and fluorescence properties of catalytic intermediates formed by the ␣ 2 ␤ 2 complex and ␤ 2 subunit. We find that ␤-mercaptoethanol and ethanol both bind or interact preferentially with the conformer of the enzyme that predominates when the aldimine of L-serine is formed (E-Ser) and shift the equilibrium in favor of this conformer. We propose that a nonpolar region of the ␤ subunit, possibly the hydrophobic indole tunnel, becomes less exposed to solvent in the conformational transition that accompanies formation of E-AA and activates the ␣ 2 ␤ 2 complex. * 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.

MATERIALS AND METHODS
Chemicals and Buffer-Indole-3-propanol phosphate (IPP) was a generous gift of Kasper Kirschner. Buffer B (50 mM sodium N,N-bis(2hydroxyethyl)glycine containing 1 mM EDTA at pH 7.8) was used throughout.
Spectroscopic and Analytical Methods-Absorption spectra were made using a Hewlett-Packard 8452 diode array spectrophotometer at 23°C. Fluorescence emission of the L-serine external aldimine was determined at 510 nm with excitation at 420 nm using a Perkin-Elmer model MPF-44B fluorimeter. Circular dichroism measurements (mean residue ellipticity in degrees cm 2 /dmol) of the ␤ 2 subunit and ␣ 2 ␤ 2 complex were made in a Jasco J-500C spectropolarimeter, equipped with a DP-500N data processor (Japan Spectroscopic Co., Easton MD) and thermostated by a circulating water bath. Gel filtration chromatography used a Superose 12 HR 10/30 column (Pharmacia Biotech Inc.) in a LCC-501 Plus fast protein liquid chromatography system (Pharmacia-LKB) equilibrated with buffer B containing 0.15 M NaCl in the presence or absence of 0.5 M ␤-mercaptoethanol. The ␣ 2 ␤ 2 complex (0.6 -1.0 mg/ml) in the equilibration buffer was injected and eluted at a flow rate of 0.5 ml/min with the same buffer. The ␣ 2 ␤ 2 complex eluted as a single peak under both conditions. Ultrafiltration of the ␣ 2 ␤ 2 complex (0.6 -1.0 mg/ml in buffer B containing 0.5 M ␤-mercaptoethanol) utilized Centricon-100 (Amicon) filters with molecular weight cutoff ϭ 100,000. In the presence or absence of 0.5 M ␤-mercaptoethanol, less than 5% of the protein was detected in the filtrate, whereas the isolated ␣ subunit passes freely through the filter (11).
Data Analysis-The data for the reactions of the ␣ 2 ␤ 2 complex with L-serine in the presence of ␤-mercaptoethanol or ethanol were fit to the models in Schemes II and III (see "Results"). For reactions in the presence of ethanol, k 2 ϭ 0 and thus E-Q ϭ 0. The total concentration of protein, Each set of experimental data was fit separately to Equations 4 -6 to obtain the values of k 0 , k 1 , and N Ϫ M in Table I and the curves in Fig.  1, B and C, and Fig. 3, B-D (see "Results").

Ethanol Alters the Distribution of Enzyme-Substrate
Intermediates-We have first investigated the solvent effects of a model nonsubstrate, ethanol, on the enzymatic reactions with L-serine (Fig. 1). 3 The reaction of L-serine with PLP at the active sites of the ␤ 2 subunit and ␣ 2 ␤ 2 complex proceeds through the external aldimines of L-serine (E-Ser) and aminoacrylate (E-AA) (see structures in Scheme I) (12)(13)(14). In the absence of a co-substrate (e.g. indole or ␤-mercaptoethanol), the reaction cannot proceed beyond the E-AA and the catalytic steps are reversible (15,16). E-Ser exhibits maximum absorbance at 424 nm and strong fluorescence emission at 510 nm upon excitation at 410 nm (17). E-Ser is the predominant intermediate that accumulates upon addition of L-serine to the ␤ 2 subunit (17,18). The spectrum observed upon addition of L-serine to the ␣ 2 ␤ 2 complex in aqueous solution (Fig. 1A, spectrum 0) exhibits an absorption maximum near 350 nm and a broad envelope of absorbance that extends out to ϳ525 nm as described previously (17). The band at 350 nm has been attributed to E-AA (see "Discussion") (12,19).
Addition of increasing ethanol concentrations decreases the absorbance at 350 nm and increases the absorbance at 424 nm (Fig. 1, A and B). The presence of an isosbestic point in the spectra in Fig. 1A indicates that ethanol shifts the equilibrium distribution of the two species from E-AA to E-Ser. The fluorescence data in Fig. 1C show that the E-Ser intermediate formed by the ␤ 2 subunit is not affected by low concentrations of ethanol, whereas the E-AA intermediate formed by the ␣ 2 ␤ 2 complex is converted to E-Ser under the same conditions.
To account for the results in Fig. 1, we postulate a simple model (Scheme II) in which both forms of the enzyme (E-Ser and E-AA) show weak "binding" 4 of ethanol (ROH), up to N moles/mol to E-Ser and up to M moles/mol to E-AA; k 0 is the intrinsic equilibrium constant for the interconversion of E-Ser and E-AA and k 1 is the apparent average binding con- 3 Ethanol and ␤-mercaptoethanol have very small effects on the spectroscopic properties of the ␣ 2 ␤ 2 complex alone in the absence of serine which do not contribute to the effects seen in the presence of L-serine. The highest concentrations of alcohols used in this work (4 M ethanol and 1 M ␤-mercaptoethanol) reduced the absorbance of the ␣ 2 ␤ 2 complex at max ϭ 412 nm by about 5%. 4 See "Discussion" for consideration of the mechanism of solvent interaction or binding to tryptophan synthase. SCHEME I. Intermediates in the reaction of the tryptophan synthase ␣ 2 ␤ 2 complex with L-serine and ␤-mercaptoethanol. Enzymebound PLP (E) reacts with L-serine to form an external aldimine (E-Ser) that is dehydrated to the external aldimine of aminoacrylate (E-AA). Addition of ␤-mercaptoethanol (␤ME) yields the quinonoid (E-Q). Protonation of E-Q yields the external aldimine of S-hydroxyethyl-L-cysteine (E-S-HEC). The enzyme and enzyme-substrate intermediates have characteristic absorption spectra with the indicated max values.
activity. We expect that solutions of ␤-mercaptoethanol may show similar nonideality and that nonideality may distort extrapolations of our results to the absence of cosolvent in both cases.
Effects of an Allosteric Ligand on the Reaction Intermediates with L-Serine and ␤-Mercaptoethanol-Ligands that bind to the active site of the ␣ subunit (indole-3-glycerol phosphate, indole-3-propanol phosphate (IPP), D-glyceraldehyde 3-phosphate, and ␣-glycerol 3-phosphate) are known to alter the equilibrium distribution of enzyme-substrate intermediates at the active site of the ␤ subunit 25 Å distant (21, 22, 24 -31). Addition of IPP results in a small increase in the K d for ␤-mercaptoethanol as a substrate and in a much larger increase in the concentration of ␤-mercaptoethanol needed to shift the equilibrium from E-Q to E-Ser ( Fig. 3C and Table I).
Structural Integrity of the ␣ 2 ␤ 2 Complex in ␤-Mercaptoethanol and Ethanol Cosolvents-The ␣ 2 ␤ 2 complex does not dissociate into ␣ subunit and ␤ 2 subunit when subjected to ultrafiltration on a Centricon-100 (Amicon) membrane or to fast protein liquid chromatography on a Superose column in the presence of 0.5 M ␤-mercaptoethanol (see "Materials and Methods") or 5 M ethanol (11). Although ethanol (2.5 M) has no effect on the near UV/visible circular dichroism spectrum of the ␣ 2 ␤ 2 complex (Fig. 4A), ethanol does decrease the ellipticity of the PLP coenzyme of the ␤ 2 subunit at 412 nm. These results indicate that the ␣ 2 ␤ 2 complex is more stable than the ␤ 2 subunit in 2.5 M ethanol and does not dissociate under these conditions.

Mechanism of Interactions of Cosolvents with Tryptophan
Synthase-The good fits of our data (Figs. 1 and 3 and Table I) to the models in Schemes II and III provide evidence that ethanol and ␤-mercaptoethanol interact preferentially with the conformer of the enzyme that predominates when E-Ser is formed and that this interaction shifts the equilibrium from E-AA and E-Q to E-Ser. We now ask what is the mechanism of interaction of these cosolvents with tryptophan synthase? Cosolvents affect the stability of proteins either by direct binding or by preferential exclusion (32)(33)(34). Preferential exclusion of the cosolvent results in preferential hydration and in stabilization of the protein, whereas preferential binding can result in protein denaturation. Certain compounds including ethanol stabilize proteins at low temperatures (i.e. are cryoprotectants), yet can also induce protein denaturation at higher temperatures. Arakawa et al. (32) have proposed that at higher temperatures, which favor hydrophobic interactions, the hydrophobic character of compounds such as ethanol predominates, leading to direct interaction with hydrophobic residues in the protein. Recent 13 C NMR studies have demonstrated the direct binding of the methyl group of ethanol to a hydrophobic pocket(s) in bovine serum albumin (35).
Although our experimental results do not distinguish between preferential exclusion and preferential binding of alcohols, we have used the site binding models described in Schemes II and III because it is more amenable to a mathematical description of the results (Table I). However, a mechanism of direct hydrophobic interaction of alcohols with tryptophan synthase is supported by our earlier finding that inhibition of tryptophan synthase by straight chain monofunctional alcohols (methanol, ethanol, 1-propanol, and 1-butanol) is proportional to the hydrophobicity and to the chain length of the alcohols (11). The apparent concentration of ethanol that inhibits half of the activity of the ␣ 2 ␤ 2 complex in the reaction of L-serine and indole (K i Ј ϭ 0.55 M) (11) is quite similar to the concentration of ethanol (C 0.5 ϭ ϳ0.8 M) that induces half maximal change in absorbance at 424 nm (Fig. 1B) and of fluorescence emission at 510 nm (Fig. 1C). The combined results of the two studies indicate that ethanol binds directly to hydrophobic sites on tryptophan synthase.
We were initially surprised to find that ␤-mercaptoethanol was a stronger perturbant of the conformational equilibrium (C 0.5 ϭ ϳ0.5, Fig. 3C) than ethanol since ethylene glycol had no inhibitory effect (11). However, using the method of Rekker and de Kort (36), we calculated log P values (a measure of hydrophobicity) for ethanol (Ϫ0.20) and ␤-mercaptoethanol (Ϫ0.16). Thus, ␤-mercaptoethanol is more hydrophobic than ethanol, consistent with the greater potency of ␤-mercaptoethanol. In agreement with our analysis, numerous hydrophobicity scales used in discussions of protein structures all assign cysteine (ϪCH 2 SH side chain) as significantly more hydrophobic than serine (ϪCH 2 OH side chain) (reviewed in Ref. 37).
Mechanism of Activation-What do our results tell us about the mechanism of catalysis and the mechanism of activation of tryptophan synthase? Our most important conclusion is that the conformer of the enzyme that predominates when the aldimine of L-serine (E-Ser) is formed binds more ethanol or ␤-mercaptoethanol and therefore has more hydrophobic character than the conformer of the enzyme that predominates when the aldimine of aminoacrylate (E-AA) is formed. What else is known about the transition between these two conformers? Our group (38,39) and Dunn's group (30,31,40,41) have developed models which propose that formation of E-AA triggers the change in conformation from an "open" to a "closed" form (Scheme IV; for abbreviations, see Scheme I legend). Recent evidence indicates that the conformer that predominates when the quinonoid intermediate E-Q is formed is also in the closed TABLE I Effects of ␤-mercaptoethanol (␤ME) and ethanol (EtOH) concentration on the equilibrium distribution of reaction intermediates Spectroscopic data from Figs. 1-3 were analyzed using the PC-MLAB program (Civilized Software, Bethesda, MD) and fit to the models shown in Schemes II and III (see "Data Analysis" under "Experimental Procedures"). The models assume that ␤-mercaptoethanol acts as a substrate at low concentrations and as a solvent at higher concentrations, perturbing the equilibria, by preferential binding to one or more forms of the protein, whereas ethanol acts only as a solvent.
The fluorescence emission curves in Fig. 3D for both ␤ 2 subunit and ␣ 2 ␤ 2 complex showed evidence for quenching by increasing concentrations of ␤-mercaptoethanol. The data for the ␤ 2 subunit were fit to the Stern-Volmer equation (F/F o ϭ (1 ϩ K sv [␤-ME]) Ϫ1 to give a value of K sv ϭ 1.22 Ϯ 0.06 M Ϫ1 . This value was used to correct data obtained with the ␣ 2 ␤ 2 complex for quenching.
form and the conformer that predominates when the aldimine intermediate E-S-hydroxyethyl-L-cysteine is formed is in the open form (Scheme IV) (41). Formation of E-AA is associated with activation of the ␤ site for reaction with nucleophiles and of the ␣ site for reaction with indole-3-glycerol phosphate (31,42,43). The results presented here indicate that activation of the ␤ subunit, which accompanies conversion of E-Ser to E-AA, must be associated with a conformational transition that removes some nonpolar region of the ␤ subunit from exposure to solvent (i.e. N Ͼ M, Table I).
What is the location of the nonpolar region removed from contact with solvent in the conformational transition? A possible location is in the hydrophobic indole tunnel that connects the active sites of the ␣ and ␤ subunits. This possibility is attractive because it would help to explain other results with tryptophan synthase. Several mutant forms of the ␣ 2 ␤ 2 complex have much faster rates than the wild type ␣ 2 ␤ 2 complex in ␤-elimination reactions of L-serine or ␤-chloro-L-alanine (38). To account for this observation, we proposed that hydrolysis of the E-AA intermediate occurs in each mutant complex because the ␤ subunit is unable to undergo the transition from open, solvent accessible conformer to a closed, less solvent accessible conformer (Scheme II) (38). Our new results provide direct chemical data to support this hypothesis.
Indirect evidence for a solvent-excluded active site in E-AA has come from recent studies that demonstrate slow exchange of the enzyme-bound 2 H abstracted from the ␣-carbon of E-Ser with 1 H from solvent water (44). This work also provides evidence for a low barrier hydrogen bond in the E-AA intermediate and suggests that this bond may play an important role in catalysis. Low barrier hydrogen bonds are stabilized by the absence of solvent.
The presence of the E-AA intermediate in a solvent-excluded active site would also explain why this intermediate has an absorption maximum at 350 nm ( Fig. 2A), whereas aminoacrylate aldimines in model and enzymatic systems absorb maximally at 450 -480 nm (45,46). Investigations of the solvent-dependent equilibria between tautomers of Schiff bases (aldimines) of amines with salicylaldehyde and with PLP (see (47,48) and references therein) have demonstrated that enolimine tautomers, which are neutral species that are favored by nonpolar environments, absorb at much lower wavelengths than ketoenamine tautomers, which are dipolar, resonance stabilized species that are favored by more polar environments. It follows that the observed absorption peak at 350 nm ( Fig. 2A) is appropriate for that of the enolimine tautomer of the aminoacrylate aldimine in a nonpolar environment.
The ligand-induced conformational transitions depicted in Scheme IV have also been investigated by use of two fluorescent probes, Nile Red (49) and 8-anilino-1-naphthalenesulfonate (ANS) (41). The former studies provide evidence that active site and allosteric ligands prevent Nile Red binding to a hydrophobic site in the indole tunnel (49). The latter studies demonstrate that the transition from the open to closed conformation prevents ANS binding. Although the location and mode of ANS binding to tryptophan synthase are unknown, ANS may bind in the hydrophobic tunnel or to disordered surface loop structures (41). The ANS results might indicate that the open conformation of tryptophan synthase has more hydrophobic binding sites than the closed form, as we propose on the basis of the cosolvent binding studies presented in the present work. How-ever, the interaction of ANS with proteins may be due not only to hydrophobic forces, but also to electrostatic forces, as ANS has a water-soluble sulfonyl group (50). ANS can also bind with ␤ structural polypeptides even in the absence of hydrophobic side chains (50). Thus our results give more direct evidence than the ANS results on the hydrophobic properties of the open and closed conformations.
Does tunnel closure play a role in catalysis by tryptophan synthase? Dunn and co-workers have argued that tunnel closure would be counterproductive for channeling of indole between the ␣ and ␤ sites (41). However, tunnel closure may be a dynamic, fluctuating process. Because the rate of indole diffusion through the tunnel (Ͼ1000 s Ϫ1 ) is much faster than the rate-limiting step in the ␣␤ reaction (8 s Ϫ1 ) (42,43,51), the tunnel could be closed most of the time without impeding the rate of indole transfer. The combined results from the cited studies (38,44,49) and the solvent studies presented here suggest that ligand-mediated tunnel closure results in exclusion of solvent from the active site of the ␤ subunit in the active conformer. Solvent exclusion from the ␤ site may prevent the unwanted side reaction, hydrolysis of E-AA to form pyruvate and ammonia (38), and stabilize an activated form of the enzyme (44).
Our results also provide conditions for trapping an alternative conformational state of the ␣ 2 ␤ 2 complex. The ability to preferentially accumulate the E-Ser intermediate in the presence of low concentrations of ethanol may be useful for future crystallographic studies, kinetic studies, and analysis by NMR. Several PLP enzymes, including aspartate aminotransferase (52, 53), 2-amino-3-ketobutyrate-CoA ligase (54), and tryptophan synthase, 5 display a series of distinct 1 H NMR resonances in the downfield 10 -18 ppm region. 600 MHz 1 H NMR spectra of 2-amino-3-ketobutyrate-CoA ligase in the presence of different substrates have been interpreted as giving evidence for low-barrier hydrogen bonds (54).
Practical Implications of the Results-Since the discovery that ␤-mercaptoethanol is a co-substrate for tryptophan synthase in a ␤-replacement reaction (Equation 3) (17) and in a thiol-dependent transamination reaction (18), several investigations of these reactions have used ␤-mercaptoethanol in the concentration range of 10 -140 mM (8,15,40,41). These high concentrations were used to avoid depletion of ␤-mercaptoethanol due to turnover and to increase the binding of ␤-mercaptoethanol.
Our model predicts that solvent effects of ␤-mercaptoethanol would be observable at concentrations as low as 10 mM, a concentration frequently used in buffers as a reducing agent in enzyme studies. For example, an investigation of the effects of a disulfide engineered into T4 lysozyme, compared thermal inactivation of the oxidized mutant enzyme with that of the reduced enzyme in the presence of 10 mM ␤-mercaptoethanol (55). A later study reported that T4 lysozyme crystals are very difficult to grow in the absence of ␤-mercaptoethanol and utilized 60 mM ␤-mercaptoethanol in the crystallization of the reduced form of T4 lysozyme having an engineered disulfide (56). An investigation of the thermal inactivation of ␣-mannosidase, an enzyme devoid of cysteine or cystine, demonstrated that a higher concentration of ␤-mercaptoethanol (0.26 M) reduced the T I from 68 to 58°C and facilitated subunit dissociation (57). These authors proposed that ␤-mercaptoethanol may act as an organic solvent. In another study, ␤-mercaptoethanol was suggested to cause the inactivation of (Na ϩ K ϩ )-ATPase by dual functions as a reducing agent and as an organic solvent via a delocalized protein denaturation mechanism (58). Our