Identification of the Geometric Requirements for Allosteric Communication between the α- and β-Subunits of Tryptophan Synthase*

The pyridoxal 5′-phosphate-dependent tryptophan synthase α2β2 complex is a paradigmatic protein for substrate channeling and allosteric regulation. The enzymatic activity is modulated by a ligand-mediated equilibrium between open (inactive) and closed (active) conformations of the α- and β-subunit, predominantly involving the mobile α loop 6 and the β-COMM domain that contains β helix 6. The α ligand-triggered intersubunit communication seems to rely on a single hydrogen bond formed between the carbonyl oxygen of βSer-178 of β helix 6 and the NH group of αGly-181 of α loop 6. We investigated whether and to what extent mutations of αGly-181 and βSer-178 affect allosteric regulation by the replacement of βSer-178 with Pro or Ala and of αGly-181 with either Pro to remove the amidic proton that forms the hydrogen bond or Ala, Val, and Phe to analyze the dependence on steric hindrance of the open-closed conformational transition. The α and β activity assays and the equilibrium distribution of β-subunit catalytic intermediates indicate that mutations do not significantly influence the intersubunit catalytic activation but completely abolish ligand-induced α-to β-subunit signaling, demonstrating distinct pathways for α-β-site communication. Limited proteolysis experiments indicate that the removal of the interaction between βSer-178 and αGly-181 strongly favors the more trypsin-accessible open conformation of the α-active site. When the hydrogen bond cannot be formed, the α-subunit is unable to attain the closed conformation, and consequently, the allosteric signal is aborted at the subunit interface.

The ␣ 2 ␤ 2 complex of tryptophan synthase (TS) 1 represents a particularly interesting case (2) because the open to closed transition serves a dual role: the increase in catalytic activity and the allosteric communication between subunits (3)(4)(5)(6)(7)(8)(9). The enzyme catalyzes the biosynthesis of L-tryptophan in two steps: the ␣-subunit cleaves indole-3-glycerol phosphate to D-glyceraldehyde-3-phosphate and indole, which, intramolecularly channeled to the ␤-subunit (10,11), reacts with L-serine in a ␤ replacement reaction. The ␤-subunit contains a pyridoxal 5Јphosphate bound to Lys-87 as an internal aldimine (12). During catalysis, a series of chromophoric intermediates are formed, including an external aldimine, an ␣-aminoacrylate and quinonoid species (Scheme 1). The catalytic activity of ␣and ␤-subunits is kept in phase by a fine-tuning associated with intersubunit communication (3,13,14). This allosteric regulation between active sites that are 20 Å apart (4,10) involves alternative open and closed states of both the ␣and ␤-subunit (5,9,(15)(16)(17). The open to closed transition not only leads to an increase in the catalytic activity of the subunit but also is involved in signaling the catalytic state of one subunit to the other (3,13,14).
Structural and functional investigations on the wild type and on a variety of mutants of both subunits have allowed the unveiling of several aspects of this coordinated series of events at a molecular level (6 -9, 15, 17-31). Specifically, the comparison of the three-dimensional structure of TS in the absence (both ␣and ␤-subunits in the open state) and presence of a ligand in the ␣-active site and the catalytic intermediate ␣aminoacrylate in the ␤-active site (both ␣and ␤-subunits in the closed state) (5,15) evidenced several structural differences: (i) the formation of hydrogen bonds between ␣ loop 2 and ␤ helix 6, (ii) the stabilization of ␣ loop 6, which is crystallographically undetectable in the open form due to high mobility (5,10), in a closed form (Fig. 1), (iii) the presence of a single hydrogen bond between ␣ loop 6 and ␤ helix 6, and (iv) the movement of part of the ␤-subunit, including ␤ helix 6. This mobile portion of the ␤-subunit was called the COMM domain because of its role in communicating to the ␤-active site conformational changes occurring at the ␣-active site. The relevance of the ␣ loop 6-␤ helix 6 hydrogen bond between the carbonyl oxygen of ␤Ser-178 and the amide NH of ␣Gly-181 for allosteric signaling was tested by introducing the ␤S178P mutation (7). Indeed, the enzyme was still active but allosterically "knocked out." A structural study showed several alterations of the ␣-␤ interface and the inability of the ␣-subunit to achieve the closed state even in the presence of ␣-subunit ligands (17).
Because all mutants of ␤Ser-178 are potentially able to form a hydrogen bond with ␣Gly-181, we prepared the ␤S178A mutant to verify whether the disruption of the ␤Ser-178-␣Gly-181 hydrogen bond was due to proline geometric features. Furthermore, by taking into account that ␣Gly-181, but not ␤Ser-178, is a conserved residue among TS from several species (7, 32), we have investigated the geometric requirements for the achievement of the ␤Ser-178-␣Gly-181 hydrogen bond by preparing ␣G181A, ␣G181V, and ␣G181F, mutants characterized by a progressively increasing side chain steric hindrance, and ␣G181P.
Activity assays, distribution of catalytic intermediates, and limited proteolysis kinetics in the absence and presence of substrates and ligands concordantly indicate that the ␤Ser-SCHEME 1. Enzyme-substrate intermediates formed in the reaction of TS with L-serine and indole. The maximum of the absorbance peak, assigned to each intermediate, is shown. 178-␣Gly-181 hydrogen bond is critical for both allosteric communication and stabilization of the closed state of the ␣-subunit, but not for intersubunit activation.

MATERIALS AND METHODS
Reagents-All chemicals (Sigma-Aldrich) were of the best available quality and used without further purification. D,L-␣-Glyceraldehyde-3phosphate was prepared from the monobarium salt of the diethylacetal form. Restriction enzymes were purchased from Amersham Biosciences and New England Biolabs.
TS Expression and Purification-The bacterial strain Escherichia coli CB149 carrying plasmid pEBA10 with the gene for the TS complex from Salmonella typhimurium was kindly provided by Dr. Edith W. Miles (National Institutes of Health, Bethesda, MD). The mutant and wild type enzymes were purified as described previously (33) and stored at Ϫ80°C.
Primers contained both the desired mutation (new triplets are in bold) and the silent mutation introducing a new restriction site (underlined) for BglII (␣Gly-181 mutants) and BamHI (␤S178A) in the gene. Sequences were verified by restriction digestion and sequencing.
␤ Replacement Activity Assay-The ␤ activity assay was carried out as described previously (34), using a 1-cm-path length quartz cuvette at 37°C. When present in the assay solution, the concentration of D,L-␣glycerol-3-phosphate (GP) was 50 mM.
␤ Elimination Activity Assay-The serine deaminase reaction was measured by monitoring pyruvate formation with a coupled lactate dehydrogenase assay at 22°C (35,36). The absorbance decrease at 340 nm was recorded using 1-cm-path length cuvettes. The reaction mixture contained 50 mM bicine (sodium N,N-bis(2-hydroxyethyl)glycine), 100 mM L-serine, 500 mM NaCl, 0.2 mM NADH, pH 7.8, and 80 units of lactate dehydrogenase from bovine heart. When present in the assay solution, the concentration of GP was 80 mM.
Spectrophotometric Measurements-Absorption spectra were collected using a Cary400 spectrophotometer (Varian) for a solution containing 50 mM bicine, 500 mM NaCl, and 100 mM L-serine, pH 7.8, in the absence and presence of 80 mM GP at 22°C.
Limited Proteolysis-Solutions containing TS (1.22 mg/ml), 50 mM bicine, and 1 mM EDTA, at pH 7.8, were treated with 1 g/ml TPCKtrypsin in the presence and absence of 80 mM GP and 100 mM L-serine at 22°C (37). Digestions were stopped by boiling solutions for 3 min in the presence of SDS-PAGE loading buffer. Proteolysis products were resolved by SDS-PAGE, visualized by Coomassie Blue staining, and quantified by densitometric analysis carried out on digitalized images with Quantity One 4.2.0 Software (Bio-Rad).

Reverse ␣ and ␤ Replacement and ␤ Elimination Activities-
The rate of the reverse ␣ reaction between indole and glyceradehyde-3-phosphate was found to be almost 2-3-fold slower for the ␣ mutants and ␤S178P with respect to the wild type enzyme (Table I). Not surprisingly, the ␤S178A mutant exhibited only a 20% decrease. These findings suggest that mutations might prevent the formation of the closed, active conformation of the ␣ subunit. The ␤ replacement activity for each mutant was measured in the absence and presence of GP (Table I). All mutants are as active as the wild type, with the exception of ␤S178P and ␣G181P, which show slightly reduced ␤ activity. GP exhibits a significant inhibitory effect on ␤S178A and a much smaller one on ␤S178P, whereas no inhibition is observed for all other mutants. These data indicate that the ligandmediated intersubunit communication has been interrupted, suggesting that mutations do not allow the formation of the hydrogen bond between positions ␤178 and ␣181. In addition, these data confirm that intersubunit activation exploits a different pathway. In fact, ␣G181A, ␣G181F, and ␣G181V mutants exhibit unaltered catalytic activity despite the completely abolished ligand-induced allosteric regulation. The 1.4-and 2-fold decrease in the ␤ activity observed for ␣G181P and ␤S178P, respectively, might be ascribed to slightly altered intersubunit contacts due to different conformations imposed by this mutation to ␣ loop 6 and/or ␤ helix 6. This decrease in activity is very modest when compared with the 30-and 100fold reduction observed for the isolated ␣and ␤-subunits, respectively (12).
In addition to the specific ␤ replacement reaction, tryptophan synthase catalyzes a ␤ elimination side reaction, in which L-serine deamination yields pyruvic acid and ammonia (12). The deaminase activity of wild type TS and mutants in the absence and presence of GP (Table I) indicates that the mutation causes a decrease of activity only for ␤S178P. Furthermore, addition of GP significantly reduces the deaminase activity of the wild type and ␤S178A mutant, whereas it has no effect on ␣181 mutants. For ␤S178P, a 30% reduction in the specific activity was observed in the presence of the allosteric effector.
Equilibrium Distribution between External Aldimine and ␣-Aminoacrylate in the Absence and Presence of GP-The UVvisible spectra of the internal aldimine for wild type TS and mutants (Fig. 2) show, for some mutants, the presence of a band at 325 nm, which is likely to be attributed to a derivative of the coenzyme with active site residues. This band also appears with aging in pyridoxal 5Ј-phosphate-dependent enzymes (38). When L-serine reacts with TS, an equilibrium between the catalytic intermediates ␣-aminoacrylate and external aldimine is attained, with the former being favored by ␣-subunit ligands, cesium ions, low pH, and high temperature (39 -41). These  Fig. 2. In the absence of GP, spectra obtained for the wild type and mutants are similar, except for ␣G181P and ␤S178P. In the former, the mutation favors the accumulation of the external aldimine, whereas in the latter, the mutation stabilizes the aminoacrylate species. Furthermore, differences in the relative absorption intensities for the bands at 422, 350, and 325 nm also arise from distinct amounts of the 325 nm species. Addition of GP to the ␤S178A mutant shifts the equilibrium toward the aminoacrylate species, whereas in ␤S178P and ␣181 mutants, the distribution is unaffected by GP binding. The same pattern of effects was obtained by monitoring the change in the amount of the external aldimine as a function of GP recording the fluorescence emission at 500 nm upon excitation at 420 nm (data not shown).
Effect of ␣and ␤-Subunit Ligands on the Rate of Limited Proteolysis for Wild Type TS and Mutants-In order to detect changes in the open to closed transition of ␣ loop 6 upon mutation, the rate of tryptic cleavage of the ␣-subunit in the ␣ 2 ␤ 2 complex of wild type and mutant TS was measured in the absence and presence of ␣and ␤-subunit ligands. The time course of proteolysis was analyzed by SDS-PAGE (Fig. 3A) and quantified by densitometric analysis (Fig. 3B). Proteolysis of the wild type and ␤S178A mutant in the absence of ligands results in rapid and almost complete cleavage of the ␣-subunit at a single site, ␣Arg-188, to produce ␣-1 and ␣-2 fragments (34,42,43). Binding of GP to the ␣-subunit dramatically decreases the rate of proteolysis, as observed previously (43). Addition of L-serine in combination with GP almost totally protects the ␣-subunit, whereas L-serine alone has no effect (37,43). On the contrary, ␣G181A and ␣G181P mutants are less susceptible to proteolysis in the absence of ligands with respect to the wild type. However, a more rapid and extensive degradation occurs when ␣ and ␤ ligands are bound. The proteolysis site in ␣G181F and ␣G181V mutants remains highly accessible in both the absence and presence of bound substrates. The ␣-subunit proteolysis rate of the ␤S178P mutant in the absence of L-serine and GP is slower than that of the wild type and is substantially unchanged in the presence of GP alone, whereas the addition of both ligands slightly reduces it. For all ␣ mutants, the addition of L-serine alone makes ␣Arg-188 more susceptible to cleavage, suggesting that binding of the ␤ ligand may stabilize a more flexible and mobile conformation of ␣ loop 6.

DISCUSSION
The characterization of mutants at the ␣181 and ␤178 position of S. typhimurium TS allows detailed analysis of the mechanism of intersubunit communication. In the proposed mecha-TABLE I Reverse ␣, ␤ replacement and ␤ elimination activity of wild type and mutant TS in the absence and presence of the ␣ ligand GP Determination of the reverse ␣ activity was performed at 37°C using 1-mm-path length quartz cuvettes in 100 mM BTP, 100 mM NaCl, 2 mM indole, and 2 mM glyceraldehyde-3-phosphate, pH 7.8 (17). The ␤ activity assay was carried out as described previously (34) at 37°C. The ␤ elimination activity was measured in a coupled assay with lactate dehydrogenase, according to Crawford and Ito (35), in 50 mM bicine, 100 mM L-serine, 500 mM NaCl, and 0.2 mM NADH, pH 7.8, at 22°C. One unit of specific activity is defined as the amount of enzyme that yields 0.1 mol of product in 20 min. The percentage of activity, reported in parentheses, is in reference to the wild type in columns 1, 2, and 4 and to the enzyme activity in the absence of GP in columns 3 and 5.  nism of allosteric regulation between ␣and ␤-subunits (5, 7), ␣ ligands interact with the ␣ loop 6 residue ␣Gly-184, reducing the flexibility of the loop and favoring the formation of the hydrogen bond between ␣Gly-181 and ␤Ser-178. In turn, this hydrogen bond further reduces ␣ loop 6 mobility, stabilizing the closed conformation of the ␣-subunit. The relocation of ␣ loop 6 causes a shift in the ␤ helix 6 of the COMM domain (5). The movement of ␣ loop 6 is also propagated to ␣Asp-60 via the hydrogen bond with ␣Thr-183 (5,15). This displacement orients ␣Asp-60 for the efficient catalytic cleavage of indole-3glycerol phosphate and concomitantly allows the formation of new interactions between ␣ loop 2 and the COMM domain (5).
The present data support the hypothesis that the intersubunit hydrogen bond between the ␣Gly-181 amide proton and the ␤Ser-178 carbonyl oxygen atom is fundamental for the ␣-␤ communication and that the loss of this contact is sufficient to abolish intersubunit allosteric regulation. In fact, when the interaction cannot be formed due to either lack of the donor amide proton (␣G181P) or steric hindrance (␣G181A, ␣G181V, and ␣G181F), the ␣ ligands lose the ability to modulate ␤ activity and the equilibrium distribution of catalytic intermediates at the ␤-active site. The proteolysis experiments demonstrate that all the allosterically silent mutants fail to attain the closed conformation of ␣ loop 6 even in the presence of ␣ and ␤ ligands. These findings agree with the previously determined crystallographic structures of the ␤S178P (17) and ␣T183V mutants (30), which established that the lack of ␣ loop 6 closure is the basis for the missing intersubunit signaling. When the equilibrium between closed and open conformations of the ␣-active site is shifted in favor of the open state (5,44), the ␣ activity is low, ␣Gly-181 cannot be correctly positioned for hydrogen bonding with ␤Ser-178, the hydrogen bond cannot be formed, and the signal of ␣-active site occupancy cannot be transduced to the ␤-subunit. The loss of influence of ␣-subunit ligands on the ␤-subunit was previously reported for other mutants of ␣ loop 6 (␣T183A, ␣T183V, and ␣R179L) (21,30,45). The mutated residues are involved in the ordering and maintenance of the correct position of ␣ loop 6. Thus, these amino acid substitutions disrupt the stabilizing hydrogen bond network fundamental for loop relocation and formation of the ␣Gly-181-␤Ser-178 interaction.
When no ligands are present, ␣ loop 6 in the wild type and ␤S178A mutant is flexible and mobile, and ␣Arg-188 is highly susceptible to proteolysis. Binding of GP to the wild type ␣-active site alters the conformation and flexibility of the loop and reduces the accessibility of the site of proteolysis. The closure of the loop caused by binding of GP can be further stabilized by an L-serine-induced conformational change in the ␤-subunit. This highlights that communication between ␣ loop 6 and the ␤-active site is reciprocal. Our data indicate that the ␤S178A mutant retains ligand-induced communication between the ␣and the ␤-site. On the contrary, Phe and Val mutations at position ␣181 prevent ␣ loop 6 from attaining the closed conformation in the presence of bound ligands. In ␣G181A and ␣G181P, ␣ loop 6 in the absence of ligands likely adopts a different conformation, characterized by an impeded interaction with trypsin. Upon binding of either GP alone or GP and L-serine, the ␣ loop 6 of these mutants is not in the closed state, and its conformation seems to resemble the open state. Furthermore, the results confirm that no conformational changes in the ␣ loop 6 of ␤S178P occur after ␣ ligand binding, as evidenced by the crystal structure of the ␤S178P mutant, in which ␣ loop 6 has no definite electron density in the absence and presence of ␣ ligands (17). In addition, these experiments suggest that, similarly to the ␣G181P and ␣G181A mutants, the conformation of ␣ loop 6 in ␤S178P is endowed with different flexibility with respect to the open state.
It is important to notice that all mutants, which had lost ligand-induced allosteric regulation, were almost as active as wild type TS in the ␤ reaction, with the unique exception of S178P and ␣G181P. In these mutants, the substitution with proline is likely to distort the ␣ loop 6 and ␤ helix 6 conformation and modify contacts at the interface, as shown crystallographically for the former (17). In contrast, replacement of residues in ␣ loop 2 or residues contacting ␣ loop 2 was reported to critically reduce ␤ activity without perturbing ligand-induced communication (24,27,29). Consequently, ␣ loop 2 was proposed to be predominantly involved in the intersubunit interaction that leads to a reciprocal modulation of the activity of the ␣and ␤-subunits within the tetramer (7). These findings collectively indicate that different pathways for intersubunit signal transmission do exist, which specifically involve ␣ loop 2 and ␣ loop 6. This conclusion is also supported by the recently determined structure of the TS ␣-subunit from E. coli (31).
Multiple sequence alignment of TS ␣ and ␤ chains from different sources reveals that ␤Ser-178 is not conserved among the analyzed species (7), indicating that the side chain of Lserine is not relevant for allosteric communication and that, at this position, amino acids with different steric hindrance can be accommodated without perturbing the subunit interface. On the contrary, ␣Gly-181 is invariant (32). The invariance of ␣Gly-181 suggests that although the interaction is formed between backbone atoms, precise steric constraints exist due to the limited space to accommodate the ␣181 side chain in the closed form of ␣ loop 6. We verified that serine to alanine substitution at position ␤178, which is not likely to alter the position of the carbonyl oxygen, does not perturb hydrogen bond formation and does not interrupt signal transmission. On the contrary, the results on ␣181 mutants show that glycine at position ␣181 is necessary for the maintenance of signal transmission, suggesting that only this residue allows ␣ loop 6 closure and the correct positioning of the ␣G181 amide proton for formation of the hydrogen bond. The geometric restrictions imposed at position ␣181 were simulated by "in silico" mutagenesis. From inspection of these virtual structures, it appears that the closed conformation of ␣ loop 6 is incompatible with the presence of side chains bulkier than a hydrogen atom. A detailed molecular dynamics modeling of the mobility of ␣ loop 6 in the wild type and ␣181 mutants in the absence and presence of ␣-subunit ligands is currently under way and will be reported elsewhere. 2