Structural Basis for the Impaired Channeling and Allosteric Inter-subunit Communication in the βA169L/βC170W Mutant of Tryptophan Synthase*

We determined the 2.25 Å resolution crystal structure of the βA169L/βC170W mutant form of the tryptophan synthase α2β2 complex fromSalmonella typhimurium complexed with the α-active site substrate analogue 5-fluoro-indole-propanol-phosphate to identify the structural basis for the changed kinetic properties of the mutant (Anderson, K. S., Kim, A. Y., Quillen, J. M., Sayers, E., Yang, X. J., and Miles, E. W. (1995) J. Biol. Chem. 270, 29936–29944). Comparison with the wild-type enzyme showed that the βTrp170 side chain occludes the tunnel connecting the α- and β-active sites, explaining the accumulation of the intermediate indole during a single enzyme turnover. To prevent a steric clash between βLeu169 and βGly135, located in the β-sheet of the COMM (communication) domain (βGly102-βGly189), the latter reorganizes. The changed COMM domain conformation results in a loss of the hydrogen bonding networks between the α- and β-active sites, explaining the poor activation of the α-reaction upon formation of the aminoacrylate complex at the β-active site. The 100-fold reduced affinity for serine seems to result from a movement of βAsp305 away from the β-active site so that it cannot interact with the hydroxyl group of a pyridoxal phosphate-bound serine. The proposed structural dissection of the effects of each single mutation in the βA169L/βC170W mutant would explain the very different kinetics of this mutant and βC170F.

Tryptophan synthase (TRPS) 1 is a bifunctional tetrameric enzyme with a linear ␣␤␤␣ architecture that catalyzes the last two steps in the biosynthesis of L-tryptophan. In the ␣-subunit indole-glycerol-phosphate (IGP) is cleaved to indole and glyceraldehyde-3-phosphate (␣-reaction; for reviews see Refs. [1][2][3]. At the ␤-active site, the second substrate L-serine binds to and is then activated by a pyridoxal phosphate (PLP) cofactor to form a highly reactive aminoacrylate intermediate, which reacts with indole to form the product L-tryptophan (␤-reaction). The failure to detect indole even in rapid chemical quench experiments (4) and the existence of a ϳ25 Å-long tunnel that connects the ␣and ␤-active sites (5) strongly suggests that indole is transferred through this tunnel from the active site in the ␣-subunit to the one in the ␤-subunit. This process, the direct transfer of an intermediate or metabolite between two sequential enzymes without free diffusion through the solvent, is known as substrate channeling. Channeling is thought to play an important role in metabolite regulation and cellular modulation of enzymatic activities (1,6,7). In the case of tryptophan synthase, it prevents loss of the intermediate indole through the cell membranes.
An important requirement for enzymes exhibiting substrate channeling is the tight regulation of the two coupled reactions so that they remain in phase. In tryptophan synthase, synchronization is achieved first by the influence of the ␣-site ligand on the affinity for serine and the distribution of the intermediates formed at the ␤-active site upon binding and subsequent reaction of serine and the PLP cofactor (8) (for review see Ref. 2). Second, the IGP cleavage rate is increased ϳ150-fold upon formation of the aminoacrylate at the ␤-active site (4). Indole does not accumulate in the tunnel or at the ␤-active site, as the transfer of indole through the tunnel is fast (Ͼ1000 s Ϫ1 ) and the reaction with the aminoacrylate is largely irreversible. This kinetic model for efficient channeling predicts that obstruction or occlusion of the tunnel should lead to an accumulation of indole such that it may be detected in single-turnover rapid quench experiments. To test this hypothesis, ␤Cys 170 , which lines the tunnel, was mutated to phenylalanine or tryptophan (9,10). The latter mutant also contained the substitution ␤A169L, which introduces a restriction enzyme site for convenient verification of the mutation. The low resolution crystal structure of the double mutant shows that the introduced tryptophan, ␤Trp 170 , obstructs the tunnel (11). In agreement with the kinetic model, indole can indeed be detected in the two ␤Cys 170 mutants by single-turnover chemical quench-flow experiments (10). The rate constant for channeling, which is Ͼ1000 s Ϫ1 for wild-type TRPS, is 100 s Ϫ1 for ␤C170F and 0.2 s Ϫ1 for ␤A169L/␤C170W (Table I). It is intriguing that this difference is related to the size of the obstructing residue at position ␤170. In contrast, it was rather unexpected to find that the activation of the ␣-reaction by the aminoacrylate is the same in the wild-type and the ␤C170F mutant, whereas it is reduced 100-fold in the ␤A169L/␤C170W mutant (Table I). This observation was interpreted as an impairment of the ␣ 7 ␤ subunit communication. It was speculated that the activation of IGP cleavage at the ␣-site upon formation of the aminoacrylate complex at the ␤-active site is prevented by the presence of the indole ring of ␤Trp 170 , which may mimic an indole intermediate in the tunnel (10).
Recently, we deduced a structural model for the allosteric ␣ 7 ␤ subunit communication (12). Briefly, the substrate (or substrate analog) binding to the ␣-active site induces an ordering of the catalytically important loops ␣L2 and ␣L6 and a concerted rearrangement of both gating residues ␤Tyr 279 and ␤Phe 280 (9,13,14), resulting in tunnel blockage and thus preventing untimely passage of indole. The ordering of a rigid but movable domain (␤Gly 102 -␤Gly 189 ) is concomitant. We named it the "COMM domain" because it is central to the ␣ 7 ␤-active site communication. Rhee et al. (15) introduced the domain ␤Gly 93 -␤Gly 189 as the "mobile region." This COMM domain interacts with residues of ␣-subunit loops ␣L2 and ␣L6, ␤-subunit residues including ␤Asp 305 , which is involved in binding of the metal ion, and active site residues ␤Glu 109 and ␤Gln 114 . Changes on either active site disrupt the respective interactions with the COMM domain, which results in a rigid body movement of the COMM domain, thereby changing the interactions at the other active site and thus transmitting the signal.
We decided to revisit the ␤A169L/␤C170W mutant to see whether our model provides the structural basis for the changed kinetic properties of the mutant. Our previous study of the ␤A169L/␤C170W mutant was hampered by the low resolution (3.2 Å) of the structure determined using x-rays from a rotating anode. Therefore, we redetermined the structure of the mutant complexed with the noncleavable IGP analogue 5-fluoro-indole propanol phosphate (F-IPP) to 2.25 Å resolution using synchrotron radiation. The comparison of the kinetic properties of the single mutant ␤C170F and the double mutant ␤A169L/␤C170W, together with the comparison of the structures of the latter and the wild-type enzyme, suggests the reason for the impaired inter-subunit communication being the ␤A169L mutation, whereas the ␤C170W replacement seems to be responsible for the modified channeling characteristics.

MATERIALS AND METHODS
The ␤A169L/␤C170W mutant of TRPS was purified (9,16) and crystallized (11) as described previously. Diffraction data were collected of a crystal mounted in a capillary and kept at 4°C at beam-line X12C at the National Synchrotron Light Source using a MAR Research image plate detector. The data were integrated and scaled with the HKL suite of programs (17). Refinement to 2.5 Å resolution was started with CNS 0.9a (18) by performing rigid body and simulated-annealing steps.
The coordinates of the wild-type TRPS IPP complex (PDB code 1Q0P) were used as a starting model, omitting the coordinates of loops ␣L2 and ␣L6, IPP, the cofactor PLP, and all water molecules. The mutated residues were modeled as alanines to avoid model bias. The final model was built by cyclic rounds of manual model building with the program "O" (19) and Maximum Likelihood refinement with the program REFMAC (20) using all reflections from 20 to 2.25 Å. Water molecules were incorporated by ARP (21) using the automatic cut-off option. All waters were checked manually and removed if displaying unusual H-bonding geometry. Apart from residues at the C termini of both polypeptide chains, the only part of the ␤A169L/ ␤C170W F-IPP complex that remains too disordered to be built into electron density is the C-terminal part of loop ␣L6 (␣Arg 188 -␣Pro 192 ). As shown in Fig. 1A, two regions in the ␤-subunit show disproportionately high B-factors (Ͼ2 ϫ ϽBϾ ϭ 72 Å 2 , mean value ϽBϾ ϭ 36 Å 2 ): the first residues (␤Ala 136 -␤Ser 143 ) of helix ␤H5 and the residues (␤His 160 -␤Als 164 ) preceding helix ␤H6 (secondary structure definition according to Ref. 12). Because both amino acid stretches belong to the COMM domain and may have structural and/or communicative functions, we retained these coordinates in the final model, although the final Sigma A-weighted 2 mF o Ϫ DF c (20) and a composed-omit (18) maps show only very weak electron density. The reason for this high flexibility and/or bad model definition is discussed below.
The final model consists of 4960 protein atoms, the ␣-ligand F-IPP, the cofactor PLP, and 190 water molecules. The Rfactors are R ϭ 17.0% and R free ϭ 22.4%. Data and refinement statistics are given in Table II. The coordinates and structure factor amplitudes have been deposited with the PDB (accession code 1FUY).

RESULTS AND DISCUSSION
The structure of the double mutant ␤A169L/␤C170W is of high quality as indicated by the low R-factors and the good stereochemistry (see Table II). The side chain conformations of the mutated amino acids are very well defined by the electron density ( Fig. 2A). The mutant has the same secondary and tertiary structure as the wild-type enzyme, and the overall topology is similar to the one assigned by Schneider et al. (12) that is observed in all TRPS structures. F-IPP binds (not shown) in a depression on the surface of the ␣-subunit and is held in place by several hydrogen bonds to the phosphate at the one end and hydrophobic interactions with the indole moiety at the other end, similar to the wild-type enzyme (12). Following, we compare the structures of the ␤A169L/␤C170W F-IPP complex and the wild-type IPP complex (TRPS IPP , Protein Data Bank (PDB) code 1Q0P) (14). The reason is that the latter was determined to a much higher resolution than the TRPS F-IPP complex (PDB code 1A50) (12), and a comparison of both complexes shows high structural similarity; the superposition of  (Fig. 1B). The low r.m.s. deviation value for the common ␣-ligand atoms indicates a high similarity of the ␣-ligand binding site and also for the enzyme, excluding the COMM domain; this shows that the additional fluorine atom in F-IPP compared with IPP has no influence on the TRPS structure. In particular, it does not change the COMM domain conformation (12,14). The ␤A169L/␤C170W F-IPP and TRPS IPP structures differ significantly in the conformations of the COMM domains and residues ␤Tyr 279 and ␤Phe 280 , which are located in the middle of the tunnel. Because they have been observed in positions that result in an open (5,12) or closed (12,13) tunnel, respectively, they have been thought of as a molecular gate. As described previously (11), the ␤Trp 170 side chain points into the tunnel, thereby blocking it. In addition, because the bulky side chain of ␤C170W would clash with the ␤Phe 280 side chain, the latter rotates by about 110°in 1 , and the backbone ⌽/⌿ torsion angles change from ⌽/⌿ WT ϭ 63.5°/23.2 to ⌽/⌿ mut ϭ 55.4°/ 36.8°. The new side chain conformation of ␤Phe 280 induces a 1 rotation of about 18°for ␤Tyr 279 (see Fig. 2); leading to a number of new interactions of its hydroxyl group. Of note is the hydrogen bond (2.6 Å) to the oxygen atom O ␦1 of asparagine ␤Asn 171 , which further forms a hydrogen bond (2.7 Å) to aspartate ␣Asp 56 O ␦2 . This latter interaction is also found in wildtype TRPS structures (12,14), but the interaction mediated by the asparagine ␤Asn 171 between the side chain of a gating residue and a loop ␣Leu 2 residue is seen only in the ␤A169L/ ␤C170W mutant structure. In addition, the new conformation of the ␤Tyr 279 side chain results in an unfavorable contact between its hydroxyl group and the C ␤ carbon atom of aspartate ␣Asp 56 (2.7 Å). These interactions may influence the communication properties of the ␤A169L/␤C170W mutant, because ␣Asp 56 and ␤Asn 171 are involved in the inter-subunit communication pathway (12).
The main effect of the ␤C170W mutation on the gating residues ␤Tyr 279 and ␤Phe 280 is a loss of rotational freedom. In particular, phenylalanine ␤Phe 280 is trapped in a hydrophobic pocket built of side chains from ␣Leu 58 , ␤Trp 170 , ␤Leu 174 , ␤Tyr 279 , ␤Met 282 , ␤Met 286 , ␤Phe 306 , and ␤Phe 307 . Although the conformations of ␤Tyr 279 and ␤Phe 280 correspond to an open gate in the ␤A169L/␤C170W F-IPP complex, the tunnel is occluded by the indolyl side chain of ␤Trp 170 .
Several COMM domain regions have significantly different conformations in the ␤A169L/␤C170W F-IPP complex compared with the wild type (Figs. 1B and 2B). The changes are caused primarily by the introduction of leucine ␤Leu 169 , because its side chain would be too close to the adjacent amide nitrogen atom of ␤Gly 135 (distance 1.6 Å) and the C ␣ of ␤Met 134 (distance 2.5 Å) located at the end of strand ␤S4. Their movement prevents this backbone clash and results in a domino effect-like rearrangement of the COMM domain; because strand ␤S4 is one of the middle ␤-strands of the four-stranded parallel COMM domain ␤-sheet and forms hydrogen bonds to residues of both neighboring strands ␤S3 and ␤S5, these strands and the N-terminal part of ␤-strand ␤S6 also move. The shift of all four COMM domain ␤-strands is conveyed by the movement of helix ␤H5 and loops ␤L3 and ␤L5. The maximum C␣ atom shift of ϳ3.5 Å is found at lysine ␤Lys 137 . It is noteworthy that the catalytically important glutamate ␤Glu 109 (4, 22) moves away from the ␤-active site with its carboxylate carbon atom shifting by ϳ2.1 Å.
A further consequence of the changed COMM domain conformation in the ␤A169L/␤C170W F-IPP complex is found in the hydrogen bonding network involving aspartate ␤Asp 305 , which is positioned near the sodium binding site (13) and involved in the ␤-elimination reaction yielding the external aldimine (15,23). Although the side chain of aspartate ␤Asp 305 points toward the ␤-active site ("swing in" position (13)) in both the TRPS IPP (14) and the ␤A169L/␤C170W F-IPP complex, there are significant differences. In the wild type, the distance between the carboxylate oxygen O ␦2 of ␤Asp 305 and the amide nitrogen of ␤Gly 111 is 3.4 Å, and the ␤Asp 305 carboxylate is linked via one or two water molecules to the COMM domain residues ␤Asp 138 , ␤Gln 142 , and ␤Thr 165 , respectively. This network is broken in the ␤A169L/␤C170W F-IPP complex because the respective COMM domain residues are shifted from the core of the ␤-subunit due to the introduction of the leucine ␤Leu 169 as described above (distance between O ␦2 of ␤Asp 305 and the amide of ␤Gly 111 is 6.1 Å). This results in a loss of all water molecules involved in the hydrogen bonding network between ␤Asp 305 and the COMM domain. Although we cannot exclude the possibility that the nonobservation of the bridging water molecules is due to the lower resolution of the diffraction data, careful analysis of the final mF o Ϫ DF c electron density map at a very low cut-off shows no indication for bound water molecules in this region. In contrast, the final 2 mF o Ϫ DF c map shows clear backbone and side chain conformations for ␤Asp 305 with ⌽/⌿ torsion angles of Ϫ91.2°/16.5°compared with Ϫ90.8°/Ϫ6.2°in the TRPS IPP structure, indicating a small but significant conformational change because of the missing water contacts. Despite the changed conformation of ␤Asp 305 , the ␤A169L/ ␤C170W F-IPP complex contains a sodium ion. This fact can be attributed to the altered side chain conformation of the gating residues, because the side chain conformation of ␤Phe 280 stabilizes the metal binding loop through hydrophobic interactions with the side chain rings of ␤Phe 306 and proline ␤Pro 307 . Based on the structure of the ␤A169L/␤C170L F-IPP complex presented here, one can discuss the kinetic properties of both ␤Cys 170 mutants (see Table I) (4,10). Both should have a more or less blocked tunnel, because the side chain of residue ␤170 points into the tunnel. A manually built ␤C170F model (figure not shown) based on the TRPS IPP structure indicates that introduction of the phenylalanine side chain retains a larger conformational freedom of the tunnel residues ␤Phe 170 , ␤Tyr 279 , and ␤Phe 280 compared with the double mutant. This result is in line with the single turnover experiments on the ␤C170F mutant that showed, in contrast to the ␤A169L/ ␤C170W mutant, only little indole accumulation in the ␣␤ reaction (10). Introduction of the bulky tryptophan residue at the ␤Cys 170 position effects only the side chain conformations of the "gating residues" ␤Tyr 279 and ␤Phe 280 . The rest of the ␤-subunit, in particular the COMM domain conformation, seems not to be affected by the ␤C170W mutation. Therefore, we postulate an unchanged wild type-like COMM domain conformation in the ␤C170F F-IPP or IPP complex; this would be in line with the observed turnover number of the ␣-reaction and the activation of the ␣-reaction by L-serine, neither of which changes upon the ␤C170F mutation. Furthermore, the ␤C170F mutant has the same L-serine affinity as wild-type TRPS ( Table  I). The only (small) kinetic differences between the wild-type and the ␤C170F mutant exist in the ␤and the combined ␣/␤-reactions. Both require indole transport to the ␤-active site, but this would be hindered (and is consistent with our x-ray structure presented here) by the ␤C170F or ␤C170W mutation.
There are, however, striking differences between the ␤C170F and the ␤A169L/␤C170W mutant in the activation of the ␣-reaction upon formation of the aminoacrylate complex at the ␤-active site and in their affinity for serine (Table I). These must be caused (given our assumption that the COMM domain conformation is largely unaffected by the ␤C170 mutation) by the introduction of a leucine at position ␤169. As described in detail above, introduction of ␤Leu 169 sterically induces a major rearrangement of the COMM domain; this results in an opening of the ␤-subunit, a withdrawal of catalytically important residues such as ␤Glu 109 from the active site of the ␤-subunit, and a disruption of hydrogen bonding networks that connect the ␣and ␤-active sites or stabilize conformations of catalytically important residues such as ␤Asp 305 (15). The latter is stabilized in the wild-type TRPS IPP complex by a number of water molecules in a conformation in which its carboxylate can interact with the hydroxyl group of the PLP-bound serine (external aldimine). It is interesting to note that the backbone conformation of ␤Asp 305 is the same in the ␤A169L/␤C170W F-IPP complex and the wild-type aminoacrylate TRPS F-IPP complex (12). In this conformation, the ␤Asp 305 carboxylate is further away from PLP and cannot stabilize the bound serine, which is most  (24) and RASTER3D (25,26). B, stereo view of the superposition of the COMM domains of wild-type TRPS IPP and ␤A169L/␤C170W F-IPP . Mutated and gating residues of the ␤-subunit are shown in a ball-and-stick representation. C␣ trace and residues of wild-type TRPS IPP are red (␣-subunit), dark blue, (␤-subunit), and yellow (COMM domain, the double mutant is green, carbon atoms of the ␣-ligands and the cofactor PLP are gray, oxygen atoms red, nitrogen atoms blue, and phosphate is magenta. Panels A and B are related by an ϳ90°r otation around the axis perpendicular to the paper plane. The figure was prepared using MOLSCRIPT (27) and RASTER3D (25,26). likely the reason for the 100-fold lower L-serine affinity of the ␤A169L/␤C170W mutant.
In conclusion, the impaired channeling of the ␤A169L/ ␤C170W mutant is caused by a complete occlusion of the tunnel by the ␤Trp 170 side chain (despite a stabilization of the open conformation of the gating residues ␤Tyr 279 and ␤Phe 280 ). We predict that the situation is similar in the ␤C170F mutant. The remarkable differences in the kinetic parameters of the ␣␤reaction in both of the ␤Cys 170 mutants, in particular the lack of activation of the ␣-reaction upon formation of the aminoacrylate intermediate at the ␤-active site in the double mutant and the reduced serine affinity, seems to be caused mainly by the ␤A169L substitution. The longer ␤Leu 169 side chain induces a rearrangement of the COMM domain. This affects the ␣ 7 ␤ subunit communication as it is mediated by rigid body displacements of the COMM domain. In particular, the prerequisite for activation of the ␣-reaction is the closing of the ␤-subunit by a COMM domain movement. In addition to sterically preventing the closing of the COMM domain, the introduction of ␤Leu 169 changes the COMM domain conformation such that the water molecules that hold ␤Asp 305 in a position where it can interact with the hydroxyl group of the PLP-bound serine are destabilized. In the ␤A169L/␤C170W mutant, ␤D305 is shifted away from PLP, which explains the 100-fold reduction in serine affinity.