The ability to synthesize L-Trp (1) is essential for many organisms including enteric bacteria, yeasts, molds, and plants. L-Trp synthesis is tightly regulated in enteric bacteria, and this control begins at the level of gene expression (Yanofsky, 1955; Yanofsky, 1981; Yanofsky, 1987). In these bacteria, the trp operon is regulated by the in vivo concentration of L-Trp. Binding of L-Trp to the trp repressor protein stabilizes the formation of the trp operon-trp repressor complex, effectively blocking expression of the operon and hence shutting down the synthesis of L-Trp. When the cellular concentration of L-Trp becomes too low to bind to the trp repressor, the trp repressor dissociates from the trp operon, triggering the expression of the five proteins responsible for L-Trp synthesis, trpE, trpD, trpC, trpB, and trpA. The enzymes expressed by trpA and trpB form an αββα bienzyme complex (Figure 1A) that carries out the last two steps in the synthesis of L-Trp (the α- and β-reactions) (Figures 1B,C). This complex is designated here as tryptophan synthase, TS. The TS α-subunit has the canonical (βα)₈ TIM barrel fold (Hyde et al., 1988), first observed in triosephosphate isomerase, while the β-subunit has an unusual fold structure comprised of the COMM domain (β102—β189; Figures 1A, 2) (Schneider et al., 1998) and a scaffolding (β1–β101 and β190—β395) that supports the COMM domain and contributes residues to the 25–30 Å long tunnel that connects the α- and β-sites (Figures 3A,B). While Kirschner et al. (1975) were the first to report evidence for ligand induced interactions between the α- and β-sites, Drewe and Dunn (1986) were the first to explicitly recognize that tryptophan synthase catalysis of L-Trp synthesis is subject to control via ligand mediated allosteric interactions transmitted between the α- and β-subunits.

At this juncture, structural and mechanistic investigations of TS-allostery show that the TS α₂β₂ bienzyme complex is a multifaceted, highly nuanced molecular machine. This machine utilizes substrate channeling, pyridoxal phosphate, a monovalent cation effector, a hydrophobic nanotube and heterotropic allosteric site-site interactions, to regulate and synchronize the activities of the αβ-subunit pairs to achieve the efficient synthesis of L-tryptophan.

In Escherichia coli and Salmonella enterica serovar Typhimurium, assembly of the α and β subunits into the α₂β₂ TS complex is essential for full activity (Yanofsky and Crawford, 1972; Creighton, 1970; Miles, 1979; Bahar and Jernigan, 1999; Miles, 2001; Raboni et al., 2009; Kulik et al., 2005; Dunn, 2012; Barends et al., 2008a; Miles, 2013; Maria-Solano et al., 2019; Sakhrani et al., 2020; Buller et al., 2015). The isolated α-subunit shows a catalytic activity diminished by ∼ 1,000- to 3,000-fold while compared to the holo α₂β₂ complex the activity of the PLP-requiring β₂ dimer, is diminished by ∼100 fold (Yanofsky and Crawford, 1972; Miles, 1979; Miles, 2001; Kulik et al., 2005; Barends et al., 2008a; Dunn, 2012; Miles, 2013; Buller et al., 2015). The α-subunits catalyze the cleavage of substrate 3-indole D-glyceraldehyde 3′-phosphate (IGP), yielding indole and D-glyceraldehyde 3-phosphate (G3P, 6) (Figure 1B, the α-reaction) Yanofsky and Crawford, 1972; Miles, 1979; Miles, 2001; Barends et al., 2008a; Kulik et al., 2002), while the β-subunits catalyze replacement of the L-Ser (7) β-hydroxyl by indole yielding L-Trp and a water molecule (Figure 1C, the β-reaction) (Yanofsky and Crawford, 1972; Miles, 1979; Drewe, and Dunn, 1985, 1986; Dunn, 2012; Miles, 2013). During the catalytic cycle, indole formed at the α-site is transferred to the β-site via the 25 Å-long, tunnel (Creighton, 1970; Yanofsky and Crawford, 1972; Matchett, 1974; Miles, 1979; Hyde et al., 1988; Dunn et al., 1990; Brzovic P. S. et al., 1992; Miles, 2001; Raboni et al., 2009; Miles, 2013; Hilario et al., 2016; Teixeira et al., 2020). This channeling of the common intermediate, indole, is a key feature of the allosteric control mechanism for the synthesis of L-Trp (Dunn et al., 1990; Miles, 2001; Raboni, et al., 2009; Dunn, 2012; Miles, 2013; Hilario et al., 2016; Teixeira et al., 2020). Within α₂β₂, αβ dimers form allosteric units that work independently of each other while the allosteric interactions within each αβ unit are essential to the efficient synthesis of L-Trp (Dunn, 2012; Niks et al., 2013; Ghosh et al., 2021).

The first x-ray structures of TS were reported by Hyde et al. (1988) (viz. Figures 1A, 2). During the past 20 years a relatively large number of TS structures have been deposited in the protein data bank (PDB) (Berman et al., 2000). Most of these are structures of the Salmonella typhimurium bienzyme in complex with substrates, and with a variety of substrate analogues and/or covalent intermediates bound to the α- and β-sites (viz. Figure 2). These structures provide a rich source of information relevant to 1) the chemical structures of intermediates, 2) the reaction mechanisms within the catalytic cycles of the α- and β-reactions (Figures 1B,C), and 3) the structures of the conformational states (Figure 2) and allosteric transitions of the α- and β-subunits during these reaction cycles. The TS literature is replete with physical biochemical investigations of the α- and β-reactions (Yanofsky and Crawford, 1972; Miles, 1979; Miles, 2001; Raboni et al., 2009; Dunn, 2012; Miles, 2013). Recent publications include a variety of solution UV/Vis rapid kinetic studies (Harris et al., 2005; Ngo et al., 2007a, b; Dierkers et al., 2009; Ghosh et al., 2021; Phillips and Harris, 2021) and NMR studies employing ligands with ¹⁹F (Niks et al., 2013) and ¹⁷O probes (Young et al., 2016). Solid state magic angle spinning ssNMR investigations of TS complexes of substrate and substrate analogues and with ¹³C and ¹⁵N enriched PLP (McDowell et al., 1995; Lai et al., 2011; Caulkins et al., 2014, 2016; Holmes et al., 2022), and β-subunits enriched with ¹⁵N-Lys residues (Caulkins et al., 2014) and molecular dynamics studies (Huang et al., 2016; Maria-Solano et al., 2019; O’Rourke et al., 2019) have become especially important for establishing the predominant protonation states and tautomeric states of PLP intermediates in the β-reaction.

Catalysis at the α-site involves formation of a complex wherein the reacting substrate, IGP, undergoes a reverse aldolytic cleavage reaction catalyzed by two acid-base catalytic groups, αGlu49 and αAsp60 (Figure 3A). The pioneering site-directed mutagenesis work by Edith Miles provided the first evidence that the conversion of indole into an effective nucleophile requires the coupling of a charge-stabilizing interaction between a side chain carboxylate and the indole N-1, both in the α-reaction (αAsp60) and in the β-reaction (βGlu109) (Figures 3A,B) (Yutani et al., 1987; Miles et al., 1988; Nagata et al., 1989; Brzovic P. S. et al., 1992). Catalysis at the β-site involves the interconversion of at least nine covalent intermediates (Figure 1B) (Yanofsky and Crawford, 1972; Miles, 1979; Miles, 2001; Raboni et al., 2009; Dunn, 2012; Drewe and Dunn, 1985, 1986; Ngo et al., 2007b; Lai et al., 2011; Phillips and Harris, 2021). Recent mechanistic studies conclusively show that the εNH₂ group of βLys87 plays essential roles in the formation of the internal and external aldimines of substrates L-Ser and L-Trp (Figure 1B ) and provides the acid-base catalysis for all the various proton transfers involved in the interconversion of gem diamines with internal aldimines and external aldimines and with carbanionic intermediates along the catalytic path (Caulkins et al., 2014, 2016; Huang et al., 2016; Holmes et al., 2022). The carboxylate of βGlu109 participates in a columbic charge-charge, H-bonding interaction that stabilizes development of a partial positive charge at the indole ring N-1 nitrogen as the C-C bond is formed between the indole C₃ and the C_β of the α-aminoacrylate intermediate (Figure 3B) (Brzovic P. S. et al., 1992; Holmes et al., 2022). This interaction is essential to the chemical activation of weakly nucleophilic indole for the enamine attack via C₃ (Figure 3B). Notice that, in contrast to the classical view in which carbanionic intermediates are formed and stabilized as quinonoidal species in the catalytic cycles of many PLP-requiring enzymes, the TS PLP ring N remains unprotonated throughout the entire β-reaction cycle and therefore cannot form canonical quinonoid structures (Caulkins et al., 2016; Holmes et al., 2022) (Figure 1C). Nevertheless, carbanionic species with the negative charge delocalized over the C_α, Schiff base N, C_4’, C₄, C₃ and O₃ atoms of the PLP ring scaffolding give intermediates that are quasi-stable (Dunn, 2012; Caulkins et al., 2016) and provide spectroscopic analogues of the transiently formed L-Ser and L-Trp carbanions detected in the β-reaction (Drewe and Dunn, 1985; 1986; Roy et al., 1988a, b; Barends et al., 2008a, b; Caulkins et al., 2016; Holmes et al., 2022; Ghosh et al., 2021; Phillips and Harris, 2021) (organic structures are summarized in Figure 4).

Blumenstein et al. (2007), observed that the mutation of βGln114 to Asn facilitates a side reaction wherein the εNH₂ of βLys87 makes a nucleophilic attack at the PLP C4′ of E(A-A), giving E(Ain) and releasing α-aminoacrylate. This highly nucleophilic three-carbon enamine then reacts to form a new C-C bond with the E(Ain) PLP C4′ carbon (Figure 3C). This reaction progresses to give an inactivated TS derivative with a covalently modified PLP derivative bound to the β^R state (PDB ID: 2J9Y). The α-aminoacrylate side reaction has been reported for several other PLP-dependent enzymes that involve α-aminoacrylate Schiff base intermediates (Roise et al., 1984). It appears that bacterial systems have evolved an enzyme family, RidA, with the primary biological function of deaminating 3- and 4-carbon enamines to prevent similar side reactions from occurring in PLP-dependent enzymes (Flynn and Downs, 2013).

Atomistic MD simulations provide detailed insights into allosteric regulation, allosteric networks, and indole channeling (Fatmi et al., 2009). These MD studies sampled both open and partially closed conformations of the highly flexible α-subunit loop 6 (αL6) in the ligand-free and ligand bound states; the loop shifts to fully closed conformations when α site ligands are present. Post-MD analysis to compute energy and configuration entropy suggests that the fully closed conformations are induced by favorable protein-ligand interactions but are partly offset by configurational entropy loss (Fatmi et al., 2009). The αβ-dimeric unit stabilizes the substrate-protein conformation, which also forms new hydrogen bonds and lowers the conformation transition barrier to facilitate the conformation transition from an open/inactive form to a closed/active form (Fatmi and Chang 2010; O’Rourke et al., 2019). The allosteric motions regulate substrate catalysis, and the MD simulations identified interaction changes across the catalytic cycle of the α-reaction (Spyrakis et al., 2006; D’Amico et al., 2021; Bosken et al., 2022).

Although hydrogen atoms and water molecules may not directly contribute to allosteric regulation, MD and quantum mechanics studies suggest that the protonation states and conserved water molecules affect TS motions which are important for the catalytic process (Huang et al., 2016; Teixeira et al., 2019). Combined MD simulations and ancestral sequence reconstruction identify residues contributing to allosteric signal propagation in TS (Schupfner et al., 2020). In addition to classical MD simulations, enhanced sampling methods, such as steered MD simulation, have been applied to examine indole channeling between the α- and β-subunits to scrutinize interactions between indole and residues lining the channel (Zhang and Lazim, 2019).

Recent literature on the nature of allosteric transitions and protein conformation states hypothesize that the T and R designations for the conformational states of allosteric protein systems as proposed by Monod, Wyman and Changeux and Koshland, Nemethy and Filmer (Monod et al., 1965; Kirschner et al., 1966; Koshland et al., 1966) are insufficiently nuanced and that the discussion of allosteric mechanism should be reframed within the context of protein conformational ensembles that are optimized via evolution to achieve control of biological function (Goodey and Benkovic, 2008; Feher et al., 2014; Motlagh et al., 2014; Nussinov and Tsai, 2014; Wei et al., 2016; Buchenberg et al., 2017; Guarnera and Berezovsky, 2019; Wodak et al., 2019; Verkhivker et al., 2020). Because the TS free energy landscapes are not yet well described by experiments, herein we apply the T and R nomenclature to the TS system while implicitly recognizing that the various noncovalent and covalent complexes detected in the TS system no doubt are comprised of ensembles which exist in equilibria around their native states and that the free energy barriers separating these ensembles comprise the free energy pathways described herein as the α- and β-reactions (Figures 1B,C) (Maria-Solano et al., 2019; Teixeira et al., 2020).

As will be discussed in detail in following sections, the switching between open and closed subunit conformations plays a central role in the TS allosteric regulatory mechanism and is critically important to the efficient synthesis of L-Trp (Dunn et al., 1990; Kulik et al., 2005; Ngo et al., 2007b; Barends et al., 2008a; Lai et al., 2011; Niks et al., 2013; Caulkins et al., 2016; Ghosh et al., 2021). Many x-ray crystal structures of the open and closed states of TS are available in the Protein Data Bank archive (RCSB PDB) (Berman et al., 2000). With the high affinity IGP analogue F9 (8) bound to the α-site, the TS internal aldimine and the L-Ser external aldimine give complexes with closed α-subunits and open β-subunits (Niks et al., 2013; Ghosh et al., 2021) (viz, Figure 2). The F9 complexes with the α-aminoacrylate and carbanion intermediates almost always give structures where both subunits assume closed conformations (Ghosh et al., 2021). ¹⁹F NMR studies of the F9 complexes indicate these subunit conformations are also the predominate forms in solution (Niks et al., 2013). Our current working model for the TS allosteric mechanism is shown in Figure 2.

Mutations in the β-subunit can reverse the relative stabilities of β-subunit allosteric states. For example, the x-ray structure of the βGln114Ala mutant with α-aminoacrylate bound to the β-site and without ligand bound to the α-site shows an α^Tβ^T complex (PDB ID: 7KQ9) and the structures of the E(Aex₁) and E(Aex₂) species formed with the βLys87Thr mutant give α^Rβ^R complexes (Rhee et al., 1997).

Open and closed conformations of the TS subunits here are designated as T-state and R-state in conformity with early allosteric nomenclature (Monod et al., 1965; Niks et al., 2013; Ghosh et al., 2021). The designation of TS subunit conformations as either open or closed here has its origins in solution rapid kinetic experiments which demonstrate that the α- and β-subunits switch between conformations wherein ligands rapidly bind and dissociate, i.e., open conformations (T-state), and conformations wherein ligands slowly bind and dissociate, i.e., closed conformations (R-state) (Dunn et al., 1990; Harris et al., 2002; 2005; Ngo et al., 2007a,b; Ghosh et al., 2021). The x-ray structure database (Berman et al., 2000) confirms that TS switches between open and closed states in response to ligand binding at the α-site and to the covalent state of intermediates bound to the β-site (Figure 2) (Kulik et al., 2005; Ngo et al., 2007a, b; Barends et al., 2008a; Lai et al., 2011; Niks et al., 2013; Caulkins et al., 2016; Ngo et al., 2007a, b; Ghosh et al., 2021). Within α₂β₂, the switch of the α-subunit to the closed conformation when E(A-A) is formed activates the α-site by ∼30-fold while the β-subunit is activated about 10-fold (Anderson et al., 1991; Brzovic P. S. et al., 1992; Leja et al., 1995; Webber-Ban et al., 2001; Ngo et al., 2007b).

The channeling of small molecules via tunnels within macromolecular assemblages is a key strategy in biological systems for the selective transfer of small molecules and ions among cellular compartments and across cell membranes (Dunn et al., 1990; Dunn et al., 2008; Shaffer et al., 2009; Galdiero et al., 2012; Ziervogel and Roux, 2013; Baker and Baldus, 2014; Horn et al., 2014; Aryal et al., 2015; Hilario, et al., 2016). Typically, the macromolecular protein structures responsible for these transfers form channels which achieve high selectivity for the transferred molecule or ion by acting as molecular filters that restrict passage through the channel based on charge, molecular cross-section, and hydrophobicity (Davis, 1967; Holden et al., 1998; Miles et al., 1999; Amaro et al., 2003; Amaro and Luthey-Schulten, 2004; Raushel et al., 2003; Amaro et al., 2005; Friedrich, 2014). Channeling also is an important phenomenon for certain enzyme assemblages within some metabolic pathways (Raushel et al., 2003). Substrate channeling among enzyme complexes can play important roles within the cell in preventing deleterious side reactions of labile small molecules, enhance catalytic efficiency, and prevent the loss of hydrophobic small molecules into hydrophobic environments (Dunn et al., 2008; Hilario et al., 2016).

Direct transfer of a substrate between the two catalytic sites of a bienzyme complex is the simplest enzyme example of channeling. TS was the first enzyme system demonstrated to channel a common intermediate (indole) in a bienzyme complex. (Hyde et al., 1988; Dunn et al., 1990; Lane and Kirschner, 1991). The intermolecular tunnel in TS extends from the α-catalytic site near the α-β subunit interface to the β-catalytic site, a distance of ∼30 Å (Hyde et al., 1988) (Figures 5A,B). This tunnel is comprised of two sections (T₁ and T₂, Figure 5B). The first section (T₁) is a relatively hydrophilic region extending from the indole ring subsite of the α-subunit into the β-subunit to tunnel residues βTyr279 and βPhe280 (T₁, Figures 5A,B) (Hilario et al., 2016). This section is filled with a hydrogen-bonded network of water molecules (Hilario et al., 2016). The second section (T₂) is a very hydrophobic, dewetted nanotube and extends from βY279/βF280 into the indole ring sub-site of the β-catalytic site. There are no waters detected in this region of the tunnel (Hilario et al., 2016).

The exclusion of water from T₂ and the E(A-A) site provides an essentially non-aqueous environment for the C-C bond forming step between the indole ring C₃ and the α-aminoacrylate C_β carbons. While it previously has been speculated that βPhe280 plays a gating role in the channeling of indole (Anderson, et al., 1995), the findings of Hilario et al. (2016) indicate the transfer of indole is unhindered by βPhe280, a finding consistent with many of the x-ray structures of E(A-A). Hilario et al. (2016) examined the properties of the TS tunnel via x-ray crystallography, MD simulation and flexible docking studies. They reported the structures of complexes of TS with F6 bound to three different loci, one at the α-site and one within the β-subunit portion of the tunnel making contacts with βPhe280, and one bridging the α-β subunit interface (Figures 5A,B). The MD simulations indicated the following: 1) the hydrophobic region of the tunnel, T₂, excludes water, consistent with a dewetted state which prevents the transfer of water between the α- and β-sites (the nonpolar portion of F6 binds to this region of the tunnel, another F6 molecule binds to the hydrophilic region, T₁, of the tunnel, Figure 3), 2) in the E(A-A) intermediate the tunnel properties allow the transfer of indole from the α-site into the β-site even when βPhe280 partially restricts the cross-section of the tunnel (Figures 5A,B), 3) therefore, βPhe280 does not play a mechanistic gating role in the channeling of indole from the α-site to the β-site of E(A-A) during the β-reaction. These conclusions are further supported by rapid kinetic studies. The rate of reaction of indole with the α^Tβ^R form of E(A-A) is very fast relative to the turnover rate of the β-rection, as is the rate of reaction of IGP with E(A-A) (Lane and Kirschner, 1983; Drewe and Dunn, 1986; Brzovic P. S. et al, 1992; Kulik et al., 2002). Since these reaction rates significantly exceed the turnover rate of the β-reaction, βPhe280 appears not to have a significant gating role. Therefore, the primary role played by the βPhe280 side chain is simply to contribute to the hydrophobic environs of the tunnel. Nevertheless, Brownian dynamics simulations using a residue-based coarse-grained model suggest that the channel does not always exist, and it may be blocked before TS reaches its final substrate bound conformation. This modeling work highlights the roles of protein conformations in substrate channeling (Fatmi and Chang 2010).

As has been shown, the ligand-mediated allosteric interactions between the α- and β-subunits achieve the efficient utilization of IGP as the source of indole via synchronization of the α- and β-reactions (Dunn et al., 1987; Houben et al., 1989; Dunn et al., 1990; Houben and Dunn, 1990; Leja et al., 1995; Pan et al., 1997; Dunn et al., 2008; Dunn, 2012), the channeling of indole from the α-site to the β-site via the interconnecting tunnel (Hyde et al., 1988; Dunn et al., 1990; Kirschner, et al., 1991, Anderson et al., 1991; Brzović et al., 1992a, b; 1993), and the facilitation of chemical steps in the β-reaction (Ngo et al., 2007b). We hypothesize that the channeling of indole between the α- and β-subunits and the synchronization of the α- and β-reactions likely evolved to prevent the escape of indole and ensure the efficient utilization of IGP (Yanofsky and Rachmeler, 1958; Yanofsky and Crawford, 1972; Creighton, 1970); Miles, 1979; Pan et al., 1997; Miles et al., 1999; Yanofsky, 2007; Dunn et al., 2008; Dunn, 2012; Miles, 2001). These elements of the biosynthesis of L-Trp are important for organisms that utilize a Trp operon (Yanofsky, 2007).

The catalytic activity of the monovalent cation-free enzyme is strongly impaired (Peracchi et al., 1995; Woehl and Dunn, 1995a, b; Woehl and Dunn, 1999a, b; Mozzarelli et al., 2000; Weber-Ban et al., 2001). The activity of monovalent cation-free TS is decreased 45-fold compared to the Na⁺-activated enzyme. A wide variety of monovalent cations (MVCs) bound to the TS metal coordination site activate the α₂β₂ bienzyme complex, including Na⁺, K⁺, Cs⁺, NH₄ ⁺ and guanidinium ion (10) (Fan et al., 1999; Mozzarelli et al., 2000; Dunn, 2012). The available structural information shows a binding site within the β-subunit that can accommodate Na⁺, K⁺, NH₄ ⁺ or Cs⁺ (Rhee et al., 1996; Fan et al., 1999, 2000; Dierkers et al., 2009) (Figure 5C).

Kinetic studies have shown that monovalent cation binding is essential both for catalysis of the β-reaction and for the transmission of allosteric signaling between the β- and α-sites (Rhee et al., 1997; Woehl and Dunn, 1995a, b; Woehl and Dunn, 1999a, b; Mozzarelli et al., 2000; Weber-Ban et al., 2001; Fan et al., 1999, 2000; Dierkers et al., 2009). Interestingly, the Na⁺, K⁺, NH₄ ⁺ forms of TS activate the reaction of L-Ser in the β-reaction by ∼30, ∼26, and ∼40-fold respectively (Weber-Ban et al., 2001), while the reaction of L-Ser with monovalent cation-free TS does not activate the α-reaction even though the monovalent cation-free enzyme forms an α-aminoacrylate species (Weber-Ban et al., 2001). The different monovalent cation-bound forms of TS give different distributions of intermediates along the β-reaction path with Na⁺ favoring E(Ain)^T and E(Aex₁)^T and Cs⁺ favoring E(A-A)^R and E(C)^R (Weber-Ban et al., 2001; Dierkers et al., 2009). Inspection of the available x-ray crystal structures indicates that the variation in ionic radii of the monovalent cations causes a concomitant variation in coordination number and geometry (Woehl and Dunn, 1995b; Dierkers et al., 2009) which influences the dimensions of the β-subunit indole sub-site. The x-ray crystal structures of various monovalent cation-substituted TS enzymes identify the coordination site(s) as a cavity bounded by the backbone carbonyls of βVal231, βGly232, βGly268, βLeu304, βPhe306 and βSer308, residues that are not part of the COMM domain. Owing to the differences in ionic radius, Na⁺ only coordinates to three of the carbonyls (βGly232, βPhe306 and βSer308) and two waters, whereas Cs⁺ nearly fills this cavity and typically coordinates to 5 or 6 of the carbonyl oxygens). The K⁺ complex incorporates three of the carbonyl oxygens and a single water (Rhee et al., 1997; Dunn et al., 2008; Miles et al., 1999; Barends et al., 2008a, b; Ngo et al., 2007a b). Of special note is the involvement of βPhe306 in this cavity and the linkage of this residue to βAsp305. The x-ray structures show the side chain phenyl ring of βPhe306 is a component of the β-subunit indole sub-site while βAsp305 plays an important role in the stabilization of the R state conformation. Thus, the stabilizing effect exerted by Cs⁺ on the R state likely has its origins in the coordination of Cs⁺ by βPhe306 and the concomitant effect on the positioning of the βAsp305 side chain carboxylate in the conformation needed for formation of the salt bridge with βArg141 in the β-subunit R state conformation. This interaction also stabilizes the positioning of the βPhe306 phenyl group so that the VDW dimensions of the indole sub-site match the VDW dimensions of indole, creating a preformed sub-site (Figures 5C, 6A–D). Clearly, the movement of the COMM domain as the β-subunit switches from the T state to the R state causes the indole sub-site to expand from dimensions too small to accommodate the binding of indole (T state) to a dimension of 12.2 Å that matches the VDW surface of indole. Thus, the Na⁺ complex stabilizes the T-state (Dierkers et al., 2009) while Cs⁺ favors complexes with sub-sites that match or nearly match the dimensions of indole.

In 1996, Peracchi et al. (1996) reported that the pH dependent interconversion of E(Aex₁) with E(A-A) is modulated by proton binding and involves two groups with apparent pKa values of ∼7.8 and ∼10.3. Since this pH dependence is not observed in the β₂ dimer (Peracchi, et al., 1996), it most likely arises from the allosteric properties of α₂β₂. The microscopic origins of these ionizations are not known, but this behavior is reminiscent of the Bohr effect on dioxygen binding to hemoglobin (Signore et al., 2021). In TS, we speculate that these pH effects have origins in the ionizations of βLys87 and the salt bridging interactions linked to the T-to R-state transition (Peracchi et al., 1996; Phillips et al., 2005; 2008a, b).

The cofactor protonation states determined by solid-state NMR reflect the relative stabilities of the allosteric states observed in TS. For example, in the T state complex of E(Ain) the PLP moiety has a protonated Schiff base nitrogen (Figure 7A) (Caulkins et al., 2014; Klein et al., 2022), whereas the Schiff base nitrogen in the R state complexes of E(C₃)_2AP and E(A-A) have deprotonated Schiff base nitrogen states as the dominating tautomers (Figures 7B,C) (Caulkins et al., 2016; Holmes et al., 2022). We postulate that this difference in protonation states has its origins in the presence of active site water molecules in the E(Ain)^T conformation while the E(A-A)^R and E(C₃)^R conformations are significantly more dehydrated (Huang et al., 2016).

The PDB archive (Berman et al., 2000) shows that TS undergoes conformational transitions that alter the structures of the α- and β-subunits, and that the α-subunit switches between an ensemble where loop αL6 is disordered in the ligand-free α-site (Hyde et al., 1988; Miles, 2001; Kulik et al., 2005; Ngo et al., 2007b; Yanofsky, 2007) and an ensemble wherein loop αL6 is well ordered when G3P, or substrate analogue F9 are bound (Kulik et al., 2005; Miles et al., 1999; 2001; Ngo et al., 2007a, b; Barends et al., 2008a, b; Lai et al., 2011; Niks et al., 2013; Caulkins et al., 2016; Ghosh et al., 2021) (Figures 2, 8). This disordered-to-ordered transition of αL6 also induces an ordering of loop αL2. Both loops make contributions to the α-β subunit interface via helix βH6 of the COMM domain (Miles et al., 1999, 2001; Kulik et al., 2005; Ngo et al., 2007b; Barends et al., 2008a).

Structural and kinetic evidence indicates that the disordered αL6 loop renders the α-subunit essentially catalytically inactive (Brzovic et al., 1992a; 1993; Kulik et al., 2005; Miles et al., 1999; 2001). The NMR experiments of Sakhrani et al. (2020) on the solution structure of the α-subunit free and bound to the IGP analogue F9 show similar effects on the affinity of ligands for the α-subunit (∼20-fold in α^Tβ^T and ∼500-fold in α^Tβ^R). When the α subunit is switched to the closed conformation (R state), loop αL6 (residues α179–α193) folds down over the catalytic site and makes interlocking interactions with loop αL2 (residues α56–α60) and helix βH6, thus creating a local environment where the ligand is shielded from solution, and access from solution into the interconnecting tunnel is blocked, Figures 2, 8A (Miles 2001; Harris and Dunn, 2002; Harris et al., 2005; Kulik et al., 2005). Within these complexes, αThr183 swings into a position where the side chain hydroxyl forms a hydrogen bond to one oxygen of the αAsp60 carboxylate, stabilizing this catalytic group in the proposed position for α-site catalysis Kulik et al., 2002) (Figure 8). Notice in Figures 1B, 8A, the conformation of the αGlu49 side chain carboxylate group also becomes stabilized in the position necessary to act as an acid-base catalyst in facilitation of the cleavage of the C-C bond. So long as the α-subunit retains this conformation, indole formed via scission of IGP is prevented from dissociation into solution, and instead, is shuttled into the tunnel leading to the β-subunit catalytic site 25 Å away. Consequently, the allosteric transition of the α-subunit occurs via a transition of loop αL6 between disordered and ordered states and induces a reordering of αL2 (Miles, 2001; Kulik et al., 2002; Barends et al., 2008a; Raboni et al., 2009; Dunn, 2012; Maria-Solano et al., 2019; Teixeira et al., 2020) (Figure 8). The formation of well-ordered αL6 and αL2 loops creates a well-defined subunit interface between the α- and β-subunits, an α-site with high affinity for IGP, and configures αAsp60 for catalysis (Figure 8). While the x-ray structural evidence clearly supports the involvement of two subunit conformations, T and R, and four quaternary states, α^Tβ^T, α^Rβ^T, α^Tβ^R and α^Rβ^R (Niks et al., 2013; Ghosh et al., 2021), it is also clear that the binding of G3P or analog GP can give complexes where loop αL6 is only partially disordered and the extent of disorder depends, at least in part on the structure of the ASL. For example, Ngo et al. (2007b) reported evidence for three conformational states, open, partially closed, and closed of the α-subunit. They found that in the absence of α-site substrates or substrate analogues both the α- and β-sites reside in open conformations and αL6 is completely disordered. In structures where the β-domain is closed and the α-site is occupied by an ASL, the electron densities for loop αL6 (residues α179-193) are well-defined (with the exception of the two outermost residues in some structures). These structures clearly exhibit closed α- and β-subunits. In x-ray crystal structures of IGP bound to the α site of the internal aldimine (PDB ID: 2RH9, 2RHG, 1QOQ), the α subunit adopts a T conformation with a disordered αL6 loop and suggest that these IGP complexes represent an inactive state of the α subunit. In structures where the β-subunit is not closed and the α-site is occupied, typically parts of αL6 are visible in the electron density maps and may represent a partially closed state. What is now clear is that the partial disorder observed in some αL6 structures likely is a consequence of structural mismatching between the ASL and the α-subunit (vis. IPP). Since IPP lacks hydroxyls at the 2 and 3 carbons, the hydrogen bonding between site and substrate is not replicated in the IPP complex, Figure 8B. The complexes with the IGP analogue, F9, and the proposed transition state analogues 1-(2′-hydroxyphenylamino) 3-glycerolphosphate (2-HGP, 14) and N-(indolinyl) 3-glycerolphosphate (NGP, 15) (Figures 4, 8) provide sufficient interactions with αL6 to stabilize the interaction between all of the residues in αL6 and the surface of the α-subunit (Kulik et al., 2002, 2005; Dunn 2012).

It is interesting to note that the x-ray structure of the α₂β₂ TS complex found in Pyrococcus furiosus TS shows the α-subunits with the closed conformation and a well-ordered αL6 loop (Lee et al., 2005). Nevertheless, this homolog of the Escherichia coli and Salmonella enterica serovar Typhimurium enzymes shows many of the allosteric properties exhibited by the enteric tryptophan synthases.

In contrast to the α-subunit, the β-subunit allosteric transition involves a relatively modest conformational change wherein an 88 amino acid residue domain (the COMM domain, residues β102—189) moves as a rigid body, undergoing a slight rotation and translation. This motion causes a displacement of COMM domain residues by ∼ 2.6 Å along the αβ subunit interface and by ∼ 4.5 Å near the β-catalytic site (Figure 2). While this motion is relatively modest, the functional changes are large. The translation and rotation of the COMM domain causes the β-subunit to switch between two states, one open (the T state) the other closed (the R state) (Figure 2). The COMM domain contributes key residues and structural units essential both to catalysis and to allosteric communication (Figures 2, 3). One surface of the COMM domain forms a wall of the cleft linking the β-site and solution. Residues from the COMM domain and from loop αL2 provide most of the contacts that comprise the α-β subunit interface (Figures 2, 8). The COMM domain also contributes the following components of the allosteric response: the substrate carboxylate recognition sub-site (loop βL3) (Figures 2, 8) and forms part of the indole recognition sub-site (Figures 5, 6); a catalytic site residue, βGlu109, and a salt bridge forming residue, βArg141, forms a strong Coulombic interaction with βAsp305 that stabilizes the β^R conformation (Ferrari et al., 2001, 2003) (Figures 6, 9). In the transition of α^Tβ^T to α^Rβ^R, the combined motions of the COMM domain and loops αL6 and αL2 close the entrance into the α-site and tunnel from solution (Figure 2) and the movement of the COMM domain also closes the cleft into the β-site from solution (Figure 2) (Harris and Dunn, 2002; Harris, et al., 2005). Thus, the switch to the α^Rβ^R state generates steric constraints (among the ensemble of closed conformation states) that prevent the escape of indole from the enzyme α-site and the tunnel, creates the indole sub-site at the β-catalytic site, and blocks the transfer of L-Ser and L-Trp between the enzyme β-site and solution.

The β^T catalytic site resides in a solvent-exposed environment where three water molecules solvate one face of the PLP moiety and a lattice of waters extends to the aqueous milieu (Ghosh et al., 2021). In these complexes the β-site is accessible to bulk solvent via the narrow cleft, allowing L-Ser and product L-Trp to enter and exit the β-catalytic site. The switch of β^T to β^R closes the cleft to a narrow aperture (∼3Å diameter, Figure 2) and only a single water molecule is retained on this face of the PLP moiety, creating a more hydrophobic environment (Ghosh et al., 2021). The aperture of the portal is too small to allow the entry/exit of small molecules such as indole, L-Ser or L-Trp. Thus, the small motion of the COMM domain triggers activation of IGP cleavage at the α-site, facilitates channeling of indole from the α-site to the β-site, creates the indole subsite, orchestrates the interconversions of E(Aex₁), E(C₁), E(A-A), E(C₂), E(C₃) and E(Aex₂), synchronizes the catalytic cycles of the α- and β-reactions, and prevents the escape of indole from the confines of the α and β sites and the tunnel during the αβ catalytic cycle (Dunn, 2012; Hilario et al., 2016).

The efficient synthesis of L-Trp by TS requires the facile interconversion of T and R states during the overall αβ-rection. This is accomplished by constraining indole to the interior of the complex while allowing the ingress and exit of substrates at the α- and β-sites as needed. This balance requires that the energies of the T and R states modulate the binding of substrates and chemical intermediates so that the switching between T and R states is orchestrated with the appropriate chemical steps. Thus, binding of IGP and G3P to the α-site and formation of the α-aminoacrylate at the β-site drive the switching of α^T to α^R and β^T to β^R.

These constraints also dictate that the relative energies of the T and R conformation states be sensitive to weak intramolecular interactions within loops αL2 and αL6 of the α-subunit and within the COMM domain of the β-subunit. For example, Ghosh et al. (2021) have demonstrated that the replacement of βGln114 by Ala interferes with the catalytic activity of the β-subunit by altering the relative stabilities of intermediates in the β-reaction. The origins of the altered stabilities appear to arise from the incorporation of water molecules into the cavity by replacement of the Gln side chain with the smaller Ala side chain, and by the loss of hydrogen bonding interactions between the side chains of βGln114 with βAsn145 and βArg148 in β^R (Figure 10). Since there are no intramolecular hydrogen bonds to the βGln114 side chain in the T state, the Gln to Ala mutation destabilizes the R state. This destabilization of the R state is illustrated by the change in distribution of intermediates resulting from reaction of L-His with TS (Figure 11). Mutation of either βArg141 or βAsp305 to Ala similarly alters the activity of the β-site and alters the distribution of aldimine and carbanion intermediates (Ferrari, et al., 2001, 2003). It previously has been argued that these mutations destabilize the R state by destroying the βArg141—βAsp305 salt bridge (Ferrari et al., 2001, 2003; Ghosh et al., 2021).

In wild-type TS, the interconversion of the E(Aex₁) and E(A-A) species is accompanied by a switch from the T state to the R state. Early work (Rhee et al., 1997) demonstrated that mutation of the β-site catalytic residue βLys87 to Thr shifts the stabilities of the E(Aex₁) and E(Aex₂) intermediates strongly in favor of structures wherein both the α- and the β-subunit reside in closed conformations, i.e., α^Rβ^R. Thus, Rhee et al. (1997) determined the first two x-ray structures of the completely closed α^Rβ^R state. Studies by Ferrari et al. (2001, 2003) have established that the destruction of the βD305-βR141 salt bridge (Figures 9C,F) by mutation of either of these residues to Ala alters the thermodynamics of the conformational transitions of the β subunit to the R state. This alteration results in the destabilization of E^R(A-A), thus shifting the distribution of β reaction intermediates in favor of E^T(Aex₁) and an altered substrate specificity. Both the βAsp305Ala and βArg141Ala mutants give distributions of intermediates that are shifted in the favor of E^T(Aex₁). Recent studies (Ghosh et al., 2021) have shown that a mutation distal from the αβ interface such as replacing the βL3 residue βGln114 with an alanine have significant effects on the distribution of species in the β reaction. They have shown that in the reactions of L-His and L-Trp, ASL binding shifts the β subunit population mostly to the E^R(C) species, while, with the βGln114A mutant, these reactions are significantly impaired and ASL binding shifts the population from the E^R(C) species to a mixture of the E^T(Aex) and E^T(GD) complexes (Figure 11).

Using the standalone β subunit from P furiosus TS, Buller et al. (2015) were able to select for mutations that enhance the activity of the relatively inactive β-subunit to a level exceeding that of the wild-type complex. These findings suggest that the heterotropic allosteric activation achieved by formation of the αβ dimeric unit in the wild-type system can be mimicked by mutations that facilitate catalysis.

The complexes formed by the sodium form of TS in the reactions of L-Trp and D-Trp (17) (Lane and Kirschner, 1981; Drewe et al., 1989) and analogues of the carbanion intermediates formed with OIA (18) (Roy et al., 1988b) and DOA (19) (Phillips and Harris 2021) recently have been examined by UV/Vis spectroscopy and rapid kinetics, and the structures of these complexes with GP bound to the α-site have been solved by x-ray crystallography (viz., Figure 12) (Phillips and Harris 2021). These studies establish the stereochemistry of the OIA and DOA reactions and by analogy indicate the course of the stereochemical transformations during formation of E(C₂) and E(C₃) in the β-reaction. The x-ray structures also provide additional information about the allosteric states that predominate along the β-reaction pathway. The complexes of OIA and DOA give mixtures of both the external aldimine and the carbanion species within the same crystal while the subunit conformation is α^Rβ^R (Figure 12). L-Trp gives a non-covalently bound complex (Buller et al., 2015) with the α^Rβ^T conformation (Phillips and Harris 2021) whereas D-Trp gives a mixture of two external aldimine complexes with different orientations of the α-carboxylate, and a subunit conformation that also is α^Rβ^R. However, the β-subunit of one form of the heterodimeric unit in these mixtures fails to form the signature R-state salt bridge between βArg141 and βAsp305. These structures appear to model the interactions between the catalytic site and the reacting substrate during the interconversion of the external aldimine and carbanion intermediates (Figure 1C).

We hypothesize that owing to the presence of a significant pool of L-Ser in vivo, the TS β-site exists within the bacterial cell primarily as the quasi-stable α-aminoacrylate species in the α^Tβ^R conformation. In the absence of IGP, the α-site has the open α^T conformation, thus giving a TS resting state, α^Tβ^R, where the β-site is activated but sequestered away from small molecule nucleophiles that could cause deleterious side reactions (viz., Figure 3) (Blumenstein et al., 2007; Flynn and Downs, 2013; Hilario et al., 2016; Ghosh et al., 2021). This resting state appears primed for the binding and reaction of IGP at the α-site.

Because the conversion of E(Aex₁) via E(C₁) to E(A-A) at the β-site triggers activation of the α-site ∼30-fold (Anderson, et al., 1991; Brzovic P. S. et al., 1992), the synthesis of L-Trp via the α- and β-reactions begins when IGP binds and reacts at the α-site with the E(A-A) form of TS. The α-site remains in the α^R conformation until β^R is switched back to β^T when E(C₃) is converted to E(Aex₂) (Brzovic P. S. et al., 1992; Leja et al., 1995). Notice that the efficient transfer of indole from the α-site to the β-site via the interconnecting 25 Å tunnel requires that transfer occur within the α^Rβ^R. This closed conformation prevents the escape of indole.

One consequence of the switching between low and high activity states of the α-site in response to the switching of the β-subunit between β^T and β^R is to cause an in-phase synchronization of the α- and β-reactions (Figure 13) (Brzovic P. S. et al., 1992; Leja et al., 1995). This synchronization of the α- and β-reactions achieves the efficient utilization of the indole moiety of IGP for the biosynthesis of L-Trp by linking indole production via IGP cleavage at the α-site to L-Trp formation at the β-site. This linkage makes possible the channeling of indole between the α- and β-sites (Figure 13).

The relative stabilities of β^T and β^R are greatly affected by the binding and reaction of substrates to the catalytic sites of the αβ dimeric unit (Harris and Dunn, 2002, 2005). Dunn et al. (1990) reported clear evidence that the binding of the IGP analogue glycerol phosphate to the α-site strongly inhibits the reactions of indole and indole analogues with E(A-A) while the reactions with L-Ser or L-Trp are only slightly perturbed. These observations led to the conclusion that ASL binding switches the α-subunit to the closed conformation (the R state) preventing the entry of indole and indole analogues into the β-site via the α-site and the tunnel, while entry of L-Ser and L-Trp is unaffected, a conclusion in agreement with an α^Rβ^T structure (Dunn et al., 1990) (Figure 2). These conclusions have been confirmed and expanded in more recent work (Brzović et al., 1992a, b; 1993; Harris and Dunn, 2002, 2005; Leja et al., 1995; Ngo et al., 2007b). One interesting and not fully explained feature of this inhibition is the observation that although high affinity ligands strongly inhibit the reactions of nucleophiles with E(A-A), all give residual reaction rates that become independent of the concentration of the α-site ligand at high ligand concentrations. Based on what we now know about the T to R allosteric transitions of TS, it seems likely that these residual reaction rates have their origins in the switching of E(A-A) complexes between β^R where entry and egress of small molecules is sterically blocked and β^T where small molecules are free to exchange between solvent and site. Thus, the slow residual rates likely are due to the rate of switching between β^R and β^Tand the rate of this switching is strongly influenced by the affinity of the ASL for the α-site.

When taken together, the discoveries and advances described in the forgoing paragraphs combine to form a collage that captures at a structural and physio-chemical level the mechanism for the allosteric regulation of the tryptophan synthase bienzyme complex in enteric bacteria. The documentation in this collage shows tryptophan synthase allostery has become an important paradigm which begins to rival the hemoglobin allostery paradigm in terms of insight into the interplay between structure and function. However, the allosteric properties of tryptophan synthase and hemoglobin are very different, reflecting their very different biological functions. Allosteric interactions in TS are restricted to the regulation of substrate channeling in αβ dimeric units of the α₂β₂ complex that function to synthesize L-Trp from IGP, indole and L-Ser via an exquisite interplay of allosteric signaling that switches the αβ- and β-subunits between states of low and high reactivity. The allosteric transition of the TS α-subunit switches loop αL6 between disordered and ordered states, while the β-subunit transition causes a small motion of an 88 amino acid domain (Figure 2). Together, these motions synchronize the final two steps in the biosynthesis of L-Trp by reinforcing the correct placement and alignment of catalytic residues within an extended solvent-protected cavity that spans two active sites and an interconnecting, 25 Å-long tunnel (Figure 5). In contrast to TS, the allosteric properties of hemoglobin comprise a highly nuanced system of structure-function relationships that regulate the transport of dioxygen and nitric oxide by red blood cells into the tissues of higher organisms and the removal of CO₂ and H⁺.

The reaction scheme presented in Figure 14 summarizes our current hypothesis for the allosteric regulation of L-Trp synthesis within the TS αβ-reaction cycle. This superposing links the known protein conformational states to the established chemical transformations that occur during a single round of catalysis. The presence of an L-Ser pool in vivo likely insures the αβ-reaction begins with the β-subunits of α₂β₂ in the form of E(A-A) and the allosteric units predominately α^Tβ^R (Figures 2, 14). When IGP enters the open α^T-site from solution, binding of IGP switches the α-subunit to the activated α^R state, and IGP is cleaved to G3P and indole. With TS in the α^Rβ^R state, G3P remains bound to the α-site and indole is trapped within the confines of the α-site, the interconnecting tunnel, and the β-site. Because the tunnel functions as a selective filter that accommodates the passage of indole but rejects water and other polar molecules (viz. Figure 5), G3P remains bound to the α-site while indole is transferred via the tunnel into the indole subsite of the β-subunit as depicted in Figure 14 (Hilario et al., 2016; Ghosh et al., 2021).

Carbon-carbon bond formation at the β-site then occurs via nucleophilic attack of indole at C_β of E(A-A) giving E(C₂) which is quickly converted to E(C₃) while the allosteric unit remains in the completely closed state, α^Rβ^R (Figures 1, 2, 14). As TS is converted to E(Aex₂), the allosteric unit switches to the α^Tβ^T state releasing G3P. Synthesis of L-Trp is completed when E(Aex₂), via E(GD₂), is converted to the E(Ain)(L-Trp) complex followed by the release of L-Trp from the β^T conformation of E(Ain) (Figure 14). At this juncture, IGP could bind to the α-site giving the α^R conformation, however, the α-site only becomes activated again when the β-subunit is converted to the β^R conformation. In this scheme the conversion of β^T to β^R likely occurs when E(C₁) is formed via the reaction of E(Ain) with L-Ser to give E(GD₁) and E(Aex₁). Since E(C₁) is only detected as a fleeting intermediate (Drewe and Dunn, 1985), the conformation of E(C₁) is unknown. It is clear that the quasi-stable E(A-A) has the β^R conformation and the α-site is activated. Just as the allosteric switching in hemoglobin is essential to the transport of dioxygen, CO₂ and H⁺, the allosteric switching in tryptophan synthase is essential for the efficient synthesis of L-tryptophan from IGP and L-serine.

With the wealth of structural and functional data now available, the larger picture of allosteric control in tryptophan synthase is coming into focus. Yet questions remain, particularly on the coupling of structure to the finest level chemical details of the mechanistic transformations. At the chemical level, the linkage between α-site activation and E(A-A) formation is only partially understood. When E(Aex₁) is converted to E(A-A) via E(C₁), β^T is switched to β^R, the α-site is activated ∼30-fold and the indole sub-site of the β-site is expanded to dimensions tailored to match the VDW dimensions of Indole (Figures 5, 6, 13). This transformation is quite rapid. The β^T to β^R switch also rearranges the conformation of βAsp305 to form the R state salt bridge with βArg141, and the hydroxymethyl group of the reacting L-Ser moiety of E(Aex₁) rotates to a position where βLys87 functions as the acid catalyst to facilitate the elimination of the E(C)₁ hydroxyl to give E(A-A) (Figure 1B) (Holmes et al., 2022). In this transformation, the bonding hybridization of the C_β of the reacting substrate switches from sp³ to sp², as E(Aex₁) is converted to E(C₁). It is not clear if these conformational changes in the site residues and the reacting substrate occur in β^T or β^R. Answering this question will bring into sharper focus an understanding of the linkages between the α- and β-subunit allosteric transitions and the chemical transformations at the α- and β-sites. A second, related issue arises concerning the formation of the indole sub-site in the β-subunit. The indole sub-site of the β-subunit is too small in the β^T complexes of E(Ain) and E(Aex₁) and only expands to dimensions matching those of Indole when β^T is switched to β^R. The mechanism for this allosteric switch remains an elusive and open question, and requires further attention to determine how the allosteric switch to β^R triggers the expansion of the indole sub-site to fit the dimensions of indole, and why the sub-site dimensions are linked to the allosteric transition.