GTP-Dependent Regulation of CTP Synthase: Evolving Insights into Allosteric Activation and NH3 Translocation

Cytidine-5′-triphosphate (CTP) synthase (CTPS) is the class I glutamine-dependent amidotransferase (GAT) that catalyzes the last step in the de novo biosynthesis of CTP. Glutamine hydrolysis is catalyzed in the GAT domain and the liberated ammonia is transferred via an intramolecular tunnel to the synthase domain where the ATP-dependent amination of UTP occurs to form CTP. CTPS is unique among the glutamine-dependent amidotransferases, requiring an allosteric effector (GTP) to activate the GAT domain for efficient glutamine hydrolysis. Recently, the first cryo-electron microscopy structure of Drosophila CTPS was solved with bound ATP, UTP, and, notably, GTP, as well as the covalent adduct with 6-diazo-5-oxo-l-norleucine. This structural information, along with the numerous site-directed mutagenesis, kinetics, and structural studies conducted over the past 50 years, provide more detailed insights into the elaborate conformational changes that accompany GTP binding at the GAT domain and their contribution to catalysis. Interactions between GTP and the L2 loop, the L4 loop from an adjacent protomer, the L11 lid, and the L13 loop (or unique flexible “wing” region), induce conformational changes that promote the hydrolysis of glutamine at the GAT domain; however, direct experimental evidence on the specific mechanism by which these conformational changes facilitate catalysis at the GAT domain is still lacking. Significantly, the conformational changes induced by GTP binding also affect the assembly and maintenance of the NH3 tunnel. Hence, in addition to promoting glutamine hydrolysis, the allosteric effector plays an important role in coordinating the reactions catalyzed by the GAT and synthase domains of CTPS.

Considering the central role of CTP in metabolism, it is not surprising that CTPS is highly regulated. While the substrates ATP and UTP promote oligomerization of the enzyme from inactive monomers and dimers to active tetramers [37,[39][40][41][42][43] by binding at an interfacial active site formed by the synthase domains [32,44,45], the product CTP can also induce tetramerization [43] as well as act as a feedback inhibitor [37,44]. Eukaryotic homologues of CTPS are also regulated by phosphorylation [46][47][48][49][50][51]. More recently, CTPS has been shown to form a superhelix of tetramers, and this filamentous structure (also known as cytoophidium) constitutes an additional level of regulation [52][53][54][55][56][57][58][59][60][61][62]. Finally, guanosine-5 -triphosphate (GTP) is an allosteric regulator of CTPS activity [37,[63][64][65]. Indeed, among the Gln-dependent amidotransferases, CTPS is unique in that it is the only member of this family of enzymes wherein the GAT domain requires an allosteric effector for the efficient hydrolysis of Gln. Initially, GTP was believed to simply promote the Gln hydrolysis reaction; however, over the past two decades, kinetics, site-directed mutagenesis, and structural studies have revealed that the role of GTP is much more complex than originally thought. Indeed, GTP not only induces conformational changes that activate the glutaminase activity of the enzyme, but also plays a crucial role in inducing the conformational changes that permit the proper formation of the NH 3 tunnel and synchronizing the efficient delivery of the NH 3 from the GAT domain to the synthase domain. Herein, we review the results of site-directed mutagenesis and kinetics studies, as well as structural studies, that have led to the development of a detailed model describing the multifaceted role that GTP plays to effect the efficient catalysis of Gln-dependent CTP formation. triad reacts with the side chain of L-Gln to form the glutamyl-enzyme (route 1). The thiolate may also react with DON, leading to covalent modification of the enzyme and inactivation (route 2). The NH 3 generated from the hydrolysis of L-Gln is transferred to the synthase (amidoligase) domain via an intramolecular tunnel where it reacts with phosphorylated UTP (i.e., 4-P-UTP) to yield CTP (bottom). (B) Structure of the DmCTPS tetramer with bound GTP. The tetramer is shown in surface representation with bound 4-P-UTP (orange), ADP (yellow), DON (maroon), Mg 2+ (hot pink), and GTP (green carbon atoms) shown as spheres (PDB ID: 7DPT) [66]. The lower right protomer is colored light blue, light purple, and pink corresponding with the N-terminal synthase domain, interdomain linker, and GAT domain, respectively, as also colored in Figure 2. Those residues implicated in playing a role in GTP-dependent activation of the enzyme are shaded as shown in Figures 2 and 3.

Kinetics Studies: Laying the Groundwork
CTPS exhibits an exquisite ability to distinguish between all four ribonucleotide-5 -triphosphates in varying roles, including substrates, activators, inhibitors, and effectors of oligomerization and filament formation. Consequently, CTPS has served as a paradigm for understanding the complexities of protein regulation. Although it has been roughly 50 years since it was established that GTP serves as a positive allosteric effector of CTPS [64,65], enhancing the rate of Gln hydrolysis both in the presence and absence of substrate nucleotides, only within the last 15 years has structural information become available that provides atomic-level insights into the mechanism of allosteric activation. The first X-ray crystal structures of EcCTPS with bound CTP and ADP [32,44] and Thermus thermophilus HB8 CTPS (TtCTPS) with bound Gln [45] were reported in the mid-2000s, followed by the structures of apo-CTPS from Sulfolobus solfataricus (SsCTPS) [67], the synthase domain of human CTPS [68], EcCTPS with CTP and Gln bound [56], Mycobacterium tuberculosis CTPS (MtCTPS) with two UTP molecules bound or UTP, AMP-PCP, and 6-diazo-5-oxo-L-norleucine (DON) bound [15], and the glutaminase domain of Trypanosoma brucei CTPS (TbCTPS) with bound acivicin [8]. Subsequently, cryo-electron microscopy (EM) structures were reported for filaments of human CTPS [56,61] and Drosophila melanogaster CTPS (DmCTPS) [69]. However, none of these structures contained bound GTP and, hence, insights into the details of GTP-dependent activation at the atomistic level remained elusive until very recently [66]. In the absence of structural information delineating the exact conformational changes that GTP binding effects to stimulate the glutaminase activity, site-directed mutagenesis and kinetics studies were crucial in defining the location and functional effects of GTP binding.
Early kinetics studies conducted by Levitzki and Koshland revealed that GTP accelerates the glutaminase activity but has little or no effect on the rate of NH 3 -dependent CTP formation [64]. Although the original kinetic model for the effects of GTP on catalysis by EcCTPS appeared to be quite complex, apparently involving both negative and positive cooperativity accompanying equilibrium binding experiments [64], later studies revealed that the activation of Gln-dependent CTP formation by GTP was hyperbolic for EcCTPS [70][71][72], Lactococcus lacti CTPS (LlCTPS) [36,73], and the two yeast isozymes [74] under conditions where the enzyme was in its fully active, tetrameric state. That said, Levitzki and Koshland concluded that GTP likely exerted its effect by enhancing the rate of formation of the glutamyl-enzyme intermediate from the E·Gln Michaelis complex with little effect on the K m value for Gln. Consistent with this hypothesis, Bearne and co-workers showed that GTP binding enhanced the inhibition of EcCTPS by the intermediate/transition state analogue inhibitor glutamate γ-semialdehyde, thereby suggesting that GTP promotes the glutaminase activity by stabilizing a protein conformation that binds the tetrahedral intermediate(s) formed during Gln hydrolysis [75]. Similar results were obtained by Willemoës with LlCTPS [76]. The notion that the allosteric binding of GTP causes subtle conformation changes and re-organization of the environment at the oxyanion hole to facilitate the glutaminase reaction was consistent with the behavior of other amidotransferases [77][78][79]. Later studies revealed that at elevated concentrations, GTP behaved as a weak negative allosteric effector of EcCTPS inhibiting Gln-dependent CTP formation [71]. Interestingly, GTP did not inhibit the enzyme's intrinsic glutaminase activity. Furthermore, GTP inhibited NH 3 -dependent CTP formation. Overall, these and other kinetics studies on the activation and inhibition of CTPS by GTP have been interpreted in terms of kinetic mechanisms similar to the one shown in Scheme 1, which yields the initial velocity Equation (1), where k o and k act are the rate constants for the formation of CTP in the absence and presence of GTP, respectively, K A is the dissociation constant for the E·GTP complex, K i n is the apparent dissociation constant for the non-productive E·(GTP) n complex, and n is the Hill number [71,72]. Scheme 1. Kinetic mechanism for the activation and inhibition of CTPS by GTP. E represents the enzyme complexed with ATP and UTP. K A , K i , and K S are apparent dissociation constants and n is the Hill number.
Typically, saturating concentrations of Gln are employed so that Equation (1) reduces to Equation (2) to simplify the kinetic analysis.
The observation that GTP could inhibit the utilization of exogenous NH 3 in a concentration-dependent manner suggested that the binding of GTP could block the access of NH 3 to the enzyme and that the entry point for the exogenous NH 3 was located in the GAT domain. While GTP weakly inhibits NH 3 -dependent CTP formation by EcCTPS at all concentrations [71], the inhibition is more pronounced when the active-site Cys is covalently modified by 6-diazo-5-oxo-L-norleucine (DON, Figure 1) [64], which mimics the glutamylenzyme intermediate, or when the enzyme is inhibited by glutamate γ-semialdehyde, which mimics the tetrahedral intermediate(s) formed during hydrolysis [75]. It is possible that the negative cooperativity reported by Levitski and Koshland [64] reflected the conformational changes that accompany the inhibition of both the NH 3 -and Gln-dependent CTP formation at elevated concentrations of GTP (vide infra) [71,72].
In addition to these differences, EcCTPS and LlCTPS exhibited differences in how Gln hydrolysis was regulated by GTP. While Levitzki and Koshland [64] found that EcCTPS catalyzed the GTP-activated hydrolysis of Gln in the presence of UTP and ADPNP at a steady-state rate similar to the rate of the GTP-activated synthesis of CTP and with a similar degree of enhancement in the presence of GTP, Willemoës and Sigurskjold [36] found that GTP-dependent activation of the LlCTPS-catalyzed glutaminase reaction never attained the same levels as the overall reaction for GTP-activated Gln-dependent CTP formation, including in the presence of UTP and the non-hydrolyzable ATP analogue adenosine-5 -[γ-thio]triphosphate (ATPγS). Unlike EcCTPS, the affinity of LlCTPS for GTP was increased in the presence of ATP and UTP. Indeed, the degree of saturation with ATP and UTP, unlike ATPγS and UTP, had a pronounced influence on the K A value for GTP binding, with the value of K A decreasing with lower concentrations of the substrate nucleotides [73]. Moreover, at low GTP concentrations ranging from~10 to 100 µM, the rate of Gln hydrolysis exceeded that of CTP synthesis, resulting in uncoupling of the two reactions. However, as the concentration of GTP was increased, GTP either coordinated or coupled the two reactions [36]. In addition, the inhibition typically observed for GTP at concentrations above 1 mM was relieved in the presence of ATPγS and UTP. Overall, Willemoës and Sigurskjold [36] concluded that while the rate of the glutaminase reaction of EcCTPS is independent of the UTP-phosphorylation reaction, for LlCTPS, the 4-phosphorylated UTP intermediate appears to act as a weak activator of Gln hydrolysis, which, in a synergistic relationship, is greatly enhanced by the allosteric binding of GTP with concomitant enhancement of the enzyme's binding affinity for GTP. Thus, GTP binding serves to coordinate the phosphorylation of UTP with Gln hydrolysis to promote efficient CTP synthesis by LlCTPS.

Structure-Activity Studies
Lunn et al. [72] examined the ability of a variety of GTP analogues to act as allosteric activators of EcCTPS. The structural requirements for activation proved to be quite stringent, with only a few analogues of GTP capable of activating Gln-dependent CTP formation: GTP ≈ 6-thio-GTP > inosine-5 -triphosphate ≈ guanosine-5 -tetraphosphate > O 6 -methyl-GTP > 2 -deoxy-GTP. Overall, the 5 -triphosphate, 2 -OH, and 3 -OH groups were required for full activation, while the 2-NH 2 group was required for binding recognition, and substituents at the 6-position of the purine ring played an important role in activation. On the other hand, the structural requirements for inhibition appeared to be quite lax, with all GTP analogues able to weakly inhibit the utilization of exogenous NH 3 as a substrate by the enzyme. Nucleotide and nucleoside analogues of GTP and guanosine, respectively, inhibited both NH 3 -and Gln-dependent CTP formation to a similar extent. The inhibition appeared to be due solely to the purine base and, with the exceptions of inosine, ITP, and adenosine, was insensitive to the identity of the purine. 8-Oxoguanosine (IC 50 = 80 µM) was initially identified as a good inhibitor, binding 4-fold tighter than guanosine [72]. Subsequent studies supported the notion that only the purine heterocycle was required for inhibition, with xanthines and uric acids also exhibiting inhibition [80]. For example, 1,3,7-trimethyluric acid was identified as a particularly good inhibitor of EcCTPS-catalyzed NH 3 -and Gln-dependent CTP synthesis (IC 50 = 70 µM). Indeed, when this inhibition strategy was applied to CTPS from T. brucei, uric acid (IC 50 = 100 µM) was shown to be a reasonably effective inhibitor of that enzyme [14]. Overall, these observations suggested that the bound purine base was sufficient to inhibit the use of exogenous NH 3 and weakly impede GTP binding and activation.

Site-Directed Mutagenesis and Kinetics Studies: Mapping Regions of CTPS Contributing to GTP-Dependent Effects
Kinetics studies have provided invaluable insights into the role of GTP as an allosteric activator and have implicated several specific residues and regions of CTPS that participate in GTP-dependent activation of the enzyme. For the following discussion, the nomenclature used for structural elements (i.e., loops and lids) is that introduced by Baldwin and coworkers for EcCTPS [32].

The L13 Loop: Limited Proteolysis
Using trypsin-catalyzed proteolysis of EcCTPS, Bearne and co-workers showed that Arg 429 and Lys 432 (Table 1) reside on an exposed loop (L13, Figure 2) [81]. Subsequent site-directed mutagenesis studies revealed that R429A EcCTPS exhibited wild-type levels of NH 3 -dependent CTP formation; however, the values of k cat and k cat /K m for Gln-dependent CTP formation were reduced~10-and~20-fold, respectively, relative to the wild-type enzyme. The K m for Gln was not significantly altered. Activation of the glutaminase activity was reduced 6-fold at saturating concentrations of GTP, and the GTP-binding affinity was reduced 10-fold. These observations indicated that Arg 429 plays a role in both GTP-dependent allosteric activation of Gln hydrolysis and GTP binding [81]. Residues comprising the Cys-Glu-His triad are highlighted in red and the residues comprising the gate at the distal end of the NH 3 tunnel are highlighted in dark blue. Residues highlighted in light green, dark green, cyan, and purple play a role in GTP-dependent activation and correspond to the similarly colored residues in the structure shown in Figure 3A. The secondary structural assignments are those proposed by Baldwin and co-workers for EcCTPS [32,44]. The species include: Escherichia coli (Ec, P0A7E5), Lactococcus lacti (Ll, O87761), Thermus thermophilus (Tt, Q72IN0), Mycobacterium tuberculosis (Mt, P9WHK7), Saccharomyces cerevisiae (Sc, P28274, URA7), Drosophila melanogaster (Dm, Q9VUL1), and Homo sapiens (Hs, P17812, CTPS1).
Arg 429 and Lys 432 lie close to a conserved sequence motif [GG(TS)(ML)RLG)] (shaded purple in Figure 2) within the GAT domain that Willemoës suggested played a role in GTP-dependent activation [85]. This suggestion was based on the observation that the motif was only present in CTPSs and not in the GAT domain of other class I Gln-dependent amidotransferases, which are not regulated by GTP binding. Through site-directed mutagenesis and kinetics studies, Willemoës demonstrated that Thr 431 and Arg 433 within this motif of LlCTPS (Thr 438 and Arg 440 in EcCTPS) play a role in GTPdependent activation of Gln hydrolysis [85]. Since the T431V LlCTPS variant exhibited wild-type affinity for GTP, it was concluded that the hydroxyl group of Thr 431 plays a role only in the structural changes accompanying GTP-dependent activation. However, Willemoës did observe a 10-17-fold decrease in the GTP-binding affinity with the R433A LlCTPS variant, similar to the 10-fold reduction in GTP-binding affinity observed for the R429A EcCTPS variant. Hence, it appears that the conserved sequence motif and adjacent residues of the L13 loop are involved in both GTP binding and activation.

The L4 Loop
Iyengar and Bearne [82] conducted scanning alanine mutagenesis on a highly conserved region between residues 102 and 118 in EcCTPS (shaded light green in Figure 2) and showed that residues located centrally within this region functioned to ensure efficient coupling of Gln-dependent NH 3 formation to the synthase activity. Significantly, the D107A and L109A variants (Table 1) exhibited wild-type levels of Gln hydrolysis and NH 3 -dependent CTP formation; however, Gln-dependent CTP formation was markedly impaired. Interestingly, K A for GTP binding, the GTP-dependent glutaminase activity, the K m for Gln, and UTP-and ATP-dependent tetramerization were not altered significantly by the amino acid substitutions. These observations suggested that the D107A and L109A substitutions altered the normal structure of the NH 3 tunnel (i.e., the tunnel is leaky or constricted) or caused a structural perturbation that prevented the formation of a functioning tunnel, the effect being more pronounced for the L109A variant [82]. To test the hypothesis that the L109A substitution was in some way constricting the NH 3 tunnel, Lunn and Bearne [83] compared the kinetic parameters of Gln and NH 3 with those of the corresponding bulkier substrates L-γ-glutamyl hydroxamate (Gln-OH) and NH 2 OH, as well as the coupling ratios (Equations (4) and (5)) observed for Gln and Gln-OH with the wild-type and L109A EcCTPSs. Interestingly, the L109A variant exhibited greater uncoupling with the bulkier nascent NH 2 OH, derived from Gln-OH hydrolysis, than with NH 3 derived from Gln hydrolysis. These observations were consistent with the L109A variant possessing a constricted or malformed NH 3 tunnel. sub − saturating couling ratio = (k cat /K m ) CTP formation (k cat /K m ) glutaminase activity (4) saturating couling ratio = (k cat ) CTP formation (k cat ) glutaminase activity (5) These authors also examined the kinetic properties of the L109F variant to see if the NH 3 tunnel could be fully blocked. This substitution, however, yielded a marked reduction in the glutaminase activity arising from either a local perturbation at the Gln site such that Gln binding was not affected but its rate of hydrolysis was reduced, or a failure of GTP binding to induce the appropriate conformational change that promotes hydrolysis of Gln. The authors favored the latter scenario since the K A value for GTP binding was similar to that of the L109A variant, indicating that the loss of glutaminase activity did not arise from impaired binding of GTP. A similar kinetic profile for the R105A variant also suggested that Arg 105 was required for efficient Gln turnover [82].

The Lid L11 Loop
Based on the structures of EcCTPS (vide infra) [32] and TtCTPS [45], Willemoës and co-workers identified a loop region denoted as lid L11 (residues 354-362 of EcCTPS, shaded dark green in Figure 2) [32], which immediately preceded the amino acid residues comprising the oxyanion hole as a potential structural motif involved in the GTP-dependent activation of CTPS [84]. Indeed, some of the residues of lid L11 were not detected in electron density maps [45], indicating that this loop is highly flexible. The R359M, R359P, and G360P LlCTPS variants (R356 and G357 in EcCTPS) exhibited 10-to 50-fold reductions in GTPdependent activation of Gln-dependent CTP formation accompanied by 4-to 10-fold increases in the K A for GTP (Table 1). While the uncoupled glutaminase activities for these variants were not activated by GTP, the G360A variant was about 2-fold more active than wild-type LlCTPS [84]. Overall, these observations suggested that interaction of GTP with the lid L11 is required for the activation of Gln-dependent CTP synthesis. Indeed, deletion of the charge of the side chain in the R359M LlCTPS variant supported the notion put forward by Baldwin and co-workers [32] that Arg 359 interacts with GTP and may act as a lever to alter the conformation of the lid L11, leading to a conformation favoring Gln hydrolysis [84].
Interestingly, one of the first site-directed mutagenesis studies to probe the role of residues in the GAT domain in the region of the lid L11 was conducted by Weng and Zalkin [31]. These authors altered Gly 352 in the EcCTPS GAT domain, showing that the G352P variant (Table 1) exhibited no detectable Gln-dependent CTP formation but was almost fully active when NH 4 Cl was the ammonia source. Furthermore, the catalytic Cys 379 at the active site could not be labeled by DON. Although these authors suggested that the mutation either prevented formation of the covalent glutamyl-enzyme intermediate or disrupted NH 3 transfer, their results did not preclude the possibility that GTP-dependent activation was disrupted. Only later did the structure of EcCTPS reveal that Gly 352 resides at the junction between the rigid portion of the oxyanion hole and the lid L11 loop. Substitution of Gly by Pro at position 352 likely abrogated the activation of the glutaminase activity by hindering the flexibility of the lid L11 region [84].

"Pinching" the NH 3 Tunnel
The X-ray crystal structures of EcCTPS (with no GTP bound, vide infra) revealed that the NH 3 tunnel had a constriction formed by three residues (Pro 54, His 57, and Val 60, shaded dark blue in Figure 2) at the entryway to the synthase domain that are highly conserved among CTPSs [32,44]. To permit the passage of NH 3 , which has a molecular diameter close to 4 Å, the constriction must open since it is only~2.4 Å in diameter in the crystal structures. The proximity of this putative molecular gate to the proposed GTP-binding site suggested that GTP might play a role in controlling the opening of the gate to regulate the delivery of NH 3 to the synthase domain. By conducting kinetics and biophysical analyses on EcCTPS variants with different amino acid substitutions of the gate residue Val 60, which resides at the most constricted part of the NH 3 tunnel, McCluskey and Bearne [86] explored the role of the putative NH 3 gate in coordinating the reactions required for CTPS catalysis. All amino acid substitutions (i.e., Ala, Cys, Asp, Trp, and Phe) at position 60 appeared to cause local structural perturbations that had a detrimental effect on the ability of GTP to bind and activate the enzyme, as anticipated based on the proximity of Val 60 to the GTP-binding site [32,[82][83][84]. Most interestingly, these authors discovered that replacement of Val 60 by the bulkier Phe residue unveiled the coordinated role of both GTP binding and the glutaminase activity in facilitating the passage of NH 3 through the tunnel gate. The V60F variant exhibited a slightly reduced coupling efficiency at maximal glutaminase activity that was ameliorated by slowing the rate of Gln hydrolysis, suggesting a "bottleneck" effect. Moreover, the inability of V60F EcCTPS to use exogenous NH 3 as a substrate could be overcome in the presence of GTP, and more so if the enzyme was modified by DON. When NH 2 OH was employed as an alternative, bulkier substrate, its use was more efficient when it was concomitant with the glutaminase reaction. Hence, GTP-dependent activation appears to act in concert with the glutaminase activity to open the molecular gate for the passage of NH 3 [86].

Regulatory Effects of NADH
Baldwin, Kollman, Gitai, and co-workers observed that nicotinamide-containing compounds were modest inhibitors of EcCTPS with IC 50 values following the trend: 1-methyl 1,4-dihydronicotinamide (IC 50 = 140 µM) < NADH ≈ NADPH << NADP + < NAD + << 1-methyl nicotinamide (IC 50~4 000 µM), indicating that the reduced nicotinamide ring is sufficient for inhibition [87]. NADH inhibition was enhanced in the presence of increasing concentrations of GTP, including those concentrations for which GTP is either activating or inhibitory. Correspondingly, in the presence of a fixed concentration of NADH, the dependence of the activating and inhibitory effects of GTP were more acute (i.e., lowering both the EC 50 and IC 50 values for activation and inhibition, respectively). Hence, NADH enhanced the allosteric regulation of EcCTPS by GTP in a mutual fashion. Furthermore, these authors found that increasing the concentration of GTP enhanced the inhibition by CTP; conversely, in the presence of CTP, the EC 50 and IC 50 values for GTP were reduced. The observations that NADH, CTP, and GTP mutually enhanced each other's effects suggested that they interact with a common intermediate enzyme state. Finally, Barry et al. [53] demonstrated that CTP was a more effective inhibitor under conditions that favored filament formation by the enzyme (i.e., elevated concentrations of CTP or the enzyme). Using the E227R EcCTPS variant, which obviates the ability of the enzyme to form filaments and reduces the inhibitory effects of CTP, Baldwin, Kollman, Gitai, and co-workers observed that the EC 50 and IC 50 values were shifted to higher values for GTP-dependent activation and inhibition, respectively [87]. This intriguing observation suggested that GTP interacted with the inhibitory filament, which was an insightful claim, considering that at that point, GTP had not been observed in any of the high-resolution cryo-EM reconstructions of CTPS filaments [87]. Although NADH did not appear to bind at the UTP-and CTP-binding sites, it remains unclear where NADH binds to exert its effects on the GTP-dependent activity of EcCTPS [44,87]. Overall, these observations highlight the important regulatory role that GTP plays by inducing conformational changes that affect the binding and activity of other nucleotide ligands, which is also consistent with the notion that the 4-phospho-UTP and GTP could act as coactivators of the glutaminase reaction as pointed out by Willemoës [36].

Structural Studies Intimating the Location of the GTP-Binding Site and the Role of GTP-Dependent Activation in NH 3 Translocation
A breakthrough in the study of CTPS occurred when the first X-ray crystal structures of CTPS were reported in 2004. Baldwin and co-workers reported the structure of EcCTPS in its ligand-free form [32] and Hirotsu and co-workers reported the structure of TtCTPS with bound Gln [45]. A year later, Baldwin's laboratory reported the structure of the EcCTPS·ADP·CTP product complex [44]. These seminal structures helped to delineate the spatial arrangement of the site of Gln hydrolysis, the NH 3 tunnel, the binding of nucleotides at the synthase site, and the protein oligomerization interfaces, as well as inform the design and interpretation of site-directed mutagenesis studies. Unfortunately, no electron density for GTP was present, despite GTP being present in the crystallization solution for the EcCTPS structures [44,87]. Hence, few direct insights were provided regarding the conformational changes accompanying GTP binding. However, by comparing the GAT domain with structurally related GTP-binding proteins (e.g., EF-Tu and EF-G), Baldwin and co-workers [32] identified a potential binding site for GTP located adjacent to the entrance to the glutaminase active site in a deep cleft residing between the GAT and synthase domains close to the A-A' dimer interface (Figure 3). The cleft is formed by residues 50-55, 297-301, 353-356, 438-441, 468, and 470. Baldwin and co-workers constructed a model of GTP bound at this site ( Figure 3A) [32], which has permitted the results from site-directed mutagenesis and kinetics experiments to be put into context. Indeed, this binding site has now been confirmed with the first structure of CTPS with GTP bound (vide infra) [66].  Figure 2. The light green and magenta arrows suggest movement of the similarly colored residues upon GTP binding. Clearly, the enzyme must undergo a change in conformation (suggested by arrows) from that shown so that the protein interacts with the 6-position of the purine ring since GTP-dependent activation of Gln-dependent CTP formation requires the presence of a carbonyl group at the 6-position of GTP. (B) Conservation of residues based on a multiple sequence alignment prepared and rendered in surface representation using the CONSURF server [88] is shown for two chains of the CTPS tetramer. Highly conserved residues (purple) line the GTP-binding cleft denoted by the purple arrow. (C) Electrostatic potential surface of the CTPS tetramer (surface representation) is shown with the same orientation as in panel B. The blue arrow denotes the electropositive region that binds the 5 -triphosphate moiety of GTP.
The GTP-binding site resides in a region of highly conserved amino acid sequence ( Figure 3B), which exhibits a predominantly negative electrostatic potential surface for interacting with the 5 -triphosphate moiety of GTP ( Figure 3C). With the exception of the amino acid residues on the flexible loop L13 between Val 428 and Thr 438, all the conserved residues listed in Table 1 that were implicated in the GTP-dependent activation of Gln-dependent CTP formation by site-directed mutagenesis experiments line this cleft ( Figure 3A). In the structure, loop L4 (EcCTPS residues 105-114) resides between the GAT and synthase active sites and forms part of the rim of the GTP-binding site. Arg 468, Arg 470, and Glu 68 appear to interact with the purine moiety of GTP [32]. Significantly, Leu 109 is located on a flexible loop of an adjacent subunit, which extends over the deep cleft separating the GAT and synthase sites. This places Leu 109 above the opening that Endrizzi et al. [32] had identified as a putative entry point for NH 3 to access the solventfilled "vestibule" connecting the GAT active site and the GAT/synthase interface. Leu 109 is not interacting with GTP in the model. However, the kinetic properties of the L109A EcCTPS variant suggested that the residue is involved with GTP-dependent activation of the enzyme. Clearly, Leu 109 does not directly cause a constriction in the NH 3 tunnel, but substitution by Ala alters the kinetics by interfering with the appropriate conformational changes that accompany GTP binding and proper formation of the NH 3 tunnel. As such, the resulting tunnel may still be constricted or malformed. Baldwin and co-workers suggested that a polypeptide linker connecting the L4 loop and the symmetry-related subunit could mediate coupling between ATP-binding and Gln hydrolysis [44]. Most interestingly, the loop L4 residues Arg 105 and Leu 109 [82], as well as the loop L13 residue Arg 429 [81], do not make any key contacts with other residues in the apo-EcCTPS structure, consistent with the numerous studies that suggest that GTP binding induces substantial conformational changes in EcCTPS [34,64,71,75,89] and LlCTPS [36]. Such changes may serve to appropriately reposition the triad residues His 515 and Glu 517 for catalysis and open the NH 3 tunnel [86].
Baldwin and co-workers also noted that a 3-Å wide entrance to a solvent-filled "vestibule" that connects the GAT active site and the GAT/synthase interface is present at the base of the GTP-binding cleft [32]. That GTP binds and occludes this entryway for exogenous NH 3 was supported by the inhibition of NH 3 -dependent CTP formation by elevated concentrations of GTP [71,72,75]. Furthermore, GTP could serve as a "bung" to prevent the loss of nascent NH 3 and ensure the fidelity of NH 3 transfer to the synthase domain.
Analysis of the X-ray structure of TtCTPS [45] suggested that the binding of ATP and UTP induce a conformational change that brings the GAT and synthase domains closer together, permitting formation of the NH 3 tunnel. Indeed, the various residues that have been implicated in GTP-dependent activation of the glutaminase activity (especially L109A in EcCTPS) are brought closer together in the structure. This observation suggests that an even more prominent structural transition could be associated with GTP binding. Interestingly, Hirotsu and co-workers did not observe the presence of an NH 3 tunnel in the structures of TtCTPS or the TtCTPS·3SO 4 2− and TtCTPS·Gln complexes, leading them to conclude that TtCTPS must undergo an extensive conformational change upon binding ATP and UTP [45]. Similarly, a patent NH 3 tunnel was not observed in the structure of MtCTPS [15]. Based on the site-directed mutagenesis and kinetics studies outlined above and sequence conservation, Hirotsu and co-workers [45] identified two consensus sequences associated with GTP binding and activation: residues 111-130 (residues 102-121 in EcCTPS, partly shaded light green in Figure 2) in the synthase domain and residues 438-444 (residues 436-442 in EcCTPS, shaded purple in Figure 2) in the GAT domain. These two consensus sequences are located 15-25 Å apart in the TtCTPS structure, and changes in conformation upon the binding of ATP and UTP were suggested to move these sequences closer together to form the GTP-binding site (arrows, Figure 3A).

GTP-Dependent Activation of CTPSs from Other Species
While the GTP-dependent activation of EcCTPS and LlCTPS have been studied in the most detail as outlined above, the activation of Gln-dependent CTP formation by GTP has either been noted or partially characterized in a variety of other organisms (Table 2).     [95]. Arrows denote changes in activity.

Thermus thermophilus
Although structural information is available for TtCTPS, only limited kinetics studies have been conducted with the enzyme, revealing that it is activated by GTP with a Hill number of 2.1 [45].

Chlamydia trachomatis
Limited kinetics studies revealed that CtCTPS is activated by GTP with a K A value of 7 µM [92].

Trypanosoma brucei
Fijolek et al. [9] found that GTP activated Gln-dependent CTP formation catalyzed by TbCTPS with a K A value of 70 µM. Steeves and Bearne [14] also observed activation of TbCTPS by GTP, but inhibition at concentrations above 0.2 mM, with values of k o , k act , K A , K i , and n equal to 0.16 s −1 , 0.87 s −1 , 57 µM, 272 µM, and 4.2, respectively. Similar to EcCTPS, purine derivatives such as GTP, guanosine, caffeine, and uric acid inhibited TbCTPS with IC 50 values of 460, 380, 480, and 100 µM, respectively.

Giardia intestinalis
GiCTPS from the intestinal protozoan parasite G. intestinalis is activated by GTP (K A = 60 µM), although, whether the requirement for GTP was absolute or not was not established [94]. Additionally, GDP (K A = 0.10 mM) and dGTP (K A = 0.23 mM) were found to activate the enzyme but not to the same degree as GTP (GTP > GDP > dGTP). The concentration of GTP did not affect the apparent dissociation constants for ATP, UTP, or Gln. Interestingly, the amino acid sequence of GiCTPS contains two inserts in the synthase domain and one in the GAT domain that are not present in the yeast, human, and E. coli enzymes [12]. The role of these additional amino acid inserts is unknown.

Plasmodium falciparum
Curiously, Pf CTPS from the human parasite P. falciparum contains two inserted stretches of 42 and 233 amino acids within the GAT domain, suggesting that this novel difference from other CTPSs might be capitalized on for drug development [10]. The gene encoding Pf CTPS containing partial synthetic sequences substituted with preferred codons utilized by E. coli has been expressed heterologously in E. coli and the resulting enzyme shown to be active [93]. Although GTP was claimed to activate Pf CTPS, no data were reported.

Saccharomyces cerevisiae
There are two isoforms of CTPS in S. cerevisiae encoded by the URA7 and URA8 genes [106,107]. For ScCTPS7 encoded by URA7, which is responsible for the majority of CTP made in vivo [107], activation by GTP was not found to be an absolute requirement for CTPS activity but did enhance the activity 4-fold [96]. Double-reciprocal plots of the initial velocity data, obtained with Gln as the variable substrate with various fixed concentrations of GTP, exhibited concave-downward curves, consistent with negative cooperative kinetics with respect to Gln and GTP. Carman and co-workers conducted a series of studies examining the phosphorylation of ScCTPS7 by protein kinases A (PKA) and C (PKC) [47,[97][98][99]108]. Phosphorylation by PKA at Ser 424 [58] in the GAT domain [99] resulted in a slight decrease in the K A value for GTP from 15.3 to 12.0 µM, but did not have a significant effect on the apparent V max and K m values with respect to Gln when measured either in the absence or presence of GTP. Interestingly, the phosphorylation did eliminate the negative cooperativity exhibited with respect to Gln [98]. This behavior was also observed when the enzyme was phosphorylated by PKC, including a slight decrease in the K A value from 15.7 to 12.2 µM [97]. Finally, to assess the role of the mobile L11 loop adjacent to the allosteric GTP-binding cleft in filament formation by ScCTPS7, the abilities of the GFP-tagged variants R381M, R381P, and G382A (cf. Arg 356 and Gly 357 in EcCTPS) to form filaments in vivo were examined. Interestingly, these variants exhibited a~3.1-fold increase in the number of yeast cells forming filaments compared with strains expressing wild-type ScCTPS7. Moreover, the median length of the filaments was increased by 33% and 15% for the R381M and R381P variants, respectively, relative to the wild-type enzyme. The G382A variant also exhibited an increased number of cells with filaments; however, the median length of the filaments was reduced by 25%. To date, this is the only report of residues at the GTP-binding site contributing to regulation of both the frequency of filament formation and filament length [58].
The ScCTPS8 isoform was also activated by GTP and, again, the requirement for GTP was not absolute [74]. Binding of GTP did not alter the enzyme's affinity for Gln and, at a saturating concentration of Gln, GTP stimulated Gln-dependent CTP formation by 12-fold (K A = 26 µM).

Drosophila melanogaster
Gln-dependent CTP synthesis catalyzed by DmCTPS is activated by GTP [66]. Interestingly, the substitution of Leu 444 (Leu 435 in EcCTPS) by Ala at the GTP binding site (vide infra) markedly reduced the activating effect of GTP. In addition, substitution of Ala for Phe 50 (Met 52 in EcCTPS), which resides on the L2 loop (shaded yellow in Figure 2) and forms part of the NH 3 tunnel, also markedly reduced the activating effect of GTP [66].

Plants
The kinetic properties of CTPSs from plants have not been extensively characterized. Interestingly, the enzyme from Arabidopsis thaliana (AtCTPS) occurs as five isoforms and six putative isoforms that have been identified in rice (endospermless variant from Oryza sativa var. japonica cultivar Hwayoung) [90]. AtCTPS1-4 are expressed throughout the plant tissues, while AtCTPS5 is expressed only in developing embryos [91]. This enzyme also undergoes filament formation as previously demonstrated for the yeast, fruit fly, human, and E. coli enzymes (vide infra). GTP was shown to activate AtCTPS3 with an increase in the concentration of GTP from 0.05 mM to a saturating concentration of 1.0 mM GTP, increasing the rate of Gln-dependent CTP formation nearly 5-fold with an apparent K A value of~0.25 mM.

Mammals
Human CTPS exists as two isoforms, hCTPS1 and hCTPS2 [109,110]. Comparison of the effects of GTP concentration on hCTPS1 and hCTPS2 activity revealed that GTP stimulated Gln-dependent CTP formation with K A values of 1.6 and 8.0 µM, respectively, while concentrations of GTP exceeding 10 mM inhibited the activity of both isoforms [100]. Earlier studies with hCTPS isolated from homogenates of HL-60 cells revealed negative cooperativity with GTP (n = 0.7). However, this was a crude preparation and likely contained a mixture of isoforms [111].
Initial studies on CTPS purified 200-fold from Ehrlich ascites tumor cells (mouse) revealed that the enzyme was active in the absence of GTP and that GTP activated the enzyme~5-fold with maximal activation occurring at 1.0 mM GTP (K A = 50 µM) [104]. The K m value for Gln was not significantly altered by the presence of GTP. GTP had a slight inhibitory effect on the activity with exogenous NH 3 . Follow-up studies showed that GTP accelerated the glutaminase activity of the enzyme about 2.5-fold in the absence of other substrates and about 45-fold in the presence of UTP and ATP [65]. Curiously, these authors also reported that GTP could substitute for ATP as a substrate (K m = 2.2 mM), even after confirming by HPLC analyses that their UTP and GTP preparations were both free of any ATP impurity. GMP, GDP, and dGTP were also capable of partially activating the enzyme, with the latter compound also serving to replace ATP [105]. Finally, the non-hydrolyzable GTP analogue Gpp[NH]p was shown to activate the enzyme by stimulating the glutaminase activity only~1.4-fold or~1.8-fold in the absence or presence, respectively, of both UTP and App[NH]p. However, in the presence of UTP and ATP, the glutaminase activity with concomitant CTP formation was enhanced~32-fold. These results are not unlike those obtained by Willemoës and co-workers working with LlCTPS [36], indicating that the presence of UTP and ATP is required for the maximum activation of the glutaminase activity by GTP or Gpp[NH]p. Bovine calf liver CTPS was shown to not have an absolute requirement for GTP, but in the absence of GTP, the rate was reduced~7-fold relative to the rate observed at saturating concentrations of GTP [63,101]. The K A value was reported to be 70 µM (n = 1.1), the same as the apparent K A value accompanying GTP-dependent activation of crude preparations of rat liver CTPS (Table 2) [18]. GTP did not promote tetramerization of the enzyme [63,101,102].
Among the various species for which the kinetics of GTP-dependent activation have been examined, the K A values range from 1.6 µM to~250 µM (Table 2). Compared to the intracellular concentrations of GTP, which are typically 468 ± 224 µM [112], it appears that CTPSs would be operating at near maximal activation by GTP under typical physiological conditions. Since intracellular concentrations of GTP in normal human cells and human tumor cells are 232 ± 202 µM and 473 ± 214 µM [112], respectively, this appears to be especially true for hCTPS1 with the lowest K A value reported to date [100]. However, the following observations present a possible caveat to this generalization. Using a mutant mouse T lymphoblast cell line (FURT-1A), which contained a CTPS refractory to complete inhibition by CTP, Aronow and Ullman [103] showed that exogenous manipulation of levels of GTP between 10% and 200% of that of unperturbed cells by either mycophenolic acid or guanosine revealed that the cooperativity of the enzyme with respect to GTP activation was much greater in situ (n = 1.59) than in vitro (n = 0.76). Furthermore, the concentrations of GTP required to activate the enzyme in situ were much greater than those in vitro.
In order to explore the molecular biological basis of cytoophidia, Liu and co-workers purified DmCTPS and found that the protein can form two types of filamentous structures in vitro with different conformations in the presence of substrates ATP, UTP, or product CTP ( Figure 5A,B) [66]. In these two types of filaments, DmCTPS is present in tetrameric units, which interact with each other through an interface in the GAT domain to form the stable filamentous structures. By comparing amino acid sequences of CTPSs, Liu and colleagues discovered that these structures are conserved among multiple species including human, fly, and yeast. At the same time, cryo-EM structures of the two CTPS isoforms in human, hCTPS1 and hCTPS2 [56,61], and the two CTPS isoforms in budding yeast, URA7 and URA8 [121], revealed that these CTPSs are assembled into filamentous structures through the same region ( Figure 5C). Interestingly, there is an insertion of about 15 amino acids between β sheets 2 and 3 of the GAT domains from Drosophila and human CTPSs compared with the GAT domain from other class I Gln-dependent amidotransferases, which makes Drosophila and human GAT domains form an additional short helix and helix2 extensions. In DmCTPS, the adjacent GAT domains are used to form filamentous structures through the interaction of helix2 and surrounding amino acids. Indeed, a single point mutation yielding the H355A DmCTPS variant is sufficient to disrupt this interaction and affect the formation of filamentous structure ( Figure 5D) [66]. This role of the GAT domain in filament forming ability obtained through amino acid insertion undoubtedly provides a new level for the regulation of the enzyme's catalytic activity [56,61,66,121]; however, the role of filament formation on the GAT domain itself and its regulation of catalytic activity in the GAT domain remain unclear. In budding yeast, the GAT domains of URA7 and URA8 are bound with different strengths and interact via different amino acids at this insertion [121]. Therefore, regulation of the activity of the GAT domain through the insertion in this region may vary between species.

CTPS with Bound GTP: Confirmation of the Location of the GTP-Binding Site and Relationship to Other Cryo-EM Structures
Despite the various kinetics and site-directed mutagenesis experiments conducted to delineate the structural and functional aspects of the GTP-dependent regulation of CTPS activity, specific structural information on the GTP binding site and the conformational changes that accompanied GTP binding remained elusive. In 2021, the first structure of CTPS with bound GTP was captured using cryo-EM ( Figure 6) [66]. This structure of DmCTPS was covalently modified at the GAT domain by the Gln analogue DON and revealed a tightly closed state with the CTPS tetramers forming a filamentous structure through the amino acid insertion mentioned above. As predicted by Baldwin and coworkers [32], GTP is bound in a cleft between the GAT domain and the synthase domain ( Figure 3). The GTP molecule is tightly surrounded by the gate (which forms a part of the NH 3 tunnel), lid L11, Phe 373 (which encapsulates the catalytic triad of the GAT domain), and the L4 loop 105-110 of an adjacent protomer (Figure 7). Compared with other CTPS structures in the tightly closed state, the DON-modified tetramer with bound GTP exhibits an even tighter conformation, which may be related to the formation and tight connection of the NH 3 tunnel. Intriguingly, Liu and co-workers also observed the presence of 4-phosphorylated UTP intermediate in their cryo-EM structure of DmCTPS [66]. This observation suggests that the first step of the catalytic reaction (i.e., ATP-dependent phosphorylation of UTP) has been completed in the synthase domain, and the intermediate is waiting for NH 3 to be transferred from the GAT domain to complete the final amination reaction.  shown in space-filling representation from the cryo-EM structure solved by Liu and co-workers [66]. Shown in stick representation are residues implicated in GTP-dependent activation of CTPS, which are highlighted in light green (L4), dark green (L11), and purple (L13 or "wing") (cf. coloring in Figure 2), and orange (gate region). (B) GTP-binding pocket of DmCTPS is depicted by a transparent surface. GTP and DON are shown in ball and stick representation, as are residues that interact with GTP or play a role in GTP-dependent conformational effects: F50, H55, L107, F373, R376, L444, M448, R429, and R481 (colored as in Figure 2). The Mg 2+ ion is shown as a light green sphere in both panels.
GTP also plays an important role in the formation of the NH 3 tunnel. Upon simulating the removal of GTP from their structure, Liu and co-workers found that there is a hole in the NH 3 tunnel adjacent to where GTP binds, as originally suggested by Baldwin and co-workers [32]. This observation indicates that GTP can indeed prevent leakage of NH 3 from the ammonia tunnel and appears to account for why GTP inhibits the reaction with exogenous NH 3 when CTPS is covalently modified by DON, as shown in previous studies [36,64,86]. Interestingly, the NH 3 tunnel formed in GTP-bound DmCTPS is narrower than that in CTPS bound with the products or some other substrates.
Compared with carbamoyl-phosphate synthetase 2, which is also a class I Gln-dependent amidotransferase, there is an insertion of about eight amino acids between β sheets 5 and 6 in the GAT domain of CTPS, which has been referred to as the "wing" (residues 440-448 in DmCTPS and 431-439 in EcCTPS; L13 purple region in Figure 2) [66]. Many of the previous crystal and cryo-EM structures of CTPSs have shown that this wing is very flexible, making it very challenging to determine its precise structure and thereby limiting our understanding of its role in GTP-dependent activation. Utilizing DON-modified DmCTPS with bound GTP, Liu and co-workers were able to obtain a clear structure of the wing of DmCTPS using cryo-EM [66]. The wing directly contacts GTP, suggesting that GTP binding evokes a conformational change that stabilizes the wing region. Similar to the R429A EcCTPS variant [81] and the T431V and R433A LlCTPS variants [85], the L444A substitution obviated the regulatory effect of GTP on DmCTPS activity [66]. This further shows that the integrity of the wing is necessary for the GTP-dependent regulation of CTPS activity and furnishes a structural explanation for the results obtained from previous kinetics studies (Section 2.3.1).

A Model for GTP-Dependent Regulation of CTPS
With the report of the GTP-bound structure of DmCTPS [66], combined with the results from the site-directed mutagenesis and kinetics studies described in Sections 2 and 3, a more comprehensive model for the role of GTP in regulating CTPS catalysis emerges. While the model described here incorporates features observed for CTPSs from different organisms, subtle differences likely exist between various CTPSs. As Liu and co-workers noted, the initial binding of UTP and ATP effects the open-to-closed state transition that alters the conformation of the GTP-binding cleft to facilitate GTP binding [66]. Concomitantly, ATP and UTP binding cause a rotation of the GAT and synthase domains to align their active sites for formation of the NH 3 tunnel [56]. In LlCTPS, Willemoës and co-workers proposed that GTP binding not only closes a lid over the active site to effect proficient hydrolysis of Gln, but there is also concomitant formation of the 4-phospho-UTP intermediate, which causes allosteric activation of the glutaminase activity [36]. This reciprocal activating effect of GTP and the intermediate is consistent with the observed presence of the 4-phospho-UTP intermediate in the GTP-bound structure of DON-modified DmCTPS [66].
Within the GAT domain, a series of interactions contribute to altering the conformation of the GAT domain. Gln binding orders Phe 373 (DmCTPS numbering, 353 in EcCTPS) of the L11 lid to afford an additional stabilizing interaction with bound GTP. In addition to interacting with GTP, Leu 107 (109 in EcCTPS), from the L4 region of an adjacent protomer, interacts with Leu 444 (435 in EcCTPS) located in the unique wing structure (L13) in the GAT domain to propagate the conformational change accompanying GTP binding. Arg 376 (Arg 359 in LlCTPS) interacts with GTP and appears to act as a lever to alter the conformation of the lid L11 to favor Gln hydrolysis [84]. The interaction between Phe 50 and GTP appears to play a role in decreasing the flexibility of the L2 region, thereby ensuring the integrity of the NH 3 tunnel while GTP is bound. The hole from which the nascent NH 3 could be lost to bulk solvent is effectively "capped" by the bound GTP, ensuring the fidelity of the transfer of NH 3 to the synthase domain.
Thus, GTP not only promotes the hydrolysis of Gln through its multiple interactions with the GAT domain of CTPS, but also participates in the assembly and maintenance of the NH 3 tunnel, and thereby coordinates the reactions catalyzed by the GAT and synthase domains of the enzyme. Although direct experimental evidence on the specific mechanism for GTP-dependent regulation of catalysis at the GAT domain of CTPS is still lacking, we posit that the GAT domain in CTPS forms a unique regulatory system around GTP binding that is both efficient and sensitive.