Ammonia Channel Couples Glutaminase with Transamidase Reactions in GatCAB

See allHide authors and affiliations

Science  30 Jun 2006:
Vol. 312, Issue 5782, pp. 1954-1958
DOI: 10.1126/science.1127156


The formation of glutaminyl transfer RNA (Gln-tRNAGln) differs among the three domains of life. Most bacteria employ an indirect pathway to produce Gln-tRNAGln by a heterotrimeric glutamine amidotransferase CAB (GatCAB) that acts on the misacylated Glu-tRNAGln. Here, we describe a series of crystal structures of intact GatCAB from Staphylococcus aureus in the apo form and in the complexes with glutamine, asparagine, Mn2+, and adenosine triphosphate analog. Two identified catalytic centers for the glutaminase and transamidase reactions are markedly distant but connected by a hydrophilic ammonia channel 30 Å in length. Further, we show that the first U-A base pair in the acceptor stem and the D loop of tRNAGln serve as identity elements essential for discrimination by GatCAB and propose a complete model for the overall concerted reactions to synthesize Gln-tRNAGln.

Accurate translation of the genetic code into proteins requires high efficiency and fidelity of tRNA-aminoacylation (1). This vital process is governed by a fundamental family of enzymes called aminoacyl-tRNA synthetases (aaRS), which catalyze the direct attachment of amino acids to the 3′-ends of the corresponding tRNAs. However, Gln-tRNAGln is synthesized by two different pathways in nature (24). In the eukaryotic cytoplasm and in some bacteria, glutamine is ligated directly to tRNAGln by glutaminyl-tRNA synthetase (GlnRS), whereas a noncanonical indirect pathway is employed in the majority of bacteria and all archaea, which do not possess GlnRS, through a two-step process (2, 5). During the first step, tRNAGln is misacylated with glutamic acid by a nondiscriminating glutamyl-tRNA synthetase (GluRS), which recognizes both tRNAGlu and tRNAGln. Thereafter, the mischarged Glu-tRNAGln is transformed into Gln-tRNAGln by Glu-tRNAGln–dependent amidotransferase (Glu-AdT). Glu-AdT converts Glu-tRNAGln into Gln-tRNAGln by initially activating Glu-tRNAGln into γ-phosphoryl-Glu-tRNAGln at the expense of ATP, which is subsequently transamidated into Gln-tRNAGln using ammonia generated by hydrolysis of glutamine or asparagine. Moreover, Asn-tRNAAsn can also be synthesized through such an indirect pathway by Asp-tRNAAsn–dependent amidotransferase (Asp-AdT) in organisms lacking asparaginyl-tRNA synthetase (6, 7).

Bacterial Glu-AdTs are heterotrimeric proteins composed of A, B, and C subunits, and have therefore been named glutamine amidotransferase CAB (GatCAB), while archaea employ the heterodimeric GatDE enzymes (8, 9). Although bacterial GatCAB acts as both Glu- and Asp-AdT, GatDE specifically recognizes only Glu-tRNAGln (9). It is intriguing that the three domains in nature use distinct enzymes to synthesize Gln-tRNAGln in protein synthesis. Based on the primary sequences analysis, GatA is homologous to the amidase enzymes, whereas GatD is closer to the l-asparaginases (9, 10). Therefore, GatA and GatD are structurally different but are assumed to play the same role as the glutaminase subunit for ammonia production. On the other hand, GatB is highly related exclusively to its archaeal counterpart, GatE (9, 10). GatE alone can produce γ-phosphoryl-Glu-tRNAGln, which suggests that the activation of Glu-tRNAGln is achieved only by the GatB or GatE subunits (11). Notably, the glutaminase activity and the activation of Glu-tRNAGln into γ-phosphoryl-Glu-tRNAGln are tightly coupled upon the binding of Glu-tRNAGln (11, 12). Hence, these modules of distinct functional enzymes were assembled to form a more sophisticated protein complex and acquired novel mechanisms to expand the genetic code during evolution.

Recently, the α2β2 tetramer structure of archaeal GatDE from Pyrococcus abyssi was reported at 3.0 Å resolution (10); however, the precise mechanisms for each of the overall reactions coupling the glutaminase with the kinase and transamidase activities, regulated by Glu-tRNAGln, remain elusive. To address this issue, we initiated the crystal structure analysis of bacterial GatCAB from the pathogenic bacterium, S. aureus Mu50 strain. The crystal structure of apo GatCAB was solved at a resolution of 2.5 Å, by the single-wavelength anomalous diffraction (SAD) method, and refined to a final R-factor of 23.8% (R-free = 27.5%) using the LAFIRE program running with CNS (1315) (Fig. 1A). The asymmetric unit contained one GatCAB heterotrimeric molecule, in which the flexible C-terminal region of GatB (residues 412 to 475) was disordered and could not be modeled. The structure immediately highlights an important role of the C subunit as a stabilizer of the protein complex. GatC is featured by an unstructured extended loop flanked by two helices at its N terminus and two β strands at its C terminus (Fig. 1B). GatC wraps around the interface region as a belt and makes extensive interactions with both GatA and GatB. The amphipathic Cα1 and Cα2 helices form a helical bundle with the hydrophobic core of GatA (Aα9-11, 13), thus relieving the local hydrophobicity (Fig. 1C). The internal loop region of GatC crosses over the loop-rich side of GatB, stabilized through a hydrogen bond network by two invariant Arg64 and Asp66 residues of GatC (Fig. 1D). The association of GatAB complex is further assured by the hydrophobic interactions of Cα3 helix of GatC with GatAB. Interestingly, the C-terminal Cβ1-2 strands make an antiparallel sheet with a β hairpin (Bβ8-9) of GatB, forming a four-stranded sheet, resembling the corresponding region in GatDE (Fig. 1E).

Fig. 1.

Bacterial GatCAB complex fastens a molecular belt. (A) Front-view ribbon diagram of the overall structure of S. aureus GatCAB/glutamine complex at 2.3 Å resolution, depicted in three different colors for each subunit: blue, green, and magenta for GatA, GatB, and GatC, respectively. Glutamine in the active site of GatA is drawn as yellow stick representations, whereas the purple sphere is the magnesium ion found in the active site of GatB. ADP from the cocrystal structure with ADP–AlF4 is shown together as pink sticks. This color code is used throughout all the figures. (B) Top view of annularly shaped GatC. (C) Amphipathic helices at the N terminus of GatC form a helical bundle with the hydrophobic core of GatA (hydrophobic residues are colored gray). (D) Detailed interactions between the internal loop region of GatC and GatAB complex. Conserved residues involved are labeled, with hydrogen bonds indicated (≤3.2 Å, dashed black line). Polar interactions are prominent on the C-terminal side, whereas the hydrophobic interactions are clustered on the opposite side. (E) The C terminus of GatC tightens GatAB complex, constructing an antiparallel β sheet and a hydrophobic platform.

GatA is a single-domain protein consisting of a central, mixed 11-stranded β sheet core covered by double layers of α helices on the top and bottom (fig. S2). The highest structural similarity was found in malonamidase E2 (MAE2) from Bradyrhizobium japonicum [Protein Data Bank (PDB) code 1OCK (16)], with a root mean square deviation of 2.1 Å for 405 pairs of Cα atoms compared. GatA faces its loop-rich plane to, and caps a loop-rich side of, the cradle domain of GatB, burying 7% (2680 Å2) of the total surface area in the inter-subunit interface (Fig. 1A). There is a putative ammonia channel (described below) running through the middle of the interface, which is surrounded by two layers of interfacial interactions (fig. S1). The inner layer encircling the channel is composed of polar residues, whereas the outer layer is mostly hydrophobic; therefore, GatA is moderately attached to GatB through both hydrogen-bonding and hydrophobic interactions.

Seeking the molecular mechanism of the glutaminase reaction in GatA, we soaked GatCAB crystals in 1 mM glutamine and determined the cocrystal structure by molecular replacement. The glutamine was found in the center of GatA constructed by the amidase signature sequence (fig. S2). The amide group of glutamine is recognized by Asp425, and the carboxyl group is recognized by Arg358 (Fig. 2A). The side chain of the glutamine is located close to the conserved Ser–cis-Ser–Lys catalytic scissors: Ser178, Ser154, and Lys79. Consistent with MAE2 (16), in GatA, Ser154 is located in the loop enriched in small residues (glycine, serine, and alanine) and forms an unusual cis conformation, which enables itself to deprotonate Ser178. Surprisingly, Ser178 made a tight covalent bond (1.44 Å) with the amide carbonyl carbon atom of the glutaminyl side chain, revealing the tetrahedral covalent intermediate, which is stabilized by the oxyanion hole constructed by the backbone nitrogen atoms of Thr175, Gly176, Gly177, and Ser178 (Fig. 2A). We further observed a continuous electron density sprouting out from the amide group of the side chain of glutamine, which may correspond to amide ammonia liberated from the substrate. On the opposite site is a water molecule, which was not observed in the apo GatCAB crystal, anchored by hydrogen bonds with Ser153, Ser178, and the amide group of the substrate itself, which is in the vicinity of hydrolysis of the enzyme-acyl intermediate. These observations clearly indicated that GatA employs the same hydrolysis mechanism as the amide hydrolysis activity of amidase. Curiously, in contrast to the previous study on GatDE from Pyrococcus abyssi that showed glutaminase activity only in the presence of Glu-tRNAGln (10, 11), GatCAB intrinsically possesses basal glutaminase activity without binding of substrate tRNA [in the current structure and by kinetics studies (12)], which suggests that the conformational rearrangement of the glutaminase active site of GatA is not involved in regulation of the concerted reactions.

Fig. 2.

A 30 Å long ammonia channel connects the two remote active centers of GatCAB. (A) The active site of a glutaminase reaction in GatA is composed of the conserved Ser-cis-Ser-Lys catalytic scissors shown as magenta stick representations. Residues involved in the hydrogen-bonded network (dashed black lines) in the active site are labeled. A plausible hydrolytic water molecule is colored light blue and is on the opposite side of a supposed ammonia product (orange sphere). The Fo-Fc electron density map (contoured at 3σ, green mesh) calculated without the glutamine and Ser178 clearly demonstrates the tetrahedral covalent intermediate of the glutamine with Ser178. (B) The environment of the ADP binding site shown together with the omit Fo-Fc electron density map (2σ, blue). Residues contributing to ADP (ball-and-stick) recognition are represented as stick models with labels. Two water molecules (light blue spheres) are coordinated to a magnesium ion (purple) and to β phosphate. (C) The putative ammonia channel was calculated using the program CAVER (26), with the structure of the water-omitted GatCAB/glutamine complex. Glu125B blocking the ammonia transport route is shown in a space-filling representation for clarity. The channel was filled with a row of solvent molecules (light blue spheres), which interact with the conserved polar residues (colored sticks) along the pathway. A bound glutamine in GatA is drawn as spheres indicating the start point of the channel. (D) Schematic representation of the ammonia channel. Residues defining the channel are colored corresponding to their properties: red, negative; blue, positive; black, nonpolar side chain; gray, main chain. Hydrolyzed ammonia is colored orange. Strictly conserved residues are underlined and hydrogen-bonding distances are indicated (Å). The presumed movement of the Glu125B gate to open the ammonia channel is indicated by a black arrow.

GatB is organized into two domains connected by a characteristic ∼60 Å elongated linker loop (fig. S4). The globular “cradle” domain is topologically unique and can be distinguished from GatE by two insertions on each side of the cradle (10). GatE has an insertion domain—resembling those found in bacterial AspRS—between Bα3 and Bβ20, and an antiparallel β sheet inserted between Bβ1 and Bα1 (fig. S3). The auxiliary C-terminal helical domain is built by an α-helical bundle architecture with seven traceable helices. The disordered C-terminal region of GatB (residues 412 to 475) is presumed to form another three-helix bundle, as indicated by the sequence identity to the C terminus of uncharacterized Yqey protein (PDB code, 1NG6). An obvious density peak of Mg2+ ion was observed at the bottom of the cradle domain coordinated to three conserved residues consisting of His12, Glu124, and Glu150, and to three water molecules (Fig. 2B). To underline the molecular basis for the mechanism of the kinase activity, we cocrystallized GatCAB with the ADP–AlF4 compound as an ATP analog. The 3.35 Å 2Fo-Fc and Fo-Fc maps clearly indicated that ADP sits on strands Bβ1 and Bβ11 at the bottom of the cradle domain, with a bound water molecule (Fig. 2B). The adenosine base dips into a hydrophobic pocket, created by Val6, Phe205, and Pro155. N1 and N6 are strictly recognized by an invariant Ser196 residue in agreement with the previous report that GTP could not serve as a phosphate donor for GatCAB (17). The hydroxyl groups of ribose interact with the backbone of a loop between Bα2 and Bβ10, whereas the ribose O4 is recognized by the conserved Asn194 residue. The β phosphate is at an appropriate distance for hydrogen bonding with the conserved Glu10 and an anchored wat95 molecule.

Cautious inspection of the Fo-Fc difference electron density map from a crystal derivatized with MnCl2 revealed a pair of Mn cations in the active site of GatB (fig. S4). The prominent Mn2+ binding site is identical to the Mg2+ site in the native crystal. The second transient binding site is located adjacent to the first (a distance of 6.3 Å), coordinated to the conserved acidic residue cluster formed by Glu10, Asp192, and Glu210. This position corresponds to the wat95 water molecule in the ADP–AlF4 structure and is at an appropriate distance to make interactions with β- and γ-phosphates of ATP. These results strongly suggested that GatB (and plausibly GatE) uses a two-metal-ion mechanism of catalysis, which is reminiscent of the mechanism employed by glutamine synthetases in amino acid biosynthesis (1821). We propose that the primary Mg2+-binding site should be responsible for coordinate bonding, recognition, and positioning of the γ-carboxyl group of the glutamate attached to the 3′ end of tRNAGln (fig. S6). On the other hand, the transient Mg2+ binding site is likely to interact with β and γ phosphates of ATP and to participate in phosphoryl transfer by polarizing the γ-phosphate group of ATP, promoting the nucleophilic attack to form a γ-phosphoryl-Glu-tRNAGln (fig. S6).

The apo GatCAB structure revealed a distance of 30 Å that separates the catalytic sites at which the glutaminase reaction takes place and at which the CCA end of tRNA binds. We found that two reaction centers in GatA and GatB are connected by a penetrable tunnel lined with a continuous succession of conserved polar residues (Fig. 2, C and D). In our GatCAB/glutamine complex structure, this tunnel is thoroughly hydrophilic and is filled with a row of solvent molecules indicating the actual route of ammonia traveling from the amino acid pocket in GatA to reach the γ-phosphoryl-Glu-tRNAGln in GatB. Strictly conserved positive and negative residues, designating the ammonia channel, are located alternately along the course (Fig. 2, C and D). This may imply a “proton-relay mechanism” by which ammonia is dispersed from one site to the other by repeating protonation and deprotonation with these successive conserved polar residues. This hypothesis is reconciled by the positioning of Thr175A at the entrance of the channel and Lys79B at the exit as the proton donor and acceptor, respectively. However, additional biochemical and structural studies should be required to clarify the precise mechanism of ammonia transport in GatCAB. In addition, distinct ammonia tunnels have been described in multistep enzymatic proteins that hydrolyze glutamine to obtain ammonia as a substrate for a second reaction, such as asparagine synthetase B, imidazole glycerol phosphate synthase, or glutamine phosphoribosylpyrophosphate, as well as in the case of the transmembranal ammonia transporter proteins (2225). All of these adopt a dissimilar architecture of hydrophobic tunnel (∼20 Å in length), suggesting that ammonia (NH3), but not ammonium ion (NH4+), is transported. Therefore, GatB is idiosyncratic not only in its folding but also in its mechanism of the intramolecular ammonia transportation.

According to the program CAVER (26) and an in-house program used in calculation of the tunnel of GatCAB, we readily noticed a bottleneck in the middle of the path of ammonia travel (Fig. 2C). The channel is constricted by a completely conserved Glu125 residue residing in Bβ7 of GatB, which makes a salt bridge with the invariant Lys88 on a loop between Bβ4 and Bβ3. As the efficiency of glutamine hydrolysis is known to be stringently dependent on binding of Glu-tRNAGln, but neither uncharged tRNA nor Gln-tRNAGln (12), the Glu-tRNAGln recognition must induce as yet uncharacterized conformational changes. We postulate that there is an open-close motion of Glu125 as a gate susceptible to tRNA binding. Juxtaposed to Glu125 is a conserved Asp126 residue, which forms a tight salt bridge with the invariant Arg190 on the opposite Bβ11 strand, resulting in a twist in part of the Bβ7 strand. This Asp126–Arg190 salt bridge is located precisely at the entrance of the active site where the CCA end of tRNA will approach. It is not unreasonable to assume that the CCA end of the incoming tRNA would stimulate the movement of Arg190, resulting in disruption of the salt bridge, relaxing the distortion of strand Bβ7, and finally opening the glutamic gate. This scenario might be a shared feature between GatCAB and GatDE, because these relevant residues are definitely conserved in both molecules.

Our structural data, however, leave the nature and specificity of binding the tRNA substrate unsolved. We constructed a series of the in vitro transcribed S. aureus (Sav) tRNAGln mutants based on the differences between tRNAGln and tRNAGlu and investigated the tRNA-binding activity with the gel-shift assay (Fig. 3, A and B). GatCAB bound quite specifically to tRNAGln but not tRNAGlu. Unexpectedly, unlike GluRS or GlnRS, GatCAB did not bind or recognize the anticodon of tRNA, whereas swapping the upper part of the acceptor stem of tRNAGln to tRNAGlu diminished the binding activity to GatCAB. Changing only the first U1-A72 base pair of tRNAGln almost extinguished the binding capacity. This is consistent with the fact that the first U1-A72 base pair is absolutely conserved and unique to tRNAGln and tRNAAsn in bacteria (27) (fig. S5). No binding activity could be detected in the case of the D-arm swapped mutant. The differences in the D arms between tRNAGln and tRNAGlu are the size of the D loop and the base pair in the middle of the D stem (Fig. 3A). Further mutational analysis of the D arm revealed that the U insertion in the D loop was lethal to tRNA recognition. These results suggested that the morphology of the D loop is important for discrimination between tRNAGln and tRNAGlu by GatCAB. Taken together, GatCAB recognizes and discriminates two identity elements on tRNAGln; the U1-A72 base pair is exploited as the positive determinant to discriminate tRNAGln from the other tRNA species, whereas the U insertion in the D loop serves as the negative determinant for distinguishing tRNAGlu from tRNAGln.

Fig. 3.

Mutational analyses of tRNAGln and GatCAB. (A) Cloverleaf representations of tRNAGln (black) and tRNAGlu (gray) from S. aureus. Indicated bases or stem loops (magenta) were altered to those of tRNAGlu to verify the binding activity. Bases involved in specific recognition by GatCAB are emphasized and marked with blue stars. (B) Gel-shift assay with tRNAGln variants as specified in (A). Bands corresponding to 150 pmol tRNAGln and 50 pmol GatCAB are shown in the first and second lanes, respectively. The molecular ratios of the enzyme to tRNAGln were fixed at 1:3. Uncharacterized bands above tRNA were probably responsible for the tRNA dimer suggested by the results of size-exclusion chromatography in our experiments. Gels were stained with Coomassie brilliant blue and sequentially with toluidine blue to visualize proteins and RNA, respectively. (C) A series of GatB deletion mutants (50 pmol GatCAB complex per lane) were subjected to gel-shift assay in the absence (–) or presence (+) of 150 pmol of the wild-type tRNAGln. Only Leu472 harboring Δ3 mutant (deletion of three residues from C terminus of GatB) retained the same binding activity as wild-type (WT) GatCAB. Deletion of Leu472 from the C terminus of GatB (Δ4 mutants) abolished the tRNA-binding capacity. Further deleted mutant (Δ10), as well as a mutant lacking the helical bundle domain (ΔCHD) of GatB, did not show any tRNA-binding activity.

It is unclear how GatCAB discriminates the U1-A72 base pair in tRNAGln from the G-C base pair. Interestingly, in the crystal structure of Escherichia coli GlnRS-tRNAGln complex, the first U1-A72 base pair of tRNAGln was disrupted by the Leu136 residue at the tip of a β turn in the acceptor-binding domain of GlnRS (28). GatCAB may use a similar mechanism to distinguish A-U from G-C base pairs by the smaller free energy cost of base-pair denaturation. If this is the case with GatCAB, two turn loops, one between Bβ11 and Bα2 and the other between Bβ12 and Bα3, are good presumable candidates. On the other hand, we found that the C-terminal helical domain (CHD) of GatB has a crucial role in binding and discrimination of tRNAGln (Fig. 3C). The mutant with deletion of the entire domain completely failed to bind to tRNAGln. Notably, deletion of the conserved Leu472 residue, the fourth residue from the C terminus, of GatB also caused loss of tRNA-binding activity, whereas a GatB deletion mutant harboring Leu472 did not impair binding activity. Because dozens of hydrophobic residues responsible for interhelical interactions are highly conserved in CHD, we concluded that elimination of Leu472 disrupted the intact configuration of the last three helices (which were not visible in the electron density map) of CHD. This idea is corroborated by the unusually slow migration of Leu472-deleted mutant protein in our gel-shift experiment. This distal three-helix bundle is connected to the Bα11 helix by a long, flexible linker partially seen in the present structure and expected to be able to reach and bind to the D loop of bound tRNAGln. Discrimination of the size or configuration of the D loop of tRNAGln may be achieved by fitting to the concave made by helical bundle as a means of an “indirect readout” mechanism (29).

On the basis of our crystal structures combined with the evident biochemical data, we can extract a simple rationale for the operation of GatCAB controlling the Glu-tRNAGln–dependent amidotransferase process (fig. S6). Glu-tRNAGln is successfully accommodated by GatB through two checkpoints on the body of tRNAGln; the U1-A72 base pair and the D loop. Subsequent precise coordination of the carboxylic acid moiety of glutamate, attached to the terminal adenosine of tRNA, to primary Mg2+ could induce local conformational remodeling of the active site, shifting the Glu125 gate to open the ammonia channel. This motion may slightly accelerate the glutaminase activity of GatA because of the occurrent flow of ammonia. However, the full activation depends strictly on the recruitment of ATP from the opposite site of the active center. Replacement of the wat95 molecule with the secondary Mg2+ makes the active center ready for the transamidase reaction by preventing γ-phosphoryl-glutamate from hydrolysis. Consumption of the ammonia by the transamidase reaction would robustly drive the total potential activity of GatCAB by pumping ammonia downward with a concentration gradient. Our proposed model provides a detailed hypothesis to explain the previous biochemical and kinetic studies of the coupling reaction cycle of GatCAB (11, 12); however, further biochemical and structural works on GatCAB/Glu-tRNAGln complex may be needed to complete the scenario of the reaction scheme of adding glutamine to the genetic code.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article