A novel carbamoyl-phosphate synthetase from Aquifex aeolicus.

Aquifex aeolicus, an extreme hyperthermophile, has neither a full-length carbamoyl-phosphate synthetase (CPSase) resembling the enzyme found in all mesophilic organisms nor a carbamate kinase-like CPSase such as those present in several hyperthermophilic archaea. However, the genome has open reading frames encoding putative proteins that are homologous to the major CPSase domains. The glutaminase, CPS.A, and CPS.B homologs from A. aeolicus were cloned, overexpressed in Escherichia coli, and purified to homogeneity. The isolated proteins could catalyze several partial reactions but not the overall synthesis of carbamoyl phosphate. However, a stable 124-kDa complex could be reconstituted from stoichiometric amounts of CPS.A and CPS.B proteins that synthesized carbamoyl phosphate from ATP, bicarbonate, and ammonia. The inclusion of the glutaminase subunit resulted in the formation of a 171-kDa complex that could utilize glutamine as the nitrogen-donating substrate, although the catalytic efficiency was significantly compromised. Molecular modeling, using E. coli CPSase as a template, showed that the enzyme has a similar structural organization and interdomain interfaces and that all of the residues known to be essential for function are conserved and properly positioned. A steady state kinetic study at 78 degrees C indicated that although the substrate affinity was similar for bicarbonate, ammonia, and glutamine, the K(m) for ATP was appreciably higher than that of any known CPSase. The A. aeolicus complex, with a split gene encoding the major synthetase domains and relatively inefficient coupling of amidotransferase and synthetase functions, may be more closely related to the ancestral precursor of contemporary mesophilic CPSases.

The enzyme from Escherichia coli consists of a 120-kDa synthetase subunit (CPS) and a 40-kDa amidotransferase or glutaminase subunit (GLN) (1). Upon dissociation of the heterodimer (2,3), the GLN subunit was found to catalyze the hydrolysis of glutamine, whereas the CPS subunit catalyzed the synthesis of carbamoyl phosphate from ammonia, bicarbonate, and ATP. The GLN and CPS domains are fused in CAD (4 -6), a mammalian multifunctional protein that catalyzes the first three steps of the de novo pyrimidine biosynthetic pathway. Although glutamine hydrolysis is the usual source of ammonia for carbamoyl phosphate synthesis, mitochondrial CPSase I, the enzyme that catalyzes the first step in the urea cycle (1), has an inactive homolog of the GLN subunit fused to the amino end of the synthetase subunit. Consequently, CPSase I cannot hydrolyze glutamine, and instead it uses ammonia directly as the nitrogen-donating substrate.
Although the CPSases have a diverse structural organization, they share a common catalytic mechanism (1) that proceeds through a complex series of partial reactions. HCO 3 Ϫ ϩ ATP ¡ carboxy phosphate ϩ ADP REACTION 2 carboxy phosphate ϩ NH 3 ¡ carbamate ϩ P i REACTION 3 carbamate ϩ ATP ¡ carbamoyl phosphate ϩ ADP REACTION 4 Lusty and Nyunoya (7,8) were the first to clone and sequence both subunits of a member of this family of enzymes, E. coli CPSase. They noted that the sequence of the synthetase subunit consists of two highly homologous halves that probably arose from a duplication, translocation, and fusion of an ancestral kinase gene. Similarly, they found that the GLN subunit consists of two domains, one of which shared homology to other triad-type amidotransferases (9). This structural organization was confirmed when the x-ray structure of E. coli CPSase was solved (10). One of the most interesting aspects of the x-ray structure was the presence of intramolecular tunnels connecting the active sites ensuring that the labile intermediates are sequestered.
In studies of several different CPSases (10 -18), each of the two halves of the synthetase subunit, designated CPS.A and CPS.B, were found to have an ATP binding site and to catalyze different ATP-dependent partial reactions. Site-directed mutagenesis of the E. coli enzyme showed (15) that CPS.A catalyzes the activation of bicarbonate, whereas CPS.B is responsible for the phosphorylation of carbamate to form carbamoyl phosphate. Although the two domains have a specialized function in the native molecule, they are functionally equivalent (19,20) when cloned separately and expressed in E. coli. The isolated domains form homodimers that catalyze both ATP-dependent partial reactions and the overall synthesis of carbamoyl phosphate from bicarbonate, NH 3 , and two ATP molecules. The only apparent functional difference between the CPS.A and CPS.B dimers is that the latter is subject to allosteric control because effectors bind to the regulatory subdomain (21)(22)(23)(24)(25)(26)(27) at the extreme carboxyl end of CPS.B.
An entirely different strategy for the synthesis of carbamoyl phosphate may be employed by some hyperthermophilic archaebacteria. In Pyrococcus abyssi (28,29) and Pyrococcus furiosus (30,31), carbamoyl phosphate can be synthesized by the ATP-dependent phosphorylation of carbamate formed spontaneously in solution from ammonia and bicarbonate. Their sequence and structure (32) closely resemble those of the catabolic carbamate kinases found in several eubacterial species, and they have been designated carbamate kinase-like CPSases.
With the advent of genome sequence projects, it became apparent that some hyperthermophilic organisms, including Methanococcus jannaschii (33) and Aquifex aeolicus (34), do not have an enzyme homologous to the CPSases found in mesophilic organisms. Instead, these genomes have open reading frames that encode proteins with deduced amino acid sequences that closely resemble the GLN, CPS.A, and CPS.B domains of mesophilic CPSases. We report, for the first time, the cloning and expression of these genes from A. aeolicus and the reconstitution of functional complexes.

Cloning and Expression
The genes encoding the carbamoyl-phosphate synthetase subunits were obtained by polymerase chain reaction using Pfu DNA polymerase (Stratagene), 180 ng of the template A. aeolicus chromosomal DNA, and 100 pmol of primers. The 5Ј and 3Ј primers used for amplification were the following: for carA (GLN), 5Ј-CGGAACTCGAGCATTTTGGCGCT-TGAGGACGG-3Ј and 5Ј-ACTTCTGCAGCTCCTC ATCCCTGAGCCAT-3Ј; for carB2 (CPS.A), 5Ј-CTTTGGGATCCAAAAGGACGGACATCAA-G-3Ј and 5Ј-TAAACAAGATCTTTAATCTTCATCAAGGATTTC-3Ј; and for carB1 (CPS.B), 5Ј-TTCTTGGATCCTAAAAAGGTTGTAATACTCG-GA-3Ј and 5Ј-ATAAAGATCTCTAGGTCCATAAGAATTTGTA-3Ј. Each primer also included a restriction site to facilitate subcloning: XhoI/PstI for carA, BamHI/BglII for carB1, and BamHI/BglII for carB2. The amplified DNA fragments were cleaved with the appropriate restriction enzymes and inserted into the corresponding sites of the pRSETB expression vector and co-transformed with the plasmid pSJS1240 into E. coli BL21 (DE3). The helper plasmid pSJS1240 (35) encodes two tRNA synthetases that recognize the codons AUA for isoleucine and AGA for serine. These codons occur infrequently in E. coli but are highly represented in the A. aeolicus genome.

Purification
Each of the proteins is expressed as a fusion protein with a 3-kDa His-tagged polypeptide attached to the amino end of the enzyme. The cells harvested from a 100-ml culture were resuspended in 3 ml of 50 mM Tris-HCl, pH 8, 10 mM 2-mercaptoethanol and disrupted by six 30-s bursts of sonication. The cell extract was centrifuged at 17,000 ϫ g for 20 min, and the supernatant was applied to a 1.5-ml Ni 2ϩ -Probond column pre-equilibrated with 50 mM Tris-HCl, pH 8, 10 mM 2-mercaptoethanol, and 200 mM NaCl. The column was washed with 20 ml of the same buffer. The A. aeolicus enzymes were then eluted with 1 ml of increasing concentrations of imidazole up to 200 mM in this buffer. The 1-ml fractions were analyzed by electrophoresis on 12.5% SDS-polyacrylamide gels, and those containing pure proteins were dialyzed at 4°C against 50 mM potassium phosphate buffer, pH 8, and 10 mM 2-mercaptoethanol. The proteins were stored in the same buffer except that 10% glycerol was added to the GLN subunit storage buffer.

Enzyme Assays
Glutaminase Assay-The glutaminase activity was measured by coupling the formation of L-glutamate to the production of ␣-ketoglutarate using L-glutamate dehydrogenase (15). The assay mixture, consisting of varying concentrations of glutamine in 50 mM potassium phosphate buffer, pH 8, was pre-equilibrated at 78°C for 1.5 min before starting the reaction with 30 g of the enzyme. The reaction was quenched after 1.5 min with 50 l of 1 M HCl, and the reaction mixture was then placed on ice for 15 min. The pH was neutralized by the addition of 270 l of 1 M Tris-HCl, pH 10. The sample was then added to 600 l of 0.1 M Tris-HCl, pH 8, 0.8 mM acetylpyridine adenine dinucleotide (APAD), and 20 units of L-glutamate dehydrogenase. The absorbance at 363 nm of reduced APAD was measured after a 45-min incubation at 25°C, and the concentration of reduced APAD was calculated from a standard curve.
Carbamoyl-Phosphate Synthetase Assay-For measuring the ammonia-dependent CPSase, the assay mixture consisted of 200 mM ammonium chloride, 50 mM sodium [ 14 C]bicarbonate (100,000 -120,000 dpm/ mol), 30 mM ATP, 32 mM MgCl 2 , 100 mM KCl, 50 mM potassium phosphate, pH 8, 3 g of purified A. aeolicus ATCase, 6 mM aspartate, and 30 g of the enzyme (unless specified otherwise) in a total volume of 0.5 ml. The assay mixture was the same for the glutamine-dependent CPSase except that the ammonium chloride was replaced with 2 mM glutamine. The assay mixture without enzyme and bicarbonate was equilibrated at 78°C for 1.5 min. The reaction was initiated by the addition of enzyme and bicarbonate, allowed to proceed for 1.5 min, and quenched by the addition of 0.5 ml of 10% trichloroacetic acid. The samples were processed for counting as described previously (36). The saturation curves were obtained by varying the concentration of one substrate while fixing the others at the saturating concentrations given above.
Partial Reactions-The bicarbonate-dependent ATPase was assayed by measuring the rate of ADP formation in the presence and absence of ammonium chloride using a pyruvate kinase/lactate dehydrogenase coupled assay (15). The assay mixture contained 50 mM bicarbonate, 100 mM KCl, 30 mM ATP, 32 mM MgCl 2 , 10 mM 2-mercaptoethanol, and, when present, 200 mM ammonium chloride in 50 mM potassium phosphate, pH 8, with the coupling substrates and enzymes. Carbamoyl phosphate-dependent ATP synthetase activity was assayed similarly by measuring the rate of ATP formation from ADP and carbamoyl phosphate using a hexokinase/glucose-6-phosphate dehydrogenase-coupled assay (15). The partial reactions were only assayed at 25°C because of the limited stability of the coupling enzymes at elevated temperatures.

Gel Filtration Chromatography
To determine whether A. aeolicus proteins form stable complexes, 0.5-ml aliquots (3-4 mg) of the purified proteins were applied to a 1.5 ϫ 62-cm Sephacryl S-300 High Resolution column equilibrated with 50 mM potassium phosphate buffer, pH 8, 200 mM sodium chloride, and 10 mM 2-mercaptoethanol. The column was eluted with the same buffer at a flow rate of 0.36 ml/min, and fractions (0.5 ml) were analyzed by measuring the absorbance at 280 nm, SDS-gel electrophoresis, and CPSase assays. Bovine serum albumin (66 kDa), alcohol dehydrogenase (150 kDa), apoferritin (443 kDa), and blue dextran were used to calibrate the column.

Data Analysis
The kinetic parameters were obtained by least squares fit of substrate saturation curves to the Michaelis-Menten or Hill equations using the program Scientist (Micromath). Sequence comparison and multiple alignments were performed by using the GAP and Pileup programs of the GCG software package (Genetics Computer Group, University of Wisconsin) (37). BLAST software was used to search the A. aeolicus genome for open reading frames encoding homologs of mesophilic enzymes. Modeling of the three-dimensional structure and energy minimization was performed using the program SwissModel (38 -40) from the ExPASy server. The structure was visualized and analyzed using the Rasmol and Swiss Protein Database, Viewer 3.7.

Identification and Sequence Analysis of A. aeolicus CPSase-An
exhaustive search of the A. aeolicus genome failed to reveal any open reading frame encoding a protein homologous to the 120-kDa synthetase subunit of E. coli and other eubacterial CP-Sases. Similarly, a small 35-kDa carbamate kinase-like CP-Sase, responsible for carbamoyl phosphate synthesis in some archaeal species, was not detected when the genome was searched with seven query sequences representing highly conserved segments of carbamate kinases. However, two open reading frames encoding proteins homologous to the major structural domains of E. coli CPSase had been identified (34). The two A. aeolicus proteins exhibited 39.5% sequence identity to each other (Fig. 1C), a value similar to that obtained (28.9 -50.6%) when the two halves of mesophilic CPSases were aligned. One of these genes, designated carB2 (CPS.A in Fig.  1C) was found to have significantly greater sequence similarity to the CPS.A domain (49.6 -63.0%) than to the CPS.B domain (25.6 -33.6%) of other CPSases. The converse was true for the second gene, carB1 (CPS.B), which was located 652 kilobase pairs upstream of carB2 (Fig. 1A). Thus, the proteins encoded by the A. aeolicus carB2 and carB1 were designated CPS.A and CPS.B, respectively. An alignment of the A. aeolicus CPS.A and CPS.B subunits at the junction between the two fused domains in the E. coli enzyme is shown in Fig. 1B. Similarly, the carA gene could be identified unambiguously by its strong sequence similarity (45.3-50.6% identity) to the amidotransferase or glutaminase domain of other well characterized prokaryotic CPSases. CarA is located about 543 kilobase pairs upstream of carB1 and overlaps pyrB, the gene encoding ATCase, by three base pairs. PyrC, which encodes dihydroorotase, was located approximately midway between carA and carB1 ( Fig. 1).
Expression and Purification of the Recombinant Proteins-The genes encoding CPS.A, CPS.B, and GLN were amplified by polymerase chain reaction using A. aeolicus chromosomal DNA as the template. The oligonucleotides incorporated convenient restriction sites that allowed each DNA fragment to be inserted in frame into the pRSETB expression vector. This vector appends a His tag to the amino end of the recombinant protein to facilitate purification. The resulting constructs pAAGLN, pAACPSA, and pAACPSB encoding GLN, CPS.A and CPS.B, respectively, were sequenced to confirm the fidelity of the amplification process. Co-transformation of the E. coli BL21(DE3) strain with the helper plasmid pSJS1240 resulted in the expression of high levels of the soluble proteins ( mg/liter of culture for CPS.A, CPS.B, and GLN, respectively). Each protein could be purified (Fig. 2) to homogeneity by Ni 2ϩ affinity chromatography.
Catalytic Activity of the Recombinant Proteins-The CPS.A and CPS.B subunits had no detectable NH 3 -dependent CPSase activity. In contrast, the GLN domain hydrolyzed glutamine to glutamate and ammonia at a rate of 0.87 nmol/min/mg at 25°C. Although neither CPS.A nor CPS.B alone could catalyze the overall reaction, a stoichiometric mixture of CPS.A and CPS.B catalyzed the formation of carbamoyl phosphate from ammonia, ATP, and bicarbonate at a rate of 5.7 nmol/min/mg (25°C). Under the same conditions, the rate of ATP hydrolysis was 10.7 nmol/min/mg. 3 The ratio of 1.9:1 of ATP consumed to carbamoyl phosphate formed is consistent with the 2:1 stoichiometry of the classical CPSase mechanism.
The isolated CPS.A and CPS.B subunits were found to catalyze both ATP-dependent partial reactions as assayed by established procedures ("Experimental Procedures"). The first partial reaction, the activation of bicarbonate, was measured as a bicarbonate-dependent ATPase in the absence of a nitrogendonating substrate. Equivalent amounts of CPS.A, CPS.B, and CPS.A-CPS.B had a similar bicarbonate-dependent ATPase activity (Table I). When ammonia was present, allowing the overall reaction to proceed, the ATPase activity of CPS.A and CPS.B remained for the most part unchanged, but the rate of ATP hydrolysis by the CPS.A-CPS.B complex increased 3-fold to 41 nmol/min/mg.
The second ATP-dependent partial reaction, the phosphorylation of carbamate, is assayed in the reverse direction as carbamoyl phosphate-dependent ATP synthesis ("Experimental Procedures"). Again, both CPS.A and CPS.B could catalyze this reaction (Table I), although it was surprising that the rate of ATP formation by CPS.A was significantly higher than that observed for either CPS.B or the CPS.A-CPS.B complex.
Oligomeric Structure of the Recombinant Proteins-SDS-gel electrophoresis on calibrated polyacrylamide gels (Fig. 2) showed that the molecular mass of the CPS.A, CPS.B, and GLN proteins was 64 Ϯ 1.2, 63 Ϯ 1.6, and 45 Ϯ 1.0 kDa, respectively. These values are close to the molecular weights of 62,289, 60,037, and 41,682 as predicted from the deduced amino acid sequence when the 3-kDa His tag on the recombinant proteins is taken into consideration. The size of each protein was determined under nondenaturing conditions by size exclusion chromatography on a calibrated Sephacryl S-300 column (Fig. 3A). CPS.A eluted as a 66-kDa species, indicating that it is monomeric. In contrast, CPS.B was more heterogeneous. The major species had a molecular mass of 170 kDa, but there was an appreciable fraction that eluted in the void volume, suggesting that the isolated CPS.B aggregates. These results were confirmed by electrophoresis on nondenaturing polyacrylamide gels (results not shown). The stoichiometric mixture of CPS.A and CPS.B eluted as a 124-kDa dimer, although there was a shoulder on the leading edge that may represent the presence of larger species. Assay of CPSase activity in the column fractions (Fig. 3A) showed that only the CPS.A-CPS.B dimer had catalytic activity.
A mixture of equimolar amounts of CPS.A, CPS.B, and GLN was found to elute as a single 171-kDa species that could catalyze glutamine-dependent carbamoyl phosphate synthesis (Fig. 3B). A complex consisting of one copy of each subunit would be expected to have a mass of 173 kDa. SDS-polyacrylamide gel electrophoresis of the peak fractions (Fig. 3B, insert) showed that the molar ratio of CPS.A-CPS.B:GLN was 1.2, 4 consistent with a species comprising equivalent molar amounts of CPS.A-CPS.B and GLN subunits. It is interesting that the major species visualized on the SDS gel had a mass of 170 kDa when the samples were not heated at 100°C prior to electrophoresis (data not shown), suggesting that the complex is difficult to dissociate even in strong detergents. Taken together,  these results show that the GLN, CPS.A, and CPS.B associate to form a stable 1:1:1 heterotrimer.
Thermostability of CPS.A-CPS.B-The thermostability of the CPS.A-CPS.B complex was assessed by preincubating the protein at increasing temperatures for 10 min, rapidly cooling the sample, and assaying the CPSase activity at 37°C. The activity remained unchanged between 25°C and 60°C and then abruptly increased 54% at 70°C (Fig. 4). The precipitous loss of catalytic activity above 80°C is indicative of thermal denaturation. The temperature at which half the maximum activity was lost is 81°C. The GLN-CPS.A-CPS.B complex had appreciably greater thermostability, showing no loss of activity up to 90°C. Interestingly, the large increase in thermal stability at temperatures prior to thermal denaturation of CPS.A-CPS.B was not observed in the GLN-CPS.A-CPS.B complex.
Steady State Kinetics-The catalytic activity of CPS.A-CPS.B was found to increase with increasing temperature from a value of 0.02 mol/min/mg at 25°C to 0.5 mol/min/mg at 78°C. Because A. aeolicus grows optimally at elevated temper-ature, the steady state kinetic parameters for CPS.A-CPS.B were measured at 78°C, as close to the thermal denaturation temperature (Fig. 4) as was practically possible. Saturation curves were obtained (Fig. 5, A-C) with NH 4 Cl, NaHCO 3 or ATP as the variable substrate. Although the NH 4 Cl and NaHCO 3 curves exhibited Michaelis-Menten kinetics, the ATP saturation curve was sigmoidal and was fit to the Hill equation (Hill coefficient 1.4). The kinetic parameters, summarized in Table II, indicate that the K m values for bicarbonate (7.9 mM) and ammonia (3.2 mM) 5 for the A. aeolicus complex were comparable with the values obtained (41) for the E. coli CPSase synthetase subunit (K m (HCO 3 Ϫ ) ϭ 10.8 mM, K m (NH 3 ) ϭ 5.2 mM). However, the K m for ATP was 6-fold higher than that of the E. coli subunit (K m (ATP) ϭ 1.3 mM). The k cat of the A. aeolicus synthetase complex (1.6 s Ϫ1 ), determined from an ATP saturation curve, was appreciably lower than the value obtained for the E. coli CPSase synthetase subunit (7.3 s Ϫ1 ) (41).
The isolated GLN subunit has a relatively high affinity for glutamine (K m ϭ 58 M), but the k cat is very low (9 ϫ 10 Ϫ3 s Ϫ1 ), indicating a poor catalytic efficiency (Table II). However, when the GLN subunit forms a complex with the CPS.A-CPS.B dimer, the rate of glutamine hydrolysis increases 18-fold (k cat ϭ 0.196 s Ϫ1 ). The increased rate is partially offset by a 12-fold increase in the K m for glutamine. The k cat and K m values for the GLN-CPS.A-CPS.B complex (Fig. 5, D-F) using ammonia as the nitrogen-donating substrate are similar (Table II) to the values observed for the isolated CPS.A-CPS.B dimer. However, a comparison of the k cat values for the heterotrimer (Table II) for the overall synthesis of carbamoyl phosphate indicates that the catalytic efficiency of the heterotrimer is 4 -9-fold lower when glutamine rather than ammonia serves as the nitrogendonating substrate.
Allosteric Regulation-UMP is an allosteric inhibitor and ornithine is an activator of A. aeolicus CPSase. The same effectors modulate the activity of E. coli CPSase. When the ATP saturation curve (Fig. 6A) for CPS.A-CPS.B was determined in the presence of 2 mM UMP, the K m was found to increase 5 The K m for ammonia was calculated from the value given in Table  II for ammonium chloride assuming a pK a of 8.88 for the dissociation of the proton from the ammonium ion (38).  Table II). Lineweaver-Burk plots for each saturation curve are shown as inserts. The kinetic parameters are summarized in Table II. 2.3-fold and the V max decreased 30% (Table III). Together these changes in the kinetic parameters resulted in a 3-fold decrease in the apparent second order rate constant, k cat / K m (ATP). Ornithine had little effect on the K m for ATP but increased the V max 1.8-fold, giving a k cat / K m (ATP) that was about 2-fold higher than the corresponding value for the enzyme in the absence of allosteric ligands. UMP and ornithine had a more pronounced effect on the glutamine-dependent CPSase activity of GLN-CPS.A-CPS.B. The presence of UMP increased the K m 360% and decreased the V max 80%, whereas ornithine decreased the K m 40% and increased the V max 320%. Thus, effectors can alter the apparent second order rate constant for the glutamine-dependent CPSase activity over a 100-fold range (Table III).
The effect of increasing concentrations of the allosteric ligands on the catalytic activity of CPS.A-CPS.B and GLN-CPS.A-CPS.B (Fig. 6, B and C) showed that half-maximal activation occurs at 0.1 mM ornithine for CPS.A-CPS.B and at about 2 mM for GLN-CPS.A-CPS.B. Very low concentrations of UMP activate ammonia-and glutamine-dependent CPSase activity, but higher concentrations inhibit in a concentration-dependent manner. Half-maximal inhibition occurred at 0.5 and 0.05 mM UMP for CPS.A-CPS.B and GLN-CPS.A-CPS.B, respectively. The maximum ornithine activation and UMP inhibition were appreciably greater for GLN-CPS.A-CPS.B than for the complex lacking the GLN subunit. Other allosteric effectors known to modulate the catalytic activity of other CPSases, including UTP, IMP, PRPP (phosphoribosyl 5Ј-pyrophosphate), and N-acetyl-L-glutamate, had no effect on the catalytic activity of A. aeolicus CPSase.
Molecular Modeling of the A. aeolicus GLN, CPS.A, and CPS.B Subunits-The high degree of sequence identity between E. coli CPSase and the GLN, CPS.A, and CPS.B subunits of A. aeolicus (49,63, and 55% identity, respectively) makes the alignment of the proteins unambiguous. Compared with E. coli CPSase, the A. aeolicus CPS.A subunit is nine residues longer on the carboxyl end of the polypeptide and has three short deletions (residues 340, 423-425, and 483-485 in the E. coli polypeptide) and a single one-residue insertion (between residues 410 and 411 in the E. coli sequence). The A. aeolicus CPS.B subunit is four residues shorter on the amino end and has one additional residue on the carboxyl end. There are three insertions in the A. aeolicus CPS.B subunit, 13 residues (874 -875 in the E. coli enzyme), four residues (885-886), and one residue (1056 -1057). Similarly, the GLN subunit, which is six residues shorter at the carboxyl end and has only three deletions (residues 131, 263, and 191-193 in the E. coli enzyme) can be aligned with the E. coli GLN subunit sequence with a high degree of confidence.
The tertiary structure of the A. aeolicus CPS.A-CPS.B, and GLN subunits was modeled using the E. coli CPSase (Protein Data Bank Code 1BXR) x-ray structure (42) as the tertiary template. A superposition of the template and model structures showed that the backbone is virtually identical. There are few insertions and deletions, all of which lie on the surface of the molecule in unobtrusive locations that would not be expected to disrupt function. The additional five residues (Fig. 1B) at the junction between CPS.A and CPS.B are located in a region far from all of the active sites. The other major insertion of 13 residues in CPS.B is a loop with an unusual sequence consisting of seven charged and six very hydrophobic residues. Although the conformation of this loop was not modeled, it is clearly located on the CPS.B surface in what appears to be a noncritical region.
An examination of the model (Table IV) showed that all of the residues involved in substrate binding and catalysis in both the GLN and CPS subunits of the E. coli enzyme are conserved and properly positioned (42,43), as are all of the residues that constitute the proposed intermolecular tunnels (10) between the GLN-CPS.A domains and between CPS.A and CPS.B.
The residues at the interdomain interfaces (Table V) (Table V) indicate that, of 24 interdomain contacts found in the E. coli enzyme, 17 are present in the A. aeolicus model (14 identical residues). Of 20 residues on CPS.B in close proximity to the CPS.A domain in the E. coli enzyme, 17 are found in the model structure. An analysis of the interactions between the GLN subunit and CPS.A of the model indicated that this intersubunit interface was also similar. Thus, it is likely that the intersubunit interactions and arrangement of the A. aeolicus CPSase subunits are the same as in E. coli CPSase. DISCUSSION A functional A. aeolicus CPSase was assembled from three subunits homologous to the CPSase domains of other organisms. The isolated CPS.A and CPS.B subunits catalyze the partial reactions, but ammonia-dependent carbamoyl phosphate synthesis requires the concerted action of both subunits. The addition of the GLN subunit confers the ability to utilize glutamine as a nitrogen-donating substrate. The complex is a 171-kDa heterotrimer composed of one copy of each type of subunit. Sequence comparisons and molecular modeling indicate that the structure of the A. aeolicus heterotrimer closely resembles the enzyme from E. coli and other mesophilic organisms.
In principle, it is possible that the enzyme catalyzes carbamoyl phosphate synthesis, via a carbamate kinase-like mechanism, from carbamate formed spontaneously in solution from ammonia and bicarbonate. This has been documented (28 -31) in some hyperthermophilic archaebacterial enzymes. However, there are several lines of evidence that argue conclusively against this possibility: 1) the CPS.A and CPS.B subunits catalyze both ATP-dependent partial reactions; 2) the stoichiometry of the reaction, in accordance with the classical mechanism (Reactions 2-4), is 2 mol of ATP consumed/mol of carbamoyl phosphate synthesized; and 3) the complete complex catalyzes the glutamine-dependent CPSase reaction in the absence of ammonia, an observation that is difficult to reconcile with a carbamate kinase-like mechanism. Thus, we conclude that carbamoyl phosphate is synthesized in A. aeolicus by the classical mechanism involving the concerted action of two ATP binding sites that catalyze a complex series of sequential partial reactions. The intramolecular tunnels, a prerequisite for sequestering the unstable intermediates, are present in the model structure.
Although the overall turnover number is low, the other steady state kinetic parameters of the A. aeolicus complex are close to the values reported for E. coli CPSase. The only notable exceptions are: 1) the K m for ATP is unusually high in both the CPS.A-CPS.B and the GLN-CPS.A-CPS.B complexes, and 2) the k cat for glutamine-dependent carbamoyl phosphate synthesis is about 7-fold lower than the value obtained using ammonium chloride as the nitrogen source. Thus, unlike the situation in other CPSases (44,45), the rate of the reaction appears to be limited by the rate of glutamine hydrolysis.
The coupling between glutamine hydrolysis and the activation of bicarbonate (Reaction 2) is an important aspect of the coordination of these parallel reactions in glutamine-dependent carbamoyl phosphate synthesis. The reactions must occur in phase to avoid the wasteful hydrolysis of glutamine and ATP. The isolated CAD GLN domain, obtained by cloning and expression, has a very low k cat and a very high K m for glutamine (45,46). When combined with the CPS domain, the k cat for glutamine hydrolysis increases 17-fold and the K m decreases 47-fold, indicating that interdomain interactions appreciably improve the catalytic efficiency of glutamine hydrolysis. The k cat /K m was found to increases 800% upon association. In the case of the isolated A. aeolicus GLN subunit, the k cat is low (Table II), but the K m is also low, indicating high affinity for the substrate. When associated with CPS.A-CPS.B, the k cat increases 18-fold, but the affinity for glutamine is 12-fold lower; therefore, the apparent second order rate constant increases by only 50%.
In E. coli and mammalian CPSases, the activity of the GLN domain is very low in the absence of the other substrates needed for carbamoyl phosphate synthesis, but the k cat increases 14 -100-fold when these substrates bind to the synthetase domain (1,47,48). This coupling mechanism is thought to ensure that reactions occurring on distinct domains remain in phase. Preliminary experiments with the A. aeolicus GLN-CPS.A-CPS.B complex indicated that the presence of ATP and bicarbonate stimulate the glutaminase activity less than 2-fold, 6 indicating that the coupling mechanism is relatively inefficient in the A. aeolicus complex.
There is good evidence (11)(12)(13)(14)(15)(16)(17)(18) that CPS.A and CPS.B in the native complex have specialized functions. CPS.A catalyzes the activation of bicarbonate (Reaction 2), whereas CPS.B phosphorylates carbamate (Reaction 4). However, the isolated CPS.A and CPS.B subdomains of mammalian, yeast, and E. coli CPSase were found (19,20) to form homodimers that could catalyze ammonia-dependent carbamoyl phosphate synthesis. If the homodimer was dissociated by high pressure (49), the monomers reversibly lost CPSase activity but retained the ability to catalyze both partial reactions. Similarly, the isolated A. aeolicus CPS.A and CPS.B subunits each effectively catalyzed both partial reactions (Table I). This result is not unexpected given the similarity of the reactions catalyzed and the isosteric nature of the substrates, HCO 3 Ϫ and carbamate. Although the isolated A. aeolicus subunits catalyze the partial reactions, they do not catalyze the overall synthesis of carbamoyl phosphate, probably because they cannot form functional dimers. The CPS.A subunit is a monomer, whereas CPS.B tends to aggregate. The failure of the subunits to dimerize is a necessary constraint for this type of molecular organization. Because CPS.A and CPS.B are expressed as individual polypeptides that must be assembled in vivo, if functional homodimers could form, a substantial fraction of the complexes (25%) would be of the type (CPS.A) 2 and would thus be unregulated.
The relative rates of the partial reactions catalyzed by the A. aeolicus CPS.A and CPS.B subunits do not agree with the functions assigned to these domains in other CPSases. CPS.A catalyzes carbamoyl phosphate-dependent ATP synthesis three times faster than does CPS.B. A similar trend is observed for the bicarbonate-dependent ATPase, although the difference is not as large. One might argue on this basis that our assignment is wrong and that the subunit that we have designated CPS.B is in reality CPS.A and vice versa. However, comparison of the CPSase sequences ( Fig. 1) strongly suggests that this explanation is incorrect. Experiments are planned to directly assess the function of the two domains in the intact A. aeolicus complex.
In E. coli, the CPS.A and CPS.B domains are fused to form a single synthetase subunit, whereas the genes encoding the CPS and GLN subunits are part of an operon that ensures their coordinated expression. In contrast, the A. aeolicus genes encoding the CPSase subunits are scattered throughout the genome. Moreover, the individual subunits fold independently into functional units capable of catalyzing the partial reactions, and thus, unproductive futile cycles would occur unless the individual subunits are rapidly incorporated into a fully assembled complex. Consequently, coordinated regulation of the expression of these genes must be a major factor in ensuring that stoichiometric amounts of the CPSase subunits are simultaneously synthesized. An examination of 200 base pairs of the sequences upstream of these genes did not reveal any common transcription factor binding sites, but a detailed study of the regulation of the expression of the subunits would be informative. One especially interesting aspect of the organization of the pyrimidine biosynthetic genes in A. aeolicus is that the ATCase and GLN coding sequences overlap. This gene arrangement may suggest the coordinated expression of the GLN, CPS, and ATCase genes, a potentially important control mechanism because A. aeolicus CPSase functions in concert with ATCase to facilitate the channeling 2 of the unstable carbamoyl phosphate, thus preserving it from thermal degradation.
Mesophilic CPSases are thought (7, 50 -52) to have arisen by gene duplication, translocation, and fusion of an ancestral kinase followed by the acquisition of an amidotransferase subunit that conferred the ability to utilize glutamine. A. aeolicus CPSase with split genes encoding separate CPS.A and CPS.B subunits and relatively inefficient coupling of amidotransferase and synthetase functions may be more closely related to the ancestral precursor of contemporary mesophilic CPSases.