Function of the major synthetase subdomains of carbamyl-phosphate synthetase.

The amidotransferase domain (GLNase) of mammalian carbamyl-phosphate synthetase II hydrolyzes glutamine and transfers ammonia to the synthetase domain where carbamyl phosphate is formed in a three-step reaction sequence. The synthetase domain consists of two homologous subdomains, CPS.A and CPS.B. Recent studies suggest that CPS.A catalyzes the initial ATP dependent-activation of bicarbonate, whereas CPS.B uses a second ATP to form carbamyl phosphate. To establish the function of these substructural elements, we have cloned and expressed the mammalian protein and its subdomains in Escherichia coli. Recombinant CPSase (GLNase-CPS.A-CPS.B) was found to be fully functional. Two other proteins were made; the first consisted of only GLNase and CPS.A, whereas the second lacked CPS.A and had the GLNase domain fused directly to CPS.B. Remarkably, both proteins catalyzed the entire series of reactions involved in glutamine-dependent carbamyl phosphate synthesis. The stoichiometry, like that of the native enzyme, was 2 mol of ATP utilized per mol of carbamyl phosphate formed. GLN-CPS.B is allosterically regulated, whereas GLN-CPS.A was insensitive to effectors, a result consistent with evidence showing that allosteric effectors bind to CPS.B. These properties are not peculiar to the mammalian protein, because the separately cloned CPS.A subdomain of the E. coli enzyme was also found to catalyze carbamyl phosphate synthesis. Gel filtration chromatography and chemical cross-linking studies showed that these molecules are dimers, a structural organization that may be a prerequisite for the overall reaction. Thus, the homologous CPS.A and CPS.B subdomains are functionally equivalent, although in the native enzyme they may have different functions resulting from their juxtaposition relative to the other components in the complex.

Carbamyl-phosphate synthetase catalyzes the formation of carbamyl phosphate from glutamine, bicarbonate, and 2 mol of ATP (1)(2)(3). The mechanism has been established for Escherichia coli CPSase 1 by measuring the partial reactions, trapping intermediates, and positional isotope exchange (4 -10) and is thought to be general for all members of this family of enzymes (1). The biosynthesis involves four partial reactions.
With the exception of mitochondrial CPSase I, which uses ammonia as the physiological substrate, all known carbamylphosphate synthetases use glutamine as the nitrogen donor (11,12). Glutamine is hydrolyzed (Reaction 1) on a separate amidotransferase domain or subunit and ammonia is transferred to the synthetase component of the complex, which catalyzes all of the other partial reactions. Bicarbonate is activated by ATP (Reaction 2), forming the transient intermediate carboxy phosphate, which reacts with ammonia (Reaction 3) to form carbamate. In the last step (Reaction 4), carbamyl phosphate is formed by phosphorylation of carbamate by a second mole of ATP.
E. coli CPSase consists of a 42-kDa small subunit and a 118-kDa large subunit (13). Separation of the subunits (13,14) showed that the small subunit hydrolyzes glutamine, whereas the large subunit catalyzes ammonia-dependent carbamyl phosphate synthesis and binds allosteric effectors. Mammalian pyrimidine-specific CPSase (CPSase II) is part of a multifunctional protein, CAD (15)(16)(17), which also catalyzes the second and third steps in the de novo pyrimidine biosynthetic pathway. The CPSase component (18,19) of CAD has a 40-kDa amidotransferase or glutaminase domain fused via a 30-residue linker to the amino end of a 120-kDa synthetase domain. Despite the differences in structural organization, the bacterial and mammalian proteins have a high degree of sequence similarity. When the E. coli CPSase large subunit was sequenced (20), it was found to consist of two homologous halves or subdomains. CAD CPSase (18,19) and all other known CPSase molecules (21)(22)(23)(24)(25)(26)(27)(28) have this internal sequence homology, suggesting that the synthetase domain arose by an ancestral gene duplication.
Sequence analysis (18,20,29) revealed that the two CPS-ase synthetase subdomains, CPS.A and CPS.B, contain regions that are homologous to the nucleotide binding site of enzymes that utilize ATP. Chemical modification studies of E. coli (30), CAD (31), and the mitochondrial urea cycle enzymes (29,32) have shown that ATP analogs bind to both halves of the synthetase domain. Moreover, these affinity reagents differentially affect the two ATP-dependent partial reactions. Recently, Post et al. (33) found that the two ATPdependent partial reactions could be selectively inhibited by introducing site-specific mutations into the two halves of the molecule. Affinity cleavage of rat mitochondrial CPSase (36) showed that CPS.B catalyzes the phosphorylation of carbamate. Thus, the current view holds that the two homologous halves of the CPSase synthetase subunit or domain are specialized to carry out the two different ATP-dependent partial reactions. CPS.A catalyzes the initial ATP-dependent activation of bicarbonate, whereas the synthesis of carbamyl phosphate from carbamate occurs on CPS.B. To obtain further evidence for this model, we have separately cloned and expressed the two CPSase subdomains of CAD and obtained the unexpected result that CPS.A and CPS.B are functionally equivalent.

EXPERIMENTAL PROCEDURES
Plasmids and Strains-The plasmid pHG-17 is an intermediate in the construction (35) of the full-length CAD cDNA clone, pCK-CAD10. The plasmid contains the CAD entire coding sequence in a vector (36) derived from pEK81. The insert is located immediately downstream of the pyrBI promoter, which controls the expression of E. coli aspartate transcarbamylase.. The 9.1-kb plasmid, pCAD142 (37), was a gift of George Stark (Cleveland Clinic, Clevelend, OH), whereas pHN12, encoding the large subunit of E. coli carbamyl-phosphate synthetase, was kindly provided by Dr. Carol Lusty (The Public Health Research Institute of the City of New York, New York) as was the E. coli strain, L673 (38), which is defective in the carA and carB genes, encoding both E. coli carbamyl-phosphate synthetase subunits, as well as the Lon protease.
Cell Growth and Recombinant DNA Methods-Cells were grown, and the expression of the recombinant proteins under the control of the pyrBI promoter was induced as described (35) previously. Transformation was carried out by the Hanahan procedure (39). Restriction digests, ligations, and other DNA methods were carried out using standard protocols (40).
Protein Methods and Enzyme Assays-The procedures for the isolation of native CAD (17,41) from an overproducing strain of Syrian hamster cells (165-23) and of the E. coli CPSase large subunit (42) from pHN12 transformants were described previously. Protein concentrations were determined by the Bradford dye-binding method (43). SDS gel electrophoresis was carried out on 7.5-15% linear polyacrylamide gradients using the Laemmli procedure (44). Carbamyl-phosphate synthetase activity was assayed using the radiometric procedure (17,45), in which carbamyl phosphate was trapped as acid stable carbamyl aspartate. Bicarbonate-dependent ATPase activity, carbamyl phosphatedependent ATP synthetase activity, and glutaminase activity were measured spectrophotometrically as described (33) previously. The only modification was that the glutaminase assay contained 15% Me 2 SO and 2.5% glycerol.
Gel Filtration Chromatography-The molecular mass of the recombinant proteins was determined by gel filtration chromatography (46) on a 1.9 ϫ 42-cm Sephacyl S-300 column equilibrated in 0.05 M Tris-HCl, pH 7.4, 1 mM dithiothreitol, and 5% glycerol. A 0.50-ml sample containing 14 -120 g of the protein was applied to the column. The column was calibrated with 1 mg of commercial standard proteins (Sigma), carbonic anhydrase (29 kDa), bovine serum albumin (66 kDa), and alcohol dehydrogenase (150 kDa), and the void volume was determined with blue dextran. The column was eluted with the equilibration buffer, 1.0-ml fractions were collected, and the elution position was determined by measuring the CPSase activity of the recombinant proteins or by determining the absorbance at 280 nm for the standard proteins.
Chemical Cross-linking-The recombinant proteins were crosslinked with dimethyl suberimidate following the procedure (17,47) used for intact CAD. The reaction mixture consisted of 0.3 mg of cross-linking reagent and 100 -200 g of protein in a total volume of 1.0 ml of 0.05 M HEPES, pH 9.2, and 5% glycerol. The reaction, which was carried out at room temperature, was initiated by the addition of the cross-linking reagent. Samples (0.05 ml) were taken at 5-min intervals over a period of 180 min, and the reaction was quenched by the addition of 1 M glycine to a final concentration of 0.1 M. The extent of crosslinking was determined by SDS-polyacrylamide gel electrophoresis.

Construction of Recombinant Mammalian Proteins-The
contiguous GLNase and CPSase domains of CAD ( Fig. 1) span residues 1-1456 of the CAD polypeptide. The CPSase domain (residues 395-1456) is connected to the GLNase domain (residues 1-366) by a 30-amino acid linker. The junction between the CPSase subdomains was determined (18) by alignment of the sequence of the two halves of mammalian CPSase and four other known CPSase sequences. The CPS.A subdomain includes residues 395-933, whereas CPS.B extends from residues 934 to 1456. There is no obvious linker connecting the two CPSase subdomains.
The plasmid pHL-1, which encodes the entire mammalian GLNase CPSase (Fig. 2), was constructed from pHG-17 (35), which contained GLNase, CPS.A, and most of the CPS.B coding sequences. The missing CPS.B coding sequences were obtained by the polymerase chain reaction using the partial CAD cDNA clone, pCAD142 (37), as a template. The 5Ј primer incorporated a XhoI site at nucleotide 3211 in the CAD coding sequence, whereas the 3Ј primer contained an EcoRI site at nucleotide 4362 and a stop codon. The recombinant protein starts at the first residue in CAD and extends to the exact end of the CPSase domain (residue 1456) defined by sequence homolgy.
The plasmid pHG-GCA, which has GLNase and CPS.A coding sequences, was also made from pHG-17 by deleting the CPS.B coding sequences. The plasmid was restricted with Bsp-mII and EcoRI. The large fragment was isolated, reacted with Klenow, and religated. The resulting construct encoded a protein that began at the CAD initiation codon and included all of the GLNase and CPS.A domains and 28 amino acids of CPS.B. In addition, two amino acid residues derived from the vector are added to the carboxyl end of the domain (-Asn-Ser-COO Ϫ ).
Preliminary characterization of pHG-GCA prompted us to construct another plasmid lacking CPS.A. The plasmid, pHG-GCB, was derived from pHL-1. The CPS.A coding sequences and all of the residues in the GLN CPS linker were deleted by restriction with BspmII and reaction with Mung Bean nuclease, followed by SmaI restriction. The 6.0-kb fragment was gel purified and religated giving a plasmid in which the sequences encoding the GLNase domain (residues 1-366) were fused in frame with CPS.B coding sequences (residues 964-1456). This protein lacks 30 residues of the amino end of the previously defined CPS.B domain.
Expression and Purification-The plasmids were initially transformed into the E. coli strain L673, a uridine auxotroph that lacks the E. coli carA and carB genes encoding the subunits of carbamyl-phosphate synthetase. The untransformed cells are unable to grow on minimal medium (Table I). Cells transformed with pHG-17 encoding the truncated mammalian carbamyl-phosphate synthetase grew very slowly. In contrast, transfomation with pCKCAD-10, pHL1, and pHG-GCA fully complemented the E. coli defect, and the transformants grew vigorously. Although pCKCAD-10 and pHL1 encode the entire GLNase CPSase region of CAD and would be expected to be functional, complementation by pHG-GCA was not anticipated because this plasmid encodes an incomplete CPSase lacking CPS.B.
Measurements of the CPSase activity in extracts of cells transformed with pHL-1 indicated that the level of expression was very low, and attempts to purify the full-length mammalian CPSase were unsuccessful. Both GLN-CPS.A and GLN-CPS.B were expressed at moderate levels (1-2% of the total soluble protein) and could be readily isolated by chromatography (Fig. 3) on a DEAE-Sephacyl column. GLN-CPS.A was nearly homogeneous (not shown), whereas GLN-CPS.B had three small contaminates (Fig. 3) in the 25-35-kDa molecular mass range, which together represented about 15% of the total purified protein.  2. Construction of the recombinants. The plasmid pHG-17, a previously described intermediate (35) in the construction of a full-length CAD cDNA clone, encodes CAD residues 1-1232. The remaining CPS.B coding sequences were introduced into pHG-17 by inserting a polymerase chain reaction product obtained using pCAD142, another partial cDNA CAD clone (37), as a template. The 5Ј primer incorporated the XhoI site at nucleotide 3211 in the CAD coding sequence, whereas the 3Ј primer contained the EcoRI site at 4362 and a stop codon. The polymerase chain reaction product was cleaved with both of these enzymes and ligated to the 5.6-kb XhoI EcoRI fragment of pHG-17 to give a construct, pHLI, which encoded the entire mammalian glutamine-dependent carbamyl-phosphate synthetase. The GLN-CPS.A recombinant, pHG-GCA, was obtained by restricting pHG-17 with BspmII at 2885, close to the junction between CPS.A and CPS.B and with EcoRI. The 6.3-kb fragment was gel purified, and the 5Ј extensions were filled using the Klenow fragment of DNA polymerase prior to blunt end ligation with T4 DNA ligase. The GLN-CPS.B recombinant, pHG-GCB, was obtained by restricting pHL1 with BspmII and SmaI. The 5Ј extensions of the gel-purified 6.0-kb fragment were removed with Mung bean nuclease prior to ligation. Two constructs encoding the CPS.A domain of E. coli carbamyl-phosphate synthetase were obtained from pHN12 (38), which encodes the large subunit of the E. coli enzyme. The first, pHG-ECA1, was obtained by restriction with EcoRI, Mung bean nuclease treatment, and religation of the purified 4.9-kb fragment. To produce the second construct, pHN12 was cleaved with AccI and EcoRI. The 5Ј extensions of the purified 4.9-kb fragment were filled in by reaction with the Klenow fragment and religated to give pHG-ECA2.  (Table II). The rate of carbamyl-phosphate synthetase was measured by initiating the reaction with [ 14 C]NaHCO 3 and trapping the radiolabeled carbamyl phosphate. Both GLN-CPS.A and GLN-CPS.B catalyzed carbamyl phosphate synthesis, and both molecules could use either ammonia or glutamine as a nitrogen donor. Thus, both recombinant proteins could utilize ammonia produced endogenously by the GLNase domain, although under these assay conditions, the activity was somewhat higher with exogenous ammonia. Both GLN-CPS.A and GLN-CPS.B catalyze ATP hydrolysis in the absence of a nitrogen donor, a reaction that is taken as a measure of ATP-dependent activation of bicarbonate. In the absence of a nitrogen donor, the reaction stops with the formation of the unstable carboxy phosphate intermediate. When glutamine is included in the reaction mixture, the second ATP-dependent partial reaction, the phosphorylation of carbamate, also occurs, and ATP hydrolysis increases 2-fold from 0.169 to 0.328 mol/min/mg. Thus, all of the carboxy phosphate intermediate is quantitatively converted to carbamate and sub-sequently to carbamyl phosphate. As for the native enzyme, the ratio of ATP hydrolyzed to carbamyl phosphate synthesized is 2 for both GLN-CPS.A and GLN-CPS.B, reflecting the stoichiometry of the overall reaction. The hydrolysis of glutamine was also directly measured by a coupled enzyme assay. In the absence of ATP and bicarbonate, glutamine hydrolysis is relatively low but increases 2.5-3.0-fold when these substrates are present.
Steady State Kinetics-GLN-CPS.A exhibited typical hyperbolic ATP saturation curves for the glutamine-dependent carbamyl-phosphate synthetase activity (Fig. 4). The Allosteric Regulation-It is interesting that although GLN-CPS.A is not affected by UTP and PRPP, both allosteric effectors modulate the activity of GLN-CPS.B (Fig. 3). The effects appear to be significantly smaller than those observed for the purified native enzyme. For GLN-CPS.B, the maximum inhibition by 2 mM UTP, observed at 0.25 mM ATP, was 50%. At the same ATP concentration, 50 M PRPP activates 243%.
The ATP saturation curves of GLN-CPS.A in the presence of 2 mM UTP or 50 M PRPP are virtually indistinguishable from the curve obtained in the absence of allosteric effectors (Fig. 4). Substrate inhibition complicated the analysis of the ATP saturation curves of GLN-CPS.B. As observed for the glutaminedependent CPSase activity of CAD, the inhibitor UTP shifts the observed K m , the concentration of ATP that gave half-maximal activity, to higher ATP concentrations, whereas PRPP has the opposite effect. A more detailed kinetic study would be required to determine whether the effectors truly alter the V max . The insensitivity of GLN-CPS.A to allosteric effectors is consistent with previous studies of CPSase (48 -51), which showed that the locus of regulation lies at the carboxyl end of the CPS.B subdomain.
E. coli Carbamyl-phosphate Synthetase CPS.A Subdomain-Because the observation that the individual CPSase subdomains can catalyze the overall synthesis of carbamyl phosphate was unexpected, we were curious as to whether carbamyl-    . 2) was used in the construction of the E. coli CPS.A recombinant. The plasmid pHN12, which encodes the carB gene, was restricted with EcoRI, and the 4.9-kb fragment was reacted with Mung bean nuclease to ensure that an adjacent stop codon would be in frame and then purified and religated. The resulting plasmid, pHG-ECA1, encoded a protein extending from the CPSase start codon to Ile 576 and thus contained all of the CPS.A and 16 residues of CPS.B. SDS gel electrophoresis revealed an abundant new protein of the expected molecular mass, 63.4 kDa, in extracts of pHG-ECA1-transformed L673 cells, and preliminary assays showed that these extracts contained carbamyl-phosphate synthetase activity. The recombinant protein was partially purified by fractional ammonium sulfate precipitation (35-65%) and chromatography on a Sephacyl S-300 column. The protein catalyzed the overall synthesis of carbamyl phosphate from bicarbonate, ATP, and ammonia (Table III). The stoichiometry was 2 mol of ATP used for the synthesis of 1 mol of carbamyl phosphate. The carbamylphosphate synthetase activity of E. coli CPS.A (Table III) is unaffected by the inhibitor UMP or the activator ornithine. After this work was completed, we learned of a similar experiment (52) that was carried out with E. coli CPSase. Both CPS.A and CPS.B were separately cloned, but the resulting proteins lacked carbamyl-phosphate synthetase activity and were unable to catalyze either partial reaction. In the case of CPS.A, a different restriction site, AccI at nucleotide 1666 was used to construct the recombinant. Consequently, we made a second E. coli CPS.A clone to determine whether this difference was significant. The plasmid pHG-ECA2 was constructed (Fig.  2) by restricting the parental plasmid pHN12 with AccI and EcoRI. The large fragment was isolated and the ends made blunt by filling in the extensions with DNA polymerase Klenow fragment and religated. Both pHG-ECA1 and pHG-ECA2 were transformed into E. coli L673 cells, and the transformants were grown to stationary phase. No ammonia-dependent CPSase activity could be detected (Fig. 5) in extracts of pHG-ECA2 transformed cells, confirming that the smaller protein was inactive.
Subunit The molecular mass of these proteins under nondenaturing conditions was determined by gel filtration chromatography (Fig. 6) on a calibrated Sephacyl S-300 column. GLN-CPS.A eluted as a single species (Fig. 6A) with a calculated molecular mass of 185 kDa, suggesting that the protein is dimeric. Similarly GLN-CPS.B had a calculated molecular mass of 250 kDa.
There were two peaks in the elution profile of E. coli CPS.A corresponding to species with molecular masses of 55 and 127 kDa. Thus, this protein appears to be a mixture of monomers and dimers. When the fractions containing the monomer were pooled, concentrated, and reapplied to the column, the protein eluted exclusively as a dimer, suggesting that the monomeric and dimeric species are in equilibrium. No monomer was detected when E. coli CPS.A was chromatographed on the same column equilibrated in a buffer containing the substrates, 5 mM sodium bicarbonate, 10 mM ATP, and 100 mM ammonium chloride.
The dimeric structure of GLN-CPS.B and E. coli CPS.A was confirmed by cross-linking with dimethyl suberimidate as described under "Experimental Procedures." Initially, only the 94-kDa GLN-CPS.B monomer was present, but as the crosslinking reaction proceeded, the monomer gradually disappeared and a 188-kDa dimeric species appeared in the reaction mixture (data not shown). Complete conversion to the crosslinked dimer occurred within 30 -40 min. Similar results were obtained with E. coli CPS.A.
The results reported here, which show that CPS.A and CPS.B, as well as the E. coli CPS.A, can be individually expressed as stable, functional proteins, clearly demonstrate that  the homologous halves of the synthetase represent distinct, autonomously folded subdomains. The most remarkable discovery is that CPS.A and CPS.B are not only functional, but they are functionally equivalent. Both subdomains carry out the two ATP-dependent partial reactions and the overall synthesis of carbamyl phosphate. These reactions are catalyzed by both mammalian and bacterial subdomains, suggesting that this capability may be a general characteristic of all CPSase molecules. Measurements of the steady state kinetic parameters showed that both GLN-CPS.A and GLN-CPS.B catalyze the overall reaction nearly as well as the native enzyme. The turnover numbers are of the same order of magnitude and the K m values are significantly lower, suggesting that the recombinants have a higher affinity for ATP than the intact enzyme. Rubio et al. (48) proposed that an ancestral carbamate kinase is the evolutionary precursor of carbamyl-phosphate synthetase. This hypothesis is based on the observation that carbamate kinase from Pseudomonas aeruginosa (53) is homologous to the first 300 amino acids of the CPS.B domain of E. coli CPSase. A possible objection to our interpretation is that the separately expressed CPS.A and CPS.B subdomains function as carbamate kinases, converting carbamate formed nonenzymatically in situ by reaction of ammonia and bicarbonate to carbamyl phosphate. Carbamate formation would be favored by the high concentrations of both bicarbonate and ammonia present when the ammonia-dependent carbamyl-phosphate synthetase is assayed. However, the reaction proceeds at about the same rate when glutamine is used as the nitrogen donor under conditions where ammonia does not accumulate, making this explanation unlikely. Moreover, the stoichiometry of the reactions for both glutamine-and ammonia-dependent reactions clearly indicates that two molecules of ATP are utilized in each catalytic cycle. Thus, either the CPS.A or CPS.B recombinant catalyzes the same overall stoichiometric reaction as the entire synthetase domain in the native enzyme.
This conclusion has several interesting implications. In addition to providing confirmation that both CPS.A and CPS.B bind ATP, these studies show that each subdomain must also bind both the substrate bicarbonate and the intermediate carbamate. The latter conclusion is perhaps not surprising because both molecules are substrates for ATP-dependent phosphorylation reactions and have very similar structures. Except for the -OH and the -NH 2 groups, which are isosteric, bicarbonate and carbamate are identical. Biotin carboxylase, which exhibits several mechanistic similarities (54) to carbamyl-phosphate synthetase, has one ATP binding site and has been found to catalyze both bicarbonate-dependent ATP hydrolysis (55) and carbamyl phosphate-dependent ATP synthesis (56). A more remarkable implication is that the interaction of both CPS.A and CPS.B with the GLNase domain must be nearly identical because both efficiently utilize glutamine as a nitrogen donor. The ammonia released by glutamine is transferred to or produced near the site of formation of carboxy phosphate, and both subdomains must have an ammonia binding site.
The gel filtration chromatography and cross-linking studies showed that both mammalian and bacterial CPSase subdomains associate to form dimers. Although in the case of E. coli CPS.A, a significant fraction of the protein is monomeric in the absence of substrates, E. coli carbamyl-phosphate synthetase (57)(58)(59) exists in equilibrium between monomer, dimer, and tetramers. The state of association is dependent on the concentration of the protein and the presence of MgATP, K ϩ , and allosteric effectors. However, Anderson (60) demonstrated that FIG. 5. Activity of truncated E. coli CPSase recombinants. Cultures (100 ml) of L673 cells transformed with pHG-ECA1 or pHG-ECA2 were grown to stationary phase. The cells were harvested, and a 1-ml cell extract was prepared as described under "Experimental Procedures." The ammonia-dependent CPSase activity was measured as a function of the volume of extract in the assay. The fractions (2.0 ml) were assayed for glutamine-dependent CPSase activity. The molecular mass, M r , was estimated from plots of ln M r against K av for five standard proteins. The curve for GLN-CPS.A was normalized by adjusting the highest value to correspond to the highest observed activity obtained for GLN-CPS.B. B shows the elution profile of E. coli CPS.A (1.7 ml, 0.82 mg/ml) on the same column. In the absence of substrates (q), two peaks, corresponding to the monomer and dimer were observed. When the monomer peak was isolated, concentrated to the same initial protein concentration, and rechromatographed on the same column (f), the protein eluted as a dimer. The experiment was then repeated with the column equilibrated and eluted in the same buffter containing 10 mM ATP, 5 mM sodium bicarbonate, and 100 mM ammonium chloride. The corresponding elution profile (E) indicated that the protein was dimeric. The molecular mass was calculated as described above for A. The curves were normalized to the highest value observed for E. coli CPS.A in the absence of substrates. catalytic activity of the enzyme is not directly dependent on the oligomeric state. Thus, self-association of the bacterial subdomains and the homologous mammalian subdomains might be expected. The presence of substrates promoted the complete conversion of the monomeric form of E. coli CPS.A into dimers.
The interpretation of these results depends on the total number and location of the ATP binding sites on the molecule. Sequence analysis of E. coli and yeast arginine specific CP-Sases (20) revealed two short conserved sequences in each half of the synthetase subunit, which may represent either a single composite ATP binding site or possibly two different ATP binding sites in each half. Subsequently, two alternative consensus sequences (29) corresponding to different regions of the ATP binding site found in many proteins (61) were identified in each duplicated half of the synthetase domain. Large regions homologous to yeast pyruvate carboxylase (62) and biotin carboxylase subunit of chicken acetyl CoA carboxylase (63) have been noted in E. coli (62) and CAD (18) CPSase. These postulated ATP binding subdomains, A2 and B2 (18), encompass large segments in the middle of CAD CPS.A and CPS.B subdomains (Fig. 1) and are flanked by protease-sensitive sites (64) commonly found at subdomain junctions. Because the carboxylases bind only one ATP molecule, which is used to activate bicarbonate, it is probable, but not certain, that there is a single ATP binding site on CPS.A and on CPS.B.
The two-ATP dependent partial reactions catalyzed by E. coli CPSase exhibit different catalytic and regulatory properties (4) and chemical modification (30,66) and inhibitor studies (67), suggest that they occur at distinct sites. Moreover, pulse-chase experiments (68 -70) with the E. coli CPSase and steady state kinetic studies of both E. coli (71) and mammalian (72) enzymes suggest that CPSase can bind 2 mol of ATP simultaneously. Both ATP-dependent partial reactions of CAD are inactivated by the reactive ATP analog, 5Ј-p-fluorosulfonylbenzoyl adenosine (FSBA), but at different rates (31), providing further evidence that the partial reactions occur at distinct sites. When CAD was reacted (31) with ( 14 C) FSBA, all of the radiolabeled peptides map at or near the postulated ATP binding subdomains of CPS.A and CPS.B. Similarly, FSBA modification of mammalian mitochondrial CPSase I (30,73,74) showed that inactivation was accompanied by the incorporation of 2 mol of the reagent and that the radiolabeled peptides mapped in both halves of the synthetase domain. Taken together, these studies ( Fig. 1) strongly suggest that there are at least two distinct ATP binding sites that catalyze different partial reactions. Moreover, there is convincing evidence that ATP binds to both CPS.A and CPS.B. The results reported here are fully consistent with this conclusion and provide the most direct proof that there is an ATP binding site on each synthetase subdomain. Although each subdomain very probably binds a single ATP, the possibility of two ATP molecules binding to each subdomain cannot be conclusively discounted.
In the most definitive study of the function of the CPSase subdomains, Post et al. (33) have shown that replacement of residues in the CPS.A domain of E. coli CPSase by site-directed mutagenesis results in 90% inactivation of the bicarbonatedependent ATPase activity with only 10% inhibition of carbamyl phosphate-dependent ATP synthesis. Conversely, modification of exactly homologous residues in CPS.B abolishes the ATP synthesis activity without significantly altering the ATPase activity. These authors concluded that CPS.A catalyzes the activation of bicarbonate, whereas CPS.B phosphorylates carbamate. Subsequent mutagenesis studies (65) support this conclusion. Moreover, studies of rat mitochondrial CPSase I by Alonso and Rubio (34) provided convincing evidence that the subdomains of this enzyme are also specialized. Oxidative cleavage mediated by FeATP, bound to the site that catalyzes the phosphorylation of carbamate, occurs exclusively within CPS.B.
Since the discovery of the internal duplication of the CPSase synthetase, three suggestions have been advanced regarding the function of the two CPSase subdomains: 1) the subdomains have specialized functions with each one catalyzing a different ATP-dependent partial reaction, 2) the synthetase domain is a covalently linked dimer with each monomer catalyzing all of the reactions involved in carbamyl phosphate synthesis, and 3) one of the subdomains catalyzes the entire overall reaction, whereas the other half has evolved into a regulatory subdomain.
The results reported here conclusively disprove the third proposal, which was based on analogy to phosphofructose kinase. This hypothesis predicts that CPS.A would have catalytic activity but that CPS.B should be inactive. Our findings are consistent with previous results (48 -51) that showed that the allosteric sites are located at the carboxyl end of CPS.B. However, CPS.B catalyzes the overall reaction as well as CPS.A and thus cannot be solely involved in regulation. The second proposal, that each subdomain catalyzes the entire reaction, is unlikely to be correct because it requires that either both ATPdependent partial reactions occur at the same site or that there are two ATP binding sites on each subdomain. The evidence cited above clearly shows that the two partial reactions occur at different sites and strongly suggests that there is only one ATP binding site in each subdomain. Moreover, the site-directed mutagenesis studies strongly favor the first proposal, which has now become the generally accepted view, that CPS.A and CPS.B in the native enzyme have specialized functions.
Although our results on the surface would seem to be more in accord with the proposal that CPSase is a covalently linked dimer, this interpretation is at odds with the compelling mu- Only GLN-CPS.A is shown, but GLN-CPS.B is presumed to function in an identical way. In the wild type protein, CPS.A activates bicarbonate, whereas CPS.B phosphorylates carbamate and forms carbamyl phosphate. For the recombinants, the model assumes that: 1) the noncovalently associated dimer is required for activity, 2) both subunits in the dimer, A and AЈ in this schematic drawing, are functionally equivalent, and 3) in some catalytic cycles glutamine binds to the GLNase domain associated with the A subunit, which then catalyzes bicarbonate activation, and carbamyl phosphate formation is then catalyzed by subunit AЈ. In other cycles, glutamine first binds to the GLNase domain associated with subunit AЈ, and the role of subunits A and AЈ is reversed. tagenesis studies of the E. coli enzyme. Therefore, we propose the following model in an attempt to reconcile the evidence that suggests that CPS.A and CPS.B subdomains catalyze different partial reactions with the observations reported here. The model, depicted in Fig. 7, assumes that 1) the monomer may or may not catalyze the partial reactions, but the overall reaction can only be catalyzed by the dimer, 2) the two monomers in the CPS.A dimer (or CPS.B dimer) are functionally equivalent in the sense that they catalyze both partial reactions and can each hydrolyze glutamine and efficiently transfer ammonia for carbamyl phosphate synthesis, and 3) in any given catalytic cycle, either of the subunits may catalyze the activation of bicarbonate, but once carboxy phosphate is formed by one subunit, the other assumes the role of carbamate phosphorylation by default. If this explanation is correct, the two associated CPS.A or CPS.B subdomains in the homodimer are analogous to the covalently linked CPS.A CPS.B subdomains in native CPSase. The mutagenesis studies suggest that in the native enzyme, the function of CPS.A and CPS.B are fixed and do not alternate in this fashion between bicarbonate activation and carbamate phosphorylation. A possible explanation is that the GLNase domain may interact more intimately with CPS.A, to which it is covalently linked in the parent mammalian molecule, and deliver ammonia directly to the site of formation of carboxy phosphate. If so, carbamyl-phosphate synthetase may represent an interesting example in which, although the constituent domains are functionally equivalent, the reaction each catalyzes depends on the topology of the complex.