The Smallest Carbamoyl-phosphate Synthetase

Escherichia coli carbamoyl-phosphate synthetase (CPSase) is comprised of a 40-kDa glutaminase (GLN) and a 120-kDa synthetase (CPS) subunit. The CPS subunit consists of two homologous domains, CPS.A and CPS.B, which catalyze the two different ATP-dependent partial reactions involved in carbamoyl phosphate synthesis. Sequence similarities and controlled proteolysis experiments suggest that the CPS subdomains consist, in turn, of three subdomains, designated A1, A2, A3 and B1, B2, B3 for CPS.A and CPS.B, respectively. Previous studies of individually cloned CPS.A and CPS.B from E. coli and mammalian CPSase have shown that homologous dimers of either of these “half-molecules” could catalyze all three reactions involved in ammonia-dependent carbamoyl phosphate synthesis. Four smaller recombinant proteins were made for this study as follows: 1) A1-A2 in which the A3 subdomain was deleted from CPS.A, 2) B1-B2 lacking subdomain B3 of CPS.B, 3) the A2 subdomain of CPS.A, and 4) the B2 subdomain of CPS.B. When associated with the GLN subunit, A1-A2 and B1-B2 had both glutamine- and ammonia-dependent CPSase activities comparable to the wild-type protein. In contrast, the 27-kDa A2 and B2 recombinant proteins, which represent only 17% of the mass of the parent protein, were unable to use glutamine as a nitrogen donor, but the ammonia-dependent activity was enhanced 14–16-fold. The hyperactivity suggests that A2 and B2 are the catalytic subdomains and that A1 and B1 are attenuation domains which suppress the intrinsically high activity and are required for the physical association with the GLN subunit.

(CPS) subunit, whereas in its mammalian counterpart (5)(6)(7), the GLN and CPS domains are fused and are part of CAD, a multifunctional protein that also has aspartate transcarbamoylase and dihydroorotase activities. Despite the differences in structural organization, the amino acid sequences are clearly similar (8 -21) suggesting that all of these molecules are comprised of homologous domains and subdomains with analogous functions.
The isolated GLN subunit of E. coli CPSase (3,4) and the separately cloned GLN domain of CAD (22) hydrolyze glutamine (Reaction 1) and transfer ammonia to the CPS domain.
glutamine ϩ H 2 O 3 glutamate ϩ NH 3 (Eq. 1) All of the other partial reactions (Reactions 2-4) occur on the CPS domain or subunit (3,4). The determination of the amino acid sequence of CPSase from many different organisms (8 -21) revealed that the CPS domain of these molecules invariably consists of two highly homologous halves, designated CPS.A and CPS.B. The two ATP-dependent partial reactions (Reactions 2 and 4) occur at different sites, and there is now convincing evidence (23)(24)(25)(26) that CPS.A catalyzes the activation of bicarbonate, whereas the phosphorylation of carbamate to form carbamoyl phosphate is catalyzed by CPS.B. The site of carbamate formation (Reaction 3) is not known, but the rate of production of NH 3 and carboxy phosphate is precisely coordinated, and once formed, these intermediates probably react spontaneously.
Although CPS.A and CPS.B have different specific functions when fused together in the wild-type protein, the isolated domains are functionally equivalent. When CPS.A and CPS.B were separately cloned (27), the half-molecules could each catalyze both partial reactions and the overall synthesis of carbamoyl phosphate from ammonia. When the mammalian domains are fused to the GLN domain, glutamine is hydrolyzed and the ammonia thus produced is used for carbamoyl phosphate synthesis. Thus, the catalytic sites are located within both CPS.A and CPS.B. * This research was supported by a United States Public Health Service Grant GM47399. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
There is extensive evidence that CPS.A and CPS.B domains are in turn comprised of subdomains. The CPSase model (15,28,29), illustrated in Fig. 1, is based in part on sequence homology (8, 15, 30 -32) between regions of the carbamoylphosphate synthetase and several different kinases. Moreover, controlled proteolysis of CAD (29,(33)(34)(35), the mammalian urea cycle enzyme, CPSase I (36 -38), and more recently E. coli CPSase (32,39) showed that cleavage occurs at or near the junctions of many of the putative subdomains.
The regions designated A2 and B2 have been implicated in ATP binding. Consensus sequences for the nucleotide binding sites (8,40) and the active site of the homologous kinases (8, 15, 30 -32) are located in the central region of CPS.A and CPS.B. Chemical modification of CAD (41) and mammalian CPSase I (40,(42)(43)(44)(45) as well as site-directed mutagenesis of E. coli CPSase (23,46,47) showed that ATP binds to A2 and B2. The conclusions of the biochemical studies are consistent with the recently solved x-ray structure (65) of E. coli carbamoyl-phosphate synthetase.
To establish whether the subdomains of CPS.A and CPS.B correspond to autonomously folded substructural elements with specific functions, we have separately cloned and expressed some of the individual E. coli CPSase subdomains and combinations of subdomains. encoding only the B2 subdomain (residue 658 -892), was constructed by reacting pHGB2B with AvaI-(4339) and SstI (in the vector). The 3.5-kb fragment was then reacted with the Klenow fragment and religated. The recombinant protein has the 7 amino acid residues of the amino end of B3 and 4 residues from the vector appended to the carboxyl end.
All of the plasmids, except pHGB2, also contain the carA gene so that the GLN subunit is also expressed as a separate polypeptide. The four recombinant plasmids constructed for this study also encoded the first 15 amino acids of the E. coli CPS subunit which are appended to the amino end of the expressed protein. It is very unlikely that the presence of these residues alters the function of the recombinant proteins, since the amino end of the CPSase synthetase domain is poorly conserved among members of this family of proteins, both in sequence and in the number of residues preceding the core A1 domain. Moreover 12 of these same 15 residues (50) were deleted from the amino end of the CPS.A domain without effect on the properties of the E. coli enzyme. Nevertheless, we are now refining these constructs to remove these additional residues and more precisely define the limits of the functional domains.
Protein Methods and Enzyme Assays-The procedures for the isolation of the wild type E. coli CPSase (28) and the CPS subunit (51) from pLLK12 and pHN12, respectively, were described previously. Protein concentrations were determined by the Bradford dye-binding method (52) and by scanning Coomassie Blue-stained polyacrylamide gels. SDS-gel electrophoresis was carried out on 7.5-15% linear polyacrylamide gradients (53). The gels were scanned, and the concentration of the individual proteins was determined by measuring the ratio of peak density to total density in a given lane and the total amount of protein applied to the gel. The background density was subtracted, and all measurements were made within the linear range of the densitometer. Carbamoyl-phosphate synthetase activity was assayed using a radiometric procedure (7,33), whereas the GLNase activity was measured (23) spectrophotometrically. The only modification was that the GLNase assay buffer also contained 15% Me 2 SO and 2.5% glycerol.
Gel Filtration Chromatography-The molecular mass was determined by gel filtration chromatography using an open 1.9 ϫ 42-cm Sephacryl S-300 column equilibrated with 0.5 M Tris-HCl, pH 7.4, 1 mM dithiothreitol, and 5% glycerol. Alternatively, a Pharmacia FPLC system fitted with a 1.6 ϫ 50-cm Sephacryl S-300 high resolution (HR 16/50) Pharmacia column was used. The column was equilibrated in 0.1 M KH 2 PO 4 , 1 mM EDTA, pH 7.5, and eluted at a flow rate of 0.4 ml/min at room temperature. Column fractions were analyzed by measuring the absorbance at 280 nm, SDS-gel electrophoresis, and CPSase assays. The column was calibrated with carbonic anhydrase (29 kDa), bovine serum albumin (66 kDa), alcohol dehydrogenase (150 kDa), and blue dextran.

Construction and Expression of the Recombinant Plasmids-
Extensive sequence identity (15) suggests that E. coli and mammalian CPSases have common subdomain structures. We elected to work with E. coli CPSase, rather than the mammalian enzyme, because the level of expression of the bacterial protein in E. coli tends to be higher. The assignment of the E. coli CPS subdomain junctions, shown in Fig. 1, was based on the CAD substructural analysis (29) by aligning the sequences of the mammalian and bacterial proteins. The construction (Fig. 2) of the four recombinant plasmids used in this study is described under "Experimental Procedures." Plasmid pHGGA12 encodes the fused A1-A2 subdomains; pHGGB12 encodes B1-B2, and pHGGA2 encodes the A2 subdomain. These three plasmids also contain the intact carA gene so that the GLN subunit is also expressed as a separate polypeptide chain. The plasmid pHGB2, encoding the B2 subdomain, does not include the carA gene.
In L673 cells transformed with pHGGA12 and pHGGB12, the 42-kDa GLN subunit was expressed along with the 40-and a 44-kDa fragment, respectively. A1-A2 and B1-B2 were produced ( Fig. 3) as soluble, stable proteins representing 21 and 14% of the total cellular protein, respectively. Transformation of the same strain with pHGGA2 and pHGB2 produced much lower levels, 0.4% of the soluble 31-kDa A2 and 28-kDa B2 subdomains. The function of the recombinant proteins was assayed in cell extracts prepared as described under "Experimental Procedures." Function of A1-A2 and B1-B2 Subdomains-The A1-A2 and B1-B2 subdomains were both found to catalyze ammonia-dependent carbamoyl phosphate synthesis. The ATP saturation curves ( Fig. 4A) of A1-A2 exhibit typical Michaelis-Menten kinetics with a K m for ATP ( Table I) that was about 3-fold higher than the value obtained for native E. coli CPSase. The k cat 2 values of A1-A2 are approximately the same as that of the parent molecule. For B1-B2 (Fig. 4B and Table I), the k cat values are lower. This result is probably due to appreciable substrate inhibition exhibited by B1-B2, which made it difficult to obtain accurate values for the kinetic parameters.
Surprisingly, these proteins also catalyze the glutamine-de-pendent reactions suggesting that the GLN subunit forms a functional complex with both the A1-A2 and the B1-B2 proteins. Compared with E. coli CPSase, the K m for ATP of A1-A2 is 3-fold higher, whereas the k cat of A1-A2 is not significantly different. Again B1-B2 is inhibited at high ATP concentrations, so that the apparent turnover number and K m are lower than the corresponding values for the E. coli enzyme. ATP saturation curves (data not shown) were also obtained for the glutamine-dependent CPSase activity of A1-A2 and B1-B2 in the presence of 5 mM UMP and 12 mM ornithine, and neither protein was found to be allosterically regulated.
The gene encoding the GLN subunit, carA, is part of the same operon that encodes either A1-A2 or B1-B2, so that the proteins are transcribed from a single mRNA molecule. The ratio GLN/A1-A2, determined by measuring the amount of each polypeptide in cell extracts, was 0.95. The corresponding value of GLN/B1-B2 was found to be 1.05. To test for the formation of a complex of the GLN subunit with A1-A2 and B1-B2, cell extracts were chromatographed (Fig. 5A) on a calibrated Sephacryl S-300 column. A single species having Glndependent CPSase activity was eluted, indicating the formation of a stable complex of GLN and A1-A2 domains. Complex formation was confirmed by SDS-gel electrophoresis of the pooled column fractions (not shown). Similar results were obtained using extracts containing GLN and B1-B2, although this species eluted somewhat earlier as expected because B1-B2 is larger, 44 kDa, than A1-A2, 40 kDa.
Coupling of the Glutaminase and Synthetase Activities-The GLNase activity of E. coli CPSase is low unless ATP and bicarbonate are bound to the CPS subunit. With saturating bicarbonate, the rate of glutamine hydrolysis increases with increasing ATP concentration. The plot (Fig. 6) exhibits typical saturation kinetics, and when extrapolated to saturating ATP (Table II) showed a 12-fold increase in the GLNase activity. This coupling mechanism (26,54,55) ensures that the rate of glutamine hydrolysis matches the rate of production of carboxy phosphate.
The GLNase activity of the complex of the GLN subunit with either A1-A2 or B1-B2 is comparable to that observed for the native protein (Table II). The V max for the GLNase activity of GLN A1-A2 in the presence of saturating ATP and bicarbonate FIG. 1. Subunit and domain structure of E. coli CPSase. This scheme shows the structural organization of E. coli CPSase, a protein that consists of two subunits, a glutaminase subunit, GLN, and a synthetase subunit, CPS, consisting of two subdomains, CPS.A and CPS.B. Previous studies showed that the isolated CPS subunit (3,4) and each of the separately cloned domains (27) catalyzes ammonia-dependent carbamoyl phosphate synthesis. Each of the domains consists of three subdomains A1, A2, and A3 and B1, B2, and B3 for CPS.A and CPS.B, respectively. Four additional recombinants have been constructed for this study. The A3 and B3 regions have been deleted giving A1-A2 and B1-B2. In addition, A2 and B2, the catalytic subdomains (dark shaded area), have been cloned and expressed.

FIG. 2. Construction of the recombinant plasmids.
Four recombinant plasmids were constructed as described under "Experimental Procedures," 1) pHGGA12 encoding the GLN subunit and the fused A1 and A2 subdomains, 2) pHGGA2 encoding the GLN subunit and the A2 subdomain, 3) pHGGB12 encoding the GLN subunit and the fused B1 and B2 subdomains, and 4) pHGB2 encoding the B2 subdomain. (Table II) is somewhat lower than the value obtained for Glndependent CPSase activity (Table I) of this species. We believe that this discrepancy can be attributed to differences in the assay conditions 3 since the rate of glutamine hydrolysis must be at least as fast as the rate of carbamoyl phosphate synthesis.
In contrast to the native enzyme, the GLNase activity of GLN A1-A2 and GLN B1-B2 is relatively insensitive to ATP and bicarbonate. Saturating bicarbonate and ATP resulted in only a 2-fold increase in the rate of glutamine hydrolysis. Thus, although the GLN domain can catalyze the hydrolysis of glutamine and transfer the ammonia to the CPS domain, the functional linkage is impaired.
Function of the A2 and B2 Subdomains-Although the A2 and B2 subdomains represent only 22% of the mass of the CPSase synthetase subunit, both catalyze ammonia-dependent carbamoyl phosphate synthesis. The K m for ATP for both proteins is comparable to the value for E. coli CPSase, but the k cat is appreciably higher. A2 has a k cat value that is 14-fold higher than E. coli CPSase and A1-A2. Similarly, the k cat value of B2 is 16-and 45-fold higher than E. coli CPSase and B1-B2, respectively. The hyperactivity of A2 and B2 with respect to the parent molecule suggests that they are the catalytic subdomains of CPS.A and CPS.B.
The A2 domain was co-expressed with the GLN subunit but did not exhibit detectable glutamine-dependent CPSase activity, indicating that the subdomain was not able to form a functional complex with the GLN subunit. The plasmid encoding the B2 subdomain did not contain the carA gene, but the addition of the purified GLN subunit to B2 did not restore glutamine-dependent CPSase activity, a result which suggests that B2 is also unable to interact with the GLN domain.
Oligomeric Structure of B2-Previous studies (27) suggested that while monomeric CPS.A or monomeric CPS.B could catalyze both ATP-dependent partial reactions, the formation of a homodimer, (CPS.A) 2 or (CPS.B) 2 , is required to catalyze the overall synthesis of carbamoyl phosphate. Both CPS.A and CPS.B were dimeric. The molecular mass of B2 was determined by gel filtration chromatography (Fig. 5B) on a calibrated Sephacryl S-300 high resolution fast protein liquid chromatography column. The subdomain eluted with an apparent molecular mass of 58 kDa, a value consistent with a dimer comprised of two copies of the 28.2-kDa B2 subdomain. Smaller amounts of a higher molecular weight species were also observed in the elution profile.

DISCUSSION
Many lines of evidence, cited above, suggest that ATP binds to the central region of each homologous half of the CPS domain or subunit. These binding sites are part of the active sites that catalyze the formation of either carboxy phosphate by CPS.A or carbamoyl phosphate by CPS.B. To test the idea that these regions correspond to separately folded catalytic subdomains, we have cloned and expressed the segments of the carB gene encoding regions A2 and B2. In addition, recombinant  FIG. 3. Expression of GLN A1-A2 and GLN B1-B2 proteins. The plasmids pHGGA12 and pHGGB12, which encoded the A1-A2 protein and B1-B2 proteins, respectively, were transformed into L673. Both plasmids also encoded the GLN subunit on a separate polypeptide. The cells were grown and cell extracts prepared as described under "Experimental Procedures." The extracts were centrifuged at 29,000 ϫ g for 20 min, and the whole extract (Ex), pellet (Pl), and supernatants (Su) were analyzed by SDS-gel electrophoresis. The major band in the extracts and supernatants can clearly be seen on the original gel as a doublet of the 42-kDa GLN subunit and the 40-kDa A1-A2 protein or the 44-kDa B1-B2 protein.
FIG. 4. ATP saturation curve of A1-A2 and B1-B2 proteins. The ATP saturation curves were obtained for 10.8 g of the A1-A2 protein (A), and 7.5 g of the B1-B2 protein (B) were measured using either 3 mM glutamine (q) or 87 mM ammonium chloride (E) as a nitrogen donor. The assay conditions, preparation of cell extracts, and determination of the protein concentration are described under "Experimental Procedures." The A1-A2 saturation curves were fit by least squares analysis to the Michaelis-Menten equation. The corresponding curves of B1-B2 did not conform to this expression because of substrate inhibition. The kinetic parameters are given in Table I. proteins, A1-A2 and B1-B2, have been made in which the amino end of the CPS.A or CPS.B domains, A1 or B1, respectively, remained fused to their respective catalytic subdomains. All four recombinants were expressed as soluble, stable, catalytically active proteins confirming that they are functional subdomains of the CPS domain.
The junction between the subdomains defined in the model (Fig. 1) is somewhat arbitrary. Although sequence homology and susceptibility to proteolytic cleavage provide a useful guide, our experience (56) with other CAD domains suggested that these criteria cannot solely be relied upon to establish the beginning and end of a functional domain. Consequently, in designing the constructs, the domains were defined conservatively in the sense that additional residues 4 were appended to the beginning and end to maximize the chances of obtaining a functional protein. However, having now shown that these regions of the molecule fold autonomously and have specific functions, we are in a position to systematically delete residues and precisely define the core subdomains.
It is interesting that the isolated catalytic subdomains, A2 and B2, are hyperactive. The large increase in k cat compared with A1-A2 (or B1-B2), CPS.A (or CPS.B), and the native enzyme strongly suggests that the active sites are contained entirely within the A2 and B2 subdomains. Association of A2 with A1 or B2 with B1 dramatically suppresses the activity of the catalytic subdomain. The modulation of catalytic activity, observed here, is reminiscent of the functional interactions that occur within the GLN domain, which also consists of distinct subdomains. The catalytic subdomain, corresponding to the carboxyl-half of the CAD GLN domain, and the interaction subdomain, the amino half, were separately cloned and expressed in E. coli (58). The isolated catalytic subdomain was extraordinarily active. The GLNase activity was much higher than exhibited by either the isolated GLN domain or native CAD, but it did not form a stable, functional complex with the CPS domain or subunit. Thus, the GLN interaction domain is 4 Most of the constructs encoded 4 -12 additional residues from the adjoining subdomain on either the amino or carboxyl end of the segment. The only exception was B1-B2, the first construct made, which is appreciably larger than defined in the model (Fig. 1). The junction between the CAD CPS.A and CPS.B was not cleaved (29) by controlled proteolysis. Moreover, the region of homology to carbamate kinase, identified (57) in the E. coli protein, extends into the A3 domain by 23 residues and is preceded by a sequence of 30 highly conserved residues. Consequently, the B1-B2 protein was designed with an additional 53 residues derived from A3 appended to the amino end of the polypeptide. Subsequent experiments with A1-A2, which conforms much more closely to the model (Fig. 1), suggest that these additional residues are not required.  a Calculated from the V max assuming a molecular mass of 120 kDa for E. coli CPSase synthetase subunit, and 40, 44, 31, and 28 kDa for A1-A2, B1-B2, A2 and B2, respectively. For these calculations it was assumed that two copies of the CPS subdomains interact to produce carbamoyl phosphate (see "Discussion"), one catalyzes the ATP dependent activation of bicarbonate and the other catalyzes the phosphorylation of carbamate.
b The saturation curves exhibited inhibition at high ATP concentrations and thus did not conform to the Michaelis-Menten equation. The values given for the K m correspond to the concentration of ATP at half of the maximum observed activity, and V max corresponds to the maximum observed activity.
c Not determined. required for complex formation and attenuates the intrinsically high activity of the catalytic subdomain. The suppression of the activity of the CPS catalytic subdomains suggests that A1 and B1 subdomains have a role that is analogous to the GLN interaction subdomain. Moreover, A1 and B1 are also clearly required for the functional interactions that occur between the GLN and CPS domains in the native enzyme. Both E. coli CPS.A and CPS.B domains form stable complexes with the GLN subunit that can be isolated by gel filtration chromatography. 5 Moreover, these complexes can catalyze the formation of carbamoyl phosphate from glutamine. Similarly, both A1-A2 and B1-B2 can also associate with the GLN subunit and use glutamine as a substrate for carbamoyl-phosphate synthetase. Although the formation of a physical complex between A2 or B2 and the GLN subunit has not been ruled out, the isolated A2 and B2 catalytic subdomains can only use ammonia and are thus not functionally coupled to the GLN subunit. This result is consistent with previous studies (28) that showed that deletions from the amino end of the E. coli CPS.A and CPS.B, respectively, abolishes physical association with the GLN subunit. Thus, a second function of A1 and B1 is to allow physical association and functional interactions with the GLN subunit.
One of the most interesting aspects of the enzyme is that carbamoyl phosphate synthesis involves parallel and consecutive reactions that occur on different domains within the complex. The coordination of these reactions requires that the participating domains not function at their full catalytic potential but rather are constrained into less active conformations. The controlled release of these constraints in response to the binding of substrates or intermediates to one of the other domains within the complex allows the rates of these partial reactions to be matched. In the GLN, CPS.A, and CPS.B domains the suppression of the activity of the catalytic subdomains is mediated by a distinct attenuation subdomain. As might be expected, these attenuation subdomains also participate in interdomain interactions. The attenuation subdomains are just one element of a coupling mechanism that is likely to require participation of all of the major functional domains to establish a fully functional linkage. This conclusion is based on the observation that, although A1-A2 and B1-B2 associate with the GLN domain, the activity of the GLN domain does not respond to the same extent to the binding of ATP and bicarbonate to the CPS catalytic subdomains. The GLN A1-A2 and the GLN B1-B2 complexes have higher GLNase activity in the absence of ligands and are only activated 2-fold by saturating ATP and bicarbonate. The functional linkage was previously found (27) to be impaired to the same extent in complexes of the GLN domain with CPS.A or CPS.B.
We have proposed (27) that, although the isolated CPS.A and CPS.B domains of CAD and E. coli CPSase can catalyze both partial reactions, the overall synthesis of carbamoyl phosphate requires the formation of a homodimer. In this model, one monomer assumes the role of bicarbonate activation and the other catalyzes the phosphorylation of carbamate to form carbamoyl phosphate. If this interpretation is correct, then one would predict that, since the A2 and B2 subdomains also catalyze the overall reaction, they would also be dimeric. This prediction was confirmed in the case of B2. Thus, the subunit interface responsible for forming a dimer of the major functional domains must be contained, in whole or in part, within the catalytic subdomains.
Our results are consistent with the elegant x-ray structure of E. coli carbamoyl-phosphate synthetase which was published (59) after this work had been completed. The authors identified four subdomains in each half of the synthetase subunit. Subdomain A corresponds to A1 or B1 in our nomenclature and subdomain D is equivalent to A3 or B3. A2 and B2 appear to be comprised of two subdomains, B and C, each of which participate in catalysis and bind ADP. The domain junctions that we have defined are close to those observed in the x-ray structure. It is interesting that the A1 and B1 subdomains (A) are rather extended and appear to be in contact with A2 and B2 (BC), 5 A. Bouvier, H. I. Guy, and D. R. Evans, unpublished observations. FIG. 6. Coupling of GLNase and synthetase activities. The GLNase activity of the GLN subunit was measured for 5.0 g of wild type E. coli CPSase (Ⅺ), 2.7 g of the A1-A2 protein (q), and 2.6 g of B1-B2 protein (E) in the presence of 10 mM glutamine, 15 mM sodium bicarbonate, and a variable concentration of ATP. MgCl 2 was present in a 2 mM excess relative to the ATP concentration. Other assay conditions are described under "Experimental Procedures." The data are expressed as percent of the values obtained for each protein in the absence of ATP (Table II) and fit, by least squares analysis, to Equation 5, where V o and V a represent the GLNase activity in the absence of ATP and in the presence of saturating ATP, respectively, and K a is the concentration of ATP at half-maximal activation. a The activities are expressed as mol/min/mg of the synthetase subunit and as the turnover number, s Ϫ1 , assuming (see Table I footnotes) that the subdomain complexes contain two copies of A1-A2 or B1-B2. Although the glutamine concentration was effectively saturating, the latter values are not true k cat values since glutamine saturation curves were not obtained.
b The -fold activation (V a /100) was calculated as described in the legend to Fig. 6.
respectively. Thus, it is easy to imagine a role of A1 and B1 in modulating the activity of their respective catalytic subdomains. Moreover, the B2 dimeric interactions reported here were observed in the x-ray structure. Carbamate kinase, an enzyme in the arginine degradation pathway of several prokaryotic organisms (57,60,61), catalyzes the reversible formation of carbamate and ATP from carbamoyl phosphate and ADP, the same reaction that is catalyzed by the CPS.B domain of CPSase. Sequence comparisons showed that carbamate kinase is homologous to a 40-kDa segment at the amino end of CPS.B, a region corresponding to the B1-B2 recombinant protein. The hypothesis was advanced (45,61) that an ancestral carbamate kinase is an evolutionary precursor of all CPSase molecules. In this scheme, CPSase originated by the fusion of a 20-kDa domain to the carboxyl end of the gene for carbamate kinase, followed by gene duplication and fusion. The finding that B1-B2 is a separately folded functional protein lends support to this proposal. Carbamate kinase also catalyzes bicarbonate-dependent ATP hydrolysis (62), a partial reaction that reflects bicarbonate activation in CPSase. This result is consistent with recent studies (27) which showed that CPS.A and CPS.B are functionally equivalent. The defining difference between the reactions catalyzed by carbamoylphosphate synthetase and carbamate kinase is that the former uses 2 mol of ATP in the synthesis of carbamoyl phosphate, whereas the latter catalyzes the phosphorylation of carbamate formed chemically in solution using a single ATP. In contrast to carbamate kinase, the A1-A2 and B1-B2 recombinant proteins constructed for this study are authentic CPSases in the sense that they can use either glutamine or ammonia as the nitrogendonating substrate and thus catalyze the same reaction as the much large parent molecule. Moreover, two hyperthermophilic archaebacterial CPSases (63,64) have recently been found to be much smaller than CPSases from other prokaryotic and eukaryotic species. Pyrococcus abyssi CPSase (63) and Pyrococcus furiosus CPSase (64) are composed of 38-kDa polypeptide chains. Both enzymes use ammonia and not glutamine as a nitrogen donor. These molecules are CPSases, not carbamate kinases, since two ATP molecules are required for carbamoyl phosphate synthesis. By analogy to the results obtained with the separately cloned CPS.A and CPS.B CPSase domains (27), catalysis of the overall reaction by the small archaebacterial enzymes would be expected to require two monomers working in concert. In this regard, the P. furiosus CPSase has been reported (64) to be dimeric. These enzymes are approximately the same size as A1-A2 and B1-B2 and may be their naturally occurring analogs. A comparative study of the CPSase domains, carbamate kinase, and the hyperthermophilic enzymes should be informative with regard to both the functional relationships between the domains and the evolutionary origin of these enzymes.
In summary, these results support the model of the structural organization of E. coli (28) and mammalian (15,29) CPSase. The CPSase subunit or domain is organized into two major domains, CPS.A and CPS.B, which are in turn comprised of smaller subdomains. The active sites are localized entirely within the 25-kDa central catalytic domains, A2 or B2. These catalytic subdomains have intrinsically high catalytic activity that is suppressed by interactions with an 11-kDa attenuation subdomain, A1 or B1. These interactions are probably a component of the functional linkage that coordinates the two ATPdependent partial reactions occurring on CPS.A and CPS.B. Subdomains A1 and B1 are also necessary for the physical association and functional linkage between the GLN and CPS domains. The 20-kDa A3 and B3 subdomains are not required for catalytic activity, but B3 is known to bind allosteric effectors and modulates the activity of the catalytic domains.