(Received for publication, February 11, 1997, and in revised form, May 22, 1997)
From the Department of Biochemistry and Molecular Biology, Wayne State University School of Medicine, Detroit, Michigan 48201
The amidotransferase or glutaminase domain (GLN domain) of mammalian carbamyl-phosphate synthetase II (CPSase II) catalyzes glutamine hydrolysis and transfers ammonia to the synthetase domain (CPS domain), where carbamyl phosphate formation is catalyzed in three consecutive reactions. The GLN and CPS domains are part of a single polypeptide and are connected via a 29-amino acid chain segment (GC linker). In contrast, the two comparable domains of Escherichia coli CPSase are not fused, but are separate, noncovalently associated subunits. To establish the function of the GC linker in mammalian CPSase, it was deleted, and the two domains were directly fused. The deletion mutant not only catalyzed glutamine-dependent carbamyl phosphate synthesis, but was activated 10-fold relative to its wild-type counterpart. However, ammonia-dependent synthesis of carbamyl phosphate was abolished, indicating that ammonia no longer had access to the active site on the CPS domain. The mutant was still sensitive to inhibition by the allosteric effector UTP, but was no longer activated by the allosteric effector phosphoribosyl pyrophosphate, although evidence indicated that the latter could bind to the enzyme. The linker appears to serve as a spacer that allows the complex to cycle between two conformations, an open low activity form in which the ammonia site on the CPS domain is accessible and an activated conformation in which the ammonia generated in situ from glutamine is directly channeled to the CPS active site and access to exogenous ammonia is blocked.
Carbamyl-phosphate synthetase (CPSase1; EC 6.3.5.5), like many biosynthetic enzymes that use glutamine as a nitrogen donor (1, 2), consists of an amidotransferase or glutaminase domain (GLN domain) and a synthetase domain (CPS domain) (3). Glutamine is hydrolyzed on the GLN domain, and the ammonia is transferred to the CPS domain, where it is used along with bicarbonate and ATP to make carbamyl phosphate in a series of three consecutive partial reactions. This mechanism, which is thought to be general for all members of this family of enzymes, was established for Escherichia coli CPSase by measuring the partial reactions, trapping intermediates, and positional isotope exchange (4-14). Mitochondrial CPSase I, which directly uses ammonia rather than glutamine (15, 16), is the only known exception. The pH dependence of the biosynthetic reaction indicates that ammonia, not ammonium ion, is the substrate, and thus, the ammonia generated in situ by glutamine hydrolysis must be sequestered within the complex to prevent protonation (3).
The GLN and CPS activities can be associated with separate subunits as in E. coli CPSase (17) or can be carried on a single polypeptide chain, sometimes in association with other enzymatic activities, as in the mammalian multifunctional protein CAD (18-20). E. coli CPSase consists (17, 21) of a 42-kDa subunit that hydrolyzes glutamine and a 118-kDa subunit that catalyzes ammonia-dependent carbamyl phosphate synthesis and binds the allosteric effectors. The enzyme is allosterically regulated by UMP, IMP, and ornithine. Mammalian pyrimidine-specific CPSase II is part of a multifunctional protein, CAD (18-20), which also catalyzes the second and third steps in the de novo pyrimidine biosynthetic pathway. The CPSase component (22, 23, 60) of CAD has a 40-kDa GLN domain fused via a 29-residue linker to the amino-terminal end of a 120-kDa CPS domain. Mammalian CPSase is allosterically (24-28) inhibited by UTP and activated by PRPP. Despite the differences in the structural organization and mode of regulation, the bacterial and mammalian CPSases have a high degree of sequence similarity (22, 23).
Kern et al. (29) made the interesting observation that fusion of the E. coli GLN and CPS subunits via the CAD GC linker did not alter the ammonia- or glutamine-dependent activity of the bacterial enzyme. The authors proposed that the linked domains found in CAD arose by the fusion of genes encoding ancestral monofunctional subunits analogous to those of E. coli CPSase.
To further elucidate the function of the interdomain GC linker, the GLN and CPS domains of the mammalian protein were fused directly in frame without the linker. The unexpected result was that glutamine-dependent carbamyl phosphate synthesis was activated 10-fold, whereas the ammonia-dependent activity was abolished or severely curtailed. We conclude that the linker allows the complex to cycle between inactive and active conformations and that deletion of the linker essentially traps the molecule in an activated conformation.
The plasmid pHL1 (30) encodes the CAD GLN and CPS domains in a vector derived from pEK81 (31). The 9.1-kilobase plasmid pCAD142 (32) was a gift of George Stark (Cleveland Clinic, Cleveland, OH), and Carol Lusty (The Public Health Research Institute of the City of New York, New York) kindly provided RC50 and L673 cells (33), which are each defective in both the carA and carB genes encoding the E. coli CPSase subunits. Strain L673 also lacks the lon protease. E. coli strain EK1104 (31) was a gift of Evan Kantrowitz.
Construction of the Recombinant PlasmidsThe 87-base pair
sequence encoding the GC linker (Fig. 1) was deleted
from pHL1 (30), a plasmid that encodes the mammalian CPS and GLN
domains. The linker-coding sequence, which extends from the
SmaI site (residue 1095) to residue 1182, was deleted (Fig.
2) by cleaving pHL1 with SmaI and
AvrII (at base pair 1194). Following cleavage, the large
fragment was isolated and religated after removal of the 5-extensions
with the Klenow fragment of DNA polymerase. The resulting construct,
pHL1
GClnk, lacks the coding sequence for the 29-residue linker and 4 residues from the amino-terminal end of the CPS
domain.2 pCK
GClnk, a plasmid that
encodes the entire CAD polypeptide but lacks the GC linker, was made
using the approach used to construct the full-length CAD clone (34).
The fragment containing the aspartate transcarbamylase- and
dihydroorotase-coding sequences, obtained by cleaving pCAD142 with
XbaI, was ligated to pHL1
GClnk, which had been linearized
with the same enzyme. Another construct, pHG-GLNGClnk, which encodes
the mammalian GLN domain with the linker attached, was made by deleting
the CPS domain (Fig. 2) from pHL1 by cleavage with AvrII and
EcoRI. The ends of the large fragment were made flush by
treatment with the Klenow fragment and mung bean nuclease prior to
ligation. The fidelity of these constructs was verified by restriction
mapping. Restriction digests, ligations, and other recombinant DNA
methods were carried out using standard protocols (35).
Expression of the Recombinant Proteins
The recombinant
plasmids were transformed (36) into E. coli strains L673 and
RC50, uridine auxotrophs that lack the E. coli carA and
carB genes encoding the CPSase subunits. Cells transformed with pHG-GLNGClnk were unable to grow on minimal medium (see Table I).
In contrast, transformation with pHL1, which encodes the entire
mammalian GLN-CPS protein, fully complemented the E. coli defect, whereas the pHL1GClnk and pCK
GClnk transformants grew more vigorously than cells expressing the wild-type proteins. Western blotting and measurements of the CPSase activity in
extracts of the transformants showed that the level of expression of
all of the recombinants was low. The proteins were partially purified by growing 100-ml cultures (37) of the transformed L673 cells for
20 h. The cells were harvested; resuspended to a final volume of 1 ml in 50 mM Tris-HCl, pH 7.4, 1 mM
dithiothreitol, and 5% glycerol; and broken by sonication. The
extracts were then centrifuged at high speed (16,000 × g) for 20 min, and the supernatant fraction was passed over
a Sephadex G-50 column equilibrated in the same buffer. The
concentration of the recombinant proteins was determined by analysis of
0.05-ml aliquots by enzyme-linked immunosorbent assay.
|
Native CAD was isolated (20, 38) from an overproducing strain of Syrian hamster cells, BHK 165-23. Protein concentrations were determined by the Bradford dye binding method (39). The concentration of CAD and CAD constructs was determined by enzyme-linked immunosorbent assay (40) using CAD-specific antiserum. CAD concentrations determined by these two method were comparable. SDS-gel electrophoresis was carried out on 10% polyacrylamide gels using the procedure of Laemmli (41), and the proteins were either stained or electroblotted onto nitrocellulose for immunoblotting (40) using CAD antibodies. The CPSase activity of the partially purified proteins was assayed at 37 °C by a radiometric procedure (42, 43). Saturation curves were obtained using ATP, sodium bicarbonate, glutamine, and ammonium chloride as the variable substrate. Assays were carried out in the sonication buffer supplemented with 25 mM MgCl2. When fixed, the concentrations of ATP, bicarbonate, and glutamine were held at 20, 5, and 3.5 mM, respectively. Nonlinear least-squares fitting of the saturation curves to the Michaelis-Menten equation was carried out using the program Scientist (MicroMath Scientific Software). Sequence analysis was performed using the Wisconsin Package; version 9.0, Genetics Computer Group (GCG), Madison, WI.
The
contiguous GLN and CPS domains of CAD (Fig. 1) span residues 1-1455 of
the polypeptide. The CPS domain (residues 395-1455) is connected to
the GLN domain (residues 1-365) by a 29-residue chain segment (GC
linker) defined as the region of the polypeptide that is homologous to
neither the GLN or CPS subunits of the monofunctional proteins. Two
constructs were made as described under "Experimental Procedures."
Plasmid pHL1GClnk encodes mammalian GLN-CPS lacking the GC linker.
In this protein, GLN-CPS
GClnk, the GLN domain is directly fused to
the CPS domain. Plasmid pCK
GClnk encodes CAD
GClnk, which is the
entire CAD polypeptide with the GC linker deleted.
Transformation of the recombinant plasmids into E. coli strains lacking CPSase was able to complement the defect (Table I), indicating that functional mammalian proteins were expressed. However, the level of expression was low, and attempts to purify the full-length mammalian proteins were unsuccessful. All assays were carried out using extracts prepared as described under "Experimental Procedures," and the concentration of the mammalian proteins was determined by enzyme-linked immunosorbent assay.
Catalytic Activity of the Recombinant ProteinsAssays of the
glutamine- and ammonia-dependent CPSase activities of the
recombinant proteins showed that the CPS component of CAD (GLN-CPS)
catalyzed carbamyl phosphate synthesis with both nitrogen donors at a
rate comparable to that observed with full-length CAD. However, removal
of the GC linker had two unanticipated consequences in that 1) the rate
of glutamine-dependent CPSase activity increased 10-fold,
and 2) ammonia could no longer serve as a nitrogen donor. Very similar
results were obtained with the full-length CAD deletion mutant.
Enhanced stability of GLN-CPSGClnk relative to GLN-CPS as an
explanation for the higher activity of the species lacking the GC
linker was ruled out by measuring the activity of these two proteins
during a 4-h incubation under the assay conditions.
The kinetic parameters (Table II) of CAD and of the separately expressed CPSase component (GLN-CPS) were similar. Deletion of the GC linker from GLN-CPS did not appreciably alter the Km for glutamine and bicarbonate. While the ATP saturation curve did not conform to Michaelis-Menten kinetics because of apparent substrate inhibition (Fig. 3), the concentration of ATP that produced half-maximal activity was also close to the value obtained for the proteins that had the GC linker. However, the Vmax of the deletion mutant was appreciably higher than that of the wild-type proteins. The activation was especially apparent at low ATP concentrations, where, for example, at 1.5 mM ATP, the activity was 48-fold higher than that observed with intact CPSase. At saturating ATP, the activity of the deletion mutant was still 16-fold higher despite significant substrate inhibition. Similarly, saturation curves using glutamine or bicarbonate as the variable substrate and saturating ATP gave Vmax values that were 10.0- and 11.2-fold higher, respectively, than the CPSase molecule that had the GC linker. Moreover, saturation curves obtained with ammonium chloride as a variable substrate showed that the ammonia-dependent CPSase activity of the protein lacking the GC linker was effectively abolished even at 100 mM, the highest concentration of ammonium chloride tested. Thus, removal of the GC linker yielded a functional CPSase that could efficiently utilize ammonia produced endogenously by the GLN domain, but not exogenously supplied ammonia.
|
Allosteric Regulation
UTP inhibited the activity of the
GLN-CPSGClnk deletion mutant to the same extent (Fig.
4) as the CAD or GLN-CPS activity. However, the deletion
mutant was no longer sensitive to the allosteric activator PRPP under
conditions in which the intact proteins are activated 170%. Although
we have shown (44) that the allosteric sites that bind UTP and PRPP are
localized at the extreme carboxyl-terminal end of the CPS domain (B3),
it could be argued that the GC linker may somehow be essential for PRPP
binding. To test this possibility, the effect of PRPP on the
glutamine-dependent CPSase activity of the mutant was
measured (Fig. 5) in the presence of 8 mM
UTP. Under these conditions, the response of the deletion mutant to PRPP was similar to that observed for CAD and intact CPSase. All three
proteins were inhibited at low PRPP concentrations, but were activated
up to 200% as the concentration of PRPP was increased. This result
clearly showed that PRPP binds to the deletion mutant.
Direct Effect of the GC Linker on the Synthetase Activity
To determine whether specific residues or a sequence of residues within the linker interferes directly with the biosynthetic reaction, the mammalian GLN domain was recloned (see "Experimental Procedures") with the GC linker attached. GLN-GClnk was expressed at moderate levels (1-2% of the total soluble protein) and could be readily isolated in nearly homogeneous form in a single step by chromatography on a DEAE-Sephacyl column. We have found (37) that the mammalian GLN domain forms a fully functional 1:1 stoichiometric hybrid complex with the large subunit of E. coli CPSase. If the mammalian GC linker interferes with the synthesis of carbamyl phosphate, then a hybrid consisting of the mammalian GLN domain with the linker attached and the E. coli CPS subunit might be expected to have an impaired function. The glutamine saturation curve (not shown) of a stoichiometric mixture of purified GLN-GClnk and the purified large subunit of E. coli CPSase showed that the glutamine-dependent CPSase activity had been restored, and therefore, a functional hybrid complex was formed. The Km was essentially the same, but the Vmax was reduced by 3-fold compared with the hybrid formed without the linker.
Carbamyl phosphate synthesis involves a fascinating series of parallel and consecutive reactions (7-9,12-14) that occur on different domains and subdomains (21, 45, 46) and require precise coordination. One of the most interesting aspects of the mechanism is that two of the intermediates initially formed are very labile (3) and thus must be sequestered within the complex. Carboxy phosphate has a half-life of only a few seconds in aqueous solution, whereas ammonia, if released from the complex, would be rapidly protonated. On the other hand, ammonia can serve as a nitrogen donor in the absence of glutamine, suggesting that the ammonia-binding site must be accessible to the aqueous milieu at some point during the catalytic cycle. The ammonia-binding site is known to be located on the synthetase domain or subunit since the isolated E. coli CPS subunit can catalyze the ammonia-dependent reaction (21, 45).
Further evidence of ammonia sequestration is provided by measurements of the stoichiometry of the overall reaction, which showed that 1 mol of glutamine is hydrolyzed per mol of carbamyl phosphate synthesized, indicating that the ammonia generated in situ does not escape from the complex. The effect of ammonia on the glutaminase activity of CAD (47) is interesting in this regard. In the absence of ATP and bicarbonate, ammonia can completely inhibit glutamine hydrolysis, in accordance with the principle of microscopic reversibility. Thus, ammonia has access to the active site on the GLN domain, either directly in the same way as glutamine or indirectly via the NH3-binding site on the CPS domain. The same measurements made in the presence of saturating bicarbonate and ATP suggest that ammonia is entering the complex by way of its binding site on the CPS domain. In the presence of these substrates, ammonium chloride no longer completely inhibits the CAD glutaminase activity, suggesting that access to the ammonia-binding site on the CPS domain is restricted.
The implication of these results is that CPSase exists in two conformational states which differ in that the ammonia site is either sequestered within the complex or accessible to the solvent. These conformational changes may well involve changes in the juxtaposition of the GLN and CPS domains as well as intradomain changes in tertiary structure. If the GLN and CPS domains are separate subunits, as is the case with the E. coli enzyme, there would be no constraints on the relative movement of the domains. However, if the GLN and CPS domains are part of the same polypeptide chain, as occurs in CAD, then one would expect that the domains would be connected by a linker that was long enough and flexible enough to accommodate any changes in quaternary structure that may occur.
The linker between the GLN and CPS domains in CAD is 29 amino acids
long (23, 60). An analysis of the sequence of other fused CPSases (Fig.
6) showed that all of these complexes have a GC linker
ranging in length from 24 residues in the Dictyostelium protein to 29 residues in CAD, Drosophila, and yeast. Even
mammalian CPSase I, which has a nonfunctional GLN domain, has a
26-residue linker. It is notable that there is little or no sequence
similarity between the GC linkers even among proteins from closely
related species, although all are uniformly flexible throughout their entire length. Secondary structure predictions revealed no consistent patterns. The CAD GC linker has a highly hydrophobic segment (Fig. 6),
residues 15-23, that is flanked by hydrophilic regions. Despite the
presence of these hydrophilic segments and potential target sites for
trypsin, the GC linker is very resistant to proteolytic cleavage (48)
and thus may be buried at the interface between the domains. While the
hydropathic profiles of the Drosophila and CPSase I GC
linkers are somewhat similar, this feature is not universal since the
Dictyostelium linker is largely hydrophilic, whereas the
corresponding segment of the yeast protein is hydrophobic. The similar
length and lack of sequence conservation are consistent with the
proposed role of the GC linker as a spacer.
We found here that deletion of the linker and direct fusion of the GLN and CPS domains dramatically alter the catalytic and regulatory properties of the mammalian enzyme. Contrary to our expectations when these experiments were planned, neither the channeling of ammonia nor the ability to utilize glutamine as a nitrogen donor was disrupted. Rather, glutamine-dependent CPSase was activated, and the entry of exogenous ammonia to the binding site on the CPS domain was completely blocked. Activation involves a 10-16-fold increase in kcat, with no apparent effect on the Km for any of the substrates.
Despite the 3-fold reduction in Vmax of the CPS-GLN-GClnk hybrid, we do not believe that direct interactions between the linker and the synthetase domain that suppress catalytic activity account for the activation that occurs upon deletion of the linker. First, the effect is appreciably smaller than the observed 10-16-fold activation. Second, the hybrid does not faithfully mimic the native multidomain protein since the linker is only tethered at one end. Third, a multidomain protein created (29) using the mammalian GC linker to connect the E. coli GLN and CPS subunits had normal catalytic activity, suggesting that no unfavorable interactions occur. Finally, direct fusion of the GLN and CPS subunits of E. coli CPSase (49), which naturally lacks the GC linker, results in a comparable activation.
These results can be explained by a model (Fig. 7) that
assumes that the complex exists in two different conformations during the catalytic cycle. One conformation is an open, substrate-binding state in which the ammonia site on the CPS domain is accessible, but
the configuration of the active-site residues is not optimal. In the
second closed, catalytically active conformational state, access to the
ammonia-binding site is restricted, but the active-site residues have
assumed the proper juxtaposition to promote efficient catalysis. This
interpretation is consistent with previous studies of E. coli CPSase. Chemical modification and binding studies of the
bacterial enzyme (50, 51) revealed the existence of several different
conformational states and led to the formulation of a model in which
the binding of substrates to one conformational state promotes the
transition to a catalytically active form of the enzyme. Moreover, the
binding of ATP and bicarbonate to both the E. coli (5) and
mammalian (47) CPSases appreciably enhances the glutaminase activity.
Conversely, modification of the active site on the GLN subunit of
E. coli CPSase (51-53) stimulates the activity of the CPS
subunit. These observations may reflect the transition to the
catalytically active conformation upon substrate binding.
The conformational change is postulated to occur following the binding of ATP and bicarbonate to the CPS domain and the binding of either glutamine to the GLN domain or NH3 to the CPS domain. Once the enzyme has assumed the catalytically active conformation, the ammonia produced in situ from glutamine is efficiently channeled to the CPS domain and cannot escape from the complex since the ammonia-binding site on the CPS domain is no longer accessible to the solvent. Similarly, in the case of the ammonia-dependent reaction, the ammonia is sequestered within the complex in the catalytically active conformation.
According to this model, deletion of the GC linker introduces constraints that in effect trap the molecule in the catalytic conformation. Access to exogenous ammonia is restricted, but glutamine can still bind since its site on the GLN domain is distinct from the ammonia-binding site on the CPS domain. ATP and bicarbonate can also bind to both conformations. Pre-steady-state kinetic and isotope exchange measurements (13) have shown that the activation of bicarbonate is the rate-limiting step in the overall synthesis of carbamyl phosphate. It is possible that the transition from the open inactive conformation to the catalytically active conformation, an essential element of this partial reaction in the proposed model, is the slowest step. In that event, bypassing this conformational transition could account for the observed activation of the deletion mutant. An equilibrium between an active and inactive conformation could provide an alternative, but perhaps less likely, explanation of these results. In this mechanism, the activity of the complex reflects the population of these two states, and substrate binding stabilizes the catalytically active conformation.
Deletion of the GC linker also alters the allosteric regulation of the enzyme. Whereas UTP inhibits the deletion mutant normally, PRPP activation is abolished. PRPP binds to a regulatory subdomain at the extreme carboxyl-terminal end of the CPS domain of CAD (44), but deletion of the linker could introduce structural changes that interfere with PRPP binding. For example, there is some evidence (54) that the inactive GLN domain of mitochondrial CPSase I may be close to the binding site of the allosteric activator N-acetyl glutamate. However, the observation that the same concentration of PRPP reverses UTP inhibition to a comparable extent in the deletion mutant and wild-type enzyme suggests that PRPP binds with nearly the same affinity to the deletion mutant. Although PRPP cannot further stimulate the CPSase activity of the protein activated by deletion of the GC linker, the conformations of these two activated forms of the enzyme are probably distinct. PRPP activates by decreasing the Km for ATP, but has no effect on kcat. In contrast, the kcat of the deletion mutant is appreciably higher, but ATP binding is unaltered. Moreover, the inhibited conformation of the enzyme induced by UTP is still accessible in the deletion mutant, and PRPP can reverse this allosteric transition.
In summary, we conclude that the GC linker in CAD and other multifunctional biosynthetic complexes allows the GLN and CPS domains to assume two distinctly different conformational states that are an essential part of the catalytic cycle. The accompanying report (49), describing studies of an E. coli CPSase fusion protein, suggests a possible functional advantage of these conformational changes.
We thank Drs. Carol Lusty and Evan Kantrowitz for the generous gifts of plasmids and strains.