Pig heart CoA transferase exists as two oligomeric forms separated by a large kinetic barrier.

Pig heart CoA transferase (EC 2.8.3.5) has been shown previously to adopt a homodimeric structure, in which each subunit has a molecular weight of 52 197 and consists of N- and C-domains linked by a hydrophilic linker or "hinge". Here we identify and characterize a second oligomeric form constituent in purified enzyme preparations, albeit at low concentrations. Both species catalyze the transfer of CoA with similar values for k(cat) and K(M). This second form sediments more rapidly than the homodimer under the conditions of conventional sedimentation velocity and active enzyme centrifugation. Apparent molecular weight values determined by sedimentation equilibrium and gel filtration chromatography are 4-fold greater than the subunit molecular weight, confirming that this form is a homotetramer. The subunits of both oligomeric forms are indistinguishable with respect to molecular mass, far-UV CD, intrinsic tryptophan fluorescence, and equilibrium unfolding. Dissociation of the homotetramer to the homodimer occurs very slowly in benign solutions containing high salt concentrations (0.25-2.0 M KCl). The homotetramer is fully converted to homodimer during refolding from denaturant at low protein concentrations. Disruption of the hydrophilic linker between the N- and C-domains by mutagenesis or mild proteolysis causes a decrease in the relative amount of the larger conformer. The homotetramer is stabilized by interactions involving the helical hinge region, and a substantial kinetic barrier hinders interconversion of the two oligomeric species under nondenaturing conditions.

Phosphorylated and dephosphorylated structures of pig heart, GTP-specific succinyl-CoA synthetase.

Succinyl-CoA synthetase (SCS) catalyzes the reversible phosphorylation/dephosphorylation reaction:¿¿¿rm succinyl ¿hbox ¿-¿CoA+NDP+P_i¿leftrightarrow succinate+CoA+NTP¿¿where N denotes adenosine or guanosine. In the course of the reaction, an essential histidine residue is transiently phosphorylated. We have crystallized and solved the structure of the GTP-specific isoform of SCS from pig heart (EC 6.2.1.4) in both the dephosphorylated and phosphorylated forms. The structures were refined to 2.1 A resolution. In the dephosphorylated structure, the enzyme is stabilized via coordination of a phosphate ion by the active-site histidine residue and the two "power" helices, one contributed by each subunit of the alphabeta-dimer. Small changes in the conformations of residues at the amino terminus of the power helix contributed by the alpha-subunit allow the enzyme to accommodate either the covalently bound phosphoryl group or the free phosphate ion. Structural comparisons are made between the active sites in these two forms of the enzyme, both of which can occur along the catalytic path. Comparisons are also made with the structure of Escherichia coli SCS. The domain that has been shown to bind ADP in E. coli SCS is more open in the pig heart, GTP-specific SCS structure.

ADP-binding site of Escherichia coli succinyl-CoA synthetase revealed by x-ray crystallography.

Succinyl-CoA synthetase (SCS) catalyzes the following reversible reaction via a phosphorylated histidine intermediate (His 246alpha): succinyl-CoA + P(i) + NDP <--> succinate + CoA + NTP (N denotes adenosine or guanosine). To determine the structure of the enzyme with nucleotide bound, crystals of phosphorylated Escherichia coli SCS were soaked in successive experiments adopting progressive strategies. In the first experiment, 1 mM ADP (>15 x K(d)) was added; Mg(2+) ions were omitted to preclude the formation of an insoluble precipitate with the phosphate and ammonium ions. X-ray crystallography revealed that the enzyme was dephosphorylated, but the nucleotide did not remain bound to the enzyme (R(working) = 17.2%, R(free) = 22.8% for data to 2.9 A resolution). Catalysis requires Mg(2+) ions; hence, the "true" nucleotide substrate is probably an ADP-Mg(2+) complex. In the successful experiment, the phosphate buffer was exchanged with MOPS, the concentration of sulfate ions was lowered, and the concentrations of ADP and Mg(2+) ions were increased to 10.5 and 50 mM, respectively. X-ray diffraction data revealed an ADP-Mg(2+) complex bound in the ATP-grasp fold of the N-terminal domain of each beta-subunit (R(working) = 19.1%, R(free) = 24.7% for data to 3.3 A resolution). We describe the specific interactions of the nucleotide-Mg(2+) complex with SCS, compare these results with those for other proteins containing the ATP-grasp fold, and present a hypothetical model of the histidine-containing loop in the "down" position where it can interact with the nucleotide approximately 35 A from where His 246alpha is seen in both phosphorylated and dephosphorylated SCS.

Probing the nucleotide-binding site of Escherichia coli succinyl-CoA synthetase.

Succinyl-CoA synthetase (SCS) catalyzes the reversible interchange of purine nucleoside diphosphate, succinyl-CoA, and Pi with purine nucleoside triphosphate, succinate, and CoA via a phosphorylated histidine (H246alpha) intermediate. Two potential nucleotide-binding sites were predicted in the beta-subunit, and have been differentiated by photoaffinity labeling with 8-N3-ATP and by site-directed mutagenesis. It was demonstrated that 8-N3-ATP is a suitable analogue for probing the nucleotide-binding site of SCS. Two tryptic peptides from the N-terminal domain of the beta-subunit were labeled with 8-N3-ATP. These corresponded to residues 107-119beta and 121-146beta, two regions lying along one side of an ATP-grasp fold. A mutant protein with changes on the opposite side of the fold (G53betaV/R54betaE) was unable to be phosphorylated using ATP or GTP, but could be phosphorylated by succinyl-CoA and Pi. A mutant protein designed to probe nucleotide specificity (P20betaQ) had a Km(app) for GTP that was more than 5 times lower than that of wild-type SCS, whereas parameters for the other substrates remained unchanged. Mutations of residues in the C-terminal domain of the beta-subunit designed to distrupt one loop of the Rossmann fold (I322betaA, and R324betaN/D326betaA) had the greatest effect on the binding of succinate and CoA. They did not disrupt the phosphorylation of SCS with nucleotides. It was concluded that the nucleotide-binding site is located in the N-terminal domain of the beta-subunit. This implies that there are two active sites approximately 35 A apart, and that the H246alpha loop moves between them during catalysis.

A detailed structural description of Escherichia coli succinyl-CoA synthetase.

Succinyl-CoA synthetase (SCS) carries out the substrate-level phosphorylation of GDP or ADP in the citric acid cycle. A molecular model of the enzyme from Escherichia coli, crystallized in the presence of CoA, has been refined against data collected to 2.3 A resolution. The crystals are of space group P4322, having unit cell dimensions a=b=98.68 A, c=403.76 A and the data set includes the data measured from 23 crystals. E. coli SCS is an (alphabeta)2-tetramer; there are two copies of each subunit in the asymmetric unit of the crystals. The crystal packing leaves two choices for which pair of alphabeta-dimers form the physiologically relevant tetramer. The copies of the alphabeta-dimer are similar, each having one active site where the phosphorylated histidine residue and the thiol group of CoA are found. CoA is bound in an extended conformation to the nucleotide-binding motif in the N-terminal domain of the alpha-subunit. The phosphoryl group of the phosphorylated histidine residue is positioned at the amino termini of two alpha-helices, one from the C-terminal domain of the alpha-subunit and the other from the C-terminal domain of the beta-subunit. These two domains have similar topologies, despite only 14 % sequence identity. By analogy to other nucleotide-binding proteins, the binding site for the nucleotide may reside in the N-terminal domain of the beta-subunit. If this is so, the catalytic histidine residue would have to move about 35 A to react with the nucleotide.

A dimeric form of Escherichia coli succinyl-CoA synthetase produced by site-directed mutagenesis.

Succinyl-CoA synthetase (SCS) catalyzes the substrate-level phosphorylation step of the citric acid cycle. The enzyme from Escherichia coli is an (alphabeta)2-heterotetramer with two active sites, one in each alphabeta-dimer. To determine whether the two active sites could function independently, mutations were made to split the tetramer into alphabeta-dimers. Because two choices for the tetramer (I and II) were possible from the X-ray crystallographic analyses, mutations were made at two different interfaces. All mutations based on tetramer I resulted in an intact tetramer. Of the two mutants based on tetramer II, one was insoluble and the other, where M156beta, Y158beta, R161beta and E162beta were changed to D, D, E and R, respectively, was a dimer. This quaternary structure was confirmed by fast protein liquid chromatography, blue native PAGE and ultracentrifugation. The DDER mutant has kinetic parameters similar to the tetrameric E. coli enzyme. Like the tetrameric enzyme, it shows ATP-facilitated dethiophosphorylation, proving that this property is a single-site effect. The ATP-facilitated dethiophosphorylation is inhibited by phosphate. It is concluded that dimerization of alphabeta-dimers is not a prerequisite for catalytic competency nor for alternating sites cooperativity in the tetramer. The rationale behind the dimer-of-dimers in E. coli SCS is still not known, but increased solubility, increased stability and in vivo interactions of the tetramer with other proteins are still possibilities.

Mutually exclusive splicing generates two distinct isoforms of pig heart succinyl-CoA synthetase.

We have identified two distinct cDNAs encoding the alpha-subunit of pig heart succinyl-CoA synthetase. The derived amino acid sequence of one of these, PHalpha57, is highly similar to the alpha-subunit of the rat liver precursor enzyme. The second cDNA, PHalpha108, was identical throughout its sequence with PHalpha57 except for a stretch of 108 nucleotides which replaced a 57 nucleotide sequence in PHalpha57. Coexpression of either alpha-subunit cDNA with a common pig heart beta-subunit cDNA produced isozymes with GTP-specific enzyme activity. The enzyme produced by the combination of PHalpha57 and the beta-subunit cDNA resembled the "native" enzyme purified from pig heart tissue. In contrast, the expressed enzyme from the combination with PHalpha108 was clearly distinguishable from the native enzyme by, for example, hydroxyapatite chromatography. Moreover, it was now apparent that this isoform had been observed in previous preparations of the native enzyme, but always in very low amounts and, thus, disregarded. We have shown further that the two mRNA transcripts arise from a single gene and are generated by mutually exclusive splicing. The production of the PHalpha108 message involves the use of a non-canonical splice site pair, AT-AA. Finally, we provide evidence for tissue specific regulation in the splicing of the PHalpha108 message.

Productive interactions between the two domains of pig heart CoA transferase during folding and assembly.

The enzyme CoA transferase from porcine heart (EC 2.8.3.5) is a homodimer; each subunit consists of two domains linked by a hydrophilic "hinge" region. We have prepared separate DNA segments encoding each of these domains. Incorporation of these two DNA segments within an operon or within two separate transcription units does not preclude the synthesis and assembly of CoA transferase in Escherichia coli. When the two domain fragments are produced and purified individually from separate cultures and subsequently mixed, enzyme activity accumulates to near wild-type levels only after a lengthy incubation. Each domain is more susceptible to aggregation than wild-type CoA transferase. Circular dichroism shows that, prior to mixing, the domains possess a different secondary structural profile compared to their counterparts in the native enzyme. However, mixing and incubation of the domains produces a complex with far-UV CD, fluorescence, and ultracentrifugation properties similar to those of wild-type CoA transferase. Finally, we show that the intact hydrophilic peptide which links the two domains is essential for the recovery of activity observed upon refolding of the denatured enzyme in vitro. These results indicate that the folding and assembly of pig heart CoA transferase require a productive interaction between its two domains, involving a substantial conformational rearrangement.

Identification of glutamate 344 as the catalytic residue in the active site of pig heart CoA transferase.

The enzyme CoA transferase (succinyl-CoA:3-ketoacid coenzyme A transferase [3-oxoacid CoA transferase], EC 2.8.3.5) is essential for the metabolism of ketone bodies in the mammalian mitochondrion. It is known that its catalytic mechanism involves the transient thioesterification of an active-site glutamate residue by CoA. As a means of identifying this glutamate within the sequence, we have made use of a fortuitous autolytic fragmentation that occurs at the active site when the enzyme-CoA covalent intermediate is heated. The presence of protease inhibitors has no effect on the extent of cleavage detectable by SDS-PAGE, supporting the view that this fragmentation is indeed autolytic. This fragmentation can be carried out on intact CoA transferase, as well as on a proteolytically nicked but active form of the enzyme. Because the resulting C-terminal fragment is blocked at its N-terminus by a pyroglutamate moiety, it is not amenable to direct sequencing by the Edman degradation method. As an alternative, we have studied a peptide (peptide D) generated specifically by autolysis of the nicked enzyme and predicted to have an N-terminus corresponding to the site of proteolysis and a C-terminus determined by the site of autolysis. This peptide was purified by reversed-phase HPLC and subsequently characterized by electrospray mass spectrometry. We have obtained a mass value for peptide D, from which it can be deduced that glutamate 344, known to be conserved in all sequenced CoA transferases, is the catalytically active amino acid. This information should prove useful to future mutagenesis work aimed at better understanding the active-site structure and catalytic mechanism of CoA transferase.

The crystal structure of succinyl-CoA synthetase from Escherichia coli at 2.5-A resolution.

The x-ray crystal structure of succinyl-CoA synthetase (SCS) from Escherichia coli has been determined by the method of multiple isomorphous replacement to a resolution of 2.5 A. Crystals of SCS are tetragonal with a space group of P4(3)22 and unit cell dimensions of a = b = 98.47 A and c = 400.6 A. One molecule of SCS (142 kDa) is contained in the asymmetric unit. The current model has been refined to a conventional R factor of 21.6% with root mean square deviations from ideal stereochemistry of 0.022 A for bond lengths and 3.25 degrees for bond angles. The quaternary organization of the E. coli enzyme is an alpha 2 beta 2 heterotetramer. In this tetramer, the alpha-subunits interact only with the beta-subunits, whereas the beta-subunits interact to form the dimer of alpha beta-dimers. The two active site pockets are located at regions of contact between alpha- and beta-subunits. One molecule of coenzyme A is bound to each alpha-subunit at a typical nucleotide-binding motif, and His-246 of each alpha-subunit is phosphorylated. This phosphohistidine, a catalytic intermediate, is stabilized by two helix dipoles (the "power" helices), one from each of the two subunit types. A short segment of the beta-subunit from one alpha beta-dimer is in close proximity to the CoA-binding site of the other alpha beta-dimer, providing a possible rationale for the overall tetrameric structure.

Cloning, characterization, and expression of the beta subunit of pig heart succinyl-CoA synthetase.

The form of succinyl-CoA synthetase found in mammalian mitochondria is known to be an alpha beta dimer. Both GTP- and ATP-specific isozymes are present in various tissues. We have isolated essentially identical complementary DNA clones encoding the beta subunit of pig heart succinyl-CoA synthetase from both newborn and adult tissues. These cDNAs include a 1.4-kb sequence encoding the cytoplasmic precursor to the beta subunit comprised of 417 amino acid residues including a 22-residue mitochondrial targeting sequence. The cDNA encoding the 395-amino acid, 42,502-Da mature protein was confirmed to be the succinyl-CoA synthetase beta subunit by agreement with the N-terminal protein sequence and by high homology to prokaryotic forms of the beta subunit that were previously cloned (about 45% identical to beta from Escherichia coli). In contrast to a previous report (Nishimura, J.S., Ybarra, J., Mitchell, T., & Horowitz, P.M., 1988, Biochem. J. 250, 429-434), we found no tryptophan residue to be encoded in the sequence for the mature beta subunit, and this finding is corroborated by the fact that highly purified pig heart succinyl-CoA synthetase shows no tryptophan fluorescence or tryptophan content in amino acid compositional analysis. The cDNA clones encoding the mature pig heart beta subunit and its counterpart alpha subunit were coexpressed in a deletion mutant strain of E. coli. Recovery of succinyl-CoA synthetase activity demonstrated that this combination of subunits forms a productive enzymatic complex having GTP specificity.

Folding and assembly of the Escherichia coli succinyl-CoA synthetase heterotetramer without participation of molecular chaperones.

Succinyl-CoA synthetase of Escherichia coli (alpha 2B2 subunit structure) has been shown to fold and assemble without participation by molecular chaperones. Renaturation experiments showed that purified bacterial chaperone GroEL has no effect on the folding and assembly of the active tetrameric enzyme. When isolated 35S-labeled alpha or beta subunits were incubated with GroEL in the absence of ATP, there was no complex formation between the subunits and GroEL. These in vitro results were confirmed by in vivo analysis of the folding and assembly of newly synthesized succinyl-CoA synthetase subunits. When expression of the subunits was induced in E. coli strains that bear GroEL or GroES temperature-sensitive mutations, the assembly of active succinyl-CoA synthetase was not affected as the temperature was raised to 43 degrees C. These and other observations are discussed that indicate that folding and assembly of succinyl-CoA synthetase may be independent of assistance by any chaperone.

Sequence of a cDNA clone encoding pig heart mitochondrial CoA transferase.

We have isolated a full-length cDNA clone encoding the cytoplasmic precursor to pig heart mitochondrial CoA transferase (succinyl-CoA:3-ketoacid coenzyme A transferase (3-oxoacid CoA transferase, EC 2.8.3.5], a key enzyme for ketone body catabolism. The deduced amino acid sequence indicates the presence of a 39-residue mitochondrial signal sequence and a 481-residue mature protein of molecular weight 52,197. CoA transferase is known to be susceptible to proteolytic cleavage to produce a nicked but active enzyme. We have identified the site of proteolysis, and analysis of the sequence in its vicinity suggests that the polypeptide may fold into two domains connected by a highly hydrophilic bridge.

Bacterial expression of rat liver succinyl-CoA synthetase alpha-subunit. Factors that contribute to blocked translation of transcripts encoding a mitochondrial signal sequence.

This study comprises a detailed evaluation of factors that are necessary to achieve high levels of expression of eukaryotic proteins in bacterial systems. We attempted to express a rat liver cDNA clone encoding the precursor to the alpha-subunit of succinyl-CoA synthetase in an Escherichia coli expression system, without success. Removal of the region encoding the mitochondrial signal peptide (115 nucleotides) allowed efficient expression of the mature protein. This nucleotide sequence was shown to block expression at the level of translation. Two regions within this fragment were able to block the expression of other genes such as E. coli lacZ. Inhibition of expression was due to the close proximity of these inhibitory sequences with the translation initiation region (TIR). Insertion of a spacer between the inhibitory sequence and the TIR relieved the block in translation. Analysis of the 115-nucleotide fragment identified sequences capable of extensive base-pairing with the Shine-Dalgarno and surrounding sequences. Such secondary structures are capable of blocking the formation of competent translation initiation complexes.

Functional consequences of substitution of the active site (phospho)histidine residue of Escherichia coli succinyl-CoA synthetase.

Succinyl-CoA synthetase (EC 6.2.1.5, succinate:CoA ligase (ADP-forming] of Escherichia coli is an alpha 2 beta 2 tetramer, with the active site believed to be located at the point of contact between the two subunit types. It has been previously established that the reaction involves the intermediate participation of a phosphorylated enzyme form in the process of catalysis. The site of phosphorylation (His-246) and the binding sites for the substrates ADP and ATP are located in the alpha subunit, and the succinate and CoA binding sites are in beta. A mutant form of this enzyme, with the active site histidine residue replaced by aspartate, has been produced in large quantities and purified to homogeneity. This form appears to be indistinguishable from the native enzyme with respect to its subunit assembly, but has no ability to catalyze the overall reaction. As expected, the His-246 alpha----Asp mutant is incapable of undergoing phosphorylation. We have developed an assay based upon the arsenolysis of succinyl-CoA that effectively isolates the partial reaction that occurs in the portion of the active site contributed by the beta subunit; this reaction does not involve covalent participation of His-246 alpha. We have found that the His-246 alpha----Asp mutant is also devoid of activity in this arsenolysis reaction, indicating that an intact His-246 alpha is required for the establishment of the microenvironment in this portion of the active site that is required for the corresponding step of the overall reaction.

Mitochondrial translocation and processing of the precursor to the alpha-subunit of rat liver succinyl-CoA synthetase.

Succinyl-CoA synthetase functions in the mitochondrial matrix as an alpha beta-dimer. Its constitutive subunits are thus expected to be encoded in the nucleus and synthesized in the cytoplasm as precursors containing signal sequences for mitochondrial translocation. We have previously reported the isolation and sequence of a rat liver cDNA clone (lambda SCS19) that apparently encodes the cytoplasmic precursor to the alpha-subunit. Here we report the preparation of mRNA transcripts of this cDNA insert and their in vitro translation to produce labeled protein that can be translocated across the membranes of subsequently added rat liver mitochondria. Translocation is accompanied by proteolytic processing to convert the 34.5-kilodalton precursor to the 32-kilodalton mature form of the subunit. The N-terminal sequence of the mature alpha-subunit from the GTP-specific isozyme has been determined by sequential Edman degradation and compared with the amino acid sequence deduced from the cDNA. This confirms that the cloned sequence encodes the GTP-specific alpha-subunit, and establishes that the point of cleavage is between histidyl and glycyl residues and that the signal sequence consists of 27 residues. The signal sequence shares characteristics of other mitochondrial targeting sequences that have been elucidated (largely of yeast mitochondrial precursors), including the potential to form an amphiphilic helix. Import is dependent upon the presence of ATP and is inhibited by compounds that diminish mitochondrial membrane potential. Translocation of the precursor is effective for precursor produced by the reticulocyte translation system, but is not seen for the product that is translated by a wheat germ extract, indicating that the latter may lack a factor or component that is necessary for the targeting and import process.

Cloning and sequencing of the cytoplasmic precursor to the alpha subunit of rat liver mitochondrial succinyl-CoA synthetase.

Succinyl-CoA synthetase [succinate-CoA ligase (GDP-forming); EC 6.2.1.4] of rat liver, an alpha beta dimer, is a component of the enzymology of the tricarboxylic acid cycle and functions within the mitochondrial matrix. We have isolated and determined the sequence of a cDNA clone containing the coding sequence of the cytoplasmic precursor to the alpha subunit of this enzyme together with stretches of nontranslated sequence at the 5' and 3' ends. The translated amino acid sequence indicates the presence of a 27-residue N-terminal signal sequence for mitochondrial targeting. The amino acid sequence of the mature alpha subunit shows an extraordinary degree of homology to the alpha subunit of Escherichia coli succinyl-CoA synthetase, with greater than 70% of the residues identical. This suggests that the fundamental differences in the quaternary structures and catalytic functions of the mammalian and bacterial enzymes must be attributable to differences in the beta subunits. mRNA that hybridizes to the cloned DNA is approximately equal to 1800 nucleotide residues in length, confirming that each of the two subunits is encoded separately and does not arise by proteolysis of a primary gene product containing both subunits of the mature protein.

Phosphorylation by cyclic AMP-dependent protein kinase does not affect the association of ATP citrate-lyase with isolated mitochondria.

ATP citrate-lyase is known to be a substrate for various protein kinases, but the functional role, if any, of kinase-directed phosphorylation of this enzyme has not been identified. Recently, Strålfors [(1987) J. Biol. Chem. 262, 11486-11489] has suggested that effects on the association of this enzyme with mitochondria may account for the observed ability of isoproteronol or insulin to promote immobilization of ATP citrate-lyase in permeabilized cells. Here we report studies involving phosphorylation of the pure enzyme in vitro using cyclic AMP-dependent protein kinase. We show that phosphorylation has no significant effect on the fraction of the enzyme that may be bound to isolated mitochondria.

Positional oxygen isotope exchange as a probe for the mechanism of catalysis by Escherichia coli succinyl coenzyme A synthetase.

Succinyl-CoA synthetase of Escherichia coli has an alpha 2 beta 2 subunit structure. The enzyme shows strict half-sites reactivity with respect to the phosphorylation of a histidine residue in the alpha subunit that represents a step in catalysis. Several lines of evidence indicate that this behavior may result from cooperative interactions between alternatingly functional active sites, so that subsequent steps in catalysis at one site may be promoted by phosphoryl transfer to the site on the neighboring half of the molecule. This study is directed toward learning more about the nature of these cooperative interactions. Here we have used positional isotope exchange (i.e., exchange of 18O between the beta, gamma bridge and the beta nonbridge position of ATP) as a test for transient bisphosphorylation. Succinyl-CoA synthetase was ATP) as a test for transient bisphosphorylation. Succinyl-CoA synthetase was prepared in which one of the two active sites was thiophosphorylated; this species thus has one of its two active-site histidine residues occupied and unavailable for further reaction with ATP. Treatment of this monothiophosphorylated enzyme with [beta, gamma-18O]ATP resulted in no significant scrambling of isotope into the nonbridge position, clearly indicating that the enzyme does not undergo even transient bisphosphorylation. We interpret the results in terms of a model of catalysis in which phosphoryl transfer to the second site occurs in concerted fashion with transfer from the first.

Studies of the process of renaturation and assembly of Escherichia coli succinyl-CoA synthetase from its alpha and beta subunits.

Succinyl-CoA synthetase catalyzes the substrate-level phosphorylation step of the tricarboxylic acid cycle. The enzyme, as isolated from Escherichia coli, has an alpha 2 beta 2 subunit structure. It is known that substrate-binding sites are distributed between both subunit types and that the active enzyme is the nondissociating tetramer. This paper describes a study of the process of assembly of the enzyme from its denatured constituent subunits. Starting with equimolar mixtures of the subunits that are prepared in denaturing conditions (6 M urea, 5% acetic acid), rapid renaturation to produce virtually a fully active enzyme occurs after neutralization and dilution under suitable conditions. This process occurs most efficiently in the presence of either ATP or Pi, indicating that occupation of the phosphoryl-binding site on the refolding alpha subunit facilitates productive intrasubunit interactions. We have determined conditions of protein concentration, pH, temperature, final urea concentration, and buffer compositions that optimize both the rate and extent of production of active enzyme. The final refolded product is indistinguishable from the native species with respect to its specific catalytic activity, size, and other physical properties. To probe further the mechanism and route of renaturation, we have shown that the rate of appearance of activity has first-order dependence on each of the two subunits. The step that determines the rate of assembly is thus bimolecular, such as the association of structural monomers to form a dimeric transient species. The highly specific mutual interactions between the refolding transient species of subunits must be essential for the correct assembly of this enzyme from the two gene products in vivo.