The mechanism of velocity modulated allosteric regulation in D-3-phosphoglycerate dehydrogenase. Cross-linking adjacent regulatory domains with engineered disulfides mimics effector binding.

D-3-Phosphoglycerate dehydrogenase (PGDH) (EC 1.1.1.95) from Escherichia coli is an allosterically regulated enzyme of the Vmax type. It is a tetramer of identical subunits and each subunit is made up of three identifiable domains, the cofactor binding domain, the substrate binding domain, and the regulatory domain. Each subunit contacts two other subunits through adjacent cofactor binding domains and through adjacent regulatory domains. L-Serine, the physiological effector, inhibits catalytic activity by apparently tethering regulatory domains from adjacent subunits together through the formation of hydrogen bonds to each subunit. This investigation demonstrates that cross-linking adjacent regulatory domains with engineered disulfides produces catalytic inhibition in the absence of inhibitor in a manner similar to that produced by the inhibitor. The inhibition due to cross-linking can be completely reversed in a concentration dependent manner by dithiothreitol. The active mutant enzyme, containing the engineered cysteines in the reduced state, retains its ability to be inhibited by L-serine, although at a 100-fold higher concentration. Hill plots of the serine inhibition of mutant and native enzyme indicate that the number of interacting sites remains at 2 in the mutant enzyme. The reversible inhibition of enzyme activity that results from tethering adjacent regulatory domains with engineered disulfides suggests that these domains move in some manner relative to one another during the active to inhibited state transition. These observations support the model which predicts that catalytic activity is regulated by the movement of rigid domains about flexible hinges and that effector binding prevents this by locking the regulatory domains in a state that produces an open active site cleft.

PGDH is a tetramer of identical subunits. Within each subunit, three distinct structural domains are evident. They are referred to as the cofactor binding domain, the substrate binding domain, and the regulatory domain. Within the tetramer, each of two pairs of subunits contact each other by way of a massive interface between their respective cofactor binding domains. These subunit pairs form the tetrameric structure by means of an additional interface between their respective regulatory domains. Thus, there are two pairs of cofactor domain contacts and two pairs of regulatory domain contacts within the tetramer (see Fig. 1). It is this second, regulatory domain interface that is thought to be the basis for inhibitory effector modulation because the physiological effector, L-serine, binds to PGDH at the interface between regulatory domains of adjacent subunits, forming hydrogen bond contacts to both domains (9,10).
A model to explain the mechanism of allosteric regulation in PGDH has been developed from a knowledge of the crystal structure and the stoichiometry of effector binding (10). The model predicts that catalysis and allosteric inhibition of PGDH by L-serine are implemented through the motion of rigid domains about flexible hinges. There appear to be at least three potential hinge regions in PGDH. One hinge region is thought to occur between the cofactor binding domain and the substrate binding domain since catalysis probably requires the regular closing and opening of the catalytic cleft formed between these domains. A second hinge appears to occur between the substrate binding domain and the regulatory domain because the crystal structure of the inhibited enzyme (with L-serine bound between regulatory domains from adjacent subunits) shows that the substrate binding domain appears to be locked to the regulatory domain resulting in an open catalytic cleft. Finally, the binding of serine at the allosteric site appears to tether the regulatory domains to each other. This tethered state completely buries the serine molecule in the interface such that it is no longer accessible to solvent (Fig. 1). In order for serine to get in and out of its binding site, it appears that the interface between regulatory domains must open to some extent. Since PGDH remains a tetramer in the active as well as inhibited state, this opening does not involve subunit dissociation. Thus, a third hinge appears to be operative here such that the association between regulatory domains can be relaxed to allow * 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.
‡ To whom correspondence should be addressed: Dept. of Molecular Biology and Pharmacology, Box 8103, Washington University School of Medicine, 660 S. Euclid Ave., St. Louis, MO 63110. Tel.: 314-362-3367; Fax: 314-362-4698. 1 The abbreviations used are: PGDH, D-3-phosphoglycerate dehydrogenase; DTT, dithiothreitol; IPTG, isopropyl-1-thio-␤-D-galactopyranoside; PCR, polymerase chain reaction. association and dissociation of effector. The model predicts that relaxation of subunit association about this hinge, which would be the case when the effector is not bound, releases the substrate binding domain from its contacts with the regulatory domain so that it is now free to close the active site cleft for catalysis to occur. This latter hinge has been referred to as a "piano hinge" because it may act along a relatively long axis at the regulatory domain interface. The motion of the piano hinge is postulated to be directly modulated by binding of the effector molecule and functions at the point where two regulatory domains contact each other. This article demonstrates that, as predicted by the model, cross-linking the regulatory domains from adjacent subunits with engineered disulfides appears to prevent this motion and mimics effector binding, resulting in the reversible modulation of enzyme activity.
DNA sequencing was performed with the Sequenase version 2.0 sequencing kit from U. S. Biochemical Corp. Restriction fragments were isolated from agarose gels with the Gene Clean kit from BIO 101 Inc.
The serA gene encoding PGDH was derived from pSAWT (11) and placed in the expression vector pTrc 99A for expression by IPTG induction by utilizing the NcoI and HindIII sites that flank the serA gene. This construct was named pTrcPGDH. Mutants were produced by PCR mutagenesis using standard procedures (13). In preparation for PCR mutagenesis, the serA gene was placed into pBluescript SKϩ, in which an NcoI site had been added between the BamHI and PstI sites of the multiple cloning site. The gene was initially cut from pSAWT with NcoI and HindIII and spliced into the altered pBluescript SKϩ at these same sites. Subsequently, the linear template for PCR was produced from this construct with an NcoI and SalI digest. The area to be mutated was between internal EcoRI and HindIII sites that were used for the outside PCR primers and for insertion of the PCR product back into pTrcPGDH. All mutations were confirmed by Sanger dideoxy sequencing.
PGDH was expressed in E. coli following IPTG induction and isolated as described previously (11). Oxidation during purification could be reduced if all buffers contained 10 mM DTT with the exception of the final buffer used to elute the enzyme from the affinity column. After elution from the affinity column, a short dialysis (2-4 h) against elution buffer (2-4 liters) can be used to remove residual DTT if necessary. The enzyme was assayed at 25°C by following the decrease in absorbance at 340 nm in the presence of saturating levels of NADH and hydroxypyruvic acid phosphate as substrate (12). One unit of activity corresponds to the conversion of 1 nmol of NADH to NAD ϩ /min. Protein concentration shows the regulatory domains from subunits B and C in yellow and purple, respectively. The residues that are to be mutagenized to cysteine, Gly 349 and Ala 359 , are shown in white and the bound L-serine molecules which are buried in the interface, are shown in red. The orientation has been shifted slightly from the 2-fold axis so that portions of both serine molecules can be seen. Panel C shows a ribbon diagram of the same structure which clearly depicts the location of the anti-parallel ␤ sheet, the ␣-helices, and the approximate location of the engineered disulfides. The L-serine molecules are not shown in this diagram, but would be located at the interface approximately midway between the ␤ sheet and the helices. (A was reproduced from Ref. 9 with permission.) was determined by absorbance at 280 nm using an E 1% ϭ 6.7. Oxidation was performed by exposing the enzyme solution to air at 4°C. Reduction was performed at 25°C by adding the specified level of dithiothreitol. Titration with Ellman's reagent was performed according to published procedures (14) using L-cysteine to generate a standard curve.
SDS-PAGE was performed in the absence of reducing agent according to the method of Laemmli (15). Gel filtration chromatography with Sepharose S-200 HR was performed in a 1.6 ϫ 110-cm column with a flow rate of 0.4 ml/min in 20 mM imidazole buffer, pH 6.2, 1 mM EDTA. The column was calibrated with blue dextran, ␤-amylase (200,000), alcohol dehydrogenase (150,000), bovine serum albumin (66,000), and carbonic anhydrase (29,000).
The structure of PGDH was viewed with a Silicon Graphics IRIS (SGI) molecular graphics system Using SYBYL (Tripos Inc.) software. The SYBYL BIOPOLYMER program was used to simulate mutagenesis.

RESULTS
Rationale for Cross-link Location-If the close association of regulatory domains and closure of the regulatory domain interface is at the heart of the regulatory mechanism of PGDH as the model predicts, then tethering them together in a covalent fashion would be predicted to produce catalytic inhibition in a manner similar to that seen when L-serine is present. Thus, placing cysteine residues on each subunit at the domain interface so disulfide bridges can be made connecting the two subunits should cause the body of the domains to remain fixed relative to one another and thus prevent movement about the "hinge." The use of a disulfide bridge to accomplish this also allows the tether to be susceptible to reduction and thus produce a reversible effect. Moreover, the ability to reverse the cross-link with a reducing agent and restore activity would provide evidence that the rest of the protein is not compromised as a result of the mutation or the oxidation process.
Inspection of the structure of PGDH indicated that residues Ala 359 and Gly 349 are appropriately located at the regulatory domain interface (Fig. 1B) at the solvent exposed surface of the tetramer and have their ␣ carbons pointing at each other in the backbone structure. Because of the 180°symmetry at the interface, two residues in each domain must be mutated so that the resultant cysteines are adjacent to one another across the interface. Mutagenesis simulation indicates that mutating these residues to cysteine would place their ␤ carbons within 3.9 Å of each other, well within the distance needed for disulfide formation. Finally, within each domain, Ala 359 and Gly 349 are separated by approximately 15 Å so that intra-domain disulfides cannot form. They are also more than 20 Å from any of the native cysteines in PGDH.
Oxidation of A359C,G349C PGDH-When the A359C,G349C PGDH mutant was exposed to air, a gradual loss of activity was observed (Fig. 2). Native enzyme under the same conditions retained nearly 100% activity over the same time period. Table  I compares the specific activity of native and mutant PGDH with respect to time and oxidation state. Full activity can be restored at any time by treatment with DTT and the rate of recovery of activity is dependent on the concentration of DTT. The activity of the native enzyme is not affected by DTT treatment. These results have been repeated four times with four independent enzyme preparations with no significant variability observed. Moreover, oxidative inhibition of mutant enzyme can be repeated on enzyme that has already been oxidized by reducing it with DTT and then removing the DTT by gel filtration or rapid dialysis.
Titration of oxidized, inhibited A359C,G349C PGDH with Ellman's reagent indicates that approximately 1.9 sulfhydryl groups are oxidized per subunit when the enzyme is 90% inhibited (Table II). This compares very well with the expected value of 2 sulfhydryl groups/subunit for completely cross-linked enzyme due to the engineered cysteine mutations. Analysis with Ellman's reagent at low levels of inhibition indicates ap-proximately 6 cysteine residues/subunit, as expected.
Effect of Oxidation and Reduction on Subunit Association-Molecular weight analyses were performed on native and A359C,G349C PGDH using gel filtration (nondenaturing) and SDS gels (denaturing) to assess the nature of the subunit cross-linking as a result of disulfide formation. These data are presented in Table III and Fig. 3.
If cross-linking due to disulfide formation is taking place across the interface between two regulatory domains as intended, analysis under denaturing conditions, in the absence of reducing agent, should produce a monomeric species for oxidized native enzyme and a dimeric species for the oxidized mutant. The dimer would result because two subunits would be covalently linked at the point where their respective regulatory domains contact each other. Furthermore, if cross-linking was FIG. 2. Activity of native and mutant PGDH with respect to time. Native (f) and A359C,G349C PGDH (q) were assayed for activity at various times after purification. The samples were allowed to stand in solution exposed to air over the duration of the experiment.  occurring only across the regulatory domain interface between two adjacent subunits of the tetramer, analysis under nondenaturing conditions, again in the absence of reducing agent, should yield the molecular weight of the tetramer for both native and mutant enzyme. If cross-linking were taking place intermolecularly from one tetramer to another, the nondenaturing analysis should yield higher molecular weight aggregates. The data in Table III clearly indicate only interdomain cross-linking between two subunits within the tetramer as anticipated. The SDS gel in Fig. 3 also shows that dimer formation in response to oxidation has occurred in the mutant (lane 3), and that reduction with DTT efficiently regenerates the monomeric species (lane 4). Serine Inhibition of the Active A359C,G349C Mutant-The ability of L-serine to inhibit native PGDH and the active A359C,G349C mutant enzyme (the mutant enzyme prior to oxidative formation of the disulfides or after treatment with DTT) is shown in Fig. 4. Reduced mutant PGDH possesses catalytic activity equal to that of native PGDH (Table I) but the replacement of Ala 359 and Gly 349 by cysteine residues causes the IC 50 for serine inhibition to go from 8 M for the native enzyme to 800 M for the mutant. However, Hill plots indicate that the number of interacting sites for serine remain unchanged. This might be explained by considering that the introduction of cysteine side chains at the locations that are usually occupied by glycine and alanine introduces additional bulk at the regulatory domain interface where serine enters. However, the number of serine sites does not change and as long as the interface is free to open, the catalytic activity of the enzyme is not affected.

DISCUSSION
The engineered cysteines were placed at specific sites in PGDH based on detailed knowledge of the structure and the model of allosteric inhibition. Only the mutated enzyme is inhibited under oxidative conditions and then activated in response to a mild reductant well known to reduce disulfide bonds. The native enzyme does not respond in this way and is not affected by DTT. The mutant enzyme, when fully reduced, shows catalytic competence equal to the native enzyme. Therefore, the observed result is specifically a consequence of the mutation and the complete retention of specific activity in the reduced form indicates no significant changes have occurred in enzyme structure or catalytic mechanism. Furthermore, Hill plots of the serine inhibition of the native and mutated enzyme, show that the Hill coefficient remains unchanged. Thus, although the concentration at which serine affects the mutant form is higher, the mutant form retains 2 interactive sites.
The SDS gel data clearly show that an intersubunit crosslink is forming and the native gel filtration data show that cross-links are not forming from tetramer to tetramer. Therefore, the intersubunit cross-link is within each tetramer. The crystal structure shows that the only cysteines within the tetramer that are close enough to each other to form intersubunit cross-links are the ones that were engineered into the molecule. All of the native cysteines are greater than 20 Å from each other across subunits.
The crystal structure of PGDH indicates that disulfides do not exist in the native form. Moreover, inhibition with serine is completely reversible with removal of serine and does not depend on the redox state of the solution. There are only 2 native cysteines in PGDH that are in close enough proximity that an intrasubunit disulfide could be formed. Mutating these two native cysteines to alanine produces an enzyme that is indistinguishable from the native form in activity and its ability to be inhibited by serine. 2 Thus, these cysteines are not operative in the native mechanism. Furthermore, titration of free thiols in A359C,G349C PGDH shows that approximately 1.9 cysteine residues are oxidized when the enzyme is 90% inhibited and that the unoxidized enzyme contains the expected 6 cysteines/ subunit. Thus, these data are entirely consistent with oxidative inhibition of PGDH being due to the engineered cysteine residues at the regulatory domain interface.
These data indicate that tethering the regulatory domains from adjacent subunits of PGDH across their interface results in a reversible inhibition of PGDH in the absence of added natural inhibitor. Since the tetrameric nature of the enzyme has been shown not to change in response to effector mediated inhibition or domain cross-linking, this suggests that the regulatory mechanism involves the movement of one regulatory domain relative to the other. The effect of disulfide cross-2 R. Al-Rabiee and G. A. Grant, unpublished observation.  linking is to convert a dynamic situation into a static situation where the domains are essentially locked in place.
However, the exact nature of the motion of adjacent regulatory domains relative to one another remains to be determined. Inspection of the structure shows that the residues that appear to bind the effector, L-serine, are located at the turns that connect the ␣-helix nearest the interface with the ␤ sheet strands (see Fig. 1). This places the binding site for L-serine above the plane of the ␤ sheet and below the helices. It is tempting to speculate that perhaps the hinging motion involves the movement of the ␣-helices at the interface away from each other with the junction of the ␤ strands in the sheet acting as some sort of fulcrum. However, this view would require flexibility in the ␤ sheet at the interfacial junction, a phenomena for which there is very little precedent. An alternative view might entail movement of the ␣-helices independently of the ␤ sheet, with the hinge region being limited to the connecting turn that contains the effector binding residues. Effector binding to this region may stabilize the hinge. This, however, would seem to require movement of the inner helices relative to the outer helices, which seems unlikely because of the proximity of adjacent helices.
While the exact nature of regulatory domain interaction and the location of the hinge remain to be elucidated, these results support the main feature of the proposed model for the allosteric mechanism. That is, catalytic inhibition results when the regulatory domains are tethered to each other by the effector molecule, L-serine, and these domains must move in some way relative to one another during the active to inhibited state transition.