Conformational changes and stabilization of inosine 5'-monophosphate dehydrogenase associated with ligand binding and inhibition by mycophenolic acid.

The effects of substrate, product, and inhibitor (mycophenolic acid) binding on the conformation and stability of hamster type II inosine 5′-monophosphate dehydrogenase (IMPDH) have been examined. The protein in various states of ligand occupancy was compared by analyzing susceptibility to in vitro proteolysis, the degree of binding of a hydrophobic fluorescent dye, secondary structure content as determined by far-UV circular dichroism spectra, and urea-induced denaturation curves. These analysis methods revealed consistent evidence that IMPDH undergoes a local reorganization when IMP or XMP bind. NAD+ produced no such effect. In fact, no evidence was found for NAD+ binding independently of IMP. It is proposed that IMPDH adopts an open conformation around its nucleotide binding sites in the absence of substrates and that binding of IMP stabilizes a closed conformation that has a higher affinity for NAD+. The data also suggest the enzyme remains in the closed configuration throughout the catalytic steps and then reverts to the open conformation with XMP release, thereby consummating the enzyme cycle. Mycophenolic acid inhibition appeared to impart even greater stability. We propose that localized conformational changes occur during the normal and mycophenolic acid-inhibited reaction sequences of IMPDH.

Inosine 5Ј-monophosphate dehydrogenase (IMPDH; 1 EC 1.1.1.205) catalyzes the conversion of IMP to XMP, the committed step in the de novo biosynthesis of guanine nucleotides (1)(2)(3)(4). Now partially characterized from many organisms, IMPDH is well conserved with 35% or more sequence identity from bacterial and protozoan to human proteins (5,6). Since the enzyme appears to be necessary for cellular replication, some view it as an attractive target for anti-microbial chemotherapy; however, others have recognized further therapeutic possibilities for IMPDH inhibition.
Human and hamster possess two closely related isoenzymes (designated types I and II). In both species the isoforms share 84% (of 514 residues) sequence identity (5). 2 The existence of similar isoenzymes in mammals suggests a subtle distinction of roles. Many studies have shown that IMPDH activity is elevated in tumors and other actively proliferating tissues (3,4,(7)(8)(9)(10)(11)(12)(13)(14)(15). In some cases the elevated activity seems to stem from induction of the type II isoenzyme (9,11,12), although similar up-regulation of both isoforms in activated T-lymphocytes has also been reported (14). These observations suggest that inhibition of IMPDH, or perhaps selective inhibition of individual isoforms, has potential for immunosuppressive or anti-neoplastic therapies. Indeed, inhibition of IMPDH in cultured cells results in decreased guanine nucleotide levels. Ultimately, this leads to the induction of differentiation in neoplastic cells (8) and the suppression of lymphocyte activation (16).
Mycophenolic acid (MPA), an uncompetitive inhibitor of IMPDH that was originally identified in certain Penicillium cultures (17), is a potent immunosuppressant, both in vitro and in vivo (for reviews, see Refs. 18 -20). Mycophenolate mofetil, a morpholinoethyl ester pro-drug of MPA, has been developed as a therapy for organ transplantation. The drug has been approved in the United States and Europe for the prevention of acute rejection in kidney transplant recipients, but broader clinical use may be complicated by drug toxicity (for reports and reviews of clinical findings, see Refs. [21][22][23][24]. Therefore, with the goal of increasing drug efficacy and decreasing toxic side effects, there is great interest in characterizing the inhibition of IMPDH by MPA and other compound classes. Many species of IMPDH have been isolated, both from natural and recombinant sources. All possess a similar homotetrameric quaternary arrangement of about 55-57-kDa subunits, and the numerous enzyme kinetic characterizations (reviewed recently by Wu (19) and (25)) suggest a common mechanism that is unusual among dehydrogenases. In most other dehydrogenases substrate binding is either random or the hydride acceptor binds first (26). This binding often induces a large conformational change that prepares the enzyme for the catalytic events (27)(28)(29)(30). In the case of IMPDH, catalysis proceeds via an ordered bi-bi kinetic mechanism wherein IMP (the hydride donor) binds first and NAD ϩ binds second. Thus, the substrate binding order is inverted relative to many other dehydrogenases. Following substrate binding and hydride transfer, IMPDH first releases NADH and, finally, XMP (31)(32)(33)(34)(35)(36). Recent independent studies, in our laboratory (37,38) and by others (39), have identified the probable mechanism of IMPDH inhibition by MPA. The inhibitor interrupts the normal enzyme turnover cycle, after both substrates have bound, after hydride transfer and after NADH release, by trapping the enzyme with a precursor of XMP still covalently bound to the enzyme active site cysteine. Our colleagues (38) have recently solved a high-resolution x-ray crystal structure of hamster type II IMPDH with MPA bound. This structure confirms that MPA traps a covalent adduct of IMPDH and a precursor of XMP. It also establishes a structural framework necessary to better understand the wealth of kinetic analyses already published.
However, no studies have yet investigated the effects of ligand binding on the conformation and/or stability of IMPDH. A better understanding of the dynamic behavior of IMPDH upon ligand binding should prove valuable for inhibitor design, especially in light of the inverted order of substrate binding.
We report the effects of substrate, product, and inhibitor binding on the conformation and stability of hamster type II IMPDH. The protein was examined in various states of ligand occupancy, using in vitro proteolysis, fluorescence quenching techniques, and circular dichroism (CD) spectroscopy. The urea-induced denaturation of such samples was also examined using CD spectroscopy. These complementary analysis methods revealed consistent evidence that IMPDH is locally reorganized and stabilized by the binding of IMP or XMP, but not by NAD ϩ . MPA inhibition appears to impart even greater stability. We propose that localized conformational changes occur during the normal and MPA-inhibited reaction sequences of IMPDH.

EXPERIMENTAL PROCEDURES
Materials-Highest grade available IMP (free acid), NAD ϩ , NADH (disodium salt), XMP (sodium salt), 6-chloro-IMP (sodium salt), and ANS (8-anilino-1-naphthalenesulfonic acid) were purchased from Sigma. MPA was obtained from Calbiochem. IMPDH Purification and Enzyme Parameters-Expression in Escherichia coli, purification, and enzymatic characterization of hamster type II IMPDH has already been described (37). The resultant protein appeared Ͼ99% pure as judged by SDS-PAGE (40) and N-terminal sequence analyses. Analytical size-exclusion chromatography and lightscattering measurements showed the protein was virtually monodisperse with population weight-average (M w ) and number-average (M n ) molecular weights of 228,300 and 227,800, respectively (M w /M n ϭ 1.002). The enzyme contained undetectable levels of other oxidoreductase activities. The observed k cat was 0.76 s Ϫ1 , and K m values for IMP and NAD ϩ and were 22 and 37 M, respectively. The observed K ii for MPA (with respect to NAD ϩ ) was about 11 nM (37).
For the present study, the enzyme was stored at Ϫ70°C at about 5 mg/ml in Buffer A (300 mM KCl, 50 mM Tris-HCl, 2 mM EDTA, 10 mM ␤-mercaptoethanol, 10% glycerol, pH 8.0, at 4°C) prior to use. Enzyme activity was determined spectrophotometrically at 22°C, by measuring the conversion of NAD ϩ to NADH via the increase in absorbance at 340 nm. Typical assays contained substrate concentrations of 300 and 250 M for IMP and NAD ϩ , respectively, and 100 nM IMPDH in Buffer A. Protein concentrations were routinely determined by UV spectroscopy, using a specific molar extinction coefficient (A 278 ) of 23,800 M Ϫ1 ⅐cm Ϫ1 (determined according to Ref. 41), using a molar extinction coefficient for Tyr of 1400 M Ϫ1 ⅐cm Ϫ1 ), expressed relative to IMPDH protomer concentration. Where this was complicated by other UV-absorbing species, concentrations were determined by the method of Bradford (42) using hamster IMPDH II (concentration determined as above) as the standard.
In Vitro Proteolysis of IMPDH-IMPDH was incubated at a protomer concentration of 40 M (2.2 mg/ml), at 22°C in Buffer B (same as Buffer A, but with 40 mM ␤-mercaptoethanol), with 22 g/ml of the following proteases ␣-chymotrypsin, Glu-C, and elastase. Proteolysis with proteinase K was performed under similar conditions on ice. Incubation times and added ligands are indicated in the figure legends. Proteolysis was stopped by the addition of an equal volume of 2 ϫ SDS-PAGE sample buffer (40) and boiling. Proteolysis with proteinase K was arrested by addition of 2 mM phenylmethylsulfonyl fluoride and trichloroacetic acid precipitation. The precipitate was dissolved in 1 ϫ SDS-PAGE sample buffer. Samples containing 2 g of IMPDH were analyzed on 4 -20% gradient gels, which were stained with Coomassie Brilliant Blue R250 or blotted onto polyvinylidene difluoride membranes for N-terminal sequencing. The proteolytic fragments were sequenced by Edman degradation using an Applied Biosystems 477A automated protein sequencer.
ANS Fluorescence Measurements-Fluorescence data were measured using a Perkin-Elmer LS50B spectrofluorimeter and 1-cm path length cells. Samples were maintained at 25°C during data collection, after a 10-min preincubation period with or without added ligands. Emission spectra were recorded from 400 to 650 nm, using an excitation wavelength of 350 nm and slit widths of 5 and 15 nm for excitation and emission, respectively. The IMPDH protomer concentration was typically 0.2 mg/ml (3.6 M) in Buffer B, with 50 M ANS. The concentration of the ANS stock solution was determined via the absorption at 350 nm using a molar extinction coefficient of 5000 M Ϫ1 ⅐cm Ϫ1 (43). The final concentration of added IMP, 6-chloro-IMP, NAD ϩ , or XMP was 50 M. These concentrations ensured the absorbance at the excitation wavelength (A 350 ) was always below 0.1, thus minimizing inner filter effects. The data shown represent unsmoothed scans, corrected for appropriate solvent fluorescence/Raman-scattering base lines in each case. The spectra were not corrected for the emission-monochromator and photo multiplier responses of the instrument.
Circular Dichroism (CD) Spectroscopy-All CD data were acquired using a Jasco J-715 spectropolarimeter, calibrated with a 0.06% (w/v) solution of ammonium D-10-camphorsulfonate. Far-UV spectra were recorded between 260 and 184 nm, using a 1-nm slit width, and a 0.05 mm path length cell thermostatted at 25°C. IMPDH at 2 mg/ml was dialyzed exhaustively into Buffer C (100 mM potassium phosphate, 10% glycerol, and 2 mM ␤-mercaptoethanol, pH 8.0), and the protein concentration was redetermined. Ligand stock solutions (IMP, XMP, and NAD ϩ each at 25 mM, MPA at 2 mM) were prepared in dialysate. Protein, dialysate, and/or ligand stocks were then diluted to the following final concentrations: IMPDH, 1 mg/ml; MPA 200 M; IMP, XMP, and NAD ϩ each 500 M, as required. For samples and solvents, 10 separate scans (scan rate 50 nm/min; sensitivity 10 millidegree; response time 2 s) were averaged and then smoothed by fast Fourier transform methods (using standard Jasco J-715 instrument software). After appropriate dialysate plus ligand spectra were subtracted from the protein spectra, the resulting data were converted from millidegree to mean-residue molar ellipticities using a mean-residue M r of 108.7 (M r 55,759, and 513 amino acid residues per IMPDH subunit). Processed in this way, the protein spectra show the effects of ligand binding but do not comprise ligand ellipticities per se. Spectra were analyzed to predict secondary structure content, using the variable selection fitting method of Johnson et al. (44,45).
Samples were prepared for urea-denaturation studies by diluting protein and/or ligand in Buffer C into the same buffer with 8.5 M urea. Ligands were added 10 min prior to the addition of urea (final protein concentration of 0.25 mg/ml, and 0 to 8 M urea in Buffer C), and the samples were then incubated at 22°C for 15 h prior to analysis. Spectra were recorded in a 1-mm path length cell, otherwise as described above. Due to the high absorbance of urea-containing buffers below 210 nm, protein unfolding was monitored by plotting the ellipticity at 225 nm as a function of urea concentration. All data were corrected for {solvent ϩ cell} base-line signals. Fig. 1A presents a comparison, by SDS-PAGE, of the sensitivity of hamster IMPDH II to proteolysis by proteinase K, elastase, ␣-chymotrypsin, and Glu-C. For each enzyme, two proteolytic fragments were formed (apparent M r about 45,000 and 12,000) more rapidly, and at lower protease concentrations, than any products generated by other cleavage events (see below). These four proteases were chosen because they cleave at the C-terminal side of quite different recognition sequences (viz. proteinase K, N-substituted, hydrophobic aliphatic and aromatic residues; elastase, residues bearing uncharged nonaromatic side chains; ␣-chymotrypsin, Tyr, Phe, Trp, Leu, Met, Ala, Asp, and Glu residues; Glu-C, specifically Glu and Asp residues). This made it unlikely that the pattern would be dominated by protease sequence preferences. Instead, it suggests that IMPDH contains a well-exposed region about 12 kDa from either terminus. This is supported by the studies of Gilbert et al. (46), who showed rapid generation of two fragments (apparent M r about 42,000 and 14,000) upon treatment of E. coli IMPDH with trypsin or Pronase.

IMPDH Sensitivity to In Vitro Proteolysis-
The protease-sensitive region of hamster IMPDH II was identified by N-terminal sequencing. Consistent with the gel pattern, each protease yielded similar fragments. The larger fragment from elastase, ␣-chymotrypsin, and Glu-C proteolysis began at the authentic N terminus of the native IMPDH sub-unit (with the N-terminal Met, resulting from E. coli expression, not present). Proteinase K digestion generated a large fragment beginning either at residue Gly-8 or Tyr-11 of the unproteolyzed subunit. To generate the small fragment (ϳ12 kDa) the proteases cleaved between Gln-427 and Asn-428 for proteinase K, Ile-437 and Lys-438 and Ala-440 and Gln-441 for elastase, Tyr-411 and Arg-412 and Tyr-430 and Phe-431 for ␣-chymotrypsin, and Glu-433 and Ala-434 for Glu-C. These sequences are all consistent with the respective enzyme recognition preferences. Also, the sizes of the fragments observed on SDS-PAGE corresponded well with the M r values calculated from the primary structure and the identified cleavage sites, assuming no further proteolysis (data not shown). This suggests that under these conditions additional C-terminal degradation, of either fragment, was minimal for all of the proteases.
Elastase was chosen for further analysis of time dependence and ligand-binding effects, because its recognition preference is for residues deemed less likely to be directly involved in ligand binding. As shown in Fig. 1B, IMPDH was much less sensitive to proteolysis by elastase in the presence of IMP (lane 3) or XMP (lane 8) than in the absence of added ligand (lane 2). By contrast, the enzyme was not protected by NAD ϩ (lane 4), and it was only marginally protected by NADH (lane 9). In equilibrium binding studies, Xiang et al. (25) saw no evidence for NAD ϩ binding to IMPDH and observed weak binding of NADH at 4°C and no binding at 37°C. When IMPDH was incubated with both IMP and NAD ϩ , and the conditions were designed to allow the enzyme to convert substrates to products throughout the elastase incubation period (data not shown), IMPDH was again protected from proteolysis (lane 5). MPA was added in addition to IMP and NAD ϩ before proteolysis was started. As demonstrated previously (37)(38)(39), under these conditions the substrates are partially turned over, leaving free NADH and only IMPDH, MPA, and covalently bound XMP in the inhibited complex, as well as excess IMP and NAD ϩ . The MPA-inhibited complex of IMPDH was considerably more resistant to elastase proteolysis than any other form of the enzyme, showing no evidence for cleavage (lane 7) under conditions that produced essentially quantitative cleavage of the unprotected protein (lane 2). MPA alone did not lead to stabilization of IMPDH against proteolysis (not shown). IMPDH that had been completely inactivated by 6-chloro-IMP (data not shown), which forms a covalent adduct with the active site cysteine (31,47), was not protected (lane 6).
Time Course of Proteolysis of IMPDH by Elastase- Fig. 2 presents the time dependence of elastase proteolysis of IMPDH, unprotected ( Fig. 2A) and protected by IMP (Fig. 2B). For both situations, appearance of cleavage products closely paralleled depletion of the full-length subunits. This indicates that there was minimal cleavage at other internal sites, although terminal degradation of a few residues (that might not be resolved by SDS-PAGE) cannot be ruled out. Fig. 2B shows that the proteolysis rate was much lower in the presence of IMP. The unprotected protein was almost completely proteolyzed in 45 min, whereas in the presence of IMP some fulllength protein was detectable even after 10 h (see Fig. 4A, below). Fig. 3 shows the results of analyses of the IMPDH enzymatic activity of the same protected and unprotected samples undergoing proteolysis. These data demonstrate that this conspicuous internal cleavage event resulted in inactive enzyme. Densitometric scanning of stained gels, monitoring the disappearance of the full-length subunit, produced virtually identical decay curves (data not shown). It was also demonstrated, by combined size-exclusion chromatography and SDS-PAGE analysis, and light scattering experiments, that the two major fragments of proteolysis remained associated (results not shown). Therefore, inactivation resulted from the localized effects of the proteolytic cleavage and not from dissociation of motifs that together constitute the catalytic machinery. Fig. 4 extends the proteolytic incubation period to show the greater stability of the MPA-inhibited sample. After 10 h about 20% of the full-length subunit was detectable in the IMPprotected sample (Fig. 4A), whereas about 70% remained uncleaved in the MPA complex (Fig. 4B). Over this time scale some band broadening of the larger fragment was seen for the unprotected protein (data not shown). Sequence analyses showed this was probably due to variable degradation of the C terminus. No further proteolysis was observed for the unprotected enzyme. By contrast, in the presence of IMP, and less prominently in the MPA-inhibited sample, three new proteolytic fragments of about 32, 26, and 13 kDa were observed. Almost no formation of the 45-and 12-kDa fragments occurred during proteolysis of the MPA-inhibited sample. The N termini of both the 32-and 26-kDa fragments corresponded either to Ala-223 or Arg-224. The 26-kDa fragment seemed to be derived from the 32-kDa fragment, but this was not shown formally because the incubation times were not long enough to measure a decrease in intensity of the 32-kDa band. The 13-kDa fragment corresponded to the extreme N terminus of native IMPDH.
The results can be explained assuming two separate proteolytic pathways. In the first, IMPDH devoid of ligands was rapidly proteolyzed around residues 412-441 (before residues 438 and 441 for elastase). This was slowed dramatically by IMP binding and virtually eliminated by MPA complexation. In a second pathway, the IMP-IMPDH and MPA-IMPDH samples were susceptible at another site (before residues Ala-223 or Arg-224 with elastase). Susceptibility to the second pathway only became evident with the ligand-induced attenuation of the first. Also, when unprotected IMPDH was rapidly proteolyzed by the first pathway, little evidence of the second pathway was later seen. Therefore, the two pathways seem mutually exclusive. The appearance of the second pathway may reflect direct conformational effects of IMP or MPA binding. However, it is also possible that cleavage via the first pathway made IMPDH unrecognizable by elastase via the second pathway. In either scenario, the IMPDH-IMP sample would be susceptible to both pathways (as observed) because of the equilibrium between free and IMP-bound protein.
ANS Fluorescence in the Presence of Different Nucleotides-IMPDH contains 17 tyrosine residues (per subunit) and no tryptophan residues, so ligand binding effects and conformational changes are not easily investigated by measuring intrinsic protein fluorescence. Hager et al. (48) analyzed the fluorescence changes in IMPDH upon binding of the competitive inhibitor mizoribine monophosphate, but the use of their approach did not result in significant fluorescence changes when MPA and the enzyme substrates and products were added (results not shown). Therefore, we analyzed the binding of ANS to IMPDH, in the presence of different ligands. The binding of ANS, a fluorescent hydrophobic dye, is often used as a measure of the amount of hydrophobic surface on proteins (49). As shown in Fig. 5, IMPDH had the same ANS fluorescence in the presence and absence of NAD ϩ , whereas in the presence of IMP or XMP the intensity decreased by about 40%. These findings are consistent with the protection of IMPDH by IMP and XMP in the proteolysis experiments and the lack of protection by NAD ϩ . Interestingly, IMPDH that had been inactivated by 6-chloro-IMP had an ANS fluorescence intermediate between unliganded IMPDH and the IMP-bound form. The ANS fluorescence of the MPA-inhibited complex could not be interpreted because control samples containing NADH exhibited very high fluorescence, and NADH is generated (by partial substrate turnover) in the course of MPA inhibition.
Circular Dichroism Analysis of IMPDH- Fig. 6 presents the far-UV CD spectra of unliganded IMPDH, IMPDH with IMP, and MPA-inhibited enzyme (incubated with IMP, NAD ϩ , and MPA). These are the first CD analyses reported for IMPDH. All of the spectra were very similar. Addition of ligands did result in some minor perturbation of ellipticities, especially below 200 nm. The XMP spectrum is not shown because of unusual behavior below 186 nm (data not shown). We are investigating the cause of this, but we are presently unable to obtain reliable secondary structure data for that complex. The results of secondary structure analyses derived from the spectra are shown in Table I. The secondary structure content for the unliganded and IMP-bound forms of IMPDH is quite similar (about 30% ␣-helix and 24% ␤-sheet/strand), whereas the MPA-inhibited complex has a slightly lower ␣-helix (27%) and higher ␤-sheet/ strand (30%) content. The proportions of the different secondary structure components, as calculated from the CD spectra, are quite consistent with the numbers calculated from the x-ray structure of the MPA complex (38) (see Table I). However, the x-ray structure has identified only about 83% of the residues in the protein.
Urea Denaturation of IMPDH-Urea-induced unfolding of IMPDH was analyzed by monitoring the magnitude of the ellipticity minimum at 225 nm, at different urea concentrations. This was done for IMPDH with and without the ligand combinations described already. After preincubating the IMPDH/ligand mixtures for 10 min at 22°C, the samples were made up to the selected urea concentrations and then incubated for a further 15 h (at 22°C) to allow complete equilibration. As shown in Fig. 7, all samples except the MPA-inhibited complex showed symmetrical, sigmoidal decreases in ellipticity as the urea concentration was increased. This behavior is usually interpreted in terms of a simple two-state denaturation process (50). Since IMPDH is a homotetramer, these curves probably describe the combination of denaturation and dissociation processes (other data not shown), so a detailed interpretation according to a two-state model may be inappropriate. Nevertheless, the urea concentration at the midway point of the transitional portion of such curves is often used as an indicator of relative protein stability (50). The midway point for native IMPDH occurred at 5 M urea. The shape and position of the curve was unaffected by the presence of NAD ϩ . In the presence of IMP or XMP the curve shapes still appeared the same, but the midway points occurred at 5.7 M urea. This suggests the protein is stabilized by the latter ligands. The MPA-inhibited complex appeared to be much more stable toward urea unfolding. Even at the highest urea concentration  a Percentages of the secondary structural elements present in the determined x-ray crystal structure of the MPA-IMPDH complex, consisting of 425 out of 514 residues (38). used (8 M) the buffer-corrected ellipticity had not plateaued at a minimum value, indicating the protein was still not fully denatured and/or dissociated. To rule out the possibility of very slow unfolding and demonstrate the attainment of equilibrium conditions, the samples were re-analyzed after a further 24 h. These curves (data not shown) were virtually identical to those shown in Fig. 7.

DISCUSSION
Conformational Changes during the IMPDH Substrate Binding Sequence-Controlled in vitro proteolysis is often used to probe protein conformation and stability, especially when direct structural information is lacking. This approach has been used in studies to map surface loops, domain boundaries, and protein-protein or protein-ligand interaction sites (e.g. bovine brain calcineurin (51)) that were later confirmed by high resolution structural studies (52).
Our studies identify the region around residues 412-441 of the IMPDH subunits as being particularly vulnerable to proteolytic cleavage. Proteolysis in this region proceeded far more rapidly than any other internal cleavage, but binding of IMP or XMP dramatically reduced the rate of proteolysis. Without a high resolution structure of the protein, the simplest interpretation for this would be that this region forms an open "loop" or "flap" on the surface of the enzyme when no ligands are bound and that the binding of IMP or XMP induces a localized conformational change. The ligand-bound configuration, in effect, closes this putative "loop" or "flap" region by causing it to adopt a more compact arrangement as it collapses around the nucleotide. No detailed structural information is available for IMPDH without ligands bound, but the x-ray crystal structure of IMPDH complexed with MPA and XMP, recently solved by our colleagues (38), supports the simple model presented above. The structure shows XMP is partially buried, with the hypoxanthine ring making multiple hydrogen bonds to a portion of the protein described as a "flap motif." This flap, as delineated by the x-ray structure (38), spans residues 400 -450 and has a pair of ␤-strands at its borders (residues 406 -413 and residues 443-448). It is striking that the portion of this flap that connects the ␤-strands (residues 414 -442) almost exactly coincides with the region (residues 412-441) we have independently identified as the part of IMPDH most vulnerable to proteolysis in the absence of ligands and that binding of IMP or XMP greatly reduces its vulnerability.
In most dehydrogenases, binding of the hydride acceptor or cofactor either occurs first or the binding can proceed in random order (26). In such cases conformational changes induced by binding of the hydride acceptor are well documented (27)(28)(29)(30). In a striking example of this, the structure of glycolate oxidase verges on a molten globule-like state in the absence of the co-factor, FMN (30). Binding of the co-factor induces widespread structural reorganization and stabilization, whereupon the enzyme active site residues are oriented optimally for catalysis. Our findings suggest that the hydride donor, IMP, induces local changes in a binding site flap of IMPDH. This may reflect a mechanism analogous to that in other dehydrogenases with an inverted order of substrate binding. NAD ϩbound enzyme is not an intermediate in the IMPDH reaction mechanism, and our present studies show NAD ϩ did not stabilize IMPDH against proteolysis or urea denaturation in the absence of IMP. Indeed, we found no evidence, using any of the dissimilar analytical methods described herein, for NAD ϩ binding to the protein independently of IMP. Recently, Xiang et al. (25) have analyzed equilibrium binding of ligands to IMPDH and saw no evidence for binding of NAD ϩ . Also, detailed sequence comparisons (37,38) suggest that the IMP binding site of IMPDH, and the NAD ϩ binding site of dehydrogenases that bind NAD ϩ first, probably evolved from a common ancestral motif. Both substrates are nucleotides, so the structural inference seems plausible. All of this considered, we propose that the binding of NAD ϩ to IMPDH is facilitated by local conformational changes induced by IMP binding. Given that IMP and XMP binding invariably resulted in identical effects, we propose that the normal enzyme cycle is consummated by a reciprocal rearrangement of the enzyme upon release of XMP.
The results of ANS fluorescence and CD analyses are consistent with these contentions. The fluorescence experiments suggest a decrease in hydrophobic surface of IMPDH results when IMP or XMP bind. This could be due to steric effects where the ligands bury hydrophobic residues lining their common binding pocket, or a conformational change in the protein, or a mixture of both effects. 6-chloro-IMP irreversibly inactivates IMPDH by binding to the IMP site and forming a stable covalent bond from the 6-position of its purine ring to the enzyme active site cysteine (47). In the present study, IMPDH inactivated by 6-chloro-IMP showed fluorescence behavior intermediate between that of the unliganded protein and the IMP-or XMP-bound forms, yet 6-chloro-IMP gave no proteolytic protection (see below). These findings suggest that a conformational change in the protein contributed to the fluorescence decrease observed when IMP or XMP bind. The CD spectra show the ligands did not alter the overall characteristics of the protein conformation. Since the proposed conformational change could occur without a change in secondary structure content, our model is consistent with the results of the CD analysis that rule out a global rearrangement of the protein upon ligand binding.
Inactivation by 6-Chloro-IMP and Inhibition by MPA-We have proposed that IMP induces a closed conformation in IMPDH which persists throughout the reaction sequence until XMP release. In the context of this model, we surmise that the binding of 6-chloro-IMP leaves the enzyme in slightly different state than the enzymatically productive IMP-bound conformation. Unlike IMP and 2-chloro-IMP, which become linked via the 2-position of the purine ring to the active site cysteine (47), 6-chloro-IMP forms a covalent link from the 6-position. Perhaps this orients the substrate analog differently, making it unable to induce the same local changes that occur when IMP binds and assists NAD ϩ binding. Previous studies have shown MPA inhibits IMPDH by trapping a transient intermediate of the normal enzyme cycle, after hydride transfer and NADH release, but before a covalent bond with XMP is severed (37)(38)(39). The extreme proteolytic stability of MPA-inhibited IMPDH can be explained by the tight binding of MPA and further stabilization of the closed IMPDH-XMP complex. Whereas a somewhat modest but similar stabilization against urea denaturation was observed for both IMP and XMP, the MPA-inhibited complex was extremely stable. Only partial denaturation was achieved after 40 h incubation at 22°C in 8 M urea. From these data we cannot determine a free energy of stabilization by binding of MPA or the other ligands, because more complete unfolding curves would be needed (50). We are currently examining the effects of stronger chaotropic denaturants. Secondary structure analyses of CD spectra of IMPDH without bound ligands and of the MPA complex predicted the latter had slightly higher ␤-sheet/strand contents and slightly lower ␣-helix content than native IMPDH. The proportions of different secondary structure components, as calculated from the CD spectra, were quite consistent with the numbers calculated from the x-ray structure of the MPA complex (38) (see Table I). However, the x-ray structure only identifies about 83% of the residues in the protein. Because of this, and because a high resolution structure of unliganded IMPDH is not available, we cannot reasonably infer which regions gain ␤-structure propensity and which lose ␣-helix propensity upon MPA-binding.
In summary, we show evidence that IMPDH adopts an open conformation around its nucleotide binding sites in the absence of substrates and that binding of IMP induces a closed conformation that enhances the affinity for NAD ϩ . The data suggest the enzyme remains in this closed configuration throughout the catalytic steps until it reverts to the open conformation with XMP release, thereby consummating the enzyme cycle. MPA appears to further stabilize the closed IMPDH-XMP* complex (IMPDH with XMP still covalently bound). Our ongoing efforts are to dissect steric and energetic effects of IMPDH ligand binding and inhibition.