Characterization of the human aldehyde reductase gene and promoter.

Aldehyde reductase (EC 1.1.1.2; AKR1A1) is involved in the reduction of biogenic and xenobiotic aldehydes and is present in virtually every tissue. To study the regulation of its expression, the human aldehyde reductase gene and promoter were cloned and characterized. The protein coding region consists of eight exons, with two additional upstream exons, separated by a large intron of 9.4 kb, that code for the 5' untranslated region of the mRNA. Two mRNA transcripts that encode the same protein and that originate from alternative splicing were identified. The shorter transcript is the major form as shown by Northern blots and reverse transcription-PCR experiments. Northern blots of multiple tissues indicate that aldehyde reductase mRNA is present in all tissues examined and is most abundant in kidney, liver, and thyroid, which is consistent with the tissue enzyme distribution. The two mRNA transcripts do not exhibit differential tissue distribution. A construct containing a promoter region insert in a pGL3 vector drives transcription of a luciferase reporter gene and is 290-fold more active than a control vector without insert in transfected HepG2 cells. The activity of the full promoter construct is comparable to that of a pGL3 vector containing the SV40 promoter with an enhancer. The promoter does not contain a TATA box, but contains multiple GC-rich islands and exhibits bidirectional activity in transfection studies. The major active promoter element was localized by nested deletions and mutations to a DNA element (TGCAAT, -59 to -54) that presumptively binds the transcription factor CHOP [CAAT enhancer binding protein (C/EBP) homologous protein]. Comparison of the aldehyde reductase gene structure to all other characterized human genes of the aldo-keto reductase superfamily (aldose reductase, bile acid binder, and type I and type II 3alpha-hydroxysteroid dehydrogenases) indicates that it is more distantly related to these genes than they are among themselves.

Osmotic response element enhancer activity. Regulation through p38 kinase and mitogen-activated extracellular signal-regulated kinase kinase.

Hypertonicity induces a group of genes that are responsible for the intracellular accumulation of protective organic osmolytes such as sorbitol and betaine. Two representative genes are the aldose reductase enzyme (AR, EC 1.1.1.21), which is responsible for the conversion of glucose to sorbitol, and the betaine transporter (BGT1), which mediates Na+-coupled betaine uptake in response to osmotic stress. We recently reported that the induction of BGT1 mRNA in the renal epithelial Madin-Darby canine kidney cell line is inhibited by SB203580, a specific p38 kinase inhibitor. In these studies we report that the hypertonic induction of aldose reductase mRNA in HepG2 cells as well as the osmotic response element (ORE)-driven reporter gene expression in transfected HepG2 cells are both inhibited by SB203580, suggesting that p38 kinase mediates the activation and/or binding of the transcription factor(s) to the ORE. Electrophoretic gel mobility shift assays with cell extracts prepared from SB203580-treated, hypertonically stressed HepG2 cells further show that the binding of trans-acting factors to the ORE is prevented and is thus also dependent on the activity of p38 kinase. Similarly, treatment of hypertonically stressed cells with PD098059, a mitogen-activated extracellular regulated kinase kinase (MEK1) inhibitor, results in inhibition of the hypertonic induction of aldose reductase mRNA, ORE-driven reporter gene expression, and the binding of trans-acting factors to the ORE. ORE-driven reporter gene expression was not affected by p38 kinase inhibition or MEK1 inhibition in cells incubated in iso-osmotic media. These data indicate that p38 kinase and MEK1 are involved in the regulation of the hyperosmotic stress response.

Factors affecting counteraction by methylamines of urea effects on aldose reductase.

The concentration of urea in renal medullary cells is high enough to affect enzymes seriously by reducing Vmax or raising Km, yet the cells survive and function. The usual explanation is that the methylamines found in the renal medulla, namely glycerophosphocholine and betaine, have actions opposite to those of urea and thus counteract its effects. However, urea and methylamines have the similar (not counteracting) effects of reducing both the Km and Vmax of aldose reductase (EC 1.1.1.21), an enzyme whose function is important in renal medullas. Therefore, we examined factors that might determine whether counteraction occurs, namely different combinations of assay conditions (pH and salt concentration), methylamines (glycerophosphocholine, betaine, and trimethylamine N-oxide), substrates (DL-glyceraldehyde and D-xylose), and a mutation in recombinant aldose reductase protein (C298A). We find that Vmax of both wild-type and C298A mutant generally is reduced by urea and/or the methylamines. However, the effects on Km are much more complex, varying widely with the combination of conditions. At one extreme, we find a reduction of Km of wild-type enzyme by urea and/or methylamines that is partially additive, whereas at the other extreme we find that urea raises Km for D-xylose of the C298A mutant, betaine lowers the Km, and the two counteract in a classical fashion so that at a 2:1 molar ratio of betaine to urea there is no net effect. We conclude that counteraction of urea effects on enzymes by methylamines can depend on ion concentration, pH, the specific methylamine and substrate, and identity of even a single amino acid in the enzyme.

Linkage analyses in type I diabetes mellitus using CASPAR, a software and statistical program for conditional analysis of polygenic diseases.

We have developed software and statistical tools for linkage analysis of polygenic diseases. We use type I diabetes mellitus (insulin-dependent diabetes mellitus, IDDM) as our model system. Two susceptibility loci (IDDM1 on 6p21 and IDDM2 on 11p15) are well established, and recent genome searches suggest the existence of other susceptibility loci. We have implemented CASPAR, a software tool that makes it possible to test for linkage quickly and efficiently using multiple polymorphic DNA markers simultaneously in nuclear families consisting of two unaffected parents and a pair of affected siblings (ASP). We use a simulation-based method to determine whether lod scores from a collection of ASP tests are significant. We test our new software and statistical tools to assess linkage of IDDM5 and IDDM7 conditioned on analyses with 1 or 2 other unlinked type I diabetes susceptibility loci. The results from the CASPAR analysis suggest that conditioning of IDDM5 on IDDM1 and IDDM4, and of IDDM7 on IDDM1 and IDDM2 provides significant benefits for the genetic analysis of polygenic loci.

Identification and characterization of multiple osmotic response sequences in the human aldose reductase gene.

Aldose reductase (AR) has been implicated in osmoregulation in the kidney because it reduces glucose to sorbitol, which can serve as an osmolite. Under hyperosmotic stress, transcription of this gene is induced to increase the enzyme level. This mode of osmotic regulation of AR gene expression has been observed in a number of nonrenal cells as well, suggesting that this is a common response to hyperosmotic stress. We have identified a 132-base pair sequence approximately 1 kilobase pairs upstream of the transcription start site of the AR gene that enhances the transcription activity of the AR promoter as well as that of the SV40 promoter when the cells are under hyperosmotic stress. Within this 132-base pair sequence, there are three sequences that resemble TonE, the tonicity response element of the canine betaine transporter gene, and the osmotic response element of the rabbit AR gene, suggesting that the mechanism of osmotic regulation of gene expression in these animals is similar. However, our data indicate that cooperative interaction among the three TonE-like sequences in the human AR may be necessary for their enhancer function.

Analysis of candidate genes for susceptibility to type I diabetes: a case-control and family-association study of genes on chromosome 2q31-35.

Recent genome searches suggest a putative linkage of many loci to susceptibility to type I diabetes. The chromosome 2q31-35 region is reported to be linked to susceptibility to type I diabetes and is thought to contain several diabetes susceptibility loci. These candidate genes include the HOXD gene cluster, BETA2, CTLA4, CD28, IGFBP2, and IGFBP5. Association studies in populations and families are required to confirm and/or identify the actual susceptibility loci. We hereby report several previously unknown DNA polymorphisms for HOXD8, BETA2, and IGFBP5, which we have used along with previously known polymorphisms of HOXD8 and CTLA4 to test whether these candidate loci are the susceptibility genes on chromosome 2q31-35. Using a case-control design with a subsequent family-association approach to confirm associations, we find no evidence that these candidate genes are associated with susceptibility to type I diabetes.

The alrestatin double-decker: binding of two inhibitor molecules to human aldose reductase reveals a new specificity determinant.

It is generally expected that only one inhibitor molecule will bind to an enzyme active site. In fact, specific drug design theories depend upon this assumption. Here, we report the binding of two molecules of an inhibitor to the same active site which we observed in the 1.8 A resolution structure of the drug Alrestatin bound to a mutant of human aldose reductase. The two molecules of Alrestatin bind to the active site in a stacked arrangement (a double-decker). This stack positions the carboxylic acid of one drug molecule near the NADP+ cofactor at a previously determined anion binding site and the carboxylic acid of the second drug molecule near the carboxy-terminal tail of the enzyme. We propose that interactions of inhibitors with the carboxy-terminal loop of aldose reductase are critical for the development of inhibitors that are able to discriminate between aldose reductase and other members of the aldo-keto reductase superfamily. This finding suggests a new direction for the introduction of specificity to aldose reductase-targeted drugs.

The C-terminal loop of aldehyde reductase determines the substrate and inhibitor specificity.

Human aldehyde reductase has a preference for carboxyl group-containing negatively charged substrates. It belongs to the NADPH-dependent aldo-keto reductase superfamily whose members are in part distinguished by unique C-terminal loops. To probe the role of the C-terminal loops in determining substrate specificities in these enzymes, two arginine residues, Arg308 and Arg311, located in the C-terminal loop of aldehyde reductase, and not found in any other C-terminal loop, were replaced with alanine residues. The catalytic efficiency of the R311A mutant for aldehydes containing a carboxyl group is reduced 150-250-fold in comparison to that of the wild-type enzyme, while substrates not containing a negative charge are unaffected. The R311A mutant is also significantly less sensitive to inhibition by dicarboxylic acids, indicating that Arg311 interacts with one of the carboxyl groups. The inhibition pattern indicates that the other carboxyl group binds to the anion binding site formed by Tyr49, His112, and the nicotinamide moiety of NADP+. The correlation between inhibitor potency and the length of the dicarboxylic acid molecules suggests a distance of approximately 10 A between the amino group of Arg311 and the anion binding site in the aldehyde reductase molecule. The sensitivity of inhibition of the R311A mutant by several commercially available aldose reductase inhibitors (ARIs) was variable, with tolrestat and zopolrestat becoming more potent inhibitors (30- and 5-fold, respectively), while others remained the same or became less potent. The catalytic properties, substrate specificity, and susceptibility to inhibition of the R308A mutant remained similar to that of the wild-type enzyme. The data provide direct evidence for C-terminal loop participation in determining substrate and inhibitor specificity of aldo-keto reductases and specifically identifies Arg311 as the basis for the carboxyl-containing substrate preference of aldehyde reductase.

Characterization of the osmotic response element of the human aldose reductase gene promoter.

Aldose reductase (EC 1.1.1.21) catalyzes the NADPH-mediated conversion of glucose to sorbitol. The hyperglycemia of diabetes increases sorbitol production primarily through substrate availability and is thought to contribute to the pathogenesis of many diabetic complications. Increased sorbitol production can also occur at normoglycemic levels via rapid increases in aldose reductase transcription and expression, which have been shown to occur upon exposure of many cell types to hyperosmotic conditions. The induction of aldose reductase transcription and the accumulation of sorbitol, an organic osmolyte, have been shown to be part of the physiological osmoregulatory mechanism whereby renal tubular cells adjust to the intraluminal hyperosmolality during urinary concentration. Previously, to explore the mechanism regulating aldose reductase levels, we partially characterized the human aldose reductase gene promoter present in a 4.2-kb fragment upstream of the transcription initiation start site. A fragment (-192 to +31 bp) was shown to contain several elements that control the basal expression of the enzyme. In this study, we examined the entire 4.2-kb human AR gene promoter fragment by deletion mutagenesis and transfection studies for the presence of osmotic response enhancer elements. An 11-bp nucleotide sequence (TGGAAAATTAC) was located 3.7 kb upstream of the transcription initiation site that mediates hypertonicity-responsive enhancer activity. This osmotic response element (ORE) increased the expression of the chloramphenicol acetyltransferase reporter gene product 2-fold in transfected HepG2 cells exposed to hypertonic NaCl media as compared with isoosmotic media. A more distal homologous sequence is also described; however, this sequence has no osmotic enhancer activity in transfected cells. Specific ORE mutant constructs, gel shift, and DNA fragment competition studies confirm the nature of the element and identify specific nucleotides essential for enhancer activity. A plasmid construct containing three repeat OREs and a heterologous promoter increased expression 8-fold in isoosmotic media and an additional 4-fold when the transfected cells are subjected to hyperosmotic stress (total approximately 30-fold). These findings will permit future studies to identify the transcription factors involved in the normal regulatory response mechanism to hypertonicity and to identify whether and how this response is altered in a variety of pathologic states, including diabetes.

The search for IDDM susceptibility genes: the next generation.

Two human chromosomal regions, the HLA region on chromosome 6p2l and the insulin gene region on chromosome 11p15, have been investigated in detail for more than 10 years for the presence of IDDM susceptibility genes. Recent genome searches indicate the possible existence of many additional susceptibility genes in IDDM. The lengthy and protracted studies to prove the linkage and identity of the susceptibility genes in the HLA and insulin gene regions provide a perspective and background for understanding the complexities and time course for characterization of the putative additional IDDM susceptibility genes uncovered by genome searches.

Human aldose reductase: rate constants for a mechanism including interconversion of ternary complexes by recombinant wild-type enzyme.

We have used transient kinetic data for partial reactions of recombinant human aldose reductase and simulations of progress curves for D-xylose reduction with NADPH and for xylitol oxidation with NADP+ to estimate rate constants for the following mechanism at pH 8.0: E<-->E.NADPH<-->*E.NADPH<-->*E.NADPH.RCHO<-->*E.NADP+.RCH2OH <-->*E.NADP+<--> E.NADP+<-->E. The mechanism includes kinetically significant conformational changes of the two binary E.nucleotide complexes which correspond to the movement of a crystallographically identified nucleotide-clamping loop involved in nucleotide exchange. The magnitude of this conformational clamping is substantial and results in a 100- and 650-fold lowering of the nucleotide dissociation constant in the productive *E.NADPH and *E.NADP+ complexes, respectively. The transient reduction of D-xylose displays burst kinetics consistent with the conformational change preceding NADP+ release (*E.NADP+-->E.NADP+) as the rate-limiting step in the forward direction. The maximum burst rate also displays a large deuterium isotope effect (Dkburst = 3.6-4.1), indicating that hydride transfer contributes significantly to rate limitation of the sequence of steps up to and including release of xylitol. In the reverse reaction, no burst of NADPH production is observed because the hydride transfer step is overall 85% rate-limiting. Even so, the conformational change preceding NADPH release (*E.NADPH-->E.NADPH) still contributes 15% to the rate limitation for reaction in this direction. The estimated rate constant for hydride transfer from NADPH to the aldehyde of D-xylose (130 s-1) is only 5- to 10-fold lower than the corresponding rate constant determined for NADH-dependent carbonyl reduction catalyzed by lactate or liver alcohol dehydrogenase. Hydride transfer from alcohol to NADP+ (0.6 s-1), however, is at least 100- to 1000-fold slower than NAD(+)-dependent alcohol oxidation mediated by these two enzymes, resulting in a bound-state equilibrium constant for aldose reductase which greatly favors the forward reaction. The proposed kinetic model provides a basic set of rate constants for interpretation of kinetic results obtained with aldose reductase mutants generated for the purpose of examining structure-function relationships of different components of the native enzyme.

Human aldose reductase: subtle effects revealed by rapid kinetic studies of the C298A mutant enzyme.

Transient kinetic data for D-xylose reduction with NADPH and NADPD and for xylitol oxidation with NADP+ catalyzed by recombinant C298A mutant human aldose reductase at pH 8 have been used to obtain estimates for each of the rate constants in the complete reaction mechanism as outlined for the wild-type enzyme in the preceding paper (Grimshaw et al., 1995a). Analysis of the resulting kinetic model shows that the nearly 9-fold increase in Vxylose/Et for C298A mutant enzyme relative to wild-type human aldose reductase is due entirely to an 8.7-fold increase in the rate constant for the conformational change that converts the tight (Ki NADP+ = 0.14 microM) binary *E.NADP+ complex to the weak (Kd NADP+ = 6.8 microM) E.NADP+ complex from which NADP+ is released. Evaluation of the rate expressions derived from the kinetic model for the various steady-state kinetic parameters reveals that the 37-fold increase in Kxylose seen for C298A relative to wild-type aldose reductase is largely due to this same increase in the net rate of NADP+ release; the rate constant for xylose binding accounts for only a factor of 5.5. A similar 17-fold increase in the rate constant for the conformational change preceding NADPH release does not, however, result in any increase in Vxylitol/Et, because hydride transfer is largely rate-limiting for reaction in this direction.(ABSTRACT TRUNCATED AT 250 WORDS)

Human aldose reductase: pK of tyrosine 48 reveals the preferred ionization state for catalysis and inhibition.

Detailed analyses of the pH variation of kinetic parameters for the forward aldehyde reduction and reverse alcohol oxidation reactions mediated by recombinant human aldose reductase, for inhibitor binding, and for kinetic isotope effects on aldehyde reduction have revealed that the pK value for the active site acid-base catalyst group Tyr48 is quite sensitive to the oxidation state of the bound nucleotide (NADPH or NADP+) and to the presence or absence of the Cys298 sulfhydryl moiety. Thus, the Tyr48 residue of C298A mutant enzyme displays a pK value that ranges from 7.6 in the productive *E.NADP+ complex that binds and reacts with alcohols to 8.7 in the productive *E.NADPH complex that binds and reacts with aldehyde substrates. For wild-type enzyme, Tyr48 in the latter complex displays a lower pK value of about 8.25. Assignment of the pK values was facilitated by the recognition and quantitation of the degree of stickiness of several aldehyde substrates in the forward reaction. The unusual pH dependence for Valdehyde/Et and DValdehyde, which decrease roughly 20-fold through a wave and remain constant at high pH, respectively, is shown to arise from the pH-dependent decrease in the net rate of NADP+ release. The results described are fully consistent with the chemical mechanism for aldose reductase catalysis proposed previously (Bohren et al., 1994) and, furthermore, establish that binding of the spirohydantoin class of aldose reductase inhibitors, e.g., sorbinil, occurs via a reverse protonation scheme in which the ionized inhibitor binds preferentially to the *E.NADP+ complex with Tyr48 present as the protonated hydroxyl form. The latter finding allows us to propose a unified model for high-affinity aldose reductase inhibitor binding that focuses on the transition state-like nature of the *E-Tyr48-OH.NADP+.inhibitor- complex.

Mechanism of human aldehyde reductase: characterization of the active site pocket.

Human aldehyde reductase is a NADPH-dependent aldo-keto reductase that is closely related (65% identity) to aldose reductase, an enzyme involved in the pathogenesis of some diabetic and galactosemic complications. In aldose reductase, the active site residue Tyr48 is the proton donor in a hydrogen-bonding network involving residues Asp43/Lys77, while His110 directs the orientation of substrates in the active site pocket. Mutation of the homologous Tyr49 to phenylalamine or histidine (Y49F or Y49H) and of Lys79 to methionine (K79M) in aldehyde reductase yields inactive enzymes, indicating similar roles for these residues in the catalytic mechanism of aldehyde reductase. A H112Q mutant aldehyde reductase exhibited a substantial decrease in catalytic efficiency (kcat/Km) for hydrophilic (average 150-fold) and aromatic substrates (average 4200-fold) and 50-fold higher IC50 values for a variety of inhibitors than that of the wild-type enzyme. The data suggest that His112 plays a major role in determining the substrate specificity of aldehyde reductase, similar to that shown earlier for the homologous His110 in aldose reductase [Bohren, K. M., et. al. (1994) Biochemistry 33, 2021-2032]. Mutation of Ile298 or Val299 affected the kinetic parameters to a much lesser degree. Unlike native aldose reductase, which contains a thiol-sensitive Cys298, neither the I298C or V299C mutant exhibited any thiol sensitivity, suggesting a geometry of the active site pocket different from that in aldose reductase. Also different from aldose reductase, the detection of a significant primary deuterium isotope effect on kcat (1.48 +/- 0.02) shows that nucleotide exchange is only partially rate-limiting. Primary substrate and solvent deuterium isotope effects on the H112Q mutant suggest that hydride and proton transfers occur in two discrete steps with hydride transfer taking place first. Dissociation constants and spectroscopic and fluorimetric properties of nucleotide complexes with various mutants suggest that, in addition to Tyr49 and His112, Lys79 plays a hitherto unappreciated role in nucleotide binding. The mode of inhibition of aldehyde reductase by aldose reductase inhibitors (ARIs) is generally similar to that of aldose reductase and involves binding to the E:NADP+ complex, as shown by kinetic and direct inhibitor-binding experiments. The order of ARI potency was AL1576 (Ki = 60 nM) > tolrestat > ponalrestat > sorbinil > FK366 > zopolrestat > alrestatin (Ki = 148 microM). Our data on aldehyde reductase suggest that the active site pocket significantly differs from that of aldose reductase, possibly due to the participation of the C-terminal loop in its formation.

The HOXD8 locus (2q31) is linked to type I diabetes. Interaction with chromosome 6 and 11 disease susceptibility genes.

Type I diabetes susceptibility genes have been identified within the major histocompatibility complex (MHC) on chromosome 6p21.3 and near the VNTR/insulin region on chromosome 11p15.5. We have used polymorphic dinucleotide repeat markers to search the human genome for additional susceptibility genes in 162 type I diabetic families with an affected sibling pair. We report that an additional susceptibility gene is located on chromosome 2q31 near HOXD8 (P < 10(-5), maximum logarithm of odds score = 4.8) in an analysis of affected sibling pairs having specific human leukocyte antigen (HLA) and hypervariable nucleotide tandem repeat (VNTR)/insulin gene haplotypes (absence of high-risk HLA-DR3/4 haplotypes and presence of homozygous high-risk class I VNTR alleles). These results suggest the interaction of a minimum of three genes in the pathogenesis of type I diabetes in humans.

Expression, crystallization and preliminary crystallographic analysis of human carbonyl reductase.

The cDNA of human placental carbonyl reductase (EC 1.1.1.184), a member of the short-chain dehydrogenase family of enzymes, was introduced into the plasmid vector pET-11a and the enzyme overexpressed in Escherichia coli. Recombinant carbonyl reductase was purified to homogeneity, characterized physically and kinetically, and crystallized for X-ray diffraction study. The recombinant protein was indistinguishable from human tissue carbonyl reductase (CR8.5 form) on the basis of partial sequence analysis, substrate specificity, susceptibility to inhibitors and immunochemical analysis. Similar to the tissue enzyme which which occurs in multiple molecular forms thought to arise from autocatalytic modification by 2-oxocarboxylic acids, a second form of the recombinant enzyme was generated under bacterial growth conditions producing high pyruvate concentrations. Purified recombinant protein, which corresponds to the smallest, most basic tissue form (CR8.5), was crystallized against 20% polyethyleneglycol 6000 in 25 mM 2-(N-morpholino)ethanesulfonic acid buffer (Mes) at pH 6.0 using the hanging drop method. Crystals of human carbonyl reductase diffract to better than 3.0 A, and the diffraction symmetry is consistent with a crystal that belongs to the tetragonal space group P4(1)(3)2(1)2 with unit cell dimensions of a = b = 55 A, c = 175 A, alpha = beta = gamma = 90.0. The asymmetric unit contains one molecule of 30.2 kDa.

Mechanism of aldose reductase inhibition: binding of NADP+/NADPH and alrestatin-like inhibitors.

Aldose reductase enfolds NADP+/NADPH via a complex loop mechanism, with cofactor exchange being the rate-limiting step for the overall reaction. This study measures the binding constants of these cofactors in the wild-type enzyme, as well as a variety of active-site mutants (C298A, Y48H, Y48F, Y209F, H110A, W219A, and W20A), and seeks to identify the binding site and mechanism of the aldose reductase inhibitor alrestatin in the recombinant human enzyme. All the mutant enzymes, regardless of their enzyme activities, have normal or only slightly elevated coenzyme binding constants, suggesting a tertiary structure similar to that of the wild-type enzyme. Binding of alrestatin was detected by fluorescence assays, and by an ultrafiltration assay which measures the fraction of unbound alrestatin. Alrestatin binds preferentially to the enzyme/NADP+ complex, consistent with the steady-state inhibition pattern. Alrestatin binding and enzyme inhibition were abolished in the Tyr48 mutants Y48F and Y48H, implicating the positively charged anion well formed by the Asp43-/Lys77+/Tyr48(0)/NADP+ complex in inhibitor binding (Harrison et al., 1994; Bohren et al., 1994). The enzyme mutant W20A severely affected the inhibitory potencies of a variety of commercially developed aldose reductase inhibitors (zopolrestat, tolrestat, FK366, AL1576, alrestatin, ponalrestat, and sorbinil). Inhibition by citrate, previously shown to bind to the positively charged anion well, was not affected by this mutation. Inhibitors with flexible double aromatic ring systems (Zopolrestat, FK366, and ponalrestat) were less affected than others possessing a single aromatic ring system, suggesting that the additional pharmacophor ring system stabilizes the inhibitor by interaction at some other hydrophobic site.(ABSTRACT TRUNCATED AT 250 WORDS)