Cellular detoxification of tripeptidyl aldehydes by an aldo-keto reductase.

Calpain inhibitor I, N-acetyl-leucyl-leucyl-norleucinal (ALLN), a cell-permeable synthetic tripeptide with an aldehyde at its C terminus specifically inhibits the activity of cysteine proteases. Since the regulated degradation of 3-hydroxy-3-methylglutaryl-CoA reductase in Chinese hamster ovary (CHO) cells is blocked by ALLN and ALLN has a cytotoxic effect on cells, we attempted to isolate ALLN-resistant cells that overproduce an ALLN-sensitive protease(s). However, we obtained an ALLN-resistant cell line that overproduced P-glycoprotein (Sharma, R. C., Inoue, S., Roitelman, J., Schimke, R. T., and Simoni, R. D. (1992) J. Biol. Chem. 267, 5731-5734). To circumvent the multidrug resistance (MDR) phenotype during selection, we have stepwise selected an ALLN-resistant cell line of CHO cells in the presence of verapamil, a competitive inhibitor of P-glycoprotein. These non-MDR ALLN-resistant cells overexpress a 35-kDa protein and have increased aldo-keto reductase activity. Partial amino acid sequences of the 35-kDa protein are highly homologous to members of the aldo-keto reductase superfamily. The aldo-keto reductases are NADPH-dependent oxidoreductases and catalyze reduction of a wide range of carbonyl compounds such as aldehydes, sugars, and ketones. Our findings support the concept that a physiological function for aldo-keto reductases may be detoxification.

in the degradation of HMG-CoA reductase and HMGal, we attempted to isolate a cell line resistant to ALLN with the anticipation of obtaining cells expressing elevated levels of the ALLN-sensitive protease@). The first ALLN-resistant cell line we obtained showed a classical MDR phenotype, resulting from amplification of the mdrl gene ( 5 ) . These results suggested that the apparent resistance of these cells to ALLN is due to the outward pumping activity of Pglycoprotein (5). To circumvent the selection for the MDR phenotype, we adapted CHO cells to grow in increasing amounts of ALLN in the presence of verapamil, a competitive inhibitor of P-glycoprotein (6, 7). Verapamil at 5-20 ~L M reverses the MDR phenotype in MDR cell lines but has no effect on cell growth rate ( 5 ) . The ALLN-resistant (SI100) cells we have obtained in this study show no detectable elevation in P-glycoprotein. Rather, these cells express elevated levels of a 35-kDa protein that belongs to the aldo-keto reductase superfamily. The aldo-keto reductases are NADPHdependent oxidoreductases that catalyze the reduction of aldehydes and ketones to corresponding alcohols (8). Elevated levels of aldo-keto reductase in SI100 cells can inactivate ALLN by reducing its active aldehyde group to corresponding alcohol (2). This may be the mechanism of ALLN resistance in these cells.
Enzymes of the aldo-keto reductase superfamily such as aldose reductase (EC 1.1.1.21) and aldehyde reductase (EC 1.1.1.2) have been classified on the basis of a preferential, but not exclusive, substrate specificity (8). Recently, several cDNA clones of aldo-keto reductases have been isolated and their primary structures were determined. Human liver aldehyde reductase (9) has strong homology to aldose reductases of mouse vas deferens (lo), rat lens ( l l ) , and human placenta (9). The calculated molecular mass of these enzymes is about 35-37 kDa, similar to 35-kDa aldo-keto reductase obtained in the present study.
Our data suggest that resistance to a tripeptide with an aldehyde group at its C terminus results from overaccumulation of an aldo-keto reductase under conditions that do not allow overexpression of P-glycoprotein. Thus, elevated levels of an aldo-keto reductase may constitute a general mechanism for acquired resistance to toxic agents whose active form requires an aldehyde or keto group. These results support the concept that members of the aldo-keto reductase family serve as a detoxifying system (12).
Cell Culture-CHO-K1 cells were grown in minimal essential medium supplemented with 5% fetal calf serum at 37 "C in a 5% CO:, atmosphere. For stepwise selection to ALLN resistance, exponentially growing cells were treated with 10 p~ ALLN in the presence of 20 p~ verapamil. The medium was changed every 2-3 days for 1-2 weeks, at which time cells were trypsinized and replated in the same medium. After about 1 month of growth, cells became resistant to 10 p~ ALLN in the presence of verapamil. The concentration of ALLN was increased in a stepwise fashion (at 10-20 p M increments) up to 100 p~ ALLN, always in the presence of 20 JIM verapamil, a process that required about 3-4 months. These cells are designated SIlOO cells. In a similar manner, we selected cells which grow in 260 p M ALLN, designated SI260 cells. Cytotoxicity assays by the clonogenic method were performed as described (5). Transfection of wild-type CHO and SIlOO cells with the gene for HMGal was carried out as described (13), and G418 at 250 pg/ml was added to the medium to maintain a stable HMGal phenotype. These cells are designated CHO-HMGal and SI100-HMGal cells, respectively.
HMGal Assay-Activity of HMGal was determined as described by Skalnik et al. (13). Specific activity is expressed as AIzo.,/h/mg of protein.
Measurement of P-glycoprotein-Amounts of P-glycoprotein was determined as described by Sharma et al. (5).
Preparation of Cell Extracts-Unless specified otherwise, the following procedures were carried out at 4 "C. About 100 mg of cells (1-2 X 108 cells) were homogenized by 40 strokes in a Dounce homogenizer in 500 pl of 100 mM sodium phosphate, pH 7.4, containing 0.5 mM EDTA and 2 mM 8-mercaptoethanol. The homogenate was centrifuged at 1,000 X g for 10 min. The supernatant (total cell extract) was centrifuged at 130,000 X g for 30 min. The resultant supernatant was removed and is designated "cytosolic fraction." The pellet was resuspended in 200 pl of the same buffer, sonicated 3 times for 10 s, and used as a "membrane fraction." Both cytosolic and membrane fractions were dialyzed against 500 ml of 10 mM sodium phosphate, pH 7.4, containing 0.5 mM EDTA and 2 mM P-mercap-g 1 8 \ i  "ALLM is not soluble in aqueous solution at a concentration higher than 500 p~. toethanol. protease inhibitors (Table  I).
ALLM is a synthetic tripeptide Assay ofAldo-Keto Reductase-Reaction mixture, in a total volume containing an aldehyde group at its terminus (l). The of 1.0 ml, consisted of 100 mM sodium phosphate, pH 7.0, 10-50 pl of substrates as indicated in Table 111. The reaction was carried out at and SI100 cells were also at least 3-fold more resistant to room temDerature and was followed sDectroDhotometrical~v bv re-ALLM than CHO cells. However. with other cvsteine Drotease cytosolic fraction of cell extracts, 0.1 mM NADPH, and various cytotoxicity of ALLM to CHO cells is less than that of ALLN, cording the decrease in absorbance at 340 nm. Blanks withbut'substrate or without enzyme were routinely included to correct for nonspecific oxidation of NADPH. Gel E&ctrophoresis-5-10% SDS-PAGE was performed according to Laemmli (14). Isoelectric focusing or nonequilibrium pH gradient electrophoresis was carried out as described by O'Farrell (15). Gels were stained with Coomassie Brilliant Blue R-250.
Amino Acid Analyses-Proteins were separated by two-dimensional electrophoresis, blotted onto a polyvinylidene difluoride membrane (Millipore Corp.), and stained with Coomassie Brilliant Blue R-250. The protein spots were cut out and used for amino acid analysis and protein sequencing. Protein was hydrolyzed in 6 N HC1, and amino acid composition analysis was performed on a Beckman 7300 amino acid analyzer using a ninhydrin-based detector system. For amino acid sequencing, the blotted protein was digested with 5% trypsin in 100 mM Tris-C1, pH 8.5,2 M urea for 24 h at 37 "C. Peptides were separated on an Applied Biosystems 140A high performance liquid chromatograph equipped with a Brownlee CIS column (0.21 X 20 cm), using a linear gradient of 0-40% acetonitrile in 0.1% trifluoroacetic acid over 45 min at a flow rate of 200 pllmin. Absorption was monitored at 210 nm. Isolated peptides were sequenced on an Applied Biosystems 473 protein sequenator in the presence of 3 mg of Polybrene.

RESULTS AND DISCUSSION
Characterization of SI100 Cells-A cell line resistant to ALLN was selected in the presence of verapamil, as described under "Experimental Procedures." These SIlOO cells are 30fold more resistant to ALLN than the parental CHO cells ( Fig. 1 and Table I). Since ALLN is a specific inhibitor of cysteine proteases such as calpains (Ca2+-dependent neutral cysteine proteases) and lysosomal cathepsins (I), we determined whether SIlOO cells are also resistant to other cysteine inhibitors (calpeptin and E-64-d), SIlOO and CHO cells showed similar toxicities (Table I). Calpeptin is a dipeptide containing an aldehyde group at its C terminus (16, 17), and E-64-d is the cell-permeable inhibitor derived from E-64 and does not contain an aldehyde group (2,18). These results show that the resistance of SIlOO cells to cysteine protease inhibitors is specific for ALLN and ALLM.
The ALLN-resistant cells that we have previously obtained exhibit the classical multidrug resistance phenotype ( 5 ) . In this study, we isolated the ALLN-resistant cells in the presence of verapamil to prevent selection of the MDR phenotype, since verapamil reverses the MDR phenotype as a competitive inhibitor of drug transport by P-glycoprotein (6). As shown in Table I, SIlOO cells have a similar sensitivity to pleiotropic drugs (doxorubicin, etoposide, colchicine) as do CHO cells. In addition, SIlOO and CHO cells had the same amounts of Pglycoprotein (data not shown) as measured by quantitative immunoblotting using the C219 anti-P170 monoclonal antibody ( 5 ) . These results confirm that SIlOO cells do not display the classical MDR phenotype.
A Cytosolic 35-kDa Protein Is Highly Overproduced in SIlOO Cells-When proteins in total cell extracts are separated by SDS-PAGE, we found that the amount of a 35-kDa protein is dramatically increased in SIlOO cells (Fig. 2). This 35-kDa protein is cytosolic, and its amount is still higher in SI260 cells, which are selected in the presence of 260 ~L M ALLN with 20 ~L M verapamil. Two-dimensional gel electrophoresis of total cell extracts shows that only one protein band is clearly overproduced in SIlOO cells (Fig. 3). After blotting to a polyvinylidene difluoride membrane, we cut out the 35-kDa 5896 Detoxification by Aldo-Keto Reductase 36" -. protein spot, and amino acid analysis of this protein was performed. Since its N-terminal amino acid was blocked, the protein was digested with trypsin, and the amino acid sequences of isolated peptides were determined. The amino acid sequences of the two peptides obtained are Ile-Leu-Asn-Lys-Pro-Gly-Leu and Pro-Val-Tyr-Asn-Gln-Val-Glu-X-X-Pro-Tyr-Leu, respectively. A search of current protein data bases revealed that these two sequences have significant homology with the aldo-keto reductase superfamily (9). An alignment of the amino acid sequence of the peptides of the 35-kDa protein with that of several members of the superfamily is shown in Fig. 4. The comparison shows that the peptide sequences of 35-kDa protein are highly homologous to chlordecone reductase (19), aldose reductase of several species (9l l ) , prostaglandin F synthase (20), and 3a-hydroxysteroid dehydrogenase, and less homologous to aldehyde reductase (9). All these enzymes are members of a large family of proteins capable of reducing carbonyl groups on a wide variety of compounds, including sugars, aldehyde metabolites, steroid hormones, as well as xenobiotic aldehydes and ketones (8). These proteins are cytosolic, NADPH-dependent oxidoreductases and have a similar molecular mass of 35-37 kDa (8). The amino acid composition of the 35-kDa protein is shown in Table 11, which indicates a strong resemblance to those of aldose and aldehyde reductases. The PI value of 35-kDa protein on isoelectric focusing gel electrophoresis (PI = 6.5, data not shown) is also similar to the PI value of aldose reductases of human (9), mouse (IO), rat (11), and aldehyde reductase of human (9) as estimated from their amino acid compositions. These results suggest that the 35-kDa protein overproduced in SIlOO cells is an aldo-keto reductase.
Aldo-Keto Reductase Actiuity of Extracts from CHO and SIlOO Cells-The specific activity of aldo-keto reductase of the cytosolic fraction of cells was determined by using several compounds as substrates (Table 111). When D-XylOSe or DLglyceraldehyde is used as a substrate, the specific activity of aldo-keto reductase in SIlOO cells is 4-5-fold higher than that of CHO cells, respectively. When ALLN, ALLM, or calpeptin is used as a substrate, the specific activity of reductase in SIlOO cells is also 5-15-fold higher than that of CHO cells. E-64, a non-aldehyde protease inhibitor, is not a substrate. These results show that ALLN, ALLM, and calpeptin, all having an aldehyde group, are reduced by the cytosolic fraction of these cells in the presence of NADPH. As shown Table  111, SI260 cells exhibit higher aldo-keto reductase activity than SIlOO cells, indicating that the specific activity seems to be roughly in proportion to the amount of 35-kDa protein in

Aldose reductase R I L N K P G L K Y K P A V N O I E C H P V L T O '~' (ral lens)
Aldose reductase (human placenta)

(rat liver)
Aldehyde reductase (human liver)  His 3.9 2.9 2.8 "Amino acid composition of 35-kDa protein was determined as described under "Experimental Procedures" and presented as apparent mol %, since Trp and Cys were not determined (ND).

I L N K P G L P V T N O V E X X P Y L M l L N K P G L K Y K P V C N O V E C H P V L N O t W R L L N K P G L K H K P V T N O I E S H P V L T O '~' M I L N K P G L K Y K P A V N O I E C H P Y L T O '~' K I L N K P G L K Y K P V C N O V E C H P
Calculated from the deduced amino acid sequences of mouse aldose reductase (10) and human aldehyde reductase (9), respectively.

TABLE 111
Specific activity of aldo-keto reductase Specific activity of NADPH-dependent aldo-keto reductase of cytosolic fraction of cell extracts was determined spectrophotometrically. The values are the means of three independent experiments. When NADH was used instead of NADPH, no activity was detected (<0.05 nmol/min/mg of protein).

Substrate
Specific activity CHO SI100  (Fig. 2). Membrane fractions of CHO, SI100, and SI260 cells show no significant activity (c0.05 nmol/min/mg of protein). When we used NADH instead of NADPH, no activity was detected in these cell extracts against all substrates tested. 1 mM valproate, an aldehyde reductase inhibitor (21), partially inhibits all these activities (20-40% inhibition in each case). Since aldo-keto reductases have a broad substrate specificity and their primary structures are highly homologous, we do not know to which member of this reductase group the 35-kDa protein belongs. These results indicate that the ALLN-resistant phenotype of SIlOO and SI260 cells is due to the elevated level of aldo-keto reductase enzyme, which reduces the toxic aldehyde group of ALLN to a nontoxic alcohol (22). Functional Assay of Aldo-Keto Reductase in Vivo-The degradation of HMGal in CHO cells is accelerated by exogenous mevalonate, and its degradation is blocked by ALLN in uiuo (3). If ALLN is inactivated by aldo-keto reductase in SIlOO cells, it would be predicted that a higher amount of ALLN is required to block the degradation of HMGal. To test this possibility, stable transfectants of HMGal gene were isolated, and the effect of ALLN on the mevalonate-accelerated degradation of HMGal was determined. Fig. 5 shows that  " ALLM is not soluble in aqueous solution at a concentration higher than 500 p~. the activity of HMGal in both CHO and SIlOO cells decreases to about 20% of control when 20 mM mevalonate is added to cells. This decrease in activity is due to accelerated degradation of HMGal (4, 13, 23, 24). Addition of ALLN to CHO cells causes a dose-dependent increase in HMGal activity, indicating that the mevalonate-accelerated degradation is inhibited by ALLN with an ICs0 of approximately 50 p~ (Table  IV). In the SIlOO cells, however, the inhibitory effect of ALLN on the degradation of HMGal is markedly less than that of CHO cells. The ICso of ALLN in SIlOO cells is about 300 PM. This higher IC,, for inhibition of the degradation of HMGal reflects the inactivation of ALLN in SIlOO cells. Since the aldehyde group of ALLN is crucial for the inhibition of cysteine proteases (2), this result is consistent with the elevated level of aldo-keto reductase in SIlOO cells. Essentially the same effect of ALLM on the degradation of HMGal was observed. Calpeptin and E-64-d also inhibit the mevalonateaccelerated degradation of HMGal in CHO cells with ICs0 of 150 and 400 p M , respectively (Table IV). However, these inhibitory effects do not differ significantly between sensitive and resistant cells even though calpeptin has an aldehyde group and is capable of being a substrate of aldo-keto reductase (see Table 111). Similarly, calpeptin is not differentially cytotoxic to sensitive CHO cells compared with SIlOO cells (Table I). One possible reason to explain the lack of an effect of calpeptin either on cytotoxicity or HMGal degradation is that it is efficiently reduced to the inactive form by aldo-keto reductase even in CHO cells. It is also possible that the cell-The existence Of mechanisms Of resistance to a 20. Watanabe, K., Fujii, Y., Nakayama, K., Ohkubo, H., Kuramitsu, S., Kaga-

1080-1087
single agent appears to be a common theme in cultured animal