Genomic Cloning of hGSTP1*C, an Allelic Human Pi Class Glutathione S-Transferase Gene Variant and Functional Characterization of Its Retinoic Acid Response Elements*

The complete hGSTP1*C, consisting of 7 exons and 6 introns contained in 3116 base pairs, was isolated from a cosmid genomic library of a glioblastoma multiforme cell line. Although the promoter of hGSTP1*C was identical to that of the previously reported GST-Pi gene, several of its structural features had not been previously described. These include several nucleotide transitions and transversions. Transitions of A → G at +1404 and C → T at +2294 in exons 5 and 6, respectively, changed codons Ile104 to Val104 and Ala113 to Val113. The gene also contained a guanine insertion at +51 in the insulin response element in intron 1 and six tandem repeats and one palindromic retinoic acid response element (RARE) consensus half-sites, A(G)GG(T)TC(G)A in intron 5. Retinoic acid (RA) treatment increased GST-Pi gene expression significantly in MGR3 cells. GST-Pi gene constructs with and without RARE deletion were used to show the RARE requirement for GST-Pi gene induction by RA. The isolation of thehGSTP1*C gene and the evidence that it contains functional RAREs should contribute to a better understanding of the molecular regulation of the GST-Pi gene in human cells.

The complete hGSTP1*C, consisting of 7 exons and 6 introns contained in 3116 base pairs, was isolated from a cosmid genomic library of a glioblastoma multiforme cell line. Although the promoter of hGSTP1*C was identical to that of the previously reported GST-Pi gene, several of its structural features had not been previously described. These include several nucleotide transitions and transversions. Transitions of A 3 G at ؉1404 and C 3 T at ؉2294 in exons 5 and 6, respectively, changed codons Ile 104 to Val 104 and Ala 113 to Val 113 . The gene also contained a guanine insertion at ؉51 in the insulin response element in intron 1 and six tandem repeats and one palindromic retinoic acid response ele-

ment (RARE) consensus half-sites, A(G)GG(T)TC(G)A in intron 5. Retinoic acid (RA) treatment increased GST-Pi gene expression significantly in MGR3 cells. GST-Pi gene constructs with and without RARE deletion were used to show the RARE requirement for GST-Pi gene induction by RA. The isolation of the hGSTP1*C gene and the evidence that it contains functional RAREs should contribute to a better understanding of the molecular regulation of the GST-Pi gene in human cells.
The glutathione S-transferases ((GSTs) 1 EC 2.5.1.18) are best known for their ability to catalyze the neutrophilic attack of the sulfur atom of glutathione by a variety of electrophilic endogenous and exogenous compounds (1)(2)(3)(4)(5), producing watersoluble and often less reactive metabolites. Much interest is currently being focused on the Pi class GST because the gene is up-regulated during the early stages of oncogenesis and it is the most significantly overexpressed GST gene in many human tumors including gliomas (6 -17). A large number of studies have also shown the high levels of GST-Pi expression to be associated directly with tumor drug resistance and with poor patient survival (9, 12, 16 -19). In gliomas, the level of GST-Pi expression correlates positively with tumor grade (15,20), and in glioma cell lines, high GST-Pi expression has been correlated with increased resistance to 1,3-bis(2-chloroethyl)-1-nitrosourea (12).
Recently, we reported the isolation of three full-length cDNAs, hGSTP1*A, hGSTP1*B, and hGSTP1*C, corresponding to closely related GST-Pi mRNAs and encoding structurally and functionally different GST-Pi proteins (21). Of the three allelic GST-Pi gene variants, hGSTP*1C was shown to be expressed at a much higher frequency in malignant gliomas than in normal brain, placenta, and lymphocytes. To facilitate studies aimed at clarifying the molecular mechanisms underlying the overexpression of the GST-Pi gene in human gliomas, we undertook in this study the isolation and characterization of the hGSTP*1C gene from cells of a human glioblastoma multiforme cell line that expresses high levels of GST-Pi gene transcripts and protein and is resistant to 2-chloroethylnitrosoureas and cisplatin. We compared the isolated gene with the previously described GST-Pi gene and examined its regulation by all-trans retinoic acid (RA). Tumor Cells-The MGR3 human glioblastoma multiforme cell line was established in our laboratory from a primary specimen, as described previously (22). It is glial fibrillary acidic protein positive by immunocytochemistry, and the cells show the typical pleomorphism of neoplastic glial cells. The cell line is routinely maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and nonessential amino acids and had undergone 11 in vitro passages before being used in these studies. MCF7, obtained from the American Type Tissue Collection, Gaithersburg, MD, is a human mammary carcinoma cell line that, under normal physiological conditions, does not express detectable basal levels of GST-Pi transcripts or protein.

Materials-Restriction
Southern and Northern Blotting-These were performed using standard techniques (23). For Southern analysis, genomic DNA was extracted from MGR3 cells using the phenol/chloroform procedure (23), and for the Northern blots, total RNA was isolated with the acid guanidinium thiocyanate/phenol/chloroform procedure (24). PCR-Where indicated, PCR was performed in a DNA thermocycler (Perkin-Elmer). The 100-l reaction mixture contained 50 ng of Super-Cos-GST-Pi or other DNA template, 500 ng each of forward and reverse primers, 10 ϫ PCR buffer, 100 nM each dATP, dCTP, dGTP, and dTTP. 2.5 units of Amplitaq polymerase was added, and the mixture was overlaid with mineral oil. After 1 cycle of denaturing (95°C for 90 s), * This work was supported by NCI, National Institutes of Health Grant CA55835 and by a research award from the Kleberg Foundation. Part of this work was performed in fulfillment of the thesis requirement for the master of science degree in biochemistry for H.-W. Lo at the University of Texas School of Biomedical Sciences, Houston, Texas. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U21689.
Construction of Cosmid Genomic DNA Library, Subcloning, and Isolation of GST-pi Gene-High molecular weight genomic DNA from MGR3 cells was partially digested with Sau3A1, and the fragments were ligated into the SuperCos 1 vector. The constructs were packaged into bacteriophage lambda particles using the Gigapack II packaging system, and the resulting phage used to infect the bacterial host strain XL-1 Blue MR. After amplification, primary and secondary colonies were screened with a ␣-32 P-labeled full-length GST-Pi cDNA probe using standard techniques (23). After secondary and tertiary screenings, one positive clone designated SuperCos-GST-Pi was selected for further characterization. Restriction mapping of SuperCos-GST-Pi was performed using standard methods and subsequently verified by computer analysis. 0.5 g of SuperCos-GST-Pi DNA was digested with NotI and HindIII, electrophoresed in 1.2% agarose, and Southern-hybridized with a GST-Pi probe. GST-Pi-positive fragments were purified, ligated into NotI and HindIII sites of the pBluescript II phagemid vector, and transfected into competent XL-1 Blue cells. Overlapping GST-Pi fragments were amplified by PCR, the PCR products were electrophoresed in 1.2% agarose, and fragments of appropriate sizes were excised, purified, blunt-ended with 0.05 unit/pmol DNA Klenow enzyme, and ligated into EcoRV-digested and calf intestinal alkaline phosphatetreated pBluescript phagemid II. Competent XL-1 Blue cells were transfected with the vector constructs, and GST-Pi-positive clones were identified by Southern hybridization and used for sequencing.
DNA Sequencing and Structural Analysis-Nucleotide sequencing was performed with the 35 S-dideoxynucleotide chain termination method (25). Each gene fragment was sequenced in both directions, and each sequencing was performed twice to exclude PCR artifacts. Sequencing primers were designed from published GST-Pi gene sequences and from the sequence of the cosmid.
Analyses of the nucleotide sequence and structural organization of the isolated GST-Pi gene were performed using a commercial DNA sequence analysis package (Macmolly Tetra, Berlin, Germany). The sequence was compared with those of the previously described GST-Pi genes from the HPB-ALL (26) and the MCF7 (27) cell lines (GenBank TM accession numbers X0894 -6 and X08058, respectively). For further analysis of introns 5 and 6 of the isolated GST-Pi gene, a 1-kb DNA fragment containing intron 5 and a 450-bp fragment containing intron 6 as well as the regions flanking these introns were amplified using the primer pairs listed in Table I. The PCR products were purified and digested to confirm the presence of SpeI and AvaII restriction sites in introns 5 and 6, respectively, both of which had been predicted by nucleotide sequence analysis to be present in the cloned GST-Pi gene. Because of previous data indicating that the GST-Pi gene and its promoter could be regulated by retinoic acid (28,29), we analyzed the cloned GST-Pi gene for the presence of sequences homologous to known retinoic acid response element (RARE) consensus sequences (30).
GST-Pi Expression Vector Construction and Expression in Cos 1 Cells-The complete GST-Pi gene was obtained by ligating a 2.2-kb NotI/BamHI fragment to a 0.9-kb BamHI/KpnI fragment, both from SuperCos-GST-Pi, into the NotI/KpnI site of the pBK-CMV expression vector (Stratagene) in which eukaryotic expression is driven by the cytomegalovirus (CMV) immediate early promoter. The resulting GST-Pi expression construct, designated pGST-Pi-CMV, contained the entire 3.1-kb GST-Pi gene consisting of 131 bp of 5Ј promoter region, 109 bp of 3Ј-untranslated region including the polyadenylation signal, and 68 bp downstream of the polyadenylation signal.
pGST-Pi-CMV was transfected into exponentially growing Cos 1 cells using the calcium phosphate method (23). After 48 h, the cells were harvested, washed twice in phosphate-buffered saline, homogenized, and centrifuged at 20,000 ϫ g for 20 min at 4°C. Protein concentrations and total GST enzyme activity in the supernatants were determined as described previously (12,31), the latter using 1-chloro-2,4-dinitrobenzene as substrate. Specific GST-Pi protein content was determined by Western blotting as we had previously described (12).
RA Effect on GST-Pi Gene Expression in Glioma Cells-These studies were performed to examine the responsiveness of the GST-Pi gene to RA. Exponentially growing MGR3 cells from which the GST-Pi gene was isolated were treated with 1 M all-trans RA, and after 24 and 48 h of incubation at 37°C, total RNA and protein were extracted from control and RA-treated cells, and GST-Pi gene transcript and protein levels were determined by Northern and Western blotting, respectively, as we described earlier.

RA Effects on Full-length and RARE-deleted Constructs of GST-Pi Expression
Vectors-To further characterize the functionality of the RAREs in the cloned GST-Pi gene, the RA response of the expression vector, pGST-Pi-CMV, containing the full-length hGSTP1*C gene and of one, pGST-Pi-CMV-RARE(Ϫ), containing the hGSTP1*C gene from which the RARE region had been deleted, were examined. The latter vector was obtained by Bsu36I digestion of the pGST-Pi-CMV to release a 476-bp fragment containing all seven RARE half-sites. After bluntending with DNA Klenow enzyme, the linear plasmid was religated to create the circular expression vector.
Exponentially growing MCF7 cells were transfected with pGST-Pi-CMV-RARE(Ϫ), pGST-Pi-CMV, and parent pBK-CMV (without the GST-Pi gene). The cells were then treated with 1.0 M all-trans RA. Control cells transfected with the vectors were similarly set up but without RA treatment. Forty-eight hours later, the cells were harvested and examined for GST-Pi gene transcripts by Northern analysis. The rationale for using MCF7 instead of MGR3 for these studies is that the GST-Pi gene is transcriptionally silent in MCF7 cells and is not induced by RA. As such, any differences in the GST-Pi gene expression following RA treatment of transfected cells can be attributed to RA effects on the transfected vectors. In contrast, MGR3 cells express high basal GST-Pi levels, and RA treatment results in a significant increase in GST-Pi gene expression, thus making it difficult to distinguish the differential effects of RA on the RARE-deleted and -undeleted vectors after transfection.

RESULTS
Construction and Screening of Genomic Library-Approximately 10 6 colonies of the MGR3 SuperCos 1 genomic library were initially screened with the GST-Pi cDNA probe. Twenty positive colonies were subjected to secondary screening, after which two were selected for tertiary screening. One positive clone designated SuperCos-GST-Pi containing the intact GST-Pi gene was selected for further analysis.
Restriction Endonuclease Mapping of SuperCos-GST-Pi-The results of Southern analysis of NotI-and HindIII-digested SuperCos-GST-Pi with a GST-Pi cDNA probe were used to construct a simplified restriction map of the SuperCos-GST-Pi clone. The map showed the GST-Pi gene to be located between two NotI sites of the SuperCos-GST-Pi clone and to overlap two HindIII fragments. The entire gene was contained within a 2.1-kbNotI-HindIII fragment and an 11.5-kb HindIII fragment. This was subsequently verified by computer analysis of the final DNA sequence.
Nucleotide Sequence and Structural Analysis-The nucleotide sequence of the isolated GST-Pi gene is shown in Fig. 1. The sequenced region was 3116 bp in size and contained the entire GST-Pi gene, consisting of 7 exons and 6 introns located within nucleotides ϩ30 and ϩ2762. Each of the exon-intron boundaries is characterized by the AG and GT splicing signals at their 5Ј and 3Ј ends, respectively. The coding region of the isolated GST-Pi gene consists of 211 codons including the ATG initiation and TGA termination codons. The 3Ј noncoding region of the gene covers nucleotides ϩ2763 to ϩ2984 and includes the polyadenylation signal AATAAA at ϩ2818 to ϩ2823. The 5Ј-flanking region upstream of the first exon of the gene, i.e. the promoter region, contains five regulatory motifs, all of which have been previously reported (26,27). Relative to the transcription initiation site (26), these were a TATAA box located at Ϫ31 to Ϫ27, two Sp1 sites at Ϫ46 to Ϫ41 and Ϫ57 to Ϫ51, and an AP-1 site at Ϫ69 to Ϫ63. Embedded in the AP-1 site at Ϫ70 to Ϫ61 was an antioxidant response element (ARE) core consensus sequence. Table II summarizes the structural differences between the GST-Pi gene isolated from the MGR3 cell line and the previously reported GST-Pi gene from the MCF7 cell line (27). Two transitions, an A 3 G at ϩ1404 and a C 3 T at ϩ2294 changed codon 104 from ATC (Ile) to GTC (Val) and codon 113 from GCG (Ala) to GTG (Val), respectively. These findings confirmed the isolated gene to be hGSTP1*C, the full-length cDNA of which was recently isolated in our laboratory (21). A silent C 3 T nucleotide transition present at nucleotide ϩ2684 did not alter the amino acid serine encoded in the affected codon 184.

Comparison of Isolated GST-Pi Gene with Previously Described Human GST-Pi Genes-
In addition to the three nucleotide transitions in exons 5, 6, and 7, several intronic differences were also observed between the MGR3 and the MCF7 GST-Pi genes. A region of high homology with the core sequence (CCCGCCTC) of the insulin response element A, IRE-A (32) was observed at ϩ45 to ϩ52. This IRE differed from that previously described at the same location in the GST-Pi gene isolated from the MCF7 (27) by having a guanine insertion at ϩ51. The insertion created two CpG dinucleotides and a cleavage site (CG2CG) for several restriction endonucleases, including AccII, AciI, Bsp50I, BstUI, and MvnI. Additionally, nucleotide transitions in introns 5 and 6 created extra endonuclease cleavage sites in the MGR3 GST-Pi gene that are absent in the MCF7 gene. A G 3 A transition at nucleotide ϩ1968 created a SpeI site, A2CTAGT, in intron 5, which was confirmed by SpeI digestion of a 1-kb PCR product containing intron 5. Interestingly, only partial cleavage was observed upon SpeI digestion of the same DNA region from human lymphocytes and MCF7 cells, indicating the existance of polymorphism of the GST-Pi gene at this position. Additionally, two transitions of A3 G and C3 T at ϩ2557 and ϩ2559, respectively, created within intron 6 new AvaII cleavage sites G2GTCC that are not present in the MCF7 gene. AvaII digestion of a 450-bp DNA fragment amplified from MGR3 cells to cover this region of intron 6 showed the expected 400-and 50-bp fragments. In contrast, the same DNA fragment from normal human lymphocytes and MCF7 cells yielded these two AvaII cleavage products and, in addition, uncleaved DNA in the case of MCF7 and multiple restriction fragments in the case of lymphocytes. Fig. 2, and the Western blot analysis for GST-Pi protein in control Cos-1 cells and in Cos-1 cells 48 h after transfection with the pGST-Pi-CMV are shown in Fig. 2b. Densitometry of the Western blots showed a 2.9-fold-increased GST-Pi protein content in the transfected cells relative to controls. Bulk GST enzyme activity was 51.9 and 22.0 nmol/min/mg protein in transfected and control cells, respectively. The similar levels of increase in total GST activity and specific GST-Pi content indicate that the increase in GST enzyme activity was due, primarily, to the overexpressed GST-Pi protein.

Expression of Cloned GST-Pi Gene in Cos-1 Cells-The structure of the pGST-Pi-CMV expression vector is shown in
RAREs in Isolated GST-Pi Gene-RAREs are direct repeat regulatory motifs to which RA-RAR complexes bind and mediate transcription of RA-responsive genes (30, 34 -36). We report for the first time the presence of RARE sequences in the GST-Pi gene. These RARE motifs are located in intron 5 of the GST-Pi gene, in a region spanning nucleotides ϩ1521 to ϩ1644 and consisted of one palindromic and six direct repeats of RARE concensus half-sites arranged in tandem. Fig. 3 shows the region of the GST-Pi gene with the six direct repeats and one palindromic RARE half-sites, and Table III compares the consensus RAREs in the isolated GST-Pi gene with known RAREs in other selected genes.

RA Effects on Cellular GST-Pi Gene Expression and on Expression of RARE-containing and RARE-deleted hGST-Pi Genes in Expression Vectors-
The results of these studies designed to examine response of the GST-Pi gene to RA are summarized in Fig. 4, a and b. Northern analysis (Fig. 4a) showed a moderate but significant increase in the level of Cos-1 cells were transfected with pGST-Pi-CMV using the calcium phosphate method. After 48 h at 37°C and 5% CO 2 , the cells were harvested, homogenized, and centrifuged at 20,000 ϫ g. 25 g of protein extracts were electrophoresed in a 10% polyacrylamide gel, transferred to nylon membranes, and probed for GST-Pi protein by the chemiluminescence method. HP, GST, control and GST-II represent human placental GSTpi, protein from untreated cells and protein from pGST-pi-CMV transfected cells.

FIG. 3. Region of intron 5 of the GST-Pi gene showing RARE concensus half-sites arranged in tandem between ؉1521 and
؉1644. The first RARE is palindromic. The regions, designated RARE-1, RARE-2, and RARE-3 contain direct repeat RARE half-sites with the number of intervening nucleotides shown to be required for functionality.

DISCUSSION
The human GST-P1 gene locus has now been shown conclusively to be polymorphic (21). Of the three allelic gene variants, however, only one, hGSTP1*A, has been isolated (26,27). We describe here the cloning and the structural and functional characterization of another human GST-Pi gene, hGSTP1*C, that in preliminary studies has been shown to be expressed at an increased relative frequency in malignant gliomas compared with normal lymphocytes. The sequenced region of the isolated clone consisted of 3116 nucleotides and contained the complete GST-Pi gene consisting of seven exons and six introns and encoding 210 amino acids. Several of the structural features previously observed in the hGSTP1*A gene isolated independently by Cowell et al. (26) and by Morrow et al. (27) were present in the isolated hGSTP1*C gene. These include a TATAA box, two SP-1 sites, and one AP-1 site in the promoter region. An anti-oxidant response element, ARE, was also observed in the AP-1 site of the isolated hGSTP1*C gene. This ARE, also recently observed in the GSTP1*A gene (37), is identical to the ARE core sequence (GTGACTCAGC) of the human NADPH:quinone oxidoreductase gene (3,38) and has a high degree of homology to the ARE (GTGACAAAGC) in the rat GST-Ya gene (3,39).
Several important differences were observed between the GST-Pi gene described here and that previously reported (26,27). These include nucleotide transitions of A 3 G at ϩ1404  4. Effect of all-trans RA on GST-Pi gene expression in MGR3 cells. Cells were treated with 1 M RA, and after 24 and 48 h, were harvested. a, Northern blotting. Total RNA was extracted from the cells, purified, and electrophoresed at 7.5 g/lane in formaldehydeagarose gels. After transfer onto nylon membranes, hybridization was with a 32 P-labeled full-length GST-Pi cDNA. b, Western blot analysis. Cell extract proteins were electrophoresed in SDS-polyacrylamide gel electrophoresis, transferred onto polyvinylidene membranes and probed with a human GST-Pi-specific monoclonal antibody using the chemiluminescent technique. Both the mRNA and protein bands were quantitated by densitometry and used to determine the degree of GST-Pi gene induction by all-trans RA. GAPDH, glyceraldehyde-3phosphate dehydrogenase. and C 3 T at ϩ2294, which confirmed the identity of the gene to be hGSTP1*C (21). The changes of ATC (Ile) 3 GTC (Val) and GCG (Ala) 3 GTG (Val) in codons 104 and 113, respectively, caused by these transitions are in the electrophile binding (H-) site of the GST-Pi peptide, as has been previously reported (21). Other differences between the hGSTP1*C gene and the GST-Pi gene reported in other studies include transitions in introns 5 and 6 that resulted in altered restriction enzyme cleavage sites in the affected regions. These changes do not involve any known regulatory motifs, and as such may have no direct functional consequences, although they may be useful in the characterization of the GST-Pi gene in different individuals. In contrast, the guanine insertion at ϩ51 in the conserved IRE (CCCGCGTC) in intron 1 is of potential functional significance. This IRE differs from the IRE in the GST-Pi gene isolated from the MCF7 cell line (27) and the IRE in the human glyceraldehyde-3-phosphate dehydrogenase gene (33), which has a C 3 G transversion at ϩ51. The hGSTP1*C IRE is, however, identical to the IRE previously described in the GST-Pi gene isolated from the HPB-ALL cell line (26) and shown in a chloramphenicol acetyltransferase plasmid construct to be insulin-responsive (28). It remains to be established whether the altered IRE is a common feature of the GST-Pi gene in human glioma cells and/or whether the different IREs have differential insulin binding and ultimately could result in variable insulin responsiveness of the GST-Pi gene. A further significance of the guanine insertion in the hGSTP1*C IRE is that it created two CpG dinucleotides, potential sites of 5-cytosine methylation, which either directly or indirectly through altered insulin binding could result in differential regulation of the GST-Pi gene in different tumors.
In this study, we report for the first time the presence of RAREs in the human GST-Pi gene. The six direct repeat and one palindromic RARE half-sites were all located in intron 5 and are highly homologous to the concensus RARE half-site, 5Ј-AGGGTTCGA-3Ј, present and functional in other RA-responsive genes, such as those encoding RAR types ␣2, ␤2, and ␥2 (Table III; 30, 34, 40 -42) and the cellular retinoic acid and retinol binding proteins (34,43,44,54,55), the alcohol dehydrogenase gene (45), and the laminin B1 gene (46). Two pairs of the hGSTP1*C RARE half-sites were separated by the numbers of nucleotides, which in previous studies were shown to be essential for protein binding and function of the RAREs (30). We performed a number of experiments to demonstrate that the RAREs in the cloned GST-Pi gene are functional and are involved in RA-mediated induction of GST-Pi gene expression. We showed that transcription of the isolated GST-Pi gene could be induced with retinoic acid after transfection into tumor cells. Furthermore, following all-trans RA treatment, expression of the GST-Pi gene was increased by approximately 3-fold in MGR3 cells from which the gene was isolated. Using eukaryotic expression vector constructs containing the GST-Pi gene with and without deletion of the RARE region, we demonstrated that the RARE motif was required for the transcriptional activation of the GST-Pi gene in cells treated with all-trans RA.
It was of interest that the RARE region in the hGSTP1*C gene, although intronic, was functional and could regulate RA response of the GST-Pi gene. This contrasts with the majority of previously described RAREs, which are cis-acting and are located in the 5Ј/promoter regions of the genes they regulate. Functional regulatory motifs within introns, as we observed for both the IRE and the RAREs in the hGSTP1*C gene, are however not unusual in eukaryotic genes and have been shown in a variety of other genes, including oncogenes, tumor suppressor genes, and genes encoding growth factors and their receptors (47)(48)(49)(50)(51)(52)(53). A number of these intronic regulatory elements are located distant from the transcription initiation sites of the regulated genes. In the hGSTP1*C gene, the RARE region is approximately 1,500 bp from the promoter site, a distance similar to that of the tissue-specific regulatory element in the p53 gene (49) and those of the two negative regulatory elements of the PDGF-A chain gene (50) from their transcription start sites.
The observations in this study that RA increases GST-Pi expression in tumor cells transfected with the hGSTP1*C gene containing the RARE but not with a gene in which the RARE is deleted contrast with the results of a previous study that showed RA-mediated suppression of the GST-Pi promoter in a chloramphenicol acetyltransferase cDNA construct (28) and suggest a complex mechanism of cellular GST-Pi gene regulation based in part on the previously described antagonism between the AP-1 site and RA gene induction (56 -58). In this model of gene regulation, suppression of the GST-Pi gene occurs via competitive inhibition of the binding of AP-1-binding transcription factors such as jun and fos to the AP-1 site by the RA-RAR complex, as has also been suggested in a previous study (28). GST-Pi gene activation, on the other hand, will be mediated through the binding of the RA-RAR complex to RAREs in the GST-Pi gene. Such a model is also consistent with a previously proposed general mechanism for RA-mediated gene regulation by ligand-activated RARs (58). It is further supported by the fact that the activation of the GST-Pi gene by all-trans RA in MGR3 cells observed in this study is a delayed process similar to the late transcriptional induction of the laminin B1 gene by RA (59) and consistent with a mechanism involving RAR-ligand binding to RAREs.
The ability of RA to activate the GST-Pi gene has significant implications for both cancer prevention and chemotherapy. It suggests that part of the molecular basis for the cancer preventive action of long term administration of retinoic acid (60) may involve induction of GST-Pi gene expression. Given the important role that GST-Pi plays in the ability of tumor cells to inactivate anticancer agents, these results also suggest that RA pretreatment is likely to increase tumor resistance to alkylating agent chemotherapy. The isolation of a complete variant GST-Pi gene is an important step in the study of this gene and should facilitate future studies on its molecular regulation in both normal and neoplastic cells.