INTRODUCTION

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 (1-5), producing water-soluble 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).

EXPERIMENTAL PROCEDURES

Restriction endonucleases, Klenow enzyme, and T4 DNA ligase were purchased from Boehringer Mannheim. Proteinase K, RNase A, and all-trans RA were from Sigma. SuperCos 1 cosmid vector, Bluescript phagemid II, Gigapack II packaging system, calf intestinal alkaline phosphate, pBK-CMV expression vector, and the methylstyrylbenzene mammalian transfection kit were purchased from Stratagene, La Jolla, CA. [³⁵S]dATP and [-³²P]dCTP were purchased from Amersham Life Science, Inc. Taq DNA polymerase was purchased from Perkin-Elmer. Dulbecco's modified Eagle's medium and fetal calf serum were from Life Technology, Inc. T7 DNA Sequenase 2.0 dideoxy DNA sequencing kit was purchased from U. S. Biochemical Corp.

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.

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).

Where indicated, PCR was performed in a DNA thermocycler (Perkin-Elmer). The 100-µl reaction mixture contained 50 ng of SuperCos-GST-Pi or other DNA template, 500 ng each of forward and reverse primers, 10 × 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), annealing (55 °C for 2 min), and chain extension (72 °C for 3 min), 35 cycles of 95 °C (1 min), 55 °C (2 min), and 72 °C (3 min) were performed.

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 -³²P-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 µg 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 phosphate-treated 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.

Nucleotide sequencing was performed with the ³⁵S-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).

Table I. Primers for PCR amplification of overlapping GST-Pi DNA fragments from SuperCos-GST-Pi clone and of specific regions of GST-Pi gene P1 and P2 represent forward and reverse primers, respectively. Amplified region of GST-Pi gene Amplification primers Fragment size (approx.) bp Exons 2-3 P1: CGCAAGCTTCGCCACCATGCCGCCCTAC 430 P2: GGAGGCTTTGAGTGAGCCCTC Exons 3-6 P1: AGATCAGGGCCAGAGCTGGAAG 1,700 P2: CTGGTTCTGGGACAGGGTCTC Exon 7-poly A site P1: CTCTGGTCTAGAGGAAGCGA 440 P2: TCTTCCTCTTCTAGTTTGTGAGG Exons 2-7 P1: TCTTTGTTCGGACCATGCCGCCC 2,700 P2: CAGAGTCCCCCCAACCCTCACTGTTT Intron 5 P1:CAGCCCTGGTGGACATGGTGAATGAC 1,000 P2:CTGGTTCTGGGACAGCAGCTC Intron 6 P1:TGGCAGCTGAAGTGGACAGGATT 450 P2:GATCAGCAGCAAGTCCAGCAG

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 5 promoter region, 109 bp of 3-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 × 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).

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 µM 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.

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(), 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 blunt-ending 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(), pGST-Pi-CMV, and parent pBK-CMV (without the GST-Pi gene). The cells were then treated with 1.0 µM 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

Approximately 10⁶ 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.

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.

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 +30 and +2762. Each of the exon-intron boundaries is characterized by the AG and GT splicing signals at their 5 and 3 ends, respectively. The coding region of the isolated GST-Pi gene consists of 211 codons including the ATG initiation and TGA termination codons. The 3 noncoding region of the gene covers nucleotides +2763 to +2984 and includes the polyadenylation signal AATAAA at +2818 to +2823. The 5-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 31 to 27, two Sp1 sites at 46 to 41 and 57 to 51, and an AP-1 site at 69 to 63. Embedded in the AP-1 site at 70 to 61 was an antioxidant response element (ARE) core consensus sequence.

[View Larger Version of this Image (110K GIF file)]

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  G at +1404 and a C  T at +2294 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  T nucleotide transition present at nucleotide +2684 did not alter the amino acid serine encoded in the affected codon 184. 

Table II. Summary of structural differences between GST-Pi gene isolated from MGR3 glioblastoma multiform cell line (GenBankTM accession number U21689) and hGSTP1*A, previously isolated from the MCF7 cell line (GenBankTM accession number X08058) Nucleotide alterations and deletions are in boldface. Cell line Intron/exon (nucleotides) Sequence Structural change, modified endonuclease cleavage site MGR3 Intron 1 (+51) CGCGTC Guanine insertion; MCF7 CGC TC CGCG: AccII, AciI, Bsp50I, BstUI, MvnI MGR3 Intron 2 (+556) TCC A  C transversion MCF7 TAC No known endonuclease site modification MGR3 Intron 5 (+1,968) ACTAG G  A transition; MCF7 ACTGG ACTAG: MaeI, RmaI; CTAG: SpeI MGR3 Intron 6 CCTGGTCC A  G and C  T transitions; MCF7 (+2,557, +2,559) CCTAGCCC  CCWGG: ApyI, AsuI; CCWGG: EcoRII GGWCC: AvaII, Bme18I GGW: SinI; CTAG: MaeI, RmaI MGR3 Exon 5 (+1,404) GTC A  G transition MCF7 ATC ATC (Ile)  GTC (Val) at codon 104  MGR3 Exon 6 (+2,294) GTG C  T transition MCF7 GCG GCC (Ala)  GTC (Val) at codon 113  MGR3 Exon 7 (+2,684) AGC T  C transition (silent polymorphism) MCF7 AGT AGT (Ser)  AGC (Ser) at codon 184

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 +45 to +52. 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 +51. The insertion created two CpG dinucleotides and a cleavage site (CGCG) 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  A transition at nucleotide +1968 created a SpeI site, ACTAGT, 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 AG and CT at +2557 and +2559, respectively, created within intron 6 new AvaII cleavage sites GGTCC 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.

The structure of the pGST-Pi-CMV expression vector is shown in 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.

[View Larger Version of this Image (26K GIF file)]

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 +1521 to +1644 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.

[View Larger Version of this Image (19K GIF file)]

Table III. Comparison of RARE consensus sequences in human GST-Pi gene with RAREs in other genes The RARE half-sites are in boldface. Spacer regions with greater than six nucleotides are designated by the number of nucleotides. N, nucleotides. Gene RARE consensus half-site sequences GST-Pi TGACCC CTTCTT GGGTCA 13N GGGTCA GCTCT GGGCCA 15N GGGACA 26N GGGACA 17N GGGTCA RAR2 GGGTCA TTCAG AGTTCA RAR2 GGTTCA CCGAA AGGTGA RAR2 GGGTCA GGAGG AGGTGA ApoAI GGGTCA AG GGTTCA ApoCIII TGGGCA A AGGTCA Phosphoenalpyruvate carboxykinase CGGCCA A AGGTCA Oxytocin GGGTCA AGGTCA Cellular retinol binding protein I AGGTCA AA AAGTCA Cellular retinol binding protein II AGGTCA C AGGTCA kC AGGTCA C AGTTCA Laminin B1 AGGTCA GCT TAGGTTA 14N GGGTCA ADH3 GGGTCA TTCAG AGTTCA 11N GGGTCA Oct-3/4 GGGCCA G AGGTCA AGGCTA Oct-4 GGGCCA G AGGTCA 28N AGGTGA 10N AGGTGA

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 GST-Pi gene transcripts in MGR3 cells exposed to 1 µM all-trans retinoic acid over 24 and 48 h, with a maximum of a 3.4-fold increase relative to control levels at 48 h. The Western blot analysis (Fig. 4b) showed a similar increase in GST-Pi protein at 24 and 48 h after exposure to 1 µM all-trans-retinoic acid. After 48 h, GST-Pi protein content of RA-treated cells increased by 4.2-fold relative to untreated controls.

[View Larger Version of this Image (32K GIF file)]

To examine whether the RA response of the hGSTP1*C gene was mediated via the RARE in the gene, pBK-CMV control vector, and the expression vector constructs, pGST-Pi-CMV and pGST-Pi-CMV-RARE(), with and without the RARE region, respectively, were transfected into cells of the MCF7 cell line, which under normal conditions does not express detectable levels of GST-Pi gene transcripts or protein. The results, summarized in the Northern blots in Fig. 5, show no detectable GST-Pi gene transcripts in tumor cells transfected with the control, pBK-CMV vector, either without (lane 1) and with a 48-h exposure to 1 µM RA (lane 2). In contrast, a modest level of GST-Pi gene transcripts was observed in cells transfected with the pGST-Pi-CMV vector (lane 3), which increased by 3.6-fold after exposure of the cells to 1 µM RA for 48 h (lane 4). Cells transfected with the pGST-Pi-CMV-RARE(), in which the RARE sequence had been deleted from the hGSTP1*C gene, showed the same levels of GST-Pi transcripts without (lane 5) and with (lane 6) RA exposure.

[View Larger Version of this Image (21K GIF file)]

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  G at +1404 and C T at +2294, which confirmed the identity of the gene to be hGSTP1*C (21). The changes of ATC (Ile)  GTC (Val) and GCG (Ala)  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 +51 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  G transversion at +51. 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, 5-AGGGTTCGA-3, 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 5/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 (47-53). 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.