Exogenous 17β-Estradiol Blocks Alpha and Mu but Not Pi Class Glutathione S-Transferase Immunoreactivity in Epithelium of Syrian Hamster Vas Deferens

Abstract

Members of the glutathione S-transferase (GST) family of detoxification enzymes play a role in chemotherapy resistance in certain cancers but have not been directly implicated as agents whose absence may predispose tissues to hormonally induced tumorigenesis. Here we report the development of a polyclonal antiserum to a hamster mu class GST, and immunohistochemical analysis of alpha, mu, and pi class GSTs to study the effects of hormone treatment on their expression in reproductive tract tissues of male golden Syrian hamsters. These animals develop leiomyosarcomas in the vas deferens after treatment with testosterone propionate (TP) and 17β-estradiol (E2). High levels of all three GST classes were detected throughout the reproductive tract epithelium of control animals. In 100% of the experimental animals, 4 weeks of treatment either with E2 alone, or with E2 plus TP promoted a complete loss of immunostaining for alpha and mu class GSTs, but not for pi class GSTs, only in the epithelial lining of the vas deferens. In contrast, treatment with TP alone resulted in moderate hyperplasia of smooth muscle in the proximal vas deferens, with no cellular atypia and no changes in immunoreactivity of any of the GST classes. The consistent and site-specific nature of these results strongly suggests a functional role for GSTs in hormonally induced carcinogenic process.

Treatment of golden Syrian hamsters for 8-9 months with testosterone propionate (TP), in combination with diethylstibesterol or 17β-estradiol (E2), induces leiomyosarcomas in the uterus and vas deferens at a frequency of 100% (Kirkman and Algard 1965). Withdrawal of both TP and E2 after treatment durations sufficient to induce tumors (8-9 months) results in tumor regression, whereas withdrawal of either hormone alone results in maintenance of tumor cells but does not stimulate further growth (Kirkman and Algard 1965). In contrast, treatment with TP alone produces a thickened, mildly hyperplastic vas deferens (Norris et al. 1992), and E2 treatment alone results in renal neoplasms, fatal within 6 months (Oberley et al. 1991).

The molecular mechanism(s) involved in androgen/estrogen-induced tumor formation have not yet been fully characterized. In studies examining the effects of a variety of hormone treatment protocols on a ductus deferens tumor cell line (DDT1) derived from explanted tumor tissue (Norris and Kohler 1976), we found that DDT1 cells overexpress a mu class glutathione S-transferase (GSTmu) as a secondary response to glucocorticoid treatment (Norris et al. 1991). Furthermore, rapidly cycling DDT1 cells express GSTmu at lower levels than quiescent cells (Norris et al. 1992), suggesting that proliferating tumor cells in vivo may express GSTmu at lower levels. These observations led us to consider the possible involvement of GSTmu in the carcinogenic process(es).

The glutathione S-transferase (GST) superfamily is a group of cytosolic and membrane-bound detoxification enzymes. In mammalian tissues, alpha, mu, and pi are the most prevalent classes. There are seven subunits (two alpha, four mu, and one pi) in Syrian hamsters, which homo- or heterodimerize to form two alpha class, five mu class, and one pi class enzyme. These eight dimeric isoenzymes are expressed in a tissue-specific pattern; specifically, all three classes are found in the kidney, whereas the liver contains only alpha and mu and the pancreas expresses pi and trace amounts of alpha (Bogaards et al. 1992). Tissue levels of GSTs are regulated by naturally occurring xenobiotics, enzyme substrates, and reactive oxygen species. GSTs catalyze the coupling of glutathione to potentially genotoxic and/or xenobiotic electrophiles, leaving them more hydrophilic, which leads to decreased partitioning in the lipid bilayer (reviewed in Hayes and Pulford 1995). This molecular “flag” also marks the substrates for removal by ATP-dependent pumps specific for glutathione conjugates (Ishikawa 1992).

GSTs often are upregulated in tumors with increased activity towards chemotherapeutic drugs, such as vinblastine and adriamycin, promoting tumor resistance (reviewed in Hayes and Pulford 1995). Despite this well-established role in certain drug-resistant tumors, these enzymes have not yet been characterized as functional agents in hormonal carcinogenesis, which is the focus of this study. The results of the cell culture experiments described earlier led to the hypothesis that proliferating smooth muscle cells stimulated by hormone treatment in vivo express GSTmu at lower levels and that the loss of this protective enzyme leaves the cells vulnerable to the genotoxic insult(s) leading to the transformed phenotype. To test this hypothesis, we treated golden Syrian hamsters with TP, E2, or both hormones simultaneously, and performed immunohistochemical analysis on reproductive tract tissues to examine the changes in expression of GSTmu in situ. Because GSTs display class-specific reactivities towards various substrates and may have overlapping enzymatic activities and expression patterns in some tissues, we also performed immunohistochemical analysis for alpha and pi class GSTs in this model of hormonally induced tumorigenesis.

Materials and Methods

GSTmu Isolation and Antiserum Production

DDT1 cells were cultured in DME/F12 medium (Gibco BRL; Grand Island, NY) supplemented with 10% bovine calf serum (Hyclone; Logan, UT) and a 1% antibiotic/antimycotic solution containing penicillin, streptomycin, and amphotericin (Gibco BRL). At maximal density (approximately 500,000 cells/ml) the cells were treated with 10⁻⁷ M triamcinolone acetonide (TA) for 24 hr, centrifuged at 1400 x g for 5 min at 4C, washed in ice-cold PBS, and lysed in 25 ml 20 mM Tris buffer, pH 7.4, 1.5 mM EDTA with 0.1% β-mercaptoethanol by three cycles of freezing and thawing. Extracts were clarified by centrifugation at 100,000 x g, and NaCl added to a final concentration of 150 mM. The supernatant was then added to 5 ml of glutathione/agarose beads (Sigma Chemical; St Louis, MO) and incubated at 4C with gentle mixing. The column was washed with 5 x column volume (CV) of a modified PBS buffer (150 mM NaCl, 16 mM Na₂HPO₄, 4 mM NaH₂PO₄, pH 7.3) and bound protein was eluted via three rinses of 30 min each with 50 mM Tris buffer, pH 9.0, containing 100 mM glutathione.

Column eluates were pooled, and dialyzed twice for 24 hr at 4C against 2 liters of 50 mM Tris buffer, pH 8.0, containing 50 mM NaCl, and proteins were then separated by perfusion chromatography on a Poros HQ column (PerSeptive Biosciences; Framingham, MA) using a Biocad Sprint chromatography system. The 1.66-ml column was equilibrated with 100 mM Tris/Bis-tris propane to pH 9.0 and elution was performed with a 0-1000 mM NaCl gradient over 30 CVs. Column eluates were electrophoresed using denaturing 4-20% gradient Tris-glycine gels, followed by silver staining of the gels or transfer to nitrocellulose membranes for western blot analysis with ECL chemiluminescent detection (Amersham; Poole, UK). A 26-kD protein was recovered in the first elution fraction, and a 28-kD protein over three fractions at 200 mM NaCl.

Elution fractions containing the 28-kD protein were dialyzed twice for 24 hr against 1 mM Tris buffer, pH 8.0, at 4C, lyophilized for 24 hr, and resuspended in 1 ml PBS, pH 8.0. After addition of 1 ml Freund's complete adjuvant and sonication for 60 sec, the protein was injected SC at four dorsal sites on a New Zealand White rabbit. The rabbit was housed and handled under a protocol approved by the Medical University of South Carolina's Animal Care and Use Committee. Four weeks after injection, 25 ml of whole blood collected from the rabbit's ear vein was centrifuged at 3000 x g for 30 min and precipitated with 50% ammonium sulfate.

Animal Treatment and Tissue Processing

Syrian hamsters were housed and handled under a protocol approved by the Medical University of South Carolina's Animal Care and Use Committee. Twenty-one 40-day-old males were used in this study. Animals were implanted subcutaneously with pellets containing 20 mg of TP, E2, or both hormones (Hormone Pellet Press; University of Kansas, Lawrence, KS) at approximately 50 days after birth and were sacrificed at 2, 4, or 6 weeks after implantation. Two animals were used for each treatment protocol and each time point, along with one age-matched untreated animal for each time point.

Animals were sacrificed with 30 μg pentobarbital, after which reproductive tract organs (testis: caput, corpus, and cauda epididymis; epididymis/vas deferens junction; and distal vas deferens) and systemic organs (kidney, liver, and pancreas) were harvested rapidly, sliced into appropriately sized pieces, and immersed for 30 min in a solution of 10% formalin containing 0.5% zinc dichromate with the pH adjusted to 5.0 immediately before use. Tissues were dehydrated in a graded series of ethanols (70% 2 hr, 80% 2 hr, 95% 2 hr, and 100% three times for 1 hr), cleared in Histoclear (twice for 1 hr) (National Diagnostics; Atlanta, GA), and embedded in paraffin (Paraplast Plus, twice for 1 hr at 58C) (Curtin Matheson; Atlanta, GA). Paraffin blocks were serially sectioned at 4-μm thickness and mounted on glass slides. Every 25th section was stained with hematoxylin and eosin and selected sections were immunostained according to the protocol outlined below.

Immunoperoxidase Staining

Deparaffinized and rehydrated sections were treated with 0.3% hydrogen peroxide to block endogenous peroxidase activity and equilibrated in 0.1 M PBS, pH 7.2, containing 1% normal goat serum (NGS). Sections were then incubated at 4C overnight with one of the following rabbit antisera: anti-hGSTalpha (NovaCastra Laboratories; Newcastle Upon Tyne, UK, catalog #NCL-GSTalpha), directed against the alpha isoform purified from postmortem human (h) liver, diluted 1:300 in PBS/NGS; anti-haGSTmu, directed against the hamster (ha) GSTmu isoform isolated from DDT1 cells, as described above, diluted 1:4000 in PBS/NGS; or anti-rGSTpi (Panvera Corporation; Madison, WI, catalog #311) directed against the pi isoform purified from postmortem rat (r) liver, diluted 1:750 in PBS/NGS. Sections were rinsed with PBS/NGS and incubated for 20 min with biotinylated goat anti-rabbit IgG (Vector Laboratories; Burlingame, CA) rinsed again with PBS, and flooded with an avidin-biotin-horseradish peroxidase complex (Vectastain ABC, Vector Laboratories) for 30 min. Sites of bound primary antibody were visualized by a 10-min development in 3,3′-diaminobenzidine/H₂O₂ peroxidase substrate medium (Sigma Fast). Sections from all treatment groups and all time points were stained with the same protocols to minimize method variability and to provide side-by-side comparison of relative differences in immunostaining intensity.

Results

GST Isolation and Antiserum Production

Individual eluates from glutathione agarose columns were electrophoresed using denaturing 4-20% gradient Tris-glycine gels. Silver staining revealed two proteins with estimated molecular weights of 28 and 26 kD (data not shown). After separation of these two proteins by perfusion chromatography, Western blot analysis with ECL chemiluminescent detection (Amersham) employing a polyclonal antiserum used to isolate haGSTmu cDNA from a phage library (Norris et al. 1991) detected the 28-kD protein but not the 26-kD protein, thus identifying the 28-kD protein as haGSTmu (data not shown).

After immunization with haGSTmu and collection of serum 4 weeks later, Western blot analysis showed weak immunoreactivity with haGSTmu and no reactivity with the 26-kD protein. Immunization was repeated using Freund's incomplete adjuvant and serum collected 2 weeks after this boost. Antibodies in this serum reacted strongly and specifically with haGSTmu and showed no crossreactivity with the 26-kD protein (Figure 1). Preimmune serum from the same rabbit showed no staining on Western blots. Western blot analysis with anti-rGSTpi identified the 26-kD protein as a pi isoform, but anti-hGSTalpha antiserum failed to react with any proteins isolated from DDT1 cells (data not shown).

Figure 1

(A) Nitrocellulose filter used for Western blot analysis, stained with 0.1% amido black for 10 min. Lane 1, protein standards for molecular weight analysis; 30-kD standard is labeled. Lane 2, whole-cell extract from DDT1 cells. Lane 3, DDT1 cell extract after purification on glutathione/agarose affinity column. (B) Western blot analysis of the nitrocellulose filter shown in A using a 1:2000 dilution of antiserum against haGSTmu in TBS buffer with 0.3% Tween 20, 1% nonfat dry milk. The 28-kD protein previously identified as haGSTmu, but not the 26-kD protein, was bound by anti-haGSTmu in both the whole-cell extract (Lane 2) and the glutathione/agarose affinity column eluate (Lane 3).

Immunohistochemical Specificity in Control Tissues

Sections of kidney, liver, and pancreas from untreated animals were stained with the three antisera to provide positive and negative tissue control data and to confirm their specificity. Staining was performed with the same protocol on serial sections, thus allowing a direct comparison of immunoreactivity in similar sites. In the kidney, anti-hGSTalpha showed moderate immunoreactivity in proximal convoluted tubules (Figure 2A). Anti-haGSTmu strongly and specifically stained a subpopulation of cells in the collecting ducts (Figure 2B) and stained distal convoluted tubules less intensely. All three cell types showed moderate immunoreactivity with anti-rGSTpi (Figure 2C). Hepatocytes also showed class specificity with the GST antisera. Anti-hGSTalpha strongly stained nuclei in a subset of hepatocytes (Figure 2D), whereas anti-haGSTmu reacted with the cytoplasm and nuclei of a distinct group of hepatocytes not recognized by anti-hGSTalpha (Figure 2E). Adjacent sections from the same region of liver showed no immunoreactivity with anti-rGSTpi (Figure 2F). In the pancreas, the islets of Langerhans showed very weak reactivity with anti-hGSTalpha (Figure 2G) and no staining with anti-haGSTmu (Figure 2H), whereas pancreatic ducts of all sizes were intensely stained with anti-haGSTmu. Both islet cells and pancreatic ducts were intensely stained with anti-rGSTpi (Figure 2I).

Figure 2

Control tissues from an age-matched untreated male stained with anti-hGSTalpha at 1:300, anti-haGSTmu at 1:4000, or anti-rGSTpi at 1:750, as indicated. (A) Section through the kidney cortex shows moderate cytosolic immunoreactivity in the proximal convoluted tubule, but not the collecting ducts or distal convoluted tubules. Anti-hGSTalpha. (B) Section adjacent to that depicted in A shows staining of cytosol and/or nuclei of selected cells in the collecting ducts. Anti-haGSTmu. (C) Section adjacent to that shown in B stained with anti-rG-STpi demonstrates expression of this class in both proximal convoluted tubules and collecting ducts. (D) Section of a liver lobule near a central vein (cv) contains a subset of hepatocytes that display intense nuclear immunoreactivity with antiserum against GSTalpha. (E) A subpopulation of hepatocytes shows variable cytosolic and/or nuclear staining in a section adjacent to that depicted in A. Anti-haGSTmu. (F) Section of liver adjacent to that shown in D and E fails to react with antiserum against GSTpi. (G) An islet of Langerhans shows only very weak reactivity with antiserum against GSTalpha. (H) The same islet fails to stain with anti-haGSTmu, whereas a small pancreatic duct shows moderate reactivity (arrow). (I) Islet cells and a duct (arrow) in section adjacent to that shown in H are stained intensely with anti-rGSTpi. Bar = 25 μm.

Effects of Hormone Treatment

In our experimental model, the first evidence of neoplasia is seen at the region of the epididymis/vas deferens junction. This was also the only region in which differences were seen in the immunostaining intensity of GST isoforms after hormone treatment.

In untreated control animals, the epithelial cells lining the epididymis and the proximal vas deferens were stained intensely with antisera against all three classes of GSTs (data not shown). In contrast, the underlying smooth muscle layer, a thin, one-cell layer proximally that thickens as it progresses down the vas deferens, failed to react with anti-hGSTalpha and anti-haGSTmu, and showed only weak, if any, staining with anti-rGSTpi (data not shown).

Treatment with TP alone resulted in mild hyperplasia in the smooth muscle layer of the proximal vas deferens. However, no changes in immunoreactivity for any of the three GST classes were observed at any site in the reproductive tract of any of the six animals treated for 2, 4, or 6 weeks with TP compared to untreated controls. Class-specific immunostaining in a region of the proximal vas deferens near the epididymis/vas deferens junction after 4 weeks of treatment with TP alone is illustrated in Figures 3A, 3D, and 3G. This staining pattern did not differ from that observed in control animals with any of the three antisera.

Treatment with E2 alone for 4 weeks promoted mild atrophic changes along the entire male reproductive tract. Immunohistochemistry revealed no discernible changes in the staining patterns in the testis or epididymis with any of the three antisera tested. A striking observation, however, was the complete loss of immunoreactive GSTalpha (Figure 3B) and GSTmu (Figure 3E) in epithelial cells along the entire length of the vas deferens in both hamsters receiving the treatment. This loss of immunostaining for GSTalpha and GSTmu was not seen in either hamster treated with E2 alone for only 2 weeks, but persisted through 6 weeks of E2 treatment, the last time period studied. In marked contrast, there was no change in the immunostaining pattern with anti-rGSTpi at any time point with E2 treatment alone (cf. Figures 3G and 3H).

Combined treatment with TP and E2 resulted in a morphological and immunohistochemical staining profile in the reproductive tract essentially identical to that seen after treatment with E2 alone. No changes were observed after 2 weeks of treatment, but 4 and 6 weeks of treatment resulted in abolition of immunostaining for GSTalpha (Figure 3C) and GSTmu (Figure 3F) in the epithelial lining of the entire vas deferens. Again, there was no change in the staining pattern for anti-rGSTpi at any time point with the combined treatment (Figure 3I).

It is important to emphasize the remarkable consistency in the immunostaining patterns among the controls and the various treatment groups used in this study. Thirteen of the hamsters (the three untreated animals, the six animals sacrificed after 2 weeks of treatment, and the four animals treated with TP alone for 4 and 6 weeks) showed strong and consistent staining of epithelial cells lining the epididymis and vas deferens with the antisera against all three classes of GSTs. In contrast, all eight animals treated with E2 (four treated with E2 only for 4 or 6 weeks, four treated with TP and E2 for 4 or 6 weeks) showed a complete loss of immunoreactivity for GSTalpha and GSTmu, but no change for that of GSTpi, only in the vas deferens epithelium.

Discussion

Although there are many examples of estrogen-induced carcinogenesis, including the development of kidney neoplasms in Syrian hamsters (Oberley et al. 1991), the molecular and cellular mechanisms involved in this process remain poorly understood. Estrogen and its metabolites are genotoxic, and exposure to these agents can promote cell proliferation along with the production of reactive oxygen species (Yager and Liehr 1996). Estrogenic quinones also have been shown to disrupt microtubule assembly and cause changes in chromosomal structure (Epe et al. 1989). Transformation is hypothesized to occur via one or all of these effects. Before this study, our one-cell model of androgen/estrogen-induced formation of leiomyosarcomas did not involve the epithelial lining of the vas deferens and was based on the following assumptions. A first step is androgen-induced smooth muscle cell division (hyperplasia). The smooth muscle cells induced to an active, cycling state express GSTmu at lower levels, as we have shown in tissue culture (Norris et al. 1992). The subsequent induction of genomic instability by estrogenic metabolites then promotes tumor formation (Norris et al. 1992).

Although the cellular localization of the different GST classes has not been mapped in the Syrian hamster, our results are in excellent agreement with the tissue distribution and relative subunit concentration of GST isoforms in this species (Bogaards et al. 1992). For example, the relative subunit concentrations from pancreatic homogenates are 38 μ/g for alpha and 235 μ/g for pi (Bogaards et al. 1992). In our study, islet cells showed only faint staining with anti-hGSTalpha and intense staining with anti-rGSTpi. Relatively abundant levels of mu class subunit(s) in the pancreatic ducts were detectable immunohistochemically, even though no mu-subunits were detected in pancreatic homogenates (Bogaards et al. 1992). Because ducts comprise such a small percentage of pancreatic tissue, this is probably due to dilution of the mu class subunit(s) below detectable levels for the chromatography system used.

Figure 3

Sections through the proximal vas deferens from hamsters treated for 4 weeks with TP alone (A,D,G), E2 alone (B,E,H), or both hormones simultaneously (C,F,I) stained with anti-hGSTalpha at 1:300 (A-C), anti-haGSTmu at 1:4000 (D-F), or anti-rGSTpi at 1:750 (G-I). ep, epithelium; sm, smooth muscle. (A) Epithelial cells show moderate to intense cytosolic staining, but underlying smooth muscle is unreactive with anti-hGSTalpha. Staining did not differ from that seen in untreated controls (data not shown). (B) Four weeks of treatment with E2 resulted in a complete loss of immunoreactivity for alpha class GST in epithelial cells. (C) Animals treated with TP and E2 for 4 weeks showed the same loss of epithelial GST alpha as those treated with E2 alone. (D) Treatment with TP alone did not alter the staining pattern for mu class GST from that seen in untreated hamsters. The epithelium is stained intensely, whereas smooth muscle fails to react. (E) Four weeks of treatment with E2 alone promotes a complete loss of affinity for anti-haGSTmu in epithelial cells. (F) Combined administration of TP and E2 also resulted in abolition of epithelial staining with anti-huGSTmu. (G) Treatment with TP alone failed to alter the strong epithelial staining with anti-rGSTpi seen in untreated males. Smooth muscle shows weak reactivity. (H,I) Treatment with E2 alone or in combination with TP also failed to alter the immunostaining pattern seen with anti-rGSTpi. Bar = 25 μm.

We did not attempt to determine the subunit isoform specificity for the antisera used here (i.e., which of the two alpha-subunits or four mu-subunits are recognized by anti-hGSTalpha or anti-haGSTmu, respectively). However, the unique cell and tissue staining patterns described above, along with their excellent agreement with earlier data concerning GST tissue distribution and relative concentration in the Syrian hamster, confirm that the three antisera used are specific for the alpha, mu, and pi classes of GST and display no crossreactivity.

The finding that E2 alone or in combination with TP blocks expression of immunoreactive alpha and mu class GSTs in the epithelium of the vas deferens after only 4 weeks of treatment has caused us to revise our working hypothesis. Unfortunately, the requirement of both TP and E2 for tumorigenesis precludes studies of mechanisms involving E2 alone because the animals die before leiomyosarcomas develop. However, it is possible that TP serves only to prevent the development of renal neoplasms, allowing the animals to live long enough for E2 to induce and/or promote leiomyosarcomas in the vas deferens, or that TP acts to prevent atrophy of the reproductive tract so that tumors can develop in the presence of E2. In the latter event, the strong proliferative effects of TP may still act as an essential component of the carcinogenic process.

The most striking and an unexpected finding in this study was the complete loss of immunostaining for GST alpha and mu classes in the proximal vas deferens epithelium after E2 treatment alone. Although it is possible that the diminished immunostaining was due to antigen masking somehow induced by hormone treatment, this is unlikely, especially in view of the fact that reactivity for GST pi was unchanged after hormone treatment. The loss of immunoreactivity more probably reflects a loss of protein expression. This conclusion therefore invokes a model for hormonal induction of leiomyosarcomas involving two cell types, the epithelium and the underlying smooth muscle. One possibility is that the epithelium serves as a protective barrier against carcinogenic compounds normally present in the lumen of the vas deferens. After hormone treatment, the loss of this protective, detoxifying enzyme would render the underlying smooth muscle cells vulnerable to the transforming agent(s). Alternatively, the smooth muscle cells may be exposed to carcinogenic compounds via the vascular supply of the vas deferens, which under normal conditions is absorbed, detoxified and eliminated by the epithelial cells, possibly via secretion into the lumen of the vas deferens. In this case, the loss of GSTs in the epithelium could still be deleterious, allowing accumulation of carcinogenic compound(s) in the smooth muscle.

E2 may act as a functional agent in the carcinogenic event(s) via one or both of two mechanisms. In the simplest case, E2 may act only to decrease GST levels in vas deferens epithelial cells. In this event, the carcinogenic agent(s) involved in the subsequent steps leading to tumor formation would be a naturally occurring compound(s) present in the lumen of the vas deferens and/or systemic circulation, which under normal conditions would be detoxified and eliminated by GSTs in the vas deferens epithelium. A second and more complex series of events would invoke a dual role for E2. In this scenario, E2 not only would act to induce a decrease in GST levels but also would promote the generation of genotoxic carcinogens via metabolism of excess exogenously administered E2 itself. Metabolism of E2 results in estrogenic quinones, and redox cycling of these molecules to the semiquinone generates reactive oxygen species, both of which are capable of forming DNA adducts (Yager and Liehr 1996). There is also evidence that estrogenic metabolites may result in transformation via epigenetic effects (Epe et al. 1989). One or more of these effects could be responsible for tumor formation by estrogenic metabolites and/or metabolic byproducts.

Regardless of which of the above mechanistic explanations applies, it is clear that carcinogenesis probably occurs in response to an initial transforming insult promoted by exposure to E2. We hypothesize that the proliferative effects of TP would then serve to fix this error in the daughter cells of the initiated progenitor cell. The combination of these initiation and promotion events would lead to the transformed phenotype, and eventually to fully developed tumors.

These results do not unequivocally demonstrate a causal relationship between the loss of alpha and mu class GSTs and carcinogenesis, nor is it clear why these two classes were selectively decreased and not GSTpi. Unfortunately, data concerning the substrate specificity of GST classes towards hormone metabolites and/or byproducts are lacking, especially for the species-specific enzymes from the Syrian hamster. It is clear, however, that some GSTs play an important role in cellular responses to oxidative stress (Hayes and Pulford 1995), which would result from the reactive oxygen species generated by exposure to exogenous estrogen (Yager and Liehr 1996). Whether this is a critical function for either the alpha or mu class GSTs in the Syrian hamster is not known, and definitive conclusions about their role in this carcinogenic model await further data. It is important to note, however, that the changes seen in GST isoform expression after exposure to exogenous estrogen for at least 4 weeks were seen in 100% of the experimental animals and only at the site of tumor development, i.e., the proximal vas deferens. Such remarkable consistency and tissue and site specificity strongly suggest causality rather than coincidence and require that these new findings must be incorporated in future models of leiomyosarcoma induction in the Syrian hamster.