GLUTATHIONE TRANSFERASE GSTn IN BREAST TUMORS EVALUATED BY THREE TECHNIQUES

RAFAEL MOLINA t, STEFFI OESTERREICH§, HAN-LIANG ZHOU+, ATUL K. TANDON§, GARY M. CLARK§ , D. CRAIG ALLRED§ , ALAN J. TOWNS END*", JEFFRY A. MOSCOW·"', KENNETH H. COWAN**", WILLIAM L. MCGUIRE§ , SUZANNE A.W. FUQUA * § tHospital Clinico Provincial, Barce lona, Spain §University of Texas Health Science Cellter, Sail Antonio, Texas, U.S.A. ':University of Maryland, Baltimore, Maryland, U.S.A. ** BolVman Gray School o{Medicine. Willston-Salem. North Carolina, U.S.A. *** National Cancer lnstitllle, Bethesda, Maryland, U.S.A.


INTRODUCTION
Resistance to chemotherapeutic agents is one of the most important problems in cancer treatment with much effort being directed at identifying the mechanisms involved in both intrinsic and acquired drug resistance.The glutathione-S-transferase (GST) enzymes mediate a variety of normal detoxification reactions in cells (Mannervik, 1985), and may represent potential drug resistance mechanisms exploited by tumor cells.The three classes of GST isoenzymes (alpha, mu, and 1t) are encoded by different genes (Clapper and Tew, 1989) and are associated with different types of drug resistance.For example, increased GST alpha expression is associated with resistance to nitrogen mustards (Lewis et al. , 1988) and mechlorethamine (Buller et ai., 1987), whereas GST1t is elevated in adriamycin-resistant human breast cancer cells (Batist et al., 1986) and human lung tumors (Yolm et af., 1992) and carcinogen-induced rat typerplastic liver nodules resistant to a variety of xenobiotics (Cowan et af., 1986).However, the exact role of the various GST isoenzymes in resistance to specific chemotherapeutic agents remains to be established; it is known that elevated expression of GST1t alone in mammalian transfectants does not confer resistance to a variety of drugs and its presence may only be a marker of the drug-resistant phenotype (Moscow et al., 1989a).
GST1t is the most prevalent of the GST isoenzymes present in many human tumors with generally higher levels in tumor as compared to matched normal controls (Moscow et al., 1989b;Kodate et al., 1986).In particular, the GST1t content in gastric cancers and colon carcinomas is elevated as compared to normal surrounding tissue suggesting that GST1t may be useful as a marker of malignant transformation (Kodate et al., 1986).Recently, it has been reported that elevated serum levels of GST1t may also be useful for monitoring patients with cancers of the stomach, esophagus, and colon (Tsuchida et al.,  1989).
The finding that GST1t RNA expression inversely correlated with estrogen and progesterone receptors in primary breast tumors (Moscow et al., 1988a; Gilbert et al.,  1993) suggested that measurement of this isoenzyme may be valuable in breast cancer prognosis because it is this group of tumors which traditionally have a poorer outcome and shorter disease-free survival (McGuire, 1978).However, since significant amounts of GST1t can be found in many normal tissues (Moscow et al., 1989b), including breast epithelium (Terrier et aI., 1990), we questioned whether normal breast tissue GST1t expression would contribute to overall GST measurements.We therefore evaluated GST1t expression in 60 primary breast cancers by three different techniques; our overall goal was to correlate these results with variables known to be associated with clinical outcome.There was a good correlation between Western blot (WB), Northern hybridization (NH), and immunohistochemical (IHC) techniques for measuring GST1t expression.However, benign breast tissues and normal lymphocytes also expressed GST1t.Therefore, GST1t expression by these normal tissues may contribute to overall GST1t measurements.Confirming earlier reports, there was a significant inverse relationship between GST1t and steroid receptor status using any of the three detection methods.

Human breast tumor specimens
Breast tumor biopsies from 60 patients with primary disease were frozen in liquid nitrogen immediately after excision, pulverized, analyzed for steroid receptors (Dressler et al., 1988), and stored in the San Antonio Breast Tumor Data Network at -70°C until required for GST1t assessment.Specimens were considered ER-positive if they contained at least 3 fmol of specific binding per milligram of cytosolic protein, and PgRpositive if they contained at least 5 fmol per milligram of cytosolic protein.DNA flow cytometry was performed using 100 mg of each specimen as previously described on an Epics IV flow cytometer (Coulter Electronics, Hialeah, Fla.) (Clark et al., 1989).

Immunohistochemistry
Fifty-five of 60 tumors utilized in this study had sufficient tissues to perform IHC.Briefly, permanent-sections were prepared as previously described (Allred et aI., 1990) by rehydrating 50 mg of frozen particulate breast tumor at room temperature in PBS, fixing in 10% formalin for 4 hours, pelleting the particles into a tissue "button" in agar by centrifugation, infiltrating and embedding the button in paraffin, and cutting sections containing an average of about 500 intact tumor cells.Histological and nuclear grading of tumors were performed using the criteria of Fisher et ai.(Fisher et aI., 1980).
Immunostaining was performed using a standard avidin-biotin-peroxidase complex (ABC) technique (Hsn et ai., 1981).Briefly, dewaxed sections were washed in phosphate buffered saline (PBS), and endogenous peroxidase was quenched with 0.1 % sodium azide/3% H 2 0 2 in PBS for 30 min at room temperature.Sections were washed in PBS and incubated for 30 min in 5% normal goat serum/l 0% ovalbumin to block non-specific protein biotin.Sections were then incubated at room temperature for 2.5 h with a GSTn specific rabbit polyclonal antiserum (Terrier et ai., 1990) at a 1/400 dilution in 3% ovalbumin/PBS.After washing in PBS, sections were then incubated with biotinylated swine anti-rabbit secondary antibody (Vector Laboratories, Burlington, GA) at a dilution of I :200 in PBS for 30 min.Following an additional PBS wash, sections were incubated with avidin-biotin-peroxidase complex (Vector Laboratories, Burlington, GA), washed again and incubated 5 min in diaminobenzidine/H 2 0 2 chromogen substrate.After washing and counter-staining with Harris hematoxylin, samples were dehydrated through graded alcohols and xylene, and mounted with Permount.Cytospins of the Hs578T human breast cancer cell line known to express GSTn (Moscow et al., 1988a) were used as positive controls.

Western blot analysis
All of the 60 breast tumors were examined by WB analysis .Approximately 10 mg of tumor powder was exposed to 5% sodium dodecysulfate (SOS) as previously described (Tandon et al., 1989).Samples were then vortexed, boiled for 5 min, and allowed to cool to room temperature for IS min.Clear supernatant was collected after centrifugation at 13,000g for 2 min at room temperature.Protein concentration was then determined by the bicinchoninic acid method (Smith et ai., 1985).

Northern hybridization analysis
Total cellular RNA sufficient for NH was isolated from 57 of the 60 specimens using a Model 340A nucleic acid extractor (Applied Biosystems Inc., Foster City, CA).Quantitation by absorption spectroscopy at 260 nm was confirmed by inspection of an ethidium bromide-stained agarose gel.Twenty Ilg of RNA per sample was electrophoresed on a 1% agarose gel containing 0.66 M formaldehyde in 20 mM 3-[N-morpholinoJpropanesulfonic acid (MOPS) buffer, and transferred to nylon membranes (Schleicher and Schuell).Hybridization was performed at 50°C overnight in 50% formamide/5X 0.15 M naCI1l5 mM trisodium citrate (SSC)/5X Denhardt' s/0/5% SDSI 100 J..lg/ml denatured salmon sperm DNA, and 32P-labeled GSTrr-1 cDNA (Moscow et al., 1988a) prepared by random primed labeling (Boehringer Mannheim, Indianapolis, IN).After hybridization the membranes were washed with O.IX SSCII % SDS at 50°C and autoradiographed.The same filter was hybridized with the pHFb-actin cDNA probe (Gunning et aI., 1983) to control for equivalent RNA loading.The intensity of the GSTrr hybridization signal relative to p-actin signal was obtained for each sample by scanning densitometry.

Evaluation of GSTrr expression by three different techniques
Sixty primary breast tumors were chosen for study where sufficient tumor material (>235 mg) was available for analysis.WB analysis was performed first using a rabbit polyclonal anti-GSTrr antibody specific for GSTrr (Terrier et al., 1990).An extract from the Hs578T cell line was also included on each gel as an arbitrary internal reference standard.A single band at approximately 23,000 daltons molecular weight was detected with the antibody (Figure I, middle panel).GSTrr concentrations by WB were found to vary widely between 0 to 3350 densitometric units/lOO J..lg of sample protein corrected for the signal obtained for 100 J..lg of control cell line extract.
Sufficient RNA to perform NH analysis was then isolated from 57 of the 60 tumors .A single band at approximately 1.1 kb was detected with the GSTrr cDNA probe (Figure 1, top panel).GSTrr mRNA levels were determined by densitometric scanning and expressed relative to the signal obtained with p-actin (results not shown); mRNA levels also varied widely in the tumors ranging from 0 to 668 densitometric units.There was a good general agreement between these two techniques, but a direct comparison of GSTrr levels required us to define a cut-off for low versus high expression.
Therefore, we undertook to further examine GSTrr protein expression by IHe.IHC results were obtained on 55 of the 60 tumors.Representative staining obtained with the GSTrr specific antibody is shown in the lower panel of Figure I. Sixty-nine percent of the specimens showed specific cytosolic staining (defined as >5% positive tumor cell staining).Thus for correlative purposes the 69th percentile was used as the cut-off for positive GSTrr expression measured by either NH or WB.The three methods were then compared using chi-square analysis; these results are shown in Table 1.There was a significant positive correlation between the three methods used for GSTrr detection.The closest correlation (p=O.OOOI) was obtained with the two methods (NH and WB) requiring homogenization of the tumor specimens, and whose levels were obtained by densitometric quantitation.There was complete concordance in 64 % of the tumors; 54% of the tumors were positive by all ofthe three methods and 10% were negative for GSTrr expression (Figure 2).

GSTrr expression in benign breast tissues and lymphocytes
The pulverized tumor samples used for tests in this study were initially manually dissected from the fresh surgical specimens.Every effort was made to select "pure" tumor and, therefore, the samples are primarily composed of malignant rather than benign tissue elements.However, twelve of 55 samples (22%) examined histologically  contained some benign breast epithelium (ducts and/or lobules) which, on average, accounted for less than 5 % of total sample cellularity.Only about 10% of benign cells within these cases showed a positive IHC signal for GSTrr (Figure 3) and this signal was generally weak, suggesting that it is unlikely that benign epithelium made a significant contribution to the GSTrr signal in WB and NH analyses of the same samples.Connective tissue elements (i.e.endothelium, fibroblasts, etc.) within the specimens never showed positive immunostaining (Figure 3, panel C).
Fifty-eight percent of our samples contained tumor infiltrating lymphocytes which, on average, accounted for about 10% of total cellularity in these samples.Most lymphocytes present showed variably intense immunostaining for GSTrr (Figure 3, panel B), suggesting that lymphocytes could make a significant contribution to the total GSTrr signal obtained in WB or NH analyses.

Relationships between GSTrr expression and other prognostic variables of known sign(ficance in breast cancer
Associations between GSTrr and other biological indicators used in breast cancer prognosis and treatment are shown in Table 2.There was a strong, significant inverse relationship between GSTrr expression and steroid receptor status using all of the three methods with higher GSTrr expression generally seen in the receptor-negative group of tumors.In addition, there was a trend toward higher GSTrr expression in poorly differentiated histological grade III and nuclear grade III tumors.No correlation between GSTrr and ploidy or %S-phase was found .

DISCUSSION
During the last two decades, substantial progress has been made in the development of more effecti ve treatments for cancer.Unfortunately, in the majority of cases, increased response rates have not translated into marked improvements in survival.Resistance to multiple chemotherapeutic agents remains a major obstacle to successful cancer chemotherapy.GST enzymes play an important role in normal cellular defense against toxic xenobiotics and carcinogens.These enzymes have also been implicated in the detoxification of many antineoplastic agents (Mannervik, 1985;Clapper et al., 1989;Lewis et al,. 1988;Buller et al., 1987;Batist et al., 1986;Evans et al., 1987;Nakagawa et al., 1988), and have been reported to be markers of neoplastic transformation.GST isoenzyme levels, including GSTn:, are elevated in many human tumors relative to the corresponding normal tissues (Moscow et ai, 1989b;Kodate et ai, 1986;Tsuchida et ai, 1989).A significant increase in GSTn: activity has been found in tumor as compared to normal adjacent or benign breast lesions (Di Ilio et ai, 1985), suggesting that GSTn: may be involved both in breast cancer development and possibly drug resistance.Therefore, we have begun to measure GSTn: in breast tumors with the eventual goal being to evaluate its expression as a marker during breast cancer progression.
Due to the availability of both specific GSTn: antibodies and cDNA probes, we analyzed GSTn: expression at both the protein and mRNA levels.Two of the methods chosen, WB and NH, share the advantage of being semiquantitative, but require homogenization of the tumor.These methods are then subject to experimental error due *** significant.p <0.05 * 0.05 < p <0.09 to the dilution of tumor extracts with surrounding normal breast, vascular elements, or inflammatory cells which are often present in breast tumor specimens.However, in spite of these limitations we wanted to use semiquantitative methods so as to fully evaluate associations between GST1t levels and known biological parameters commonly delineated in breast cancer.
Confirming earlier reports (Tsuchida et al., 1989, Howie et ai., 1989), we observed an inverse relationship between GST1t expression and the expression of estrogen and progesterone receptors using all three of the detection methods.Similar results have been reported with the epidermal growth factor receptor where elevated expression is also seen in the receptor-negative group of patients (Cappelletti et al., 1988;Sainsbury et al., 1985;Delarue et aI., 1988).Recently, it has been suggested that the estrogen receptor may exert a constitutive repressor function on estrogen-responsive genes in the absence of hormone (Tzuckerman et al., 1990).It is an intriguing hypothesis that the observed inverse relationship between GST1t expression and the presence of the estrogen receptor may be related to this repressor activity.Although, the recent analysis of the promoter elements and the posttranscriptional fate of GST1t (Morrow et al., 1992) has shown that the differential expression in ER+ versus ER-is governed by posttranscriptional processes.
There was also a trend towards higher GST1t expression in poorly differentiated tumors.Several studies have detected a relationship between the receptor-negative phenotype and the degree of cellular dedifferentiation (Fisher et ai., 1981 b), thus the interrelationships we report here are a further demonstration of the basic biological differences between receptor-positive and receptor-negative breast cancers.Future studies should be directed at identifying common regulatory factors that may underlie these associations.
The value of estrogen receptors in predicting the endocrine response of breast cancer has been appreciated for some time (Osborne et al., 1980).Steroid receptors not only are valuable for predicting response to hormonal manipulation, but also the time course of the disease (Osborne et al., 1980;Benner et ai., 1988).Therefore, combining receptor status with other parameters such as GST1t or histopathology may provide very valuable treatment guides.However, a larger study of breast cancer specimens with adequate clinical follow-up will be required to address these issues.Additionally, the clinically important question whether GST1t is directly involved in chemotherapeutic resistance remains unanswered to date.
IHC assessment of GST1t expression in breast cancer biopsies demonstrated specific cytosolic staining in two types of non-malignant cells; both normal mammary epithelium and lymphocytes.These benign cells may be present in heterogenous breast tumor specimens.Twenty-two percent of the cases in our series contained benign epithelium which, on average, accounted for less than 5% of the cells within the sample.Furthermore, only about 10% of benign cells showed positive immunostaining for GST1t.Therefore normal breast epithelium does not appear to make a major contribution to overall GST1t measurements by NH or WB.In contrast, GST1t positive lymphocytes were present in the majority of samples, which has been reported previously (Del Boccio et al. , 1986).Thus, lymphocyte infiltration may make a contribution to the overall measurement of GST1t using methods that are unable to differentiate the cellular source of GST1t in the tumor.This may be one reason why other studies have not detected correlations between GST1t and some prognostic factors (Shea et ai., 1990).We feel that IHC should probably always be included in an analysis ofGSTn content in breast cancer.By itself, IHC provides significant information regarding the relationship of GSTn to other biological characteristics of breast cancer.In addition, IHLC can discriminate the cell source of GSTn expression, enabling accurate interpretation of more quantitative WB or NH analyses.

Figure 2 .
Figure I. Evaluation of GSTrr ex press ion by three different techniques.Tumors were simultaneously analyzed by NB (upper panel), WB (middle pan el), and IHC (lower panel).Five receptor-positive and five receptor-negative tumors arc shown; the positive control cell line (C), H s578T is also included.Molecular weight standards were run and are given in kilobase pairs (kb) in the NB and kilodaltons (kDa) in the WB.

Figure 3 .
Figure 3. GSTll expression in benign breast tissues and lymphocytes.lHe staining ofGSTll in benign breast epithelium (panel A) and infiltrating lymphocytes (panel B).Breast tumors negative for GSTll staining are shown in panels B (arrow) and C.

Table I .
Comparison ofNB, WB , and IHC measurements of GSTrr expression in human breast tumors

Table 2 .
Relationship ofGSTn expression and other prognostic variables in breast cancer