No evidence of significant activity of the multidrug resistance gene product in primary human breast cancer

Annals of Oncology 9: 85-93,1998. © 1998 Kluwer Academic Publishers. Printed in the Netherlands.

Original article No evidence of significant activity of the multidrug resistance gene product in primary human breast cancer S. Hegewisch-Becker,1 F. Staib,1 T. Loning,2 U. Pichlmeier,4 N. Kroger,1 A. Reymann3 & D. K. Hossfeld1 'Department of Oncology and Hematology, 2 Department of Gynecopathology, 3Department of Pharmacology, "Institute of Mathematics and Computer Science in Medicine, University Clinic, Hamburg, Germany

Summary

72% and 75% of samples, respectively. A positive MDR1-RTPCR signal was detected in 62% of the samples. Nevertheless, Background: The discovery of the multidrug resistance (MDR1) no correlation between immunohistochemistry and RT-PCR gene product P-glycoprotein (P-gp) has been widely seen as an could be established. Furthermore, there was no correlation important milestone in our understanding of the mechanisms between HER-2/neu expression and MDR1 -RT-PCR or P-gp underlying the clinical phenomenon of the emergence of re- immunohistochemical assays. A contamination by CD8+ and sistant cells. MDR1 expression has been shown for numerous CD4+ lymphocytes was established in 100% and 84% of solid tumors and for virtually all hematologic malignancies. tumor cell suspensions, respectively. As assessed by the Rhl23 Nevertheless, results regarding MDRl/P-gp expression in hu- efflux assay CD8+ and the CD4+ lymphocytes exhibited man breast cancer have been controversial and the results of marked P-glycoprotein activity, whereas such activity was not clinical trials on modulation of P-gp activity have not been detectable in a single instance for the breast carcinoma cells. In MDR1-RT-PCR positive samples, contamination by CD8 encouraging. Patients and methods: MDRl/P-gp expression and the func- lymphocytes averaged 4.3%, while the contamination of CD8 tion of the P-gp pump were investigated in 61 tumor samples cells in the MDR1 mRNA-negative samples was only 2.4% from patients with primary breast cancers by multiparameter (P = 0.007). This signal vanished after elimination of the lymanalysis using MDR1-RT-PCR, immunohistochemistry with phocyte subpopulations byT-cell rosetting. two MAbs (UIC2 and MRK 16) and the rhodamine 123 Conclusions: In primary breast cancer detection of MDR1 (Rhl23) efflux assay. The cellular composition of the tumor gene expression by means of RT-PCR or immunohistochemcell suspension was analyzed by using specific MAbs against ical assays is not indicative for the MDR phenotype, since the P-gp expressing lymphocyte subsets CD4, CD8 and CD56, there is no evidence of significant activity of the P-gp pump. as well as against the HER-2/neu gene product, which was used to identify breast carcinoma cells. Key words: breast cancer, MDR1 gene, multidrug resistance, Results: UIC2 and MRK16 revealed a staining positivity in P-glycoprotein, rhodamine 123

Introduction Acquired or intrinsic resistance to chemotherapy is one of the biggest problems in the treatment of cancer. The multidrug resistance (MDR) phenotype, defined by a cross resistance to a number of functionally and structurally distinct drugs, is currently the best-understood mechanism of cellular resistance. The classical MDR phenotype is mediated via overexpression of a 170-kd transmembrane P-(permeability) glycoprotein (P-gp). This protein functions as a transmembrane efflux pump for different lipophilic compounds and extrudes many commonly used anticancer agents such as anthracyclines, vinca alkaloids, epipodophyllotoxins and taxus compounds as well as several fluorescent dyes such as rhodamine 123 (Rhl23) [1] from the cell. One reason for the strong interest in the P-gp mediated MDR phenotype is the possibility of modulating

the P-gp mediated MDR phenotype. Numerous noncytotoxic substances have been demonstrated to possess the capacity to compete with the cytotoxic drug for the active outward transport, thus increasing the intracellular drug accumulation. Of these, calcium channel blockers and cyclosporines are being the best-explored. In recent years the expression of the MDR1 gene has been investigated in numerous hematologic malignancies and solid tumors [2-4]. But while a correlation between the expression of the MDR1 gene product and response to induction chemotherapy has now clearly been established for the acute nonlymphoblastic leukemias, the situation with respect to solid tumors looks rather different. For these, thus far it has only been possible to clearly demonstrate such a correlation for different types of sarcoma [3, 5, 6]. Results in breast cancer, ovarian carcinoma or lung cancer are less convincing and in part controversial [4]. Thus, it is not

86 surprising that numerous clinical studies with modulators of multidrug resistance in solid tumors have so far produced disappointing results [7]. Many groups have explored the role of MDRl gene expression in breast cancer with varying results. Trock et al. [8] recently published a carefully conducted metaanalysis of studies examining MDRl /P-gp expression in breast cancer. Although the authors found that the data are consistent with a contributory role of MDRl/P-gp in the multidrug resistance in some breast tumors there remains confusion as to the exact importance of this mechanism of drug resistance in this tumor type. In comparable patient collectives the number of positive tumor samples varied between 0% and 85%. For studies using RNA hybridization methods the average percentage of tumors expressing MDRl/P-gp was 27.1% and for those using immunohistochemistry the average was 48.5%. The authors especially criticized that the results of the majority of studies based on immunohistochemical detection were not sufficiently controled. In some studies P-gp staining was found not only in breast cancer cells but also in desmoplastic stroma and normal breast tissue [9-13]. One aim of the present study was therefore to improve our understanding of the significance of the MDR-phenotype in breast cancer through comparison of standard techniques. For this purpose we chose PCR measurement of MDRl mRNA expression, immunostaining and flow cytometric assessment of P-gp function by dye efflux. To permit the determination of P-gp function in different subsets of a given tumor cell suspension the Rhl23 extrusion assay was combined with staining for different subset specific surface markers. Rhl23 is a green fluorescent dye which provides an excellent substrate for the P-gp pump. By double-labeling with specific phycoerythrin conjugated surface markers, a cell suspension can be divided into its component subpopulations, and Rhl23 efflux can be separately analyzed for each of the individual cell populations. The method has been used successfully to detect a functional multidrug resistant phenotype in MDRl gene expressing normal and malignant cells [14, 15]. Flow cytometric identification of the breast cancer cells was achieved by labeling them with an HER-2/neu-specific antibody. Various estimates for the frequency of HER-2/neu gene amplification and overexpression in primary breast cancers range from 14% to 54% [16-20]. Only membrane staining, which is detected using flow cytometry in combination with a HER-2/neu specific monoclonal antibody, has been considered to be a reliable sign for HER-2/neu overexpression in breast cancer cells [18]. Furthermore, the HER-2/neu protein is not expressed on stromal cells. Since one can assume a contamination of the tumor sample with peripheral blood lymphocytes, we chose CD56+, CD8+ and CD4+ specific monoclonal antibodies to identify NK cells, T-suppressor and T-helper cells, respectively. These cell types have been shown to demonstrate by far the strongest MDRl gene expression amongst all peripheral blood cells [14].

In the present study, MDRl expression was investigated in a series of 61 primary operable breast carcinomas. The levels of P-gp expression by means of immunohistochemistry were correlated to MDRl mRNA expression as determined by PCR and to the flow cytometric assessment of P-gp function. To investigate possible pitfalls of these methods was a special task of this study. Patients and methods Human tumor samples Tumor tissue was obtained from 61 previously untreated patients operated upon for breast cancer. After surgical removal tumor pieces were immediately transferred into RPMI-medium containing sterile tubes. Samples were always taken from the tumor core, thus contamination with normal epithelial cells was extremely low as controlled by the analysis of paraffin-embedded sections. Tumor cell suspensions were obtained by mincing the tumor tissue and disaggregating the cells in RPMI-medium containing 200 U/ml collagenase (Worthington Biochemical Corporation, Freehold, NJ) [21]. Tumor pieces were dissociated in enzyme for 16-24 hours, some as long as 36 hours, depending on the degree of disaggregation. The procedure, as assessed with the aid of inverted phase contrast microscopy resulted in a large number of single cells and a few clusters. These remaining clusters were further dispersed by pipetting. Cells were collected by filtering the suspension through a 100 urn nylon filter (Falcon, Franklin Lakes, NJ). Following disaggregation, the viability of the cells, as assessed by Trypan blue extrusion, was very good (mean 84%). Cytospins were prepared at the end of the collagenase treatment and were stained with May-Griinwald-Giemsa. The analysis of these slides revealed that the majority of the cells retrieved after this procedure were indeed breast cancer cells. The remaining material was routinely fixed in formalin and embedded in paraffin for pathologic diagnosis and immunohistochemistry. The clinical characteristics of the patients are summarized in Table 1. Table 1. Age and histologic features of the patients. Number of patients Age (yr) Median Range Tumour size Tl T2 T3 T4 Lymph nodes NO Nl N2 Histoprognostic grade GI Gil GUI Histologic type Invasive ductal carcinoma Invasive lobular carcinoma Other Oestrogen and progesterone receptors ERER+ PRPR+

61 59 23-85 8(13%) 42 (69%) 6(10%) 5 (8%) 21 (34%) 31 (51%) 9(15%) 7(11%) 27 (44%) 27 (44%) 47 (77%) 9(15%) 5 (8%) 23 38 20 41

(38%) (62%) (33%) (67%)

87 Cell lines

Flow cytometry

Three human P-gp expressing cell lines - a human breast cancer cell line (MCF-7/ADR, resistance index of 21 against adriamycin [22]) and two leukemic cell lines (K562-RADR, resistance index of five against adriamycin. kindly provided by A. Schafer, Hamburg) and CEM/ VBL, resistance index of 270 against vinblastine) were used as positive controls. The corresponding P-gp negative parent cell lines were used as negative controls.

Following the Rhl23 uptake cells were labeled with the phycoerythrin (PE) conjugated surface markers CD4 (T-helper cells), CD8 (T-suppressor cells) or CD56 (NK cells) or the IgG isotope control at saturation concentrations following the manufacturer's instructions (Becton-Dickinson, San Jose, CA). For the detection of the HER-2/ neu+ breast cancer cells, cells were first incubated with the primary moAB c-neu (AB-5) (Oncogene Science, Uniondale, NY) or the IgG isotope control (Coulter Clone, Hialeah. FL) for 45 min. This was followed by incubation with the PE-conjugated F(ab)2 fragment (Jackson Immuno Research, West Grave. PA) for 30 min. All steps were performed at 4 °C. To minimize the possibility of false positive binding, expression of HER-2-neu was considered positive if more than 20% of cells reacted with the antibody as compared to the IgG isotope control. Flow cytometric analysis was performed on a FACScan flow cytometer at an excitation wavelength of 488 nm using 530/30 nm (green fluorescence) or 585/40 nm (red fluorescence) bandpass filters. Forward and side scatter signals were collected using linear scales and fluorescent signals were collected on logarithmic scales. By means of flow cytometry, cells can be separated according to their light scattering properties. Forward and perpendicular light scattering of lymphocytes is highly reproducible, the cells always apear within the same region [24]. Thus, as illustrated in Figure 1A it was very easy to discriminate tumor cells from contaminating lymphocytes. In those patient samples negative for HER-2/neu expression, light scattering windows (forward vs. right angle) taken from HER-2 neu positive samples were used to gate on the breast cancer cell population. Percent positivity was calculated by setting a marker according to the control sample. Percent decrease in Rhl23 accumulation in the absence of verapamil was calculated by comparing the mean of the fluorescence distribution in the presence or absence of verapamil as previously described [25]. Dead cells were assessed by propidium iodide. Viable cells were identified as those with a low red fluorescence.

Immunohistochemistry 5 urn slides from paraffin-embedded blocks were dewaxed in xylene and then rehydrated in graded alkohols. Cytospin preparations were airdried and fixed in acetone. The assessment of HER2-neu expression was performed using the avidin-biotin-peroxidase technique with Mab c-neu (Ab-3) (Oncogene Science, Uniondale, NY) as previously described [16, 20]. Since cytoplasmic staining has been shown to be unspecific, only tumor cells with unequivocal membrane staining were counted as positive [20]. All samples considered positive had > 20% positive cells. For the detection of P-gp MAbs MRK.16 (100 ug/ml) and UIC2 (50 ug/ml), directed against different epitopes of the P-gp molecule were used. Both MAbs were used at a working dilution of 1 • 20 of the indicated concentrations. With this dilution we obtained a very good concordance between the two MAbs. Immunohistochemistry was performed with the alkaline phosphatase/anti-alkaline phosphatase (APAAP) method which was chosen as we and others found it more sensitive than using immunoperoxydase [12]. The primary antibodies were applied over night. Slides were then exposed to rabbit antimouse immunoglobuline (Dako, Hamburg, Germany), followed by incubation with the APAAP complex used at a working concentration of 1:20 (Dako). The color reaction was developed using a substrate based on fast red, producing a red reaction in positive cells. Finally the slides were counterstained in hematoxylin. As an internal standard the breast cancer cell lines MCF-7 and MCF-7/ADR served as negative and positive controls. K.562 and K562-RADR cells served as further controls. Immunohistochemistry was always performed in parallel on paraffin-embedded normal colon tissue, known to express a high level of P-gp in the apical surface of its columnar cells thus serving as a positive control. Negative control slides were also run in parallel, substituting an isotope matched control antibody for the primary antibody. The stained preparations were evaluated in a representative section of each slide by two experienced observers (S. HegewischBecker and T. Loning) using a three point scale (absent, moderate (+) and intermediate to strong (++)). This three-scale method was introduced to enable us to perform analysis on subgroups. All samples considered positive had numerous positive tumor cells (^20%) throughout the section. Samples with occasional weak staining ( < 2%) were considered negative.

Rhl23 efflux studies The Rhl23 efflux was performed as previously described the only modification being that cells were initially loaded with 60 ng/ml of Rhl23 (Sigma, St. Luis, MO) for 20 min in 10 ml of staining media at 37 "C [22]. The staining media was RPMI-1640 plus 10% fetal calf serum. After two washes in ice-cold staining media, the cells were allowed to efflux the dye for 120 min. in dye free media at 37 °C in the presence or absence of 30 um verapamil (Sigma, St. Luis, MO), a P-gp inhibitor. The reaction was terminated by transferring the samples to 4°C. 120 min efflux time was chosen to increase the sensitivity of the assay and to guarantee full extrusion of the dye even in cells with a low level of P-gp expression, as might be exspected to be the case in breast cancer samples. Preliminary experiments have been performed on the P-gp expressing cell lines to guarantee that collagenase pretreatment was not affecting P-gp activity. In these cell lines Rhl23 efflux remained unchanged. Furthermore, we investigated the effect of collagenase pretreatment on Rhl23 efflux in human CD8+ lymphocytes. Again, the pretreatment had no effect on P-gp activity.

MDR1-RT-PCR assay Following RNA extraction from tissue fragments similar to those used for immunodetection (RNAzol B kit; Cinna/MRC, Cincinnati, OH) the study of MDR1 gene transcription was performed by reversetranscriptase-PCR. cDNA was synthesized with a cDNA synthesis kit (Boehringer Mannheim Corp., Indianapolis, IN), according to the manufacturer's protocol, using 1 ug of total cellular RNA and random hexanucleotide primers. PCR was performed in a final volume of 100 ul with one tenth of the cDNA template using a Gene Amp polymerase reaction kit (Perkin-Elmer/Cetus, Branchbury, NJ). MDR1 and (32microglobulin specific primers were identical to those published by Noonan et al. [26]. p2-Microglobulin was used as an internal control. Fragments of 167 bp or 120 bp were generated using MDR1 and P2microglobulin (P2m) specific primer sets, respectively. A total of 34 cycles was performed. 10-microlitre aliquots of the PCR product were then run on an ethidium bromide stained 2% agarose gel. In a number of PCR positive samples the MDR1-RT-PCR was repeated following T-cell depletion by T-cell rosetting.

T-cell rosetting For T-cell rosetting 1 ml of the tumor cell suspension was mixed with 1 ml of 50% 2-aminoethylisothiouronium bromide hydrobromide (AET, Sigma, St. Luis, MO) treated sheep red blood cells and 0.2 ml fetal calf serum. T cells were separated from the tumor cell suspension by means of Ficoll hypaque density gradient centrifugation at 2000 rpm for 20 minutes. The T-cell depleted interface was collected and washed twice in phosphate buffered saline. The percentage of contaminating lymphocytes was analyzed flow-cytometrically before and after T-cell rosetting. In parallel experiments, normal peripheral blood mononuclear cells were treated accordingly to verify the success of the procedure.

50

100 150 FSC-Height

200

250

50

100

150

200

250

200

250

50

100

150

200

250

Figure 1A Dot display of perpendicular (SSC-Height) versus forward light scatter (FSC-Height) intensities of an unseparated tumor cell suspension (all cells) and its component subpopulations. Flow cytometric analysis of the subpopulations was performed after labeling the cells with specific surface markers HER-2/neu, CD8 and CD4. The given percentages are proportional to the total number of cells. Statistical methods Categorical data were described by their absolute and relative frequencies. Association between binary variables were assessed by means of chi-square test. Weighted kappa was calculated to assess the level of agreement between two methods on the same ordinal scale. Continuous variables were described with medians and their associated ranges. To test their potential differences within two subgroups of patients the Mann-Whitney U-test was performed.

Results Expression of the HER-2/neu oncogene Flow cytometric assessment of HER-2/neu expression revealed a positivity in 49% of the samples. Using immunohistochemistry strong membrane staining was seen in 44% of the samples. There was a good concordance between the two methods, leading to an overall agreement of 82% (weighted kappa 0.64, P < 0.001). Presence of lymphocytes in the tumor cell suspension Since we were interested to examine the possibility of a contamination of the tumor cell suspensions by lymphocytes known for their high MDR1 gene expression, we had initially decided to concentrate on the detection of

CD8+ and CD56+ subpopulations. Only after CD56 had been analyzed in 15 tumor samples, but could not be detected in any one of them, CD4 measurement was included in the assay. Contamination by CD8+ lymphocyte subpopulations was detected in all 61 samples analyzed (mean 3.6%), while contamination by CD4+ lymphocytes was detected in 31 of 37 (84%) of the cellular suspension samples (mean 1.9%). Separation of the tumor cell suspension into the component subpopulations Using the different surface markers CD4, CD8 and HER-2/neu it was easily possible to separate lymphocytes from breast cancer cells. Furthermore, due to the differences in cell size and granularity it was possible to use light scatter windows to gate on a specific population. Figure 1A shows a typical example. Immunohistochemical detection ofP-gp expression Both antibodies showed clear and strong staining of the apical surface of the columnar cells of the colon and of the P-gp expressing control cell lines. In breast cancer samples staining was observed in 44 of 61 samples (72%) with UIC2 with 26 of these (43%) displaying ++ positivity and in 42 of 57 samples (75%) with MRK16

89

HER2/neu+

1Cf

10

66%

10

Rh 123 fluorescence

10 u

10

10

Figure IB. P-gp function as assessed by Rhl23 efflux in the presence (solid line) or absence (dotted line) of verapamil and double labeling with the surface markers HER-2/neu, CD8 and CD4. Cells were allowed to efflux the dye for 120 minutes. A decrease in Rhl23fluorescencein the absence of the P-gp modulator verapamil was very pronounced in CD8+ cells, was also but to a lesser extend detectable in CD4+ cells, and was completely absent in HER-2/neu+ cells. Due to the small amount of contaminating CD8+ and CD4+ cells, P-gp activity in these subpopulations became only evident when Rhl23 efflux was analyzed on gated subpopulations.

with 19 (35%) displaying ++ positivity. Although staining was heterogeneous, all samples found positive had more than 20% of the total tumor area stained. Concordance between the two antibodies was good with an overall agreement of 79% (weighted kappa 0.75, P < 0.001). Occasional weak staining of stroma cells was observed with both antibodies. Further more, in some samples we observed staining of smooth muscles in small blood vessels. This has been described before [27]. Unless otherwise stated, tumor P-gp expression refers to positivity with UIC2.

demonstrated HER-2/neu expression as assessed by flow cytometry. There was also no agreement between MDR1 mRNA expression and P-gp immunohistochemical assays (weighted kappa 0.06, P = 0.68). Concordance could only be detected in 21 samples (34%). In 11 cases, the tumor samples tested positive for MDR1 mRNA whereas no P-gp could be detected by immunohistochemistry. On the other hand, in six samples with + staining positivity and in 11 samples with ++ staining positivity no MDR1 mRNA could be detected. Eight of these 11 RT-PCR negative, P-gp ++ positive samples were HER2/neu positive, which was good evidence that MDR1RT-PCR: Correlation to HER2-neu expression, these samples contained a significant number of breast lymphocyte contamination and immunohistochemical cancer cells. In contrast, there was a clear correlation detection of P-gp between lymphocyte contamination and a positive MDR1 RT-PCR signal. In those samples that tested Expression of the MDR1 gene at mRNA level was positive for MDR1 mRNA, contamination by CD8 investigated by RT-PCR and detected in 38 of 61 (62%) lymphocytes was more pronounced than in the MDR1 of the samples. The number of viable tumor cells was mRNA-negative samples with a median of 3.05% (mean equally distributed in PCR positive and PCR negative 4.3%, range 0.6%-14%, SD ± 3.5%) and a median of samples. There was no correlation between MDR1 1.5% (mean 2.4%, range 0.6%-8.8%, SD ± 2.2%), remRNA and HER2/neu expression (chi-square, P = spectively. This difference was statistically significant 0.185 and P - 0.192 for immunohistochemical and flow (Mann-Whitney U-test, P = 0.007). Those samples cytometrical detection, respectively). MDR1 mRNA where contamination by CD8 plus CD4 was analyzed could be detected in 16 of 30 samples (53%), which had demonstrated the same picture with a median of 4.4%

90 (mean 6.1%, range 1.4%-19.1%, SD ± 4.8%) and a median of 2.9% (mean 3.5%, range 1.5%-8.5%, SD ± 2.1%) for MDRl RT-PCR positive and negantive samles, respectively (P = 0.01). As shown in Figure 2, the intensity of the PCR-signal increased with the percentage of contaminating lymphocytes. In four of the MDRl mRNA-positive samples a T-cell rosetting was performed to eliminate contamination by CD4+ and CD8+ lymphocytes. In all samples where this procedure was adopted, a positive signal for MDRl mRNA expression as determined by RT-PCR could no longer be detected after the T-cell rosetting (Figure 3). In an bivariate analysis a positive MDR1-RT PCR signal could not be found to be dependent on nodal status, tumor size, histologic types, grades, oestrogen and progesteron receptor expression. Rhl23 efflux The sensitivity of the method was first established in cell lines, displaying different degrees of resistance. Mixture experiments of the parental P-gp negative cell lines K562 with K562-RADR and MCF-7 with MCF-7/ ADR showed, that a contamination as low as 3% was easily detectable (Figure 4). MCF-7 cells are HER2/neu positive. The labeling for HER2/neu did not influence the sensitivity of the assay. Since the K562-RADR cell line displays a resistance index of only five, which is

3%

4%

4,5%

probably similar to what can be seen in human tumors, the method was assumed to be sensitive enough to detect efflux activity even in breast cancer samples with only a small amount of P-gp expressing cells. Pretreatment with collagenase did also not influence the activity of the pump. The Rhl23 efflux assay established activity of the P-gp pump in 60 of 61 CD8+ subpopulations as well as in 30 of 31 CD4+ subpopulations with a significantly higher activity in CD8+ cells. The difference in mean Rhl23 fluorescence in the absence or presence of the P-gp modulator verapamil was 135% and 79% for CD8+ and CD4+ cells with a median of 99% (range 0%-573%) and 64% (range 0%-285%), respectively (Mann-Whitney U-test, P - 0.008). This is in accordance with previously published results, demonstrating a significantly lower MDRl expression in CD4+ cells as compared to CD8+ cells [13]. However, no extrusion of the dye could be detected in any of the 30 HER-2/neupositive cell populations of which 15 had displayed ++ P-gp staining (Figure IB). In tumor cell suspensions where no HER-2/neu positivity was detectable, light scatter windows were used to gate on the tumor cell population and to exclude lymphocytes from the analysis. In this group 11 samples had displayed ++ P-gp staining. Once again, not a single instance of P-gp pump activity could be detected in these populations.

7%

9%

11%

MDR1 118

p2m

Figure 2. MDR analysis using RT-PCR: the semiquantitative PCR assay demonstrates an interdependence between signal intensity for MDRl and the percentage of contaminating lymphocytes.

Figure 3. MDR analysis using RT-PCR before and after T-cell depletion by T-cell rosetting (TCR) in a tumor cell suspension (TU) and in peripheral blood mononuclear cells (PB) with the latter being used as a control to verify the success of the procedure. T-cell rosetting resulted in a loss of the MDRl-PCR signal in both cell types. The multidrug resistant cell line CEM/VBL was used as a positive control.

91

MCF-7:MCF-7/ADR

K562:K562-RADR

10

10

Rh123 fluorescence

102

10

Rh123 fluorescence

Figure 4. Sensitivity of the Rhl23 efflux assay as determined by mixture experiments of sensitive (K562 and MCF-7) and P-gp expressing (K562RADR and MCF-7/ADR) cell lines in the presence (solid line) or absence (dotted line) of verapamil. Even 3% low P-gp expressing K562-RADR cells admixed to sensitive K562 cells were clearly detectable. To increase the sensitivity of the efflux assay we used an efflux time of 120 min. to allow even low resistant cells to efflux the dye completly.

Discussion In the present study we have examined MDR1 gene expression in primary operable breast carcinomas by MDR1 RT-PCR, immunohistochemistry and a functional assay to detect activity of the P-gp pump, thus using three different methods as required by the Memphis consensus recommendations for methods to detect P-gp associated multidrug resistance in patients' tumors [15]. Based on the data presented here, which demonstrated a strong correlation between MDR1 RT-PCR and the amount of contaminating lymphocytes but failed to demonstrate a correlation between RT-PCR and immunohistochemistry or the functional assay, one must assume that a positive MDR1 RT-PCR signal in primary operable breast cancer is attributable to contamination by P-gp-expressing lymphocyte subpopulations, and is not indicative of a strong expression of a functional multidrug resistance gene product in the breast carcinoma cells. This assumption is supported by the fact that the level of MDR1 gene expression in CD4+ and CD8+ lymphocytes is as high as or even higher than the level found in 8226/DOX-40 cells. These cells are highly resistant and display a resistance index of 40 [14, 28]. Our data are also in concordance with a study of Maas et al. [29]. The authors used a immunomagnetic purification technique to purify human breast carcinoma cells from fine-needle aspirates taken from lymph node metastases. A positive MDR1-RT-PCR signal was clearly detectable in the unseparated fraction whereas the purified breast cancer cells remained negative, even after the PCR cycle number was increased from 35 to 40. Charpin et al. [10] performed RT-PCR only in tumors, in which P-gp was also detected by immunohistochemistry, thus a direct comparison to our data is not possible. Nevertheless, they failed to detect MDR1 mRNA in three of 14 samples which had demon-

strated P-gp positivity by immunohistochemistry. They also mentioned a weak MDR1 mRNA expression in two cases negative for P-gp staining. In their opinion, these discrepancies were most likely due to differences in tissue handling and a lower sensitivity of immunohistochemistry as compared to PCR, respectively. Since in our series, the handling of all samples was identical and the viability of cells was absolutely comparable in PCR positive and negative samples, technical problems cannot serve as an explanation for the absence of a correlation between immunohistochemistry and MDR1 RTPCR. O'Driscoll et al. [30] could detect a MDR1 RTPCR signal in 67% of a smaller series, which correspond well to our PCR findings. Immunohistochemistry is the method most often used to detect the MDR-phenotype and most studies have been based on the investigation of tumor tissue from untreated patients whose breast carcinomas were for the most part operable. In the majority of these studies not a panel but only one monoclonal antibody was applied to detect P-gp and the results were not confirmed by a second method of measurement [8]. For C219, the antibody most frequently used in these studies [12, 13, 31-33] and for C494 [34, 35] certain cytoplasmic cross-reacting proteins have been described [36, 37]. Furthermore, C219 is not MDRl-specific but reacts against two other members of the MDR gene family which do not mediate the MDR phenotype. Therefore a non-specific positivity cannot be ruled out. Furthermore, working dilutions for the MAbs and immunohistochemical staining techniques differ widely. When MRK16 was used in a dilution of 1:10 to 1: 20 staining positivity was described in 55% [12] and 72% (this study) of samples. Incubation time was two hours and overnight, respectively. With a dilution of 1:100 and an incubation time of 30 min. only 7% of samples stained positive [11] whereas a dilution of 1:400 (incubation

92 time not given) produced positivity in 29% of samples [9]. C219 was normally used in a dilution of 1:20 which produced positive staining in 42%, 46%, 50%, 72% and 84% of samples [12, 13, 31-33]. When used in a dilution of 1:100, only 11% of samples were positive [9]. In the present study we used two MDR1 specific MAbs which detect different epitopes of P-gp and have been shown to work well on paraffin-embedded material [38, 39]. The fact that we found so many P-gp ++ positive but MDR1 RT-PCR negative samples certainly worries us. Unspecific binding might be one possible explanation. In this regard we like to mention that all studies that found no expression of MDRl/P-gp were based on Western blots or RNA hybridization methods and with these methods the average percentage of tumors expressing MDRl/P-gp is only 27% [8]. Therefore these methods were considered to be less sensitive. Nevertheless, these methods were regularly used for the detetction of the MDR phenotype in hematological malignancies. In these malignancies they were not found to be less sensitive than immunohistochemistry [2]. Future studies have to find out whether RNA hybridization techniques provide a more appropriate picture concerning the MDR phenotype in breast cancer. It is generally acknowledged that PCR is more sensitive than immunohistochemistry but is offset by its inability to distinguish between P-gp in normal versus tumor cells. To further improve our understanding of the significance of the classical MDR phenotype in breast cancer cells, we thus decided to include Rhl23 efflux as a third method. If staining positivity was due to expression of a functional P-gp pump, efflux had to be positive in at least those samples displaying the strongest P-gp staining. HER-2-neu had to be chosen for this assay, since it is the only surface marker known to be expressed on breast cancer cells. Detection of cytokeratins specific for breast cancer epithelium may have provided more specific results regarding the composition of the tumor cell suspension but would have required membrane permeabilisation. In this series we were not able to demonstrate significant activity of the P-gp pump or other transporter proteins in tumor cells of any of those samples which had been tested positive by means of MDR1 RT-PCR or immunohistochemistry. We can not exclude that breast cancer cells are especially susceptible to handling and that the collagenase treatment might have impaired the activity of the P-gp pump. Nevertheless, this seems unlikely for two reasons: Firstly, the presence of P-gp expressing T-lymphocytes in almost all samples provided us with an internal quality control for the assay. Rhl23 efflux of the contaminating lymphocytes was not affected by the handling procedure and did not differ from Rhl23 efflux seen in lymphocytes from freshly prepared peripheral blood samples. Secondly, collagenase treatment had no effect on P-gp activity in the MCF/ADR breast cancer cell line. One could also argue that our results are due to a selection of cell clones no longer representative of the original tumor cell popula-

tion. This seems highly unlikely since we could not correlate our data to histoprognostic grade, histologic type or HER2-neu expression. Finally, one could argue that the assay is not sensitive enough to detect efflux in very low P-gp expressing cells, since the low P-gp expressing K562-RADR cell line still has a resistance index of five. Nevertheless, cell lines with a resistance index of 4 or 5 are normally considered to represent a level of resistance exspected to be of clinical relevance in human tumor cells. In the consensus recommondations regarding methods to detect P-gp in patients' tumors it was recommended to use 8226/Dox6 cells (resistance index of 6) as a control relevant to the clinical setting [15]. Although a significant number of studies has been performed, the significance of the MDR-phenotype in breast cancer remains to be determined. Trock et al. [8] in their meta-analysis found that many studies exploring the significance of MDRl/P-gp expression in breast cancer did not comply sufficiently with criteria for methodological rigor which makes it difficult to draw any definite conclusion at this stage. Although their analysis indicates that MDRl/P-gp is expressed in a significant proportion of breast tumors they conclude that the functional relevance of MDRl/P-gp is not yet established. MDRl/P-gp expression might go along with the induction of other drug resistance mechanisms such as altered topoisomerase or glutathione transferase activity. It seems likely that more than one mechanism contributes to the multidrug resistance phenotype. One could thus imagine that MDRl/P-gp expression mainly serves as a surrogate marker for drug resistance and that the activity of the P-gp pump is of subordinate importance. In order to clarify the question of the significance of the MDR phenotype in breast cancer, further studies are needed in which a sequential analysis of the expression of the MDR1 gene in tumor tissue before and during the course of anthracycline containing therapy should be performed. When evaluating results of previous studies, it is most important to keep in mind, that the detection of MDR1 expression at RNA or protein level is challenging because (1) most tumors contain normal P-gp expressing cells such as lymphocytes or endothelial cells; (2) a highly sensitive and tumor cell specific method is required for the detection; (3) direct assessment of P-gp activity in tumor cells by means of a functional assay should be a prerequisite of each study; (4) other mechanisms of drug resistance should be investigated, whenever possible. Such rigorous criteria might help to get a more precise answer regarding the significance of the MDR phenotype. References 1. Neyfakh AA. Use od fluorescent dyes as molecular probes for the study of multidrug resistance. ExpCell Res 1988; 174: 168-76. 2. Hegewisch-Becker S, Hossfeld DK. The MDR phenotype in hematologic malignancies: prognostic relevance and future perspectives. Ann Hematol 1996: 72: 105-17.

93 3. Chan HSL. De Boer G, Haddad G, Gallie LB, Ling V. Multidrug resistance in pediatric malignancies. Hematol Oncol Clin North Am 1995: 9: 275-318. 4. Leighton JC, Goldstein LJ. P-glycoprotein in adult solid tumors: Expression and prognostic significance. Hematol Oncol Clin North Am 1995; 9: 251-73. 5. Roessner A, Ueda Y, Bockhorn-Dworniczak B et al. Prognostic implication of immunodetection of P-glycoprotein in Ewing's sarcoma. J Cancer Res Clin Oncol 1993; 119: 185-90. 6. Baldini N, Scotlandi K, Barbanti-Brodano G et al. Expression of P-glycoprotein in high-grade osteosarcomas in relation to clinical outcome. N Engl J Med 1995; 333: 1380-5. 7. Fisher GA, Sikic BI. Clinical studies with modulators of multidrug resistance. Hematol Oncol Clin North Am 1995; 9: 363-82. 8. Trock BJ, Leonessa F, Clarke R. Multidrug resistance in breast cancer: a meta-analysis of MDRl/gpl70 expression and its possible functional significance. J Natl Cancer Inst 1997; 89: 917-31. 9. Linn SC, Giaccone G, van Diest PJ et al. Prognostic relevance of P-glycoprotein expression in brest cancer. Ann Oncol 1995; 6: 679-85. 10. Charpin C, Vielh P, Duffaud F et al. Quantitative immunocytochemical assays of P-glycoprotein in breast carcinomas: Correlation to messenger RNA expression and to immunohistochemical prognostic indicators. J Natl Cancer Inst 1994; 86: 1539-45. 11. Torre de la M, Larsson R, Nygren P et al. Expression of the multidrug-resistance gene product in untreated human breast cancer and its relationship to prognostic markers. Acta Oncol 1994; 33: 773-7. 12. Wishart GC, Plumb JA, Going JJ et al. P-Glycoprotein expression in primary breast cancer detected by immunocytochemistry with two monoclonal antibodies. Br J Cancer 1990, 62, 758-61. 13. Ro J, Sahin A, Ro JY et al. Immunohistochemical analysis of P-glycoprotein expression correlated with chemotherapy resistance in locally advanced breast cancer. Hum Pathol 1990; 21: 787-91. 14. Drach D, Zhao S, Drach J, Mahadevia R et al. Subpopulations of normal peripheral blood and bone marrow cells espress a functional multidrug resistance phenotype. Blood 1992; 80: 2729-34. 15. Beck WT, Grogan TM, Willman CL et al. Methods to detect P-glycoprotein-associated multidrug resistance in patients' tumors: Consensus recommendations. Cancer Res 1996; 56: 3010-20. 16. Charpin C, Devictor B, Bonnier P et al. Expression of Her-2/neu oncogene in breast cancer: Correlation of quantitative immunocytochemistry and prognostic factors. Int J Oncol 1992; 1: 815-23. 17. Paik S, Hazan R, Fisher ER, Sass RE et al. Pathologic findings from the national surgical adjuvant breast and bowel project: Prognostic significance of erbB-2 protein overexpression in primary breast cancer. J Clin Oncol 1990; 8: 102-12. 18. De Potter CR, Schelfhout AM. The neu-protein and breast cancer. Virchows Archiv 1995; 426: 107-15. 19. Iglehart JD, Kraus MH, Langton BC et al. Increased erbB-2 gene copies and expression in multiple stages of breast cancer. Cancer Res 1990; 50: 6701-7. 20. Van de Vijver MJ, Peterse JL, Mooi WJ et al. Neu-protein overexpression in breast cancer. N Engl J Med 1988; 319: 1239-45. 21. Limon J, Dal Cin P, Sandberg AA. Application of long-term collagenase disaggregation for cytogenetic analysis of human solid tumors. Cancer Genet Cytogenet 1986; 23: 305-13. 22. Stein U, Walther W, Lemme M et al. Development and characterization of novel human multidrug resistant memory carcinoma lines in vitro and in vivo. Int J Cancer 1997; in press. 23. Hegewisch-Becker S, Hannania EG, Fu S, Korbling M et al. Transduction of MDR1 into human and mouse hematopoietic progenitor cells: Use of rhodamine (Rhl23) to determine transduction frequency and in vivo selection. Br J Haematol 1995; 90: 876-83. 24. Visser JWM, Van den Engh GJ, Van Bekkum DW. Light scatter-

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

ing properties of murine hematopoietic cells. Blood Cells 1980; 6: 391-407. Hegewisch-Becker S, Faltz C. Hossfeld DK. Prolongation of medium exchange is associated with a decrease in function but not expression of the P-glycoprotein pump in leukaemic cells. Eur J Haematol 1996; 56: 12-22. Noonan KE, Beck C, HolzmayerTA. Chin JE et al. Quantitative analysis of mdrl (multidrug resistance) gene expression in human tumours by polymerase chain reaction. Proc Natl Acad Sci 1990; 87: 7160-4. Van der Valk P, Van Kalken CK, Ketelaars H et al. Distribution of multi-drug resistance-associated P-glycoprotein in normal and neoplastic human tissues. Ann Oncol 1990; 1: 56-64. Dalton WS, Grogan TM, Rybski JA et al. Immunohistochemical detection and quantitation of P-glycoprotein in multiple drug resistant human myeloma cells: Association with levels of drug resisatnce and drug accumulation. Blood 1989; 73: 747-52. Maas RA, Bruning PF, Breedijk AJ et al. Methods in laboratory investigation. Immunomagnetic purification of human breast carcinoma cells allows tumor-specific detection of multidrug resistance gene 1-mRNA by reverse transcriptase polymerase chain reaction in fine-needle aspirates. Lab Invest 1995; 6: 760— 764. O'Driscoll L, Kennedy S, McDermott E et al. Multiple drug resistance-related messenger RNA expression in archival formalin-fixed paraffin-embedded human breast tumour tissue. Eur J Cancer 1996; 32A: 128-33. Schneider J, Rubio MP, Barbazan MJ et al. P-Glycoprotein, HER-2/neu, and mutant p53 expression in human gynecologic tumors. J Natl Cancer Inst 1994; 86: 850-5. Veneroni S, Zaffaroni N, Daidone MG et al. Expression of P-glycoprotein and in vitro or in vivo resistance to doxorubicin and cisplatin in breast and ovarian cancers. Eur J Cancer 1994; 30A: 1002-7. Decker DA, Morris LW, Levine AJ et al. Immunohistochemical analysis of P-glycoprotein expression in breast cancer: Clinical correlations. Ann Clin Lab Science 1995; 25: 52-59. Rouyer N, Lambolais C, Chenard MP et al. Stromelysin-3 (ST3) and P-glycoprotein (PGP) as predictive factors in inflammatory breast carcinoma (IBC). Proc Am Soc Clin Oncol 1995; 14: 42. Verrelle P, Meissonnier F, Fonck Y et al. Clinical relevance of immunohistochemical detection of multidrug resistance P-glycoprotein in breast carcinoma. J Natl Cancer Inst 1991; 83: 111-6. Rao VV, Anthony DC, Piwnica-Worm D. MDR1 gene-specific monoclonal antibody C494 cross-reacts with pyruvate carboxylase. Cancer Res 1994; 54: 1536-40. Thiebaut F, Tsuruo T, Hamada H et al. Immunohistochemical localization in normal tissues of different epitopes in the multidrug transporter protein P170: Evidence for localization in brain capillaries and cross reactivity of one antibody with a muscle protein. J Histochem Cytochem 1989; 37: 159-64. Flynn SD, Murren JR, Kirby WM et al. P-glycoprotein expression and multidrug resistance in adrenocortical carcinoma. Surgery 1992; 112: 981-6. Kelly DJ, Pavelic ZP, Gapany M et al. Detection of P-glycoprotein in squamous cell carcinomas of the head and neck. Arch Otolaryngol Head Neck Surg 1993; 119: 411-4.

Received 20 June 1997; accepted 29 September 1997. Correspondence to: Dr. Susanna Hegewisch-Becker Department of Oncology and Hematology University Clinic Eppendorf Martinistr. 52 D-20246 Hamburg Germany