Surg D. Alexander Today et al.: Multidrug Resistance in Gastric Cancer Jpn J Surg (1999) 29:401–406
401
Original Articles Histopathological Assessment of Multidrug Resistance in Gastric Cancer: Expression of P-Glycoprotein, Multidrug ResistanceAssociated Protein, and Lung-Resistance Protein Delamou Alexander1, Tetsu Yamamoto2, Shizuo Kato2, and Shinichi Kasai1 1
Second Department of Surgery and 2 Division of Transfusion Medicine, Asahikawa Medical College, 4-5 Nishikagura, Asahikawa 078, Japan
Abstract: Because local recurrence is common after a curative resection for advanced gastric cancer, there has been significant interest in adjuvant chemotherapy. However, the overall effect of chemotherapy remains debatable regarding patients with advanced gastric adenocarcinoma. Multidrug resistance is thought to be a major cause of failure in cancer chemotherapy, and thus the expression of P-glycoprotein (P-Gp), multidrug resistance-associated protein (MRP), and lungresistance protein (LRP) in tumor cells was evaluated by immunohistochemistry. In 20 gastric adenocarcinomas, 11 (55%), 2 (10%), and 0 (0%) were positive for MRP, LRP, and P-Gp. In malignant lymphomas, only 3 out of 10 cases were positive for MRP (30%). The positive rate of MRP staining was significantly higher in well and moderately differentiated adenocarcinomas (80%) than in poorly differentiated adenocarcinomas (20%). With regard to the degree of MRP expression and histological cell type, higher grades (grade 2– 3) were observed only in well and moderately differentiated adenocarcinomas. In terms of the positive-stained cells and staining intensity, heterogeneity was observed in the staining profile of MRP. The proliferative cell nuclear antigen labeling index (PCNA LI) of MRP-positive and MRP-negative cases was 49.3% 6 11.6% and 49.4 6 6.9%, respectively. No correlation was observed between the MRP expression and PCNA LI. In conclusion, the incidence of MRP expression in gastric cancer was the highest in three different multidrug resistancerelated epitopes. An evaluation of the MRP expression thus seemed to be beneficial for determining the optimal strategy of chemotherapy. Key Words: multidrug resistance, stomach cancer, immunohistochemistry, proliferative cell nuclear antigen
Reprint requests to: T. Yamamoto (Received for publication on Dec. 15, 1997; accepted on Sept. 11, 1998)
Introduction Gastric carcinoma usually shows little or no response to chemotherapy which thus makes postoperative treatment difficult in advanced cases.1 The 5-year survival rate for stage III and stage IV disease is 13% and 3%, respectively, in the United States.2 Approximately twothirds of all patients undergoing curative resection subsequently relapse and die of locoregional failure (87%) and/or distant metastasis (30%).3–4 Previous studies have assessed postoperative chemotherapy as a means of improving these disappointing results, but such efforts have generally proven to be ineffective when investigated at several institutions.5 In fact, no survival advantage was observed for patients who received postoperative chemotherapy when compared with those who did not.6–10 Among the basic issues awaiting clarification are who will benefit from chemotherapy and which method of evaluation is ideal for determining this. Multidrug resistance is widely thought to be a major cause of the failure in cancer chemotherapy. Tumor cells that display resistance to multiple anticancer drugs are also resistant to structurally unrelated agents with different mechanisms of action.11–14 Most tumor cells with multidrug resistance are characterized by the overexpression of P-glycoprotein (P-Gp), a membraneassociated glycoprotein that decreases the intracellular drug concentrations by acting as an efflux pump.15–17 However, the role of the multidrug resistance (MDR)-1 gene and P-Gp in clinical drug resistance has yet to be clarified. Recently, both multidrug resistance-associated protein (MRP)18–22 and lung-resistance protein (LRP)23–24 have been shown to be overexpressed in non-P-Gprelated multidrug resistance. There are a number of methods by which P-Gp, MRP, and LRP can be detected.25–28 The interpretation of Northern blotting or reverse transcriptase–polymerase chain reaction (RT-
402
D. Alexander et al.: Multidrug Resistance in Gastric Cancer
PCR) studies is complicated due to heterogeneity among tumor cells in homogenized samples, due to both possible contamination with nontumor cells and a lack of data on the distribution of positive cells within the tumor. In addition, such methods also may not be sensitive enough to detect a small population of multidrugresistance cells. An evaluation of the P-Gp, MRP, and LRP expression by immunohistochemistry can circumvent these problems by allowing for the detection of these proteins in one or only a few cells as well as by displaying the distribution of expression within the tumor. The aims of the present study were therefore to predict the patients most likely to respond to chemotherapy, and also to evaluate the expression of P-Gp, MRP, and LRP in gastric cancer.
Immunostaining Immunolocalization of multidrug resistance markers was done according to the streptavidin–biotin peroxidase complex method using a Histofine Kit (Nichirei). Tissue sections were first deparaffinized in xylol, ethanol, and water, and then endogenous peroxidase activity was blocked by immersion in 3% H2O2 in methanol for 10 min to prevent any nonspecific binding. For staining with PC-10 and P-Gp, paraffin sections were pretreated by dipping in 0.01 M citrate buffer (pH 6.0) and then heating them in a microwave oven (650°–750°C) for 10 min. After preincubation with 10% normal goat serum for 10 min, the tissue sections were incubated with the primary antibodies 5B12 (1 : 20), MRPr1 (1 : 20), LRP-56 (1 : 10), and PC-10 (1 : 50) for 60 min at room temperature. For establishing the optimum staining procedure, positive controls known to contain each antigen (tubular epithelial cells of human kidney) were used. The sections were then incubated with the secondary antibody (biotin-labeled antimouse and rabbit IgG goat antibody) for 10 min at room temperature, and then were finally incubated with peroxidase-labeled streptavidin for 5 min. The reaction products were visualized with diaminobenzidine. Finally, the sections were counterstained with hematoxylin, dehydrated in ethanol, cleared in xylol, and mounted. Nonspecific reactions were ruled out by replacing the primary antibodies with normal mouse serum.
Materials and Methods Tissue Specimens Gastric tumor specimens were retrieved from the tissue bank of the Department of Pathology at Asahikawa Medical College. All tissue specimens were sampled from surgical specimens obtained during the period from 1986 to 1995. Twenty gastric adenocarcinomas and 10 malignant lymphomas (21 men and 9 women, aged 30–84 years) were used in this study. Five examples of each histological type of gastric carcinoma (well differentiated adenocarcinoma, moderately differentiated adenocarcinoma, poorly differentiated adenocarcinoma, and signet-ring cell carcinoma) were randomly selected. None of these patients had been treated with chemotherapy prior to surgery.
Monoclonal Antibodies The following four monoclonal antibodies were used for immunohistochemistry: 5B12 (IgG2a; Novocastra, Newcastle, UK) directed against membrane bound PGp, MRPr1 (IgG2a; Nichirei, Tokyo, Japan), which recognizes an external epitope of MRP, LRP-56 (IgG2b; Nichirei), which reacts with an internal epitope of LRP, and PC-10 (IgG2a; Dako, Glostrup, Denmark), which stains proliferative cell nuclear antigen (PCNA).
Immunohistochemical Techniques Tissue Preparation Paraffin-embedded tissue specimens fixed in 10% formaldehyde were cut into 5-µm sections, mounted on silane-coated glass slides, and air-dried (45°–50°C) overnight prior to staining.
Evaluation of Staining Three investigators assessed all tissue sections and reached agreement about the final evaluation. An entire section was initially observed at a low magnification (3100) and 4–5 fields were selected to evaluate the staining. After each field was photographed at a high magnification (3200), the positive-stained cell ratio was determined by counting 500 tumor cells. The percentage of stained cells for MDR markers was then assigned to one of four grades (Fig. 1): grade 0 (0%), grade 1 (,5%), grade 2 (5%–30%), and grade 3 (.30%). All of the nuclei stained by PC-10 were regarded as positive for PCNA. The PCNA labeling index (LI) was calculated as the percentage of positive tumor cells/total tumor cells by counting 500 tumor cells. The percentage of stained cells was compared using Fisher’s exact test and Student’s t-test. A probability value of less than 0.05 was considered significant.
Results Twenty gastric adenocarcinomas and 10 malignant lymphomas were evaluated for P-Gp, MRP, and LRP
D. Alexander et al.: Multidrug Resistance in Gastric Cancer
403
Fig. 1. Grading of multidrug resistance-associated protein expression in gastric carcinoma. The percentage of positive stained cells was classified as one of four grades: grade 0 (0%), grade 1 (,5%), grade 2 (5%–30%), and grade 3 (.30%)
expression (Table 1). Eleven gastric adenocarcinomas were positive for MRP (55%) and two were positive for LRP (10%), but none were positive for P-Gp (0%). Three out of 10 malignant lymphomas were positive for MRP (30%) but none were positive for LRP or P-Gp. MRP staining was more frequent and stronger than that for LRP (P , 0.05). MRP staining was predominantly cytoplasmic, but some membrane staining was also observed. The intensity of the reactions varied from weak to strong. LRP expression was detected as granular cytoplasmic staining which was generally moderate. The relationship between tumor histology and MRP expression was analyzed for gastric adenocarcinoma (Table 2). The positive rate was significantly higher in well and moderately differentiated tumors (80%) than in poorly differentiated tumors (20%) (P , 0.05), while signet-ring cell carcinoma was intermediate (40%).
Table 1. Expression of P-Gp, MRP, and LRP in gastric adenocarcinoma and malignant lymphoma Multidrug resistance markers P-Gp Negative Positive MRP Negative Positive LRP Negative Positive
Adenocarcinoma (20 cases)
Malignant lymphoma (10 cases)
20 (100%) 0
10 (100%) 0
9 (45%) 11 (55%)
7 (70%) 3 (30%)
18 (90%) 2 (10%)
10 (100%) 0
P-Gp, P-glycoprotein; MRP, multidrug resistance-associated protein; LRP, lung-resistance protein
404
D. Alexander et al.: Multidrug Resistance in Gastric Cancer
Table 2. Relationship between epitope overexpression and the histological cell types in gastric carcinoma Multidrug resistance markers P-Gp Negative Positive MRP Negative Positive LRP Negative Positive
No. of Patients
Well differentiated adenocarcinoma
Moderately differentiated adenocarcinoma
Poorly differentiated adenocarcinoma
Signet-ring cell
20 0
5 0
5 0
5 0
5 0
9 11
1 4
1 4
4 1
3 2
18 2
4 1
4 1
5 0
5 0
Table 3. Evaluation of the MRP expression in gastric carcinoma (20 cases) Histological cell type
Grade 0 (0%)
Grade 1 (,5%)
Grade 2 (5%–30%)
Grade 3 (.30%)
Well diff. ad. Moderately diff. ad. Poorly diff. ad. Signet-ring cell
1 1 4 3
2 2 1 2
1 1 0 0
1 1 0 0
diff. ad., differentiated adenocarcinoma
Regarding the degree of MRP expression, two out of four positive well differentiated adenocarcinomas were grade 1, and the others were grade 2 and grade 3 (one case each). The staining profile of moderately differentiated adenocarcinoma was almost the same as that of well differentiate adenocarcinoma. One positive poorly differentiated adenocarcinoma was assessed as grade 1, while both positive signet-ring cell carcinomas were grade 1 (Table 3). PC-10 immunostaining was almost entirely confined to the nucleus, and was diffuse, granular, or a mixture of both. The PCNA LI for gastric adenocarcinoma ranged from 25.9% to 64.9% (mean 6 SD: 49.4% 6 9.8%). The PCNA LI of well differentiated, moderately differentiated, poorly differentiated, and signet-ring cell carcinoma was 51.6% 6 13.5%, 48.1% 6 8.4%, 55.4% 6 5.4%, and 42.4% 6 2.9%, respectively. The PCNA LI of MRP-positive and MRP-negative cases was 49.3% 6 11.6% and 49.4% 6 6.9%, respectively. No significant relationship was thus observed between MRP overexpression and the rate of proliferation of gastric cancer.
Discussion Despite a curative resection of advanced gastric carcinoma, many patients relapse and die from the
disease. One obstacle to a better outcome is the chemoresistance exhibited by gastric carcinoma cells. In the present study, an evaluation of P-Gp, MRP, and LRP overexpression in gastric adenocarcinoma and malignant lymphoma was done using immunohistochemistry. Comparing the incidence of positivity of the three different multidrug resistance-related epitopes, MRP was most frequently detected in both gastric adenocarcinoma and malignant lymphoma. The incidence of MRP in gastric malignant lymphoma was lower than in gastric adenocarcinoma (not significant), a result which was compatible with the fact that malignant lymphoma responds better to chemotherapy.29 In our study, no P-Gp-positive cases were observed in either gastric adenocarcinoma or malignant lymphoma. Mizoguchi et al. reported that the tissue levels of MDR1 mRNA was 1.5 units in gastric carcinoma and 3.6 units in colorectal carcinoma.30 As a result, some P-Gp gene expression was detectable, but the absolute amount of the expressed epitope was not sufficient for an immunohistochemical evaluation. The overall incidence of MRP expression in gastric adenocarcinoma was 55%, with well and moderately differentiated tumors showing a significantly higher expression than poorly differentiated tumors (P , 0.05). This result was compatible with the report that drug resistance was higher for well differentiated gastric
D. Alexander et al.: Multidrug Resistance in Gastric Cancer
adenocarcinoma than for poorly differentiated tumors based on drug sensitivity tests.31 Recently, Endo et al.32 reported that 34 out of 75 gastric cancers were MRPpositive on histochemical staining and also that MRPpositive tumor cells were less sensitive to several anticancer agents in an MTT assay. Therefore, the relatively high drug resistance of well differentiate gastric adenocarcinoma seems to be caused by the high incidence of MRP-positive tumor cells. Izquierdo et al.33 reported that LRP was a strong and independent prognostic indicator for predicting the response to standard chemotherapy and survival in patients with advanced ovarian carcinoma. In our study, however, only 10% of all gastric adenocarcinomas were LRP-positive and the incidence seemed to be lower than that reported previously. This difference in the LRP expression may have been caused by different tissue characteristics. A similar analysis was also carried out by Ota et al.34 for MRP expression in non-small cell lung cancer, and MRP-positive tumors were thus found to have a worse prognosis, but this was not so for patients with adenocarcinoma. In the case of gastric adenocarcinoma, there seems to be no relationship between MRP expression and prognosis,32 but the reason has yet to be determined. To clarify the biological meaning of MRP expression, the correlation between PCNA LI and MRP expression was studied. The PCNA LI was slightly higher in poorly differentiated adenocarcinoma compared with well differentiated adenocarcinoma (P 5 NS), and no significant difference was seen between the PCNA LI of MRP-positive and MRP-negative tumors. This result suggests that MRP is an independent factor predicting prognosis.35 The discrepancies in MRP expression and its meaning as a prognostic factor may thus be caused by differences in drug sensitivity. Both ovarian carcinoma and non-small cell lung cancer are called “drugsensitive” cancers. In these cancers, the expression of MDR-associated proteins may cause a high level of drug resistance and a poor prognosis. The heterogeneity of the tumor staining profile was noticed in MRP-positive specimens. Regarding the overexpression of the MDR gene, variants with different levels of amplification or activation of the gene have been reported for P-Gp.36 The amount of P-Gp in a cell depends on the gene activity of MDR1 and is presumed to be related to the strength of multidrug resistance,37 since it functions as a drug efflux pump. The same heterogeneity of epitope overexpression has also been reported for MRP.38 Therefore, an assessment of MRP overexpression may be useful in determining the most appropriate chemotherapy regimen for gastric cancer. Acknowledgments. The authors would like to thank Emeritus Prof. M. Mito for his generous support and encouragement,
405 and Prof. K. Yamamura for his advice on the statistical analysis. We are also grateful to Dr. N. Miyokawa for providing clinical materials and to the doctors of the Second Department of Surgery for their kind assistance.
References 1. Charles SF (1997) Chemotherapy for advanced gastric cancer: where do we stand? J Clin Oncol 15:3299–3300 2. Charles SF, Robert JM (1995) Gastric carcinoma. N Engl J Med 333:32–41 3. Cunningham D (1988) Gastric cancer. The recognition of a chemosensitive tumour. Br J Cancer 58:695–699 4. Neri B, de Leonardis V, Romano S, Andreoli F, Pernice LM, Bruno L, Borrelli D, Valeri A, Fabbroni S, Intini C, Cini G (1996) Adjuvant chemotherapy after gastric resection in node-positive cancer patients: a multicentre randomised study. Br J Cancer 73:549–552 5. Hermans J, Bonenkamp JJ, Boon MC, Bunt AM, Ohyama S, Sasako M, van de Velde (1993) Adjuvant therapy after curative resection for gastric cancer: Meta-analysis of randomized trials. J Clin Oncol 11:1441–1447 6. Douglass HO Jr (1989) Gastric cancer. Current status of adjuvant therapy. Oncology 3:61–66 7. Allum WH, Hallissey MT, Ward LC, Hockey MS (1989) A controlled, prospective, randomised trial of adjuvant chemotherapy or radiotherapy in resectable gastric cancer: interim report. Br J Cancer 60:739–744 8. Coombes RC, Schein PS, Chilvers CE, Wils J, Beretta G, Bliss JM, Rutten A, Amadori D, Cortes-Funes H, Villar-Grimdt A (1990) A randomized trial comparing adjuvant fluorouracil, doxorubicin and mitomycin with no treatment in operable gastric cancer. International collaborative cancer group. J Clin Oncol 8:1362–1369 9. Lise M, Nitti D, Marchet A, Sahmoud T, Buyse M, Duex N, Fiorentino M, Dos Santos JG, Labianca R, Rougier P, Gignoux M (1995) Final results of a phase III clinical trial of adjuvant chemotherapy with the modified fluorouracil, doxorubicin, and mitomycin regimen in resectable gastric cancer. J Clin Oncol 13:2757–2763 10. Bleiberg H, Gerard B, Deguiral P (1992) Adjuvant therapy in resectable gastric cancer. Br J Cancer 66:987–991 11. Pastan I, Gottesman M (1987) Multiple drug resistance in human cancer. N Engl J Med 316:1388–1393 12. Endicott JA, Ling V (1989) The biochemistry of P-glycoprotein mediated multidrug resistance. Annu Rev Biochem 58:137–171 13. Kaye SB (1988) The multidrug resistance phenotype. Br J Cancer 58:691–694 14. Cornwell MM, Tsuruo T, Gottesman MM, Pastan I (1987) ATPbinding properties of P-glycoprotein from multidrug resistant KB cells. FASEB J 1:51–54 15. Gottesman MM, Pastan I (1988) The multidrug transporter, a double-edged sword. J Biol Chem 263:12163–12166 16. Fojo AT, Ueda K, Slamon DJ, Poplack DG, Gottesman MM, Pastan I (1987) Expression of a multidrug-resistance gene in human tumors and tissues. Proc Natl Acad Sci USA 84:265–269 17. Scheper RJ, Broxterman HJ, Scheffer GL, Meijer CJLM, Pinedo HM (1992) Drug-transporter proteins in clinical multidrug resistance. Clin Chim Acta 206:25–32 18. Grant CE, Valdimarsson G, Hipfner DR, Almquist KC, Cole SP, Deeley RG (1994) Overexpression of multidrug resistanceassociated protein (MRP) increases resistance to natural product drugs. Cancer Res 54:357–361 19. Cole SP, Bhardwaj G, Gerlach JH, Mackie JE, Grant CE, Almquist KC, Stewart AJ, Kurz EU, Duncan AM, Deeley RG (1992) Overexpression of a transporter gene in a multidrugresistant human lung cancer cell line. Science 258:1650–1654
406 20. Krishnamachary N, Center MS (1993) The MRP gene associated with a non-P-glycoprotein multidrug resistance encodes a 190-KDa membrane bound glycoprotein. Cancer Res 53:3658– 3661 21. Schadendorf D, Makki A, Stahr C, Van Dyck A, Wanner R, Scheffer GL, Flens MJ, Scheper R, Henz BM (1995) Membrane transport proteins associated with drug resistance expressed in human melanoma. Am J Pathol 147:1545– 1552 22. Zaman GJ, Flens MJ, van Leusden MR, de Haas M, Mulder HS, Lankelma J, Pinedo HM, Scheper RJ, Baas F, Broxterman HJ, Borst P (1994) The human multidrug resistance-associated protein (MRP) is a plasma membrane drug efflux pump. Proc Natl Acad Sci USA 91:8822–8826 23. Scheper RJ, Broxterman HJ, Scheffer GL, Kaaijk P, Dalton WS, van Heijningen TH, van Kalken CK, Slovak ML, de Vries EG, van der Valk P, Meijer CJ, Pinedo HM (1993) Overexpression of a Mr 110 000 vesicular protein in non-P-glycoprotein mediated multidrug-resistance. Cancer Res 53:1475–1479 24. Scheffer GL, Wijngaard PL, Flens MJ, Izquierdo MA, Slovak ML, Pinedo HM, Meijer CJ, Clevers HC, Scheper RJ (1995) The drug resistance relatted protein LRP is the human major vault protein. Nat Med 1:578–582 25. Noonan KE, Beck C, Holzmayer TA, Chin JE, Wunder JS, Andrulis IL, Gazdar AF, Willman CL, Griffith B, Von Hoff DD, Roninson IB (1990) Quantitative analysis of MDR-1 (multi drug resistance) gene expression in human tumors by polymerase chain reaction. Proc Natl Acad Sci USA 87:7160–7164 26. Verrelle P, Meissonnier F, Feillel V, Dionet C, Kwiatkowski F, Plagne R, Chassagne J (1991) Clinical relevance of immunohistochemical detection of multidrug resistance P-glycoprotein in breast carcinoma. J Natl Cancer Inst 83:111–116 27. Hipfner DR, Gauldie SD, Deeley RG, Cole SP (1994) Detection of the Mr 190 000 multidrug resistance protein, MRP with monoclonal antibodies. Cancer Res 54:5788–5792 28. Flens MJ, Izquierdo MA, Scheffer GL, Fritz JM, Meijer CJ, Scheper RJ, Zaman GJ (1994) Immunochemical detection of the multidrug resistance associated protein MRP in human multidrug resistant tumor cells by monoclonal antibodies. Cancer Res 54:4557–4563
D. Alexander et al.: Multidrug Resistance in Gastric Cancer 29. Yuen AR, Sikic BI (1994) Multidrug resistance in lymphomas. J Clin Oncol 12:2453–2459 30. Mizoguchi T, Yamada K, Furukawa T, Hidaka K, Hisatsugu T, Shimazu H, Tsuruo T, Sumizawa T, Akiyama S (1990) Expression of the MDR-1 gene in human gastric and colorectal carcinomas. J Natl Cancer Inst 82:1679–1683 31. Maehara Y, Anai H, Kusumoto H, Sugimachi K (1987) Poorly differentiated human gastric carcinoma is more sensitive to antitumor drugs than well differentiated carcinoma. Eur J Surg Oncol 13:203–206 32. Endo K, Maehara Y, Ishiyoshi Y, Kusumoto T, Sakaguchi Y, Ohno S, Sugimachi K (1996) Multidrug resistance-associated protein expression in clinical gastric carcinoma. Cancer 77:1681–1687 33. Izquierdo MA, van der Zee AG, Vermorken JB, van der Valk P, Belien JA, Giaccone G, Scheffer GL, Flens MJ, Pinedo HM, Kenemans P, Meijer CJ, de Vries EG, Scheper RJ (1995) Drug resistance-associated marker Lrp for prediction of response to chemotherapy and prognoses in advanced ovarian carcinoma. J Natl Cancer Inst 87:1230–1237 34. Ota E, Abe Y, Oshika Y, Ozeki Y, Iwasaki M, Inoue H, Yamazaki H, Ueyama Y, Takagi K, Ogata T, Tamaoki N, Nakamura M (1995) Expression of the multidrug resistanceassociated protein (MRP) gene in non-small-cell lung cancer. Br J Cancer 72:550–554 35. Jain S, Filipe MI, Hall PA, Waseem N, Lane DP, Levison DA (1991) Prognostic value of proliferating cell nuclear antigen in gastric carcinoma. J Clin Pathol 44:655–659 36. Shen DW, Fojo A, Chin JE, Roninson IB, Richert N, Pastan I, Gottesman MM (1986) Human multidrug-resistant cell lines: increased mdr1 expression can precede gene amplification. Science 232:643–645 37. Kakehi Y, Kanamaru H, Yoshida O, Ohkubo H, Nakanishi S, Gottesman MM, Pastan I (1988) Measurement of multidrugresistance messenger RNA in urogenital cancers; elevated expression in renal cell carcinoma is associated with intrinsic drug resistance. J Urol 139:862–865 38. Slovak ML, Ho JP, Bhardwaj G, Kurz EU, Deeley RG, Cole SP (1993) Localization of a novel multidrug resistance-associated gene in the HT 1080/DR4 and H69AR human tumor cell lines. Cancer Res 53:3221–3225