Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
CCR5 receptor antagonist wikipedia , lookup
NMDA receptor wikipedia , lookup
Cannabinoid receptor antagonist wikipedia , lookup
Discovery and development of angiotensin receptor blockers wikipedia , lookup
Theralizumab wikipedia , lookup
Discovery and development of antiandrogens wikipedia , lookup
Drug discovery wikipedia , lookup
Neuropharmacology wikipedia , lookup
Identification of Epidermal Growth Factor Receptor and c-erbB2 Pathway Inhibitors by Correlation With Gene Expression Patterns Katja Wosikowski, Danita Schuurhuis, Kathryn Johnson, Kenneth D. Paull, Timothy G. Myers, John N. Weinstein, Susan E. Bates* Background: Growth factor receptor-signaling pathways are potentially important targets for anticancer therapy. The interaction of anticancer agents with specific molecular targets can be identified by correlating target expression patterns with cytotoxicity patterns. We sought to identify new agents that target and inhibit the activity of the epidermal growth factor (EGF) receptor and of c-erbB2 (also called HER2 or neu), by correlating EGF receptor, transforming growth factor (TGF)-a (a ligand for EGF receptor), and c-erbB2 messenger RNA (mRNA) expression levels with the results of cytotoxicity assays of the 49 000 compounds in the National Cancer Institute (NCI) drug screen database. Methods: The levels of mRNAs were measured and used to generate a molecular target database for the 60 cell lines of the NCI anticancer drug screen. The computer analysis program, COMPARE, was used to search for cytotoxicity patterns in the NCI drug screen database that were highly correlated with EGF receptor, TGF-a, or c-erbB2 mRNA expression patterns. The putative EGF receptor-inhibiting compounds were tested for effects on basal tyrosine phosphorylation, in vitro EGF receptor tyrosine kinase activity, and EGF-dependent growth. Putative ErbB2-inhibiting compounds were tested for effects on antibody-induced ErbB2 tyrosine kinase activity. Results: EGF receptor mRNA and TGF-a mRNA levels were highest in cell lines derived from renal cancers, and c-erbB2 mRNA levels were highest in cells derived from breast, ovarian, and colon cancers. Twenty-five compounds with high correlation coefficients (for cytotoxicity and levels of the measured mRNAs) were tested as inhibitors of the EGF receptor or c-erbB2 signaling pathways; 14 compounds were identified as inhibitors of these pathways. The most potent compound, B4, inhibited autophosphorylation (which occurs following activation) of ErbB2 by 50% in whole cells at 7.7 µM. Conclusions: Novel EGF receptor or c-erbB2 pathway inhibitors can be identified in the NCI drug screen by correlation of cytotoxicity patterns with EGF receptor or c-erbB2 mRNA expression levels. [J Natl Cancer Inst 1997;89:1505–15] Receptor tyrosine kinases contain an extracellular ligandbinding domain, a membrane-spanning domain, and an intracellular domain with tyrosine kinase activity. The type I family of growth factor receptors includes the epidermal growth factor (EGF) receptor, and the tyrosine kinases c-erbB2, c-erbB3, and c-erbB4 (1,2). Growth factors interact with their specific receptors at the cell surface, activating the receptor tyrosine kinase. A cascade of downstream signaling proteins are activated, includ- ing Ras, Raf, mitogen-activated protein kinase (MAPK) kinase, and MAPK, eventually leading to altered gene expression and increased growth rate (3–6). The EGF receptor is a 170-kd transmembrane glycoprotein that is found on many epithelial cell types. It is activated by at least three ligands, EGF, transforming growth factor-a (TGF-a), and amphiregulin (7,8). An autocrine growth pathway has been proposed for TGF-a in normal and malignant human breast epithelium (9). ErbB2 (also called HER2 and neu) is a 185-kd transmembrane tyrosine kinase with considerable homology to the EGF receptor, although a ligand for ErbB2 has not yet been clearly identified. EGF receptor and c-erbB2 have 43% sequence homology between their extracellular domains and 82% between their tyrosine kinase domains. Both EGF receptor and c-erbB2 are overexpressed in certain types of tumors, notably, breast, ovary, bladder, colon, kidney, and head and neck cancers as well as squamous carcinomas of the lung (10–13). In addition, a relationship between overexpression of c-erbB2 or EGF receptor and poor prognosis in patients with breast cancer has been reported by several groups (12,14–17). Therefore, the EGF receptor and c-erbB2 signaling pathways are potentially important targets for anticancer therapy. Chemotherapy has proven to be successful in selected cancers, stimulating the search for other effective antineoplastic drugs. To identify new agents with potential antitumor activity, the National Cancer Institute (NCI) established a panel of 60 human tumor cell lines organized into subpanels representing leukemia, melanoma, and cancers of the lung, breast, colon, kidney, ovary, prostate, and central nervous system (CNS). The cytotoxic effects of natural products, synthetic compounds, and semisynthetic compounds are being determined using these 60 cell lines, and the data are stored in a database that now contains cytotoxicity information from assays with more than 49 000 compounds (18–20). The patterns of cytotoxicity among the cell lines in the screen are evaluated using the computer analysis program COMPARE. This program calculates the degree of similarity (Pearson correlation coefficient, or ‘‘r’’ value) between the pattern of cytotoxicity of an index compound and all other compounds in the database. Similar patterns of cytotoxic*Affiliations of authors: K. Wosikowski, D. Schuurhuis, K. Johnson, S. E. Bates (Medicine Branch, Division of Clinical Sciences), K. D. Paull (Information Technology Branch, Development Therapeutics Program, Division of Cancer Treatment, Diagnosis, and Centers), T. G. Myers, J. N. Weinstein (Laboratory of Basic Sciences), National Cancer Institute, Bethesda, MD. Correspondence to: Susan E. Bates, M.D., National Institutes of Health, Bldg. 10, Rm. 12N226, Bethesda, MD 20892. E-mail: [email protected] See ‘‘Notes’’ following ‘‘References.’’ © Oxford University Press Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 ARTICLES 1505 ity identify compounds with similar targets and mechanisms of action (19,20). For example, agents that bind tubulin are highly correlated despite diversity in structure (21). Measurement of various molecular markers has been undertaken in the 60 cell lines, using COMPARE to correlate the patterns of expression with the cytotoxicity patterns for the 49 000 compounds in the NCI database (22,23). This last approach was used to identify unknown compounds in the drug screen database that may target the EGF receptor, the c-erbB2 signaling pathway, or the action of TGF-a. The success of this approach relies on three prerequisites: that the level of a target is quantitatively related to its activity in the cell, that inhibition of that target will have an impact on cellular growth, and that compounds exist for which inhibition of that target is a predominant mechanism of action. We determined the levels of messenger RNAs (mRNAs) encoding the EGF receptor, ErbB2, and TGF-a in the 60 cell lines comprising the NCI anticancer drug screen, generated expression patterns, and used COMPARE to find compounds with correlated patterns of sensitivity. The putative EGF receptor compounds were tested for effects on basal tyrosine phosphorylation, in vitro EGF receptor tyrosine kinase activity, and EGF-dependent growth. Putative ErbB2 compounds were tested for effects on ErbB2 tyrosine kinase activity in biochemical assays and in whole cells. Methods Cell Culture The cell lines comprising the NCI anticancer drug screen panel were obtained and processed as previously described (18,24,25). A number of additional cell lines were used as positive controls and as tools in the assays testing activity of identified compounds. SK-BR-3 and MDA-MB-453 cells were obtained from the American Type Culture Collection (Rockville, MD) and A431 cells were provided by Ira Pastan (NCI, Bethesda, MD). The EGF receptor overexpressing MCF-7 TH cells are a drug-resistant subline isolated from parental MCF-7 cells by intermittent exposure to 1 mM doxorubicin. The cells were a gift from Tom Hamilton (Fox Chase Cancer Center, Philadelphia, PA). Cells were grown as monolayer cultures in Iscove’s minimal essential medium (IMEM) (Biofluids Inc., Rockville, MD) supplemented with 10% fetal calf serum (FCS) (Life Technologies, Inc. [GIBCO-BRL], Gaithersburg, MD), 2 mM glutamine, 15 mM HEPES, and 25 mg/mL gentamicin (all from Biofluids Inc.) in a humidified 5% CO2 atmosphere at 37 °C. The MCF-10A immortalized human breast epithelial cells were provided by David Salomon (NCI, Bethesda, MD) and were cultured in Dulbecco’s minimal essential medium (DMEM)–Ham’s F12 (1:1) medium containing 5% horse serum, 100 U/mL penicillin, and 100 mg/mL streptomycin (all from Biofluids Inc.) plus 20 mM HEPES, 4 mg/mL insulin, 500 ng/mL hydrocortisone, and 20 ng/mL EGF (all from Collaborative Biomedical Products, Bedford, MA) (26). RNA Extraction and Electrophoresis Total RNA was extracted by homogenization of cells in guanidine isothiocyanate buffer followed by centrifugation over a cesium chloride cushion (27). To compare the RNA quantities and quality, RNA concentration was measured by spectrophotometry and 5 mg total RNA was electrophoretically separated in 1% agarose–6% formaldehyde gels. Gels were stained with ethidium bromide and examined to confirm regular loading. RNA that was not comparably loaded was re-extracted, remeasured, and re-evaluated. RNase Protection Assay A 141-base pair (bp) fragment of the EGF receptor complementary DNA (cDNA), a 470-bp fragment of c-erbB2 cDNA, and a 603-bp fragment of TGF-a DNA were used to generate 32P-labeled antisense riboprobes by SP6, T7, and T7 polymerase transcription (Promega Corp., Madison, WI), respectively, in the presence of [a-32P]uridine triphosphate (specific activity, 3000 Ci/mmol) (Du 1506 ARTICLES Pont NEN, Boston, MA) as described previously (28). For the ribonuclease (RNase) protection assay, 30 mg of total RNA was hybridized with 3 × 105 cpm-labeled riboprobe and then digested for 30 minutes at 25 °C with 40 mg/mL RNase A and 28 mg/mL RNase T1. Following extraction, samples were separated on a 6% polyacrylamide gel and autoradiography was performed. The relative amount of mRNA was quantified using a Phosphor Imager, coupled to Image Quant software (Molecular Dynamics, Sunnyvale, CA), or a Fotoeclipse densitometer (Fotodyne Inc., Hartland, WI). COMPARE Analysis and Determination of Pearson Correlation Coefficients Three separate COMPARE analyses were performed using the relative expression of the EGF receptor, c-erbB2, and TGF-a mRNAs in the 60 cell lines of the NCI drug screen. Pearson correlation coefficients, or ‘‘r’’ values, were obtained for each expression pattern, with the cytotoxicity pattern of compounds in the NCI drug screen database. The database stores the bioactivity data of compounds as the −log10 GI50 (e.g., a GI50 of 10−4 M is stored as +4 and a GI50 of 10−9 is stored as +9) (24). GI50 is the designation for a time zero-corrected IC50, the concentration of a compound causing 50% growth inhibition. Thus, cell lines represented by larger values indicate sensitivity to the tested compound, whereas cell lines represented by smaller numbers indicate resistance. When +log10 values of the EGF receptor, c-erbB2, and TGF-a mRNA levels were used for the COMPARE analysis, then greater mRNA expression correlated with enhanced sensitivity of the cell to the database compound. A list of compounds having the highest correlation coefficients was sorted in order of descending correlation coefficients. Compounds at the top of the list are those that correlate best with the expression pattern being analyzed. Compounds were obtained from the Drug Screen repository (Drug Synthesis and Chemistry Branch, NCI, Bethesda, MD), having been contributed from diverse sources, including academia, industry, and government. For the purpose of this article, compounds chosen from the NCI database and used in further experiments were assigned the E, B, or EB numbers shown in Table 1. This nomenclature allows rapid identification of the COMPARE analysis from which the compound was obtained: E, EGF receptor; B, c-erbB2; or EB, both EGF receptor and c-erbB2. Compounds are originally contributed to the drug screen as discreet (indicating that the structure is to be held confidential) or nondiscreet). Both discreet and nondiscreet compounds were used in the study. Cluster Analysis Patterns of averaged data GI50 data were obtained as previously described (29). The pattern for each drug was centered by subtracting from each value the pattern mean and dividing by the pattern standard deviation. The Euclidean distance between patterns normalized in this fashion approximate the additive inverse of the respective Pearson correlation coefficients (Distance 4 1 − Pearson correlation coefficient) and can thus be compared with the Pearson correlation coefficient calculation used by the COMPARE program. The JMP computer software for Macintosh (version 3.1.5, SAS Institute, Cary, NC) was used to generate the dendogram by the average-linkage clustering method. Reference compounds selected from a group of 216 standard agents on the basis of their mechanism of action were added to the group before clustering. Average linkage is a commonly used technique, which is agglomerative, starting with individual compounds and ending with a single cluster containing all of the compounds. Each join of individual compounds or previously joined clusters represents a node on the tree that is plotted at a particular similarity value (mean Pearson correlation coefficient). When the node joins two compounds, the Pearson correlation coefficient is simply the average of the two. When two previously formed clusters are joined, the mean Pearson correlation coefficient is calculated by considering all possible intercluster compound pairings. For example, the topoisomerase II inhibitors, labeled with a m in Fig. 3, are all joined into a single cluster whose node has a mean Pearson correlation coefficient of .55. Thus, the topoisomerase II inhibitors in the NCI screen have cytotoxicity patterns that are correlated on average at the .55 level. EGF-Dependent Growth Assay The inhibitory effect of compounds (E1 through E11 and EB1 through EB5) on EGF-dependent growth of MCF-10A breast epithelial cells was determined in Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 Table 1. Thirty compounds among the top 56 identified by COMPARE, using the EGF receptor, c-erbB2, and TGF-a expression profiles* EGF receptor c-erbB2 No. PCC Rank Name No. PCC Rank 676495 676497 623436 671526 676498 646148 Discreet 616511 656128 Discreet 643315 682027 669365 672768 615554 646019 680581 680625 Discreet 675223 656125 656126 631229 676496 646015 669083 650706 651295 654970 668885 0.726 0.715 0.704 0.658 0.631 0.626 0.611 0.599 0.596 0.593 0.579 0.579 0.574 0.554 0.553 0.548 0.545 0.544 0.543 0.541 0.540 0.537 0.536 0.534 0.528 0.523 0.523 0.518 0.516 0.508 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 37 TP38-001 TP4EK TGFa.-PE40 Toxin.delta.53L TP4EK-K6 EB1 643315 646019 655128 Discreet 683039 648202 646148 611388 656125 657871 646081 640490 668937 640820 Discreet 668885 674987 655439 673189 669714 Discreet 682883 10121 682879 671373 Discreet 669627 178249 668419 669164 0.614 0.566 0.560 0.558 0.554 0.552 0.542 0.532 0.504 0.504 0.499 0.494 0.491 0.485 0.482 0.480 0.477 0.471 0.469 0.469 0.468 0.465 0.461 0.458 0.457 0.453 0.449 0.447 0.435 0.423 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 24 27 30 33 34 38 46 56 E1 EB2 PD153717 E2 E3 EB3 EB4 E4 TP38-279 E5 E6 E7 EB5 TGF-a Name EB2 EB3 e23(dsFV)PE38 EB1 B1 EB4 B2 B3 B4 EB5 B5 B6 B7 B8 B9 No. PCC Rank Name 671526 Discreet 643892 Discreet 643438 Discreet 682765 648200 636467 Discreet 311046 650426 Discreet 648202 643315 644673 668456 616511 676497 634308 655128 Discreet 669365 644674 641886 642550 642195 675223 673999 676495 0.451 0.451 0.443 0.426 0.424 0.422 0.421 0.417 0.415 0.4122 0.408 0.406 0.406 0.404 0.402 0.401 0.401 0.400 0.399 0.399 0.398 0.396 0.395 0.394 0.391 0.389 0.384 0.383 0.381 0.377 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 36 Toxin.delta.53L TP4EK PD153717 TP38-001 *EGF 4 epidermal growth factor; TGF-a 4 transforming growth factor; PCC 4 Pearson correlation coefficient. 96-well plates. Eight hundred MCF-10A cells were plated in the absence and presence of 20 ng/mL EGF in DMEM–Ham’s F12 (1:1) medium containing 2.5% horse serum and supplemented as described above. Cells were treated with increasing concentration of the compound and fixed and stained after 5 days in culture as described. The results were expressed as the difference (in %) from the control value (untreated wells without EGF or compound) that was plotted as 0%. Positive values indicate more growth than in control untreated cells, and negative values indicate inhibition of growth. Experiments were performed in quadruplicate. Cell Membrane Kinase Assay Cells were plated in 10-cm Petri dishes and were allowed to attach overnight. To avoid interference of EGF receptor or ErbB2 ligands possibly present in the serum, the medium was replaced with serum-free medium composed of phenol– red free IMEM supplemented with 2 mM glutamine and 15 mM HEPES (Biofluids Inc.) plus 1 mg/mL essentially fatty acid-free bovine serum albumin [BSA], 5 mg/mL insulin, 5 mg/mL transferrin, and 5 ng/mL selenium (all from Sigma Chemical Co., St. Louis, MO) for 24–48 hours. For the EGF receptor kinase assay, membrane extracts were prepared. A431 cells were washed two times in ice-cold PBS and lysed in hypotonic lysis buffer (10 mM NaCl, 15 mM EDTA, 8.5 mM NaH2PO4, 11.5 mM Na2HPO4, 20 mg/mL aprotinin [ICN Biomedicals, Aurora, OH], 20 mg/mL leupeptin, 1 mM phenylmethylsulfonyl fluoride [PMSF], and 1 mM sodium orthovanadate [all from Sigma Chemical Co.]). After sonication, lysates were cleared by centrifugation (200g, 5 °C, 15 minutes). Supernatants were centrifuged (100 000g, 5 °C, 60 minutes) and membranes were resuspended in 50 mM HEPES, pH 7.4, 125 mM NaCl, 10% (vol/vol) glycerol. Two micrograms of membrane preparation was preincubated with 120 mM compound or genistein (LC Laboratories, Woburn, MA) in the presence or absence of 2 × 10−8 M EGF for 10 minutes at room temperature. The kinase reaction was initiated by the addition of an equal volume (20 mL) of 50 mM HEPES, 125 mM NaCl, 4 mM MnCl2, 24 mM MgCl2, and 4 mM adenosine 58-triphosphate (ATP) with 0.8 mCi [g-32P]ATP (specific activity, 3000 Ci/mmol; Du Pont, NEN) per reaction for 4 minutes at 4 °C as described by Osherov (30). The reaction was stopped by the addition of sodium dodecyl sulfate (SDS)-sample buffer. For the ErbB2 assay, serum-starved cells were stimulated with 10–25 mg/mL anti-ErbB2 recombinant human monoclonal antibody (HER2, provided by Genentech, Inc., South San Francisco, CA) (31) for 5 minutes at 37 °C and harvested in TNESV (50 mM Tris, pH 7.6, 1% [vol/vol] Nonidet P-40 [NP-40], 2 mM EDTA, 100 mM NaCl, 1 mM vanadate, 20 mg/mL aprotinin, 20 mg/mL leupeptin, and 1 mM PMSF) as described (32). ErbB2 was immunoprecipitated from 1 mg of protein using anti-ErbB2 monoclonal antibody (MAb), recognizing a cell surface epitope (c-Neu; Ab-5; Oncogene Science, Inc., Manhasset, NY) and protein A-sepharose beads (Pharmacia Biotech, Inc., Piscataway, NJ). Immunoprecipitates were preincubated with 120-mM compounds or the erbstatin analog (methyl 2,5-dihydroxycinnamate, LC Laboratories) in buffer (50 mM HEPES, pH 7.4, 125 mM NaCl) for 15 minutes at 4 °C, followed by incubation with an equal volume (30 mL) of reaction mix (50 mM HEPES, 125 mM NaCl, 4 mM MnCl2, 24 mM MgCl2, and 2 mM ATP with 15 mCi [g-32P]ATP per reaction) for 5 minutes at 4 °C. The reaction was stopped by the addition of hot SDS-sample buffer. In both assays, proteins were separated by 7% SDS–polyacrylamide gel elec trophoresis (SDS–PAGE). Gels were fixed in 10% (vol/vol) acetic acid, 30% (vol/ vol) methanol, and 10% (vol/vol) glycerol, then autoradiography was performed. ErbB2 Autophosphorylation Assay In 60-mm Petri dishes, 106 MDA-MB-453 cells were plated in IMEM containing 10% FCS and allowed to attach. Cells were serum starved for 24 hours in serum-free IMEM medium before the addition of 200 mM test compound for 1 hour, followed by stimulation with 25 mg/mL anti-ErbB2 antibody (rhuMAb HER2, Genentech, Inc.) for 5 minutes and then lysis in TNESV. Twenty to 70 microgram protein extracts were separated by 7% SDS–PAGE. The proteins were transferred to Immobilon-P (Millipore Corp., Bedford, MA) membranes, blocked with 5% BSA, and probed with antiphosphotyrosine antibody (RC20H; Transduction Lab., Cincinnati, OH) to detect tyrosine phosphorylated ErbB2 or an antibody to detect actin (Ab-1; Oncogene Research Products, Cambridge, Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 ARTICLES 1507 MA) to confirm regular loading of the gel as previously described (32). IC50 determinations were made by plotting the level of phosphorylated ErbB2 obtained at four concentrations of test compound (range, 4–200 mM). whereas those to the left of the mean indicate the lowest levels. Results Selection of Compounds Based on the Correlation of Cytotoxicity With High mRNA Expression Expression of EGF Receptor, c-erbB2, and TGF-a mRNAs in the 60 Cell Lines of the NCI Anticancer Drug Screen To identify compounds that target the EGF receptor, c-erbB2, and the TGF-a pathway, the levels of the respective mRNAs in the 60 cell lines in the drug screen were determined by RNase protection assays. Fig. 1 shows representative mRNA results for kidney, colon, breast, and ovarian cell lines, along with photographs of the ethidium bromide-stained gels demonstrating RNA integrity and the comparability of quantities of RNA used in the RNase protection assays. The levels of EGF receptor, c-erbB2, and TGF-a mRNAs in all cell lines were determined in at least two independent experiments. A cell line known to have high levels of expression of these mRNAs was included in each experiment as a positive control (MCF-7 TH, MDA-MB-453, or MDA-MB-231). The expression levels were quantitated by use of a fotoeclipse densitometer or a Phosphor Imager and were related to the levels of expression in the control MCF-7 TH, MDA-MB-453, and MDA-MB-231 cells, which were assigned a value of 1000 for EGF receptor, c-erbB2, and TGF-a mRNAs, respectively. Levels for EGF receptor in the 60 cell lines ranged from 0 to 3763; for c-erbB2, from 0 to 27 973; and for TGF-a, from 0 to 10 091. EGF receptor expression was highest in renal cell lines. Highest levels of c-erbB2 expression were observed in ovarian, breast, and colon cells. TGF-a mRNA levels were highest in renal cell lines. Levels of all three mRNAs were lowest in leukemia cells. Generation of EGF Receptor, c-erbB2, and TGF-a Expression Patterns As shown previously, the cytotoxicity pattern of a given drug can predict its mechanism of action using the COMPARE program (19,21). A variety of molecular targets have been measured in the 60 cell lines and expression patterns have been generated to compare them to the cytotoxicity patterns in the database (22,23). As shown in Fig. 2, expression patterns for EGF receptor, c-erbB2, and TGF-a mRNAs in the 60 cell lines were generated using the +log10 of the quantitative values of EGF receptor, c-erbB2, and TGF-a mRNAs. In the mean graphs shown in Fig. 2, the bars pointing to the right of the mean (the vertical line in the center) represent the highest levels of EGF receptor, c-erbB2, and TGF-a mRNAs, To identify compounds to which cells expressing high levels of EGF receptor, c-erbB2, or TGF-a might be more sensitive, the COMPARE program was entered, using the +log10 values of the EGF receptor, c-erbB2, or TGF-a mRNA levels. EGF receptor, c-erbB2, and TGF-a mRNA expression in the 60 cell lines was correlated by COMPARE analysis with the sensitivity of these cell lines to the more than 49 000 drugs in the NCI database. Table 1 shows 30 compounds among the top 56 identified by COMPARE as having the highest Pearson correlation coefficients. Larger values represent a better positive correlation between high levels of EGF receptor, c-erbB2, or TGF-a mRNAs and sensitivity of the cells to the listed compound. The TGF-a-PE40 chimeric toxin, consisting of TGF-a coupled to Pseudomonas exotoxin (PE) (33) and four of its derivatives displayed the highest Pearson correlation coefficients (r 4 .726–.631) with EGF receptor expression pattern. As shown in Table 1, a total of six TGF-a-toxin chimeras were found among the top 24 compounds. TGF-a-toxin has previously been demonstrated to have increased cytotoxicity in cells that express high levels of EGF receptor on the cell surface (33). Also high on the EGF receptor COMPARE list is the tyrosine kinase inhibitor NSC-669365 (PD153717, Parke-Davis, Ann Arbor, MI), which has been shown to be a specific inhibitor of the EGF receptor tyrosine kinase. PD153717 inhibits EGF receptor autophosphorylation in A431 cells with an IC50 of 100 nM and specifically inhibits EGF-mediated mitogenesis in fibroblasts, but has no activity against PDGF- or FGF-mediated events (Fry DW: unpublished observations). Another specific EGF receptor kinase inhibitor, NSC-669364 [PD153035, Parke-Davis (34)] was identified in the top 100 compounds from the EGF receptor COMPARE analysis (r 4 .436). The highest correlation coefficient for compounds identified by the c-erbB2 COMPARE analysis was .614. A recombinant anti-ErbB2 immunotoxin (e23(dsFv)PE) (35) was fifth on the c-erbB2 COMPARE list. Recombinant immunotoxins selectively bind to and kill cells that are recognized by the antigen-binding domain. Two of the TGF-a-toxin chimeras and OVB3-PE (36), an antibody that recognizes an ovarian cellspecific antigen coupled to PE toxin, were also among the top 100 compounds in the c-erbB2 COMPARE analysis. Fig. 1. Expression of epidermal growth factor (EGF) receptor, transforming growth factor (TGF)-a, and c-erbB2 messenger RNAs in cell lines from the National Cancer Institute anticancer drug screen. Total RNA was isolated from the 60 cell lines comprising the anticancer drug screen and 30 mg were analyzed using a ribonuclease (RNase) protection assay. 32P-labeled antisense fragments from EGF receptor, TGF-a, and c-erbB2 were protected against RNase degradation by hybridization to total RNA. The protected bands were separated on a 6% polyacrylamide gel and visualized by autoradiography. Shown are autoradiograms of the same exposure time for EGF receptor, TGF-a, or c-erbB2 in ovarian, breast, renal, and colon cell lines. Ethidium bromide (EthBr)-stained agarose gels as shown were used to confirm the integrity of the RNA and to confirm that comparable amounts of total RNA were used in the assays. Both the 28S and the 18S ribosomal RNA bands are shown in the photograph. 1508 ARTICLES Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 Fig. 2. Expression patterns of epidermal growth factor (EGF) receptor, c-erbB2, and transforming growth factor (TGF)-a messenger RNA (mRNA) levels in the cell lines of the National Cancer Institute anticancer drug screen. Expression patterns were obtained by plotting positive and negative values (deltas) generated from a set of mRNA expression levels determined in each cell line. The mean +log10 of EGF receptor, TGF-a, or c-erbB2 mRNA expression level was obtained for the panel of 60 cell lines. The individual value for each cell line then was subtracted from the mean expression level for the panel to create the corresponding delta. Bars to the right identify cell lines that express high levels of EGF receptor, c-erbB2, and TGF-a mRNAs. As in the EGF receptor COMPARE analysis, the compound with the highest Pearson correlation coefficient (r 4 .451) in the TGF-a analysis was a TGF-a-chimeric toxin derivative (Toxin.delta.53L) (37). Two more TGF-a-toxin chimeras were among the top 36 compounds. The tyrosine kinase inhibitor PD153717 identified in the EGF receptor COMPARE analysis was also identified in the TGF-a COMPARE analysis (r 4 .395). Highlighted compounds were chosen, based on correlation coefficients and availability, for use in further experiments. Compounds with an E designation were derived from the EGF receptor COMPARE analysis, B compounds were from the c-erbB2 analysis, and EB compounds were found both in the EGF receptor and c-erbB2 analysis. Compounds E8 through E11 were derived from an early COMPARE analysis, performed using a preliminary dataset. Activity Patterns of Putative Kinase Inhibitors Compared With Those of Standard Anticancer Agents To evaluate whether the putative EGF receptor and ErbB2 kinase inhibitors have distinctive cytotoxicity patterns (indicative of distinct mechanisms of action), a cluster analysis was performed. The 25 compounds that were identified in Table 1 plus six toxin chimeras and two known EGF receptor tyrosine kinase inhibitors (Parke-Davis) were analyzed together with 24 standard antineoplastic agents of known mechanism of action. Much as a phylogenetic tree is used to indicate genetic sequence similarities, a cluster tree such as the one shown in Fig. 3 can be employed to indicate similarities in the cytotoxicity patterns of compounds tested in the screen. In the process of the clustering, compounds with similar cytotoxicity patterns are Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 ARTICLES 1509 correlation coefficient 4 .47), as were the group of toxin chimeras (labeled ‘‘■’’) (average correlation coefficient 4 .34). Only at correlation coefficient values of .20 to .17 do the known TGF-a-toxin chimeras join other compounds. For reference, this is about the value at which the antitubulin standard agents join the other standard agents of vastly different mechanisms of action (.14). Furthermore, the putative tyrosine kinases as a group join the standard agents at a value of −.06. If the hypothesis that differences in activity pattern indicate differences in mechanisms of action is correct, then the clustering reveals that distinct mechanisms of cytotoxicity prevail for each of the two groups: the erbB2 kinase inhibitors (●) and the TGF-a toxin chimeras (■). Furthermore, these mechanisms are different from those of the standard agents. The most distinctive difference between the standard agents and our selected compounds is that the leukemia cells were insensitive to our selected compounds. This is in contrast to most conventional cytotoxic (standard) agents, which are highly cytotoxic in leukemic cells. Identification of Compounds E1, E10, and E11 as Putative EGF Receptor Tyrosine Kinase Inhibitors Initial experiments demonstrated inhibition of overall tyrosine phosphorylation in A431 cell membranes by compounds E1 and E11, and to a lesser extent, E6, E7, and E10 (data not shown). Since basal levels of tyrosine phosphorylation reflect not only the EGF receptor pathway but also other tyrosine kinases, we directly assessed whether the compounds could inhibit autophosphorylation of the EGF receptor following stimulation by EGF using an in vitro kinase assay. Fig. 4 shows the results for EGF-induced EGF receptor autophosphorylation in membranes prepared from A431 cells. Genistein, a nonspecific tyrosine kinase inhibitor, was used as a control. Compounds E1 and E11 and, to a lesser extent, E9 and E10 inhibited EGF-induced EGF receptor autophosphorylation. Among the 15 compounds evaluated, three compounds (E1, E10, and E11) inhibited both basal tyrosine phosphorylation in A431 membranes and the basal and EGF-induced EGF receptor autophosphorylation. Fig. 3. Dendogram result of cluster analysis of compounds based on their cytotoxicity patterns in the cell lines of the drug screen. Members of clusters of particular interest were labeled m (topoisomerase inhibitors), ● (putative epidermal growth factor [EGF] receptor or ErbB2 kinase inhibitor), and ■ (transforming growth factor [TGF-a] toxin chimeras). PCC-Pearson correlation coefficient. The cluster analysis was performed as described in the ‘‘Methods’’ section. grouped together in a cluster. The compounds in each cluster are joined at a node from which branches leading to the cluster emerge. The height of the node indicates the degree of similarity between the compounds in the clusters. This is expressed in terms of a mean Pearson correlation coefficient. For example, topoisomerase II inhibitors, including the anthracyclines, mitoxantrone, and amsacrine (mAMSA) (labeled with a bold ‘‘m’’), form a cluster; the node from which the cluster emerges is shown at a correlation level of .55. Similarly, the tubulin-active agents, including the colchicines, vincas, and taxanes, form a separate cluster with a node at 0.5. A subset of the putative ErbB2 kinase inhibitors (labeled with a bold ‘‘●’’) were very similar in pattern to each other (average 1510 ARTICLES EGF-Dependent Growth Inhibition in MCF-10A Mammary Epithelial Cells To confirm the functional inhibition of the EGF receptor pathway, the compounds E1 through E11 and EB1 through EB5 and genistein were individually added to MCF-10A cells cultured in media with and without the presence of added EGF. MCF-10A cells are immortalized, nontransformed human mammary epithelial cells that display increased growth on the addition of exogenous EGF (26). As shown in Fig. 5, the addition of EGF (shaded bars) to control MCF-10A cells stimulated their growth 50%–75% above basal levels of growth (0% on the Y axis). Addition of compounds E1, E2, E8, E10, EB2, EB3, EB5, and genistein inhibited the EGF-stimulated growth without inhibiting basal growth. At higher concentrations, both basal and EGF-stimulated growth are inhibited, suggesting nonspecific cytotoxicity. The remaining eight compounds not shown had no specific effect on EGF-stimulated growth, since inhibition of EGF-dependent growth occurred concurrently with inhibition of basal growth. These results confirm the ability of the compounds to functionally inhibit the EGF receptor pathway-dependent growth of MCF-10A cells. Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 Fig. 4. In vitro effect of selected compounds on basal and epidermal growth factor (EGF)-induced EGF receptor autophosphorylation. Membrane preparations from A431 cells were preincubated with compound in the presence or absence of 2 × 10−8 M EGF (10 minutes, room temperature), followed by incubation for 4 minutes at 4 °C with reaction mix (see ‘‘Methods’’ section). Sodium dodecyl sulfate (SDS)-sample buffer was added and samples were heated and proteins were separated by gelelectrophoresis. Incorporation of [32P] ATP was visualized by autoradiography. Shown is the 170-kd band that represents the EGF receptor. Identification of B4 as the Most Potent Putative ErbB2 Tyrosine Kinase Inhibitor An anti-ErbB2 monoclonal antibody, rhuMAb HER2, was used to stimulate ErbB2 autophosphorylation in MDA-MB-453 human breast cancer cells prior to cell lysis and incubation with the test compounds or with the erbstatin analog positive control. Four of the 14 compounds evaluated (B1, B2, B3, and B5) inhibited ErbB2 autophosphorylation in membrane preparations following stimulation with rhuMAb HER2 antibody (data not shown). This assay was repeated in intact MDA-MB-453 cells that were pretreated with the test compounds and then stimulated with the rhuMAb HER2 antibody prior to cell lysis and examination for ErbB2 autophosphorylation. In this type of assay, the intracellular part of ErbB2 is not as freely accessible as in the in vitro tyrosine kinase assay. Seven compounds (EB2, EB5, B1, B2, B3, B4, and B9) at 200 mM and the erbstatin analog at 500 Fig. 5. Effect of selected compounds on epidermal growth factor (EGF)dependent growth in MCF-10A cells. MCF-10A cells were plated in medium without (dotted bars) or with (shaded bars) 20 ng/mL EGF. Cells were treated with increasing concentrations of the test compound and fixed and stained after 5 days in culture as described in the ‘‘Methods’’ section. The results were expressed as the difference (in %) from the control value (untreated wells without EGF or compound) that was plotted as 0%. Positive values indicate more growth than in control untreated cells and negative values indicate inhibition of growth. The experiment was performed in quadruplicate (mean ± standard deviation). The experiment shown is representative of three independent experiments. Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 ARTICLES 1511 mM inhibited basal and antibody-induced autophosphorylation of ErbB2, as shown in Fig. 6. We performed dose–response experiments with these compounds and identified B4 as the most potent agent. As a control for the nonspecific effect of cytotoxic agents, we examined the effect of doxorubicin, dactinomycin, or cisplatin on induced autophosphorylation of ErbB2. No inhibition was observed, even when 200 mM drug concentrations were used. Diverse Chemical Structures of the Putative EGF Receptor or c-erbB2 Pathway Inhibitors Fig. 7 presents the chemical structures of 11 of the 14 compounds that displayed an inhibitory effect in at least one of the functional assays performed (in vitro kinase, whole cell kinase, and EGF-dependent growth assay). The diversity of their chemical structures is apparent. No two compounds shown here belong to similar chemical classes. B4 has a structure that resembles the tyrphostin class of compounds. Discussion In this study, EGF receptor, c-erbB2, and TGF-a mRNA expression levels in the 60 cell lines comprising the NCI anticancer drug screen were analyzed and correlated by COMPARE analysis with sensitivity to the more than 49 000 compounds in the NCI database. High correlation coefficients, which should indicate compounds most effective in cell lines containing high levels of these mRNAs, were sought to identify potential inhibi- tors of growth factor signaling. Six compounds found among the top 30 correlated with EGF receptor mRNA level also were found in the top 30 correlated with c-erbB2 mRNA level. The putative growth factor signaling pathway inhibitors cluster in a dendogram as if they represent a mechanism of drug action distinct from those of the well-known cytotoxic antitumor agents. Three (E1, E10, and E11) of 15 compounds tested inhibited both basal tyrosine phosphorylation in A431 membranes and basal and EGF-induced EGF receptor autophosphorylation. Seven (E1, E2, E8, E10, EB2, EB3, and EB5) of 16 compounds inhibited EGF-dependent growth. Four (B1, B2, B3, and B5) and seven (B1, B2, B3, B4, B9, EB2, and EB5) of 14 compounds tested inhibited ErbB2 autophosphorylation in cell membrane preparations and in intact MDA-MB-453 cells, respectively. The most effective of these compounds (B4) inhibited antibodyinduced ErbB2 autophosphorylation, with an IC50 of 7.7 mM in the intact cell assay. These results confirmed the hypothesis that, by correlating expression patterns with cytotoxicity patterns, COMPARE analysis could identify compounds that inhibit EGF receptor and c-erbB2 pathways. Underlying this hypothesis is the prerequisite that the level of EGF receptor or c-erbB2 is quantitatively related to its activity or importance, and clinical studies have suggested this would be true. High levels of EGF receptor and c-erbB2 mRNA and protein expression have been observed in a variety of solid human cancers and have been associated with the presence of metastatic disease and an increased probability of tumor recurrence and poor patient survival (10,12–17). High levels of expression of EGF receptor or c-erbB2 mRNA or protein has also been correlated with resistance to chemotherapy (including doxorubicin, Fig. 6. Effect of selected compounds on basal and antibody-induced ErbB2 autophosphorylation in intact MDA-MB-453 cells. Cells were preincubated with 200 mM compound for 1 hour, then stimulated with rhuMAb HER2 for 5 minutes and harvested. Protein samples were run on sodium dodecyl sulfate–polyacrylamide gel electrophoresis gels and electrophoretically transferred to Immobilon membrane. The membrane was hybridized with anti-phosphotyrosine or anti-actin antibody. The 185-kd bands reflect the tyrosine phosphorylated state of the ErbB2 protein. 1512 ARTICLES Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 Fig. 7. Chemical structures of 11 from among the 14 compounds that displayed inhibitory effect in at least one of the functional assays performed. The E, B, and EB numbers assigned can be identified in Table 1. Ph 4 phenyl. tamoxifen, cisplatin, and vinblastine) in various cancer cell lines and tumors (14,38,39). Another prerequisite for identification of EGF receptor or c-erbB2 pathway inhibitors in the drug screen is that compounds exist that have inhibition of the EGF receptor or c-erbB2 pathways as the major mechanism of action. Numerous agents have been described that intercept the EGF receptor or c-erbB2 signaling pathway at binding of the ligand to its receptor or at tyrosine kinase activation. Natural tyrosine kinase inhibitors have been described, such as genistein, lavendustin A, erbstatin, and herbimycin A, which exhibit a rather broad specificity (40– 43). Since the naturally occurring tyrosine kinase inhibitors frequently require micromolar concentrations for kinase inhibition, 100–200 mM was selected as the concentration range to screen the compounds identified by COMPARE. Dose–response studies with the compounds from the c-erbB2 COMPARE analysis confirmed relatively high IC50s, ranging from 7.7 to 277 mM. More selective tyrosine kinase inhibitors have been reported, including NSC-669364 [PD153035, Parke-Davis (34)], which was identified in the EGF receptor COMPARE analysis, and numerous tyrphostins (44). These more selective inhibitors have also been shown to be very potent, inhibiting kinase activity in the nanomolar range. The compounds identified by COMPARE may be viewed as lead compounds for which structure-activity studies designed to increase potency could be conducted. New cancer drugs can be identified by looking for active chemical compounds through use of high throughput screening or can be designed by use of a rational and specific concept of drug–target interaction (30,34,44–48). The NCI drug screen has attempted to combine screening and molecular target strategies simultaneously. The chief advantage of such an approach is that compounds that interact with a target may be identified on the basis of cytotoxicity pattern, thus obviating the need to directly screen 49 000 compounds for their effect on that target. However, the target must be important, its activity in or importance to the cell must be reflected quantitatively in its level, and inhibition must have an impact on cell growth within the 48-hour assay. Thus, there will be targets for which this approach will not be satisfactory. Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 ARTICLES 1513 The COMPARE analysis has the ability to be extremely sensitive and specific, depending on the ‘‘seed’’ pattern used. For example, using paclitaxel (Taxol) cytotoxicity as a seed, approximately 50 of the top 55 compounds identified by COMPARE are taxanes (data not shown). Reduced sensitivity, specificity, and accuracy can be expected from seeds built of laboratory-measured quantities representing RNA or protein amounts or even from functional assays. The percentage of compounds that have tyrosine kinase inhibitory effects in a randomly selected group is much less than the percentage of inhibitory compounds we obtained by selecting compounds using COMPARE. It was reported that in a screen of 150 000 compounds, less than 1.8% of the compounds inhibited v-scr tyrosine kinase activity (49). From our selected compounds, four (28%) of 15 inhibited EGF receptor autophosphorylation and four (29%) of 14 inhibited ErbB2 autophosphorylation in cell membrane preparations. Fourteen of 25 compounds examined were active in at least one of the assays performed. In contrast, nonspecific tyrosine kinase inhibitors, such as genestein and herbimycin, give low correlation coefficients with the EGF receptor mRNA expression pattern, which can be attributed to their lack of specificity (r 4 .077 and −.573, respectively). On the other hand, whether the compounds that were negative in the in vitro assays represent false-positive correlations is not clear. Our confirmatory studies concentrated on the inhibitory effect of identified compounds on early events in the signal transduction pathway. By testing for inhibition of autophosphorylation, we restricted our investigation to the identification of compounds with a mechanism of action at the tyrosine kinase level or upstream. A downstream mechanism of action would not be detected by the in vitro assays and other assays would be needed. Although compounds that inhibit EGF receptor kinase have been under development by multiple investigators, the potential contribution of such an agent to anticancer therapy has not been clearly defined. The results of the studies presented here, in which the EGF receptor COMPARE analysis ranks TGF-a toxin chimeras at the top followed by compounds that include new and known EGF receptor kinase inhibitors, suggest that this pathway is of importance in cell growth. Thus, as a class of compounds that are distinct from the standard cytotoxic agents, these inhibitors have the potential to be important in anticancer therapy. References (1) Ullrich A, Schlessinger J. Signal transduction by receptors with tyrosine kinase activity. Cell 1990;61:203–12. (2) Rajkumar T, Gullick WJ. The type I growth factor receptors in human breast cancer. Breast Cancer Res Treat 1994;29:3–9. (3) Chen WS, Lazar CS, Poenie M, Tsien RY, Gill GN, Rosenfeld MG. Requirement for intrinsic protein tyrosine kinase in the immediate and late actions of the EGF receptor. Nature 1987;328:820–3. (4) Satoh T, Endo M, Nakafuku M, Akiyama T, Yamamoto T, Kaziro Y. Accumulation of p21rasGTP in response to stimulation with epidermal growth factor and oncogene products with tyrosine kinase activity. Proc Natl Acad Sci U S A 1990;87:7926–9. (5) Frye RA. Involvement of G proteins, cytoplasmic calcium, phospholipases, phospholipid-derived second messengers, and protein kinases in signal transduction from mitogenic cell surface receptors. Cancer Treat Res 1992; 63:281–99. (6) Kyriakis JM, Force TL, Rapp UR, Bonventre JV, Avruch J. Mitogen regulation of c-Raf-1 protein kinase activity toward mitogen-activated protein kinase-kinase. J Biol Chem 1993;268:16009–19. 1514 ARTICLES (7) Perez R, Pascual M, Macias A, Lage A. Epidermal growth factor receptors in human breast cancer. Breast Cancer Res Treat 1984;4:189–93. (8) Shoyab M, Plowman GD, McDonald VL, Bradley JG, Todaro GJ. Structure and function of human amphiregulin: a member of the epidermal growth factor family. Science 1989;243(4894 Pt 1):1074–6. (9) Bates SE, Davidson NE, Valverius EM, Freter CE, Dickson RE, Tam JP, et al. Expression of transforming growth factor alpha and its messenger ribonucleic acid in human breast cancer: its regulation by estrogen and its possible functional significance. Mol Endocrinol 1988;2:543–55. (10) Hendler FJ, Ozanne BW. Human squamous cell lung cancers express increased epidermal growth factor receptors. J Clin Invest 1984;74:647–51. (11) Lofts FJ, Gullick WJ. c-erbB2 amplification and overexpression in human tumors. Cancer Treat Res 1992:161–79. (12) Singleton TP, Strickler JG. Clinical and pathologic significance of the c-erbB-2 (HER-2/neu) oncogene. Pathol Annu 1992;27 Pt 1:165–90. (13) Harris AL. Epidermal growth factor receptor in human breast cancer. Recent Results Cancer Res 1989;113:70–7. (14) Slamon DJ, Clark GM, Wong SG, Levin WJ, Ullrich A, McGuire WL. Human breast cancer: correlation of relapse and survival with amplification of HER-2/neu oncogene. Science 1987;235:177–82. (15) Costa S, Stamm H, Almendral A, Ludwig H, Wyss R, Fabbro D, et al. Predictive value of EGF receptor in breast cancer [letter]. Lancet 1988;2: 1258. (16) Nicholson S, Sainsbury JR, Halcrow P, Chambers P, Fardon JR, Harris AL. Expression of epidermal growth factor receptors associated with lack of response to endocrine therapy in recurrent breast cancer. Lancet 1989;1: 182–5. (17) Sainsbury JR, Farndon JR, Needham GK, Malcolm AJ, Harris AL. Epidermal-growth-factor receptor status as predictor of early recurrence of and death from breast cancer. Lancet 1987;1:1398–402. (18) Monks A, Scudiero D, Skehan P, Shoemaker R, Paull K, Vistica D, et al. Feasibility of a high-flux anticancer drug screen using a diverse panel of cultured human tumor cell lines. J Natl Cancer Inst 1991;83:757–66. (19) Boyd MR, Paull KD. Some practical considerations and applications of the National Cancer Institute In Vitro Anticancer Drug Discovery Screen. Drug Devel Res 1995;34:91–109. (20) Paull KD, Shoemaker RH, Hodes L, Monks A, Scudiero DA, Rubinstein L, et al. Display and analysis of patterns of differential activity of drugs against human tumor cell lines: development of mean graph and COMPARE algorithm. J Natl Cancer Inst 1989;81:1088–92. (21) Paull KD, Lin CM, Malspeis L, Hamel E. Identification of novel antimitotic agents acting at the tubulin level by computer-assisted evaluation of differential cytotoxicity data. Cancer Res 1992;52:3892–900. (22) Lee JS, Paull K, Alvarez M, Hose C, Monks A, Grever M, et al. Rhodamine efflux patterns predict P-glycoprotein substrates in the National Cancer Institute drug screen. Mol Pharmacol 1994;46:627–38. (23) Alvarez M, Paull K, Monks A, Hose C, Lee JS, Weinstein J, et al. Generation of a drug resistance profile by quantitation of mdr-1/P-glycoprotein in the cell lines of the National Cancer Institute Anticancer Drug Screen. J Clin Invest 1995;95:2205–14. (24) Boyd MR, Paull KD, Rubinstein LR. Data display and analysis strategies for the NCI disease-oriented in vitro antitumor drug screen. In: Valeriote FA, Corbett T, Baker L, editors. Cytotoxic anticancer drugs: models and concepts for drug discovery and development. Boston: Kluwer Acad. Publ. 1992:11–34. (25) Stinson SF, Alley MC, Kopp WC, Fiebig H, Mullendore LM, Kenney S, et al. Morphological and immunocytochemical characteristics of human tumor cell lines for use in a disease-oriented anticancer drug screen. Anticancer Res 1992;12:1035–53. (26) Soule HD, Maloney TM, Wolman SR, Peterson WD Jr, Brenz R, McGrath CM, et al. Isolation and characterization of a spontaneously immortalized human breast epithelial cell line, MCF-10. Cancer Res 1990;50:6075–86. (27) Sambrook J, Fritsch EF, Maniatis T. Extraction, purification, and analysis of messenger RNA from eukaryotic cells. In: Nolan C, Ford N, Ferguson M, editors. Molecular cloning, a laboratory manual. Cold Spring Harbor (NY): Cold Spring Harbor Laboratory, 1989;7.3–7.83. (28) Dickstein B, Valverius EM, Wosikowski K, Saceda M, Pearson JW, Martin MB, et al. Increased epidermal growth factor receptor in an estrogenresponsive, adriamycin-resistant MCF-7 cell line. J Cell Physiol 1993;157: 110–8. Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 (29) Paull KD, Hamel E, Malspeis L. Prediction of biochemical mechanism of action from the in vitro antitumor screen of the National Cancer Institute. In: Foye WO, editor. Cancer chemotherapeutic agents. Washington (DC): American Chemical Society, 1995:9–45. (30) Osherov N, Gazit A, Gilon C, Levitzki A. Selective inhibition of the epidermal growth factor and HER2/neu receptors by tyrphostins. J Biol Chem 1993;268:11134–42. (31) Park JW, Stagg R, Lewis GD, Carter P, Maneval D, Slamon DJ, et al. Anti-p185HER2 monoclonal antibodies: biological properties and potential for immunotherapy. In: Dickson RB, Lippman ME, editors. Genes, oncogenes and hormones: advances in cellular and molecular biology of breast cancer. Boston: Kluwer Acad. Publ. 1992:193–211. (32) Sartor O, Sameshima JH, Robbins KC. Differential association of cellular proteins with family protein-tyrosine kinases. J Biol Chem 1991;266: 6462–6. (33) Fitzgerald D, Pastan I. Pseudomonas exotoxin and recombinant immunotoxins derived from it. Ann N Y Acad Sci 1993;685:740–5. (34) Fry DW, Kraker AJ, McMichael A, Ambroso LA, Nelson JM, Leopold WR, et al. A specific inhibitor of the epidermal growth factor receptor tyrosine kinase. Science 1994;265:1093–5. (35) Reiter Y, Brinkmann U, Jung SH, Lee B, Kasprzyk PG, King CR, et al. Improved binding and antitumor activity of a recombinant anti-erbB2 immunotoxin by disulfide stabilization of the Fv fragment. J Biol Chem 1994;269:18327–31. (36) Pai LH, Bookman MA, Ozols RF, Young RC, Smith JW 2d, Longo DL, et al. Clinical evaluation of intraperitoneal Pseudomonas exotoxin immunoconjugate OVB3-PE in patients with ovarian cancer. J Clin Oncol 1991;9: 2095–103. (37) Kihara A, Pastan I. Analysis of sequences required for the cytotoxic action of a chimeric toxin composed of Pseudomonas exotoxin and transforming growth factor alpha. Bioconjug Chem 1994;5:532–8. (38) Dickstein BM, Wosikowski K, Bates SE. Resistance to cytotoxic agents in ZR75B human breast cancer cells transfected with epidermal growth factor receptor. Mol Cell Endocrinol 1995;110:205–11. (39) Borg A, Baldetorp B, Ferno M, Killander D, Olsson H, Ryden S, et al. ERBB2 amplification is associated with tamoxifen resistance in steroidreceptor positive breast cancer. Cancer Lett 1994;81:137–44. (40) Akiyama T, Ishida J, Nakagawa S, Ogawara H, Watanabe S, Itoh N, et al. Genistein, a specific inhibitor of tyrosine-specific protein kinases. J Biol Chem 1987;262:5592–5. (41) Onodo T, Isshiki K, Takeuchi T, Tatsuta K, Umezawa K. Inhibition of (42) (43) (44) (45) (46) (47) (48) (49) tyrosine kinase and epidermal growth factor receptor internalization by lavendustin A methyl ester in cultured A431 cells. Drugs Exp Clin Res 1990;16:249–53. Umezawa K, Imoto M. Use of erbstatin as protein-tyrosine kinase inhibitor. Methods Enzymol 1991;201:379–85. Uehara Y, Fukazawa H, Murakami Y, Mizuno S. Irreversible inhibition of v-src tyrosine kinase activity by herbimycin A and its abrogation by sulfhydryl compounds. Biochem Biophys Res Commun 1989;163:803–9. Levitzki A, Gazit A. Tyrosine kinase inhibition: an approach to drug development. Science 1995;267:1782–8. Chabner BA. Biological basis for cancer treatment. Ann Intern Med 1993; 118:633–7. Gibbs JB, Oliff A. Pharmaceutical research in molecular oncology. Cell 1994;79:193–8. James GL, Goldstein JL, Brown MS, Rawson TE, Somers TC, McDowell RS, et al. Benzodiazepine peptidomimetics: potent inhibitors of Ras farnesylation in animal cells. Science 1993;260:1937–42. Kohl NE, Mosser SD, deSolms SJ, Giuliani EA, Pompliano DL, Graham SL, et al. Selective inhibition of ras-dependent transformation by a farnesyltransferase inhibitor. Science 1993;260:1934–7. Fry DW, Kraker AJ, Connors RC, Elliott WL, Nelson JM, Showalter HD, et al. Strategies for the discovery of novel tyrosine kinase inhibitors with anticancer activity. Anticancer Drug Des 1994;9:331–51. Notes Supported in part by the International Medical Student Exchange Program, The Netherlands (D. Schuurhuis). We thank Ed Sausville for his critical reading of the manuscript, Curtis Hose and Anne Monks for providing the cell lines, and Bob Schultz and Suzanne Radtke for their help in the development of this project. Permission to publish information relating to the following compounds is appreciated: NSC-682027 and NSC-680625, Peakdale Fine Chemicals Limited, Glossop, Derbyshire, U.K.; NSC-643892, Bionet Research Ltd, Camelford, U.K.; NSC-680581, D. Heber, Pharmaceutical Institute, Kiel, Germany; NSC636467; NSC-636467, LIPHA SODEPHARM, Chilly-Mazarin, France; and NSC-640820 and NSC-642550, Bayer AG, Wuppertal, Germany. Presented in preliminary form at the 1996 meeting of the American Association for Cancer Research, Washington, DC. Manuscript received December 6, 1996; revised June 10, 1997; accepted July 10, 1997. Journal of the National Cancer Institute, Vol. 89, No. 20, October 15, 1997 ARTICLES 1515