- Email: [email protected]
Moesin expression is associated with the estrogen receptor–negative breast cancer phenotype
Moesin expression is associated with the estrogen receptor–negative breast cancer phenotype Charles Carmeci, MD, Devon A. Thompson, PhD, Wayne W. Kuang, BS, Nina Lightdale, BS, Heinz Furthmayr, MD, and Ronald J. Weigel, MD, PhD, Stanford, Calif.
Background. Estrogen receptor (ER)–positive breast carcinomas possess a less aggressive phenotype than ER-negative breast carcinomas. We hypothesize that a set of genes exists that is expressed only in ER-negative breast carcinomas, which account for the more malignant phenotypic characteristics of these tumors. Methods. We have used a new technique of polymerase chain reaction select suppression subtractive hybridization to identify genes that are expressed only in ER-negative carcinomas. Results. Seventy-one cDNA clones generated by suppression subtractive hybridization were screened by Northern blot analysis with RNA from ER-positive MCF7 and ER-negative MDA-MB-231 breast carcinoma cell lines. Fifteen clones were differentially expressed in MDA-MB-231 cells. Five of these 15 clones were consistently found to be associated with the ER-negative phenotype in a panel of eight breast carcinoma cell lines. Sequence analysis demonstrated that three of these clones were derived from vimentin and two clones from moesin. Western blot analysis with antihuman moesin antibody confirmed that moesin protein was overexpressed in ER-negative breast carcinoma cell lines but absent from ER-positive breast carcinomas. Moesin mRNA was examined in a panel of 29 primary breast carcinomas with semiquantitative reverse transcriptase–polymerase chain reaction. Moesin expression was found to be decreased significantly in ER-positive compared with ER-negative tumors (P < .01). Conclusions. Vimentin and moesin are differentially expressed in association with the ER-negative breast cancer phenotype. Moesin is a membrane/actin filament protein involved in dynamic restructuring of the cell surface and filopodia, a cell structure needed for cell adhesion and motility. Moesin may play a role in the invasiveness and pattern of metastasis characteristic of ER-negative breast cancers. (Surgery 1998;124:211-17.) From the Departments of Surgery and Pathology, Stanford University, Stanford, Calif.
ESTROGEN RECEPTOR (ER) EXPRESSION IS AN important prognostic marker in breast cancer. Patients with tumors that express ER (ER positive) have longer disease-free intervals and improved survival compared with patients with tumors that demonstrate minimal ER expression (ER negative).1,2 In addition, ER-positive tumors are more likely to respond to hormonal therapy.3 However, the association between ER expression and hormone Supported in part by National Institutes of Health grant R29 CA63251 and Department of the Army grant DAMD 17-94-J4353. Drs Carmeci and Thompson were funded by National Research Service Award grant F32 CA69715 and F32 CA69751, respectively. Dr Weigel is supported in part by a Clowes Career Development Award from the American College of Surgeons. Presented at the Fifty-ninth Annual Meeting of the Society of University Surgeons, Milwaukee, Wis, Feb 12-14, 1998. Reprint requests: Ronald J. Weigel, MD, PhD, Department of Surgery, Stanford University, MSOB, Room X300, Stanford, CA 94305-5414. Copyright © 1998 by Mosby, Inc. 0039-6060/98/$5.00 + 0 11/6/90368
response is not absolute because 30% to 45% of ER-positive tumors do not respond to hormone therapy and 9% to 15% of ER-negative tumors demonstrate hormone responsiveness.3 Although the relationship between ER expression, hormone response, and prognosis has been studied extensively, the molecular basis for this association remains elusive. We hypothesize that the clinical differences between ER-positive and ER-negative breast carcinomas are the result of variations in the pattern of gene expression characteristic of these two tumor phenotypes. Several lines of evidence support the hypothesis that hormone-responsive and -unresponsive cancers demonstrate differences in transcriptional regulation. The abundance of ER mRNA varies by several orders of magnitude when ER-positive and ER-negative tumors are compared.4,5 Tumors with abundant ER mRNA tend to be ER positive; progesterone receptor–positive tumors have a lower nuclear grade and are more often lobular or well-differentiated ductal carcinomas than are SURGERY 211
212 Carmeci et al.
tumors with low ER mRNA levels.5 Studies of breast carcinoma cell lines have also demonstrated that ER expression is regulated by transcription of the ER gene.6,7 Analyses of the ER gene promoter have demonstrated differences in transcriptional activity comparing ER-positve and ER-negative cell lines, and a number of transacting factors have been implicated in the regulation of the ER gene.8-10 It is certainly plausible that transcriptional mechanisms regulating ER gene expression may also control expression of a set of genes characteristic of the ER phenotype. Further evidence supporting this hypothesis is the identification of several genes, including GPCR-Br, ICERE-1, DEME-6, vimentin, and GATA-3, which are not regulated by ER but demonstrate a pattern of expression that correlates directly or inversely with the ER phenotype.11-14 The recently developed technique of suppression subtractive hybridization (SSH) has proved to be highly effective for the identification of differentially expressed genes.13 In this study SSH has been used to identify genes expressed in the ERnegative breast carcinoma cell line, MDA-MB-231, that are absent or minimally expressed in the ERpositive cell line, MCF-7. The pattern of expression of these genes was subsequently examined on a panel of breast carcinoma cell lines. These experiments demonstrated that the expression of moesin is associated with the ER-negative breast cancer phenotype. Further evidence supporting this association was developed by examining moesin expression in primary breast cancers. MATERIAL AND METHODS Cell lines. Cell lines were obtained from the American Type Culture Collection and were maintained as described previously.12 Suppression subtractive hybridization. SSH was performed with the Clontech polymerase chain reaction (PCR) select cDNA subtraction kit (Clontech Laboratories, Inc, Palo Alto, Calif), as described by the manufacturer with the following modifications and details. The starting material consisted of 2 µg MDA-MB-231 poly(A) + RNA as the tester and 2 µg MCF7 poly(A)+RNA as the driver. Primary and secondary PCR conditions were optimized to increase specificity for differentially expressed sequences. Primary PCR consisted of 25 cycles of 94° C for 10 seconds, 66° C for 15 seconds, and 72° C for 90 seconds. Primary PCR products were diluted 1:10 before secondary PCR. Secondary PCR consisted of 12 cycles of 94° C for 10 seconds, 68° C for 30 seconds, and 72° C for 90 seconds. Secondary PCR products were then subcloned into the PCR II vector with the Original
Surgery August 1998
TA Cloning Kit (Invitrogen, San Diego, Calif). Seventy-one clones were selected and screened by Northern blot analysis for differential expression between MDA-MB-231 and MCF7 cell lines. Northern blot analysis. Two micrograms of poly (A)+RNA isolated from the cell lines (MCF7, T-47D, MDA-MB-361, ZR-75-1, BT-20, MDA-MB-231, HBL100, and HeLa) were electrophoresed on 1% agarose/formaldehyde gel in 1× 3-[Nmorpholino]propanesulfonic acid (MOPS) buffer and transferred to a 0.2 µm Nytran membrane (Schleicher & Schuell) with a 20× salt sodium citrate (SSC) buffer. Blots were then hybridized with cDNA probes that had been 32P labeled by random priming (Boehringer-Mannheim) in 50% formamide, 5× Denhardt’s solution, 5× (salt sodium phosphate ethylenediaminetetraacetic acid (SSPE), 1.0% sodium dodecyl sulfate (SDS), and 100 µg/mL denatured salmon sperm DNA at 42° C overnight. The membranes were washed once with 2× SSC and 0.1% SDS at 42° C for 20 minutes and subsequently twice with 0.2× SSC and 0.1% SDS at 65° C for 20 minutes. Western blot analysis. Total cellular protein was extracted from the breast cancer cell lines MCF7, T-47D, MDA-MB-231, and HBL-100. Ten micrograms of total protein was then subjected to SDSpolyacrylamide gel electrophoresis. The proteins were transferred to polyvinylidene diflouride-plus membranes (Micron Separations, Inc) with semidry electroblotting. Membranes were blocked for 1 day with a 1% (wt/vol) dried milk solution followed by incubation with the polyclonal moesin antibody15 for 1 hour at room temperature. Subsequently, the membranes were incubated with peroxidase-labeled antirabbit or antigoat immunoglobulin G and detected by enhanced chemiluminescence (Amersham Corp) per the manufacturer’s protocol. Reverse transcriptase (RT)-PCR of primary breast tumors and slot blotting. Samples of primary human breast adenocarcinomas were obtained as a gift from Dr. Feiner of the New York University Medical Center. Tumor tissue was collected fresh from mastectomy and breast biopsy specimens, snap frozen in liquid nitrogen, and stored at –80° C. Fourteen of the tumors were ER negative and 15 were ER positive as determined by standard enzyme immunoassay. Normal breast tissue RNA was obtained from mastectomy specimens in an area free of tumor. Total RNA was isolated from each tumor sample as described previously.12 The total RNA from each tumor sample was analyzed by RT-PCR for ER and β-actin messages as described previously.12 Each sample was analyzed for moesin expression
Surgery Volume 124, Number 2
Carmeci et al. 213
Fig 1. Northern blots of SSH cDNA clones. Fifteen cDNA clones generated by SSH were found to demonstrate pattern of differential expression comparing ER-negative MDA-MB-231 and ER-positive MCF7 breast carcinoma cell lines. Each panel contains RNA from two cell lines as indicated. Northern blots probed with ER, actin, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) are included as controls. Sequence analysis identified clones 29, 67, and 71 as vimentin and clones 57 and 10 as moesin. Other clones were identified as follows: clone 6, thymidylate synthase; clones 30 and 64, human prolactin receptor–associated protein; clone 47, epithelial membrane protein; and clone 50, human MAC25/prostacyclin-stimulating factor. Remaining clones did not match sequences in GenBank/EMBL database.
with PCR primers designed across the splice junction between exons 4 and 5. Primers for moesin amplification were moe-3/long 5´-CTGAGGATGTGTCCGAGGAATTGATTCAGG-3´ and moe-2/long 5´TGTTCCTCATGCCACACCTGGATCCGCTCC-3´, which generates a 280 base-pair DNA fragment. Amplification was performed as described for ER and β-actin. Initially all the PCR samples were visualized on polyacrylamide gels. Subsequently, a 5 µL aliquot from each moesin, ER, and β-actin PCR reaction was slot blotted and the blot then probed with internal oligonucleotides as described.12 The moesin transcript was detected with the internal oligonucleotide probe moesin-1,5´-GCTGTCCAGTCTAAGTATGG-3´, designed to hybridize to the 280 basepair PCR product. To quantitate the results of the slot blots, phosphorous imaging was performed and ER and moesin levels were expressed as a percentage of β-actin as described previously.12 RESULTS SSH was used to identify genes expressed in ERnegative MDA-MB-231 cells that were absent or
minimally expressed in ER-positive MCF7 breast carcinoma cells. Seventy-one cDNA clones generated by SSH were screened by Northern blot analysis. As shown in Fig 1, 15 cDNA clones were found to hybridize to mRNA, overexpressed in MDA-MB-231 compared with MCF7. Each of these cDNA inserts was sequenced. Sequence analysis demonstrated that clones 29, 67, and 71 were derived from vimentin and clones 10 and 57 were from moesin. This analysis identified a set of genes differentially expressed in an ER-negative compared with an ER-positive breast carcinoma cell line. To identify genes consistently expressed in association with the ER-negative phenotype, the expression of each of these 15 cDNA clones was examined in a panel of breast carcinoma cell lines. Moesin was expressed in the ER-negative breast carcinoma cell lines BT-20, MDA-MB-231, and HBL-100 and the additional ERnegative cell line HeLa. Moesin expression was not detected in the ER-positive cell lines MCF7, T-47D, MDA-MB-361, and ZR-75-1 (Fig 2). As reported previously, vimentin also demonstrated a pattern of
214 Carmeci et al.
Fig 2. Northern blots of moesin and vimentin. Northern blots with panel of cell lines were hybridized with probes as indicated. Panel contains four ER-positive breast cancer cell lines (MCF7, T-47D, MDA-MB-361, and ZR-75-1) and three ER-negative breast cancer cell lines (BT-20, MDAMB-231, and HBL-100) and ER-negative cell line HeLa.
Fig 3. Western blot of moesin. Western blot of protein extracts from two ER-positive cell lines (MCF7 and T-47D) and two ER-negative cell lines (MDA-MB-231 and HBL100). Moesin is identified as 78,000-dalton protein. Radixin is also identified because of cross-reactivity with talin-4.1 family member.
expression that correlated with the ER-negative phenotype.14 The remaining 10 cDNA clones failed to demonstrate a consistent pattern of expression with the ER-negative phenotype. The data in Fig 2 demonstrate the pattern of moesin mRNA expression. Moesin protein expression was examined by Western blot. As shown in Fig 3, a 78,000-dalton protein corresponding to moesin was expressed in the ER-negative cell lines
Surgery August 1998
MDA-MB-231 and HBL-100, but moesin protein was not detected in the ER-positive cell lines MCF7 and T-47D. The moesin polyclonal antibody was known to cross-react with radixin, an 80,000-dalton member of the talin-4.1 family. Radixin was also absent from MCF7 cells. These data confirm the Northern blot data and support the association between moesin expression and the ER-negative breast cancer phenotype. RT-PCR was used to examine moesin expression in a panel of primary breast cancers. Fig 4, A illustrates moesin expression in 15 ER-positive and 14 ER-negative tumors. Two normal breast tissue samples and the MDA-MB-231 and MCF7 cell lines were also included. Higher levels of moesin expression were found in the ER-negative compared with ER-positive tumors. Fig 4, B summarizes these data for all tumor samples. As reported previously, there is a statistically significantly greater abundance of ER mRNA in ER-positive tumors.5 Consistent with the cell line data, moesin expression was found to be greater in ER-negative tumors (P < .01). Taken together, these data demonstrate a correlation between moesin gene expression and the ER-negative breast cancer phenotype. DISCUSSION There is developing evidence that there are patterns of gene expression characteristic of ERpositive and ER-negative breast cancers, and many of these genes are not part of the ER-regulation pathway. The expression of vimentin and ICERE-1 (inversely correlated with estrogen receptor; Genbank/EMBL accession number AF007790) have previously been demonstrated to be associated with the ER-negative breast cancer phenotype.12,14 These results demonstrate that the expression of moesin is also associated with the ER-negative phenotype. Previous studies of moesin have demonstrated ubiquitous expression in various tissues and tumor cell lines.16 The lack of moesin expression in a panel of ER-positive breast cancer cell lines is, therefore, particularly striking and suggests a functional role for moesin in mechanisms related to hormone-responsive breast cancer. Moesin is a member of the talin-4.1 superfamily, which includes the proteins ezrin, radixin, and merlin. 15,17 Moesin contains a conserved carboxy terminal domain that binds actin and an amino terminus that is membrane associated.18 In addition, moesin is localized in filopodia and other cellular microextensions that are cell surface structures important for cell-cell and cell-substrate recognition, adhesion, and motility.15 These findings have led to the hypothesis that moesin is critical to dynamic
Surgery Volume 124, Number 2
Carmeci et al. 215
B Fig 4. A, Moesin expression in primary breast tumors. A, RT-PCR was used to obtain semiquantitative expression of moesin in primary breast carcinomas. Fifteen ER-positive and 14 ER-negative carcinomas were evaluated. Two normal breast tissue samples and MCF7 and MDA-MB-231 cell lines were also examined in parallel as indicated. Moesin and ER expression are indicated as percentage of actin. Experiments were performed in triplicate (n = 3) with SEM shown. Tumor samples are arranged in deseending order of ER mRNA abundance. B, Cumulative data for ER-positive and ER-negative tumors (phenotype designation enzyme immunoassay). Data demonstrate statistically significant increase in moesin expression in ER-negative tumors.
216 Carmeci et al.
restructuring of the cell surface by regulating binding interactions between the plasma membrane and the actin cytoskeleton. The lack of moesin expression in ER-positive breast carcinomas thus leads us to speculate that moesin may be involved in the invasiveness or metastatic potential characteristic of the ER-negative breast cancer phenotype. The association of higher levels of moesin expression with primary ER-negative breast cancers supports the findings established in our breast cancer cell line model. However, the differences in moesin expression comparing primary ER-positive and ER-negative breast tumors were not as striking as the results in cell culture. There may be several reasons for this finding. Primary tumors are composed of a heterogenous population of cells including ductal epithelium, stromal cells, endothelium, and inflammatory cells. On the other hand, breast carcinoma cell lines are composed of a homogeneous clonal population of carcinoma cells. Previous studies have demonstrated moesin expression in all of the nonductal cell types present in breast tumors.16 Therefore even if the tumors were composed mostly of tumor cells, moesin expression in other cell types present in the tumor tissue may falsely elevate the expression level of moesin. It is possible that a direct comparison of moesin expression in ER-positive and ER-negative ductal carcinoma cells may demonstrate a more significant difference than is possible by the examination of fresh tumor tissue. Alternatively, it is possible that there is a spectrum of moesin expression in cells from the same tumor, possibly reflecting differences in metastatic potential. Further studies to examine this important finding in more detail are being pursued. Only few data are available on the control of moesin expression. One recent example has been identified in the proliferative response of mesangial to glomerular injury.19 Overexpression of the homologous protein ezrin by stable transfection of LLC-PK1 cells enhances migration of the cells but also tubulogenesis induced by stimulation with hepatocyte growth factor.20 An important effect of the expression of moesin could thus be related directly to morphogenetic effects. At present one can only speculate about the functional role of moesin in hormone-unresponsive breast cancer. Moesin has been shown to be a membrane/actin filament–linking protein involved in the formation of microextensions of the cell membrane. Such structures are considered necessary for cellular adhesion and motility and depend on activation of cdc42, rac, and rho GTPases.15,17,20,21 One attractive possibility to explain expression of moesin in
Surgery August 1998
hormone-unresponsive breast cancers is that it contributes to the increased invasiveness and metastatic potential of such tumors. These properties are reflected in more extensive and dynamic cell surface activity. REFERENCES 1. Knight WA III, Livingston RB, Gregory EJ, McGuire WL. Estrogen receptor as an independent prognostic factor for early recurrence in breast cancer. Cancer Res 1977;37:466971. 2. Sigurdsson H, Baldetorp B, Borg Å, Dalberg M, Fernö M, Killander D, et al. Indicators of prognosis in node-negative breast cancer. N Engl J Med 1990;322:1045-53. 3. Jordan VC, Wolf MF, Mirecki DM, Whitford DA, Welshons WV. Hormone receptor assays: clinical usefulness in the management of carcinoma of the breast. Crit Rev Clin Lab Sci 1988;26:97-152. 4. Barrett-Lee PJ, Travers MT, McClelland RA, Luqmani Y, Coombes RC. Characterization of estrogen receptor messenger RNA in human breast cancer. Cancer Res 1987;47:6653-9. 5. Carmeci C, deConinck EC, Lawton T, Bloch DA, Weigel RJ. Analysis of estrogen receptor messenger RNA in breast carcinomas from archival specimens is predictive of tumor biology. Am J Pathol 1997;150:1563-9. 6. Weigel RJ, deConinck EC. Transcriptional control of estrogen receptor in estrogen receptor–negative breast carcinoma. Cancer Res 1993;53:3472-4. 7. Ottaviano YL, Issa J-P, Parl FF, Smith HS, Baylin SB, Davidson NE. Methylation of the estrogen receptor gene CpG island marks loss of estrogen receptor expression in human breast cancer cells. Cancer Res 1994;54:2552-5. 8. deConinck EC, McPherson LA, Weigel RJ. Transcriptional regulation of estrogen receptor in breast carcinoma. Mol Cell Biol 1995;15:2191-6. 9. McPherson LA, Baichwal VR, Weigel RJ. Identification of ERF-1 as a member of the AP2 transcription factor family. Proc Natl Acad Sci USA 1997;94:4342-7. 10. Tang Z, Treilleux I, Brown M. A transcriptional enhancer required for the differential expression of the human estrogen receptor in breast cancers. Mol Cell Biol 1997;17:1274-80. 11. Carmeci C, Thompson DA, Ring HZ, Francke U, Weigel RJ. A novel G protein–coupled receptor associated with estrogen receptor expression in breast cancer. Genomics 1997;45:607-17. 12. Thompson DA, Weigel RJ. Characterization of a gene that is inversely correlated with estrogen receptor expression (ICERE-1) in breast carcinomas. Eur J Biochem 1998; 252:169-77. 13. Kuang WW, Thompson DA, Hoch RV, Weigel RJ. Differential screening and suppression subtractive hybridization identified genes differentially expressed in an estrogen receptor–positive breast carcinoma cell line. Nucl Acids Res 1998;26:1116-23. 14. Sommers CL, Walker-Jones D, Heckford SE, Worland P, Valverius E, Clark R, et al. Vimentin rather than keratin expression in some hormone-independent breast cancer cell lines and in oncogene-transformed mammary epithelial cells. Cancer Res 1989;49:4258-63. 15. Amieva MR, Furthmayr H. Subcellular localization of moesin in dynamic filopodia, retraction fibers, and other
Surgery Volume 124, Number 2
16.
17.
18.
19.
20.
21.
structures involved in substrate exploration, and attachment, and cell-cell contacts. Exp Cell Res 1995;219:180-96. Schwartz-Albiez R, Merling A, Spring H, Möller P, Koretz K. Differential expression of the microspike-associated protein moesin in human tissues. Eur J Cell Biol 1995;67:189-98. Lankes WT, Furthmayr H. Moesin: a member of the protein 4.1-talin-ezrin family of proteins. Proc Natl Acad Sci USA 1991;88:8297-301. Pestonjamasp K, Amieva MR, Strassel CP, Nauseef WM, Furthmayr H, Luna EJ. Moesin, ezrin, and p205 are actinbinding proteins associated with neutrophil plasma membranes. Mol Biol Cell 1995;6:247-59. Hugo C, Hugo C, Pichler R, Gordon K, Schmidt R, Amieva RM, et al. The cytoskeletal linking proteins, moesin and radixin, are upregulated by platelet-derived growth factor, but not basic fibroblast growth factor in experimental mesangial prolierative glomerulonephritis. J Clin Invest 1997;97:2499-508. Crepaldi T, Gautreau A, Comoglio PM, Louvard D, Arpin M. Ezrin is an effector of hepatocyte growth factor–mediated migration and morphogenesis in epithelial cells. J Cell Biol 1997;138:423-34. Mackay DJG, Esch F, Furthmayr H, Hall A. Rho- and racdependent assembly of focal adhesion complexes and actin filaments in permeabilized fibroblasts: an essential role for ERM proteins. J Cell Biol 1977;138:927-38.
DISCUSSION Dr Paul C. Kuo (Baltimore, Md). Although you measured mRNA in your primary breast tumors, have you determined protein expression with Western blot, immunoblot, or immunohistochemistry in these primary breast tumor cell lines? Also, what happens if you reverse the tester and driver sources of mRNA in this polymerase chain reaction–select system? Last, would you speculate on the signal transduction mechanism or the pathway that prevents expression of moesisn in ER-negative cells? Dr Carmeci. First, why did we not measure protein in the primary tumors? It is a problem to look at primary tumors when specimens and the breast biopsies they come from are getting smaller. We used samples provided as a gift from the New York University Breast Tumor Bank, and they were tested for IPOX. So the ones that were labeled ER positive were positive by IPOX and the ones that were negative, as labeled in the slides, were negative by IPOX. So they were tested for tumors. They were not tested quantitatively but they were known to be ER positive/ER negative beyond just the ER message shown on the slide. We simply did not repeat that
Carmeci et al. 217
because the finite amount of tumor sample that we had was just enough to perform our RNA isolations. Second, why not reverse the tester and the driver? We did that experiment and we did find several genes that were expressed differentially in association with the ER-positive phenotype. Two of these turned out to be novel sequences that were estrogen responsive. The results of those studies are in press in Nucleic Acids Research. With regard to the last question about the mechanism preventing moesin expression in ER-negative tumors, there are a couple of things that could be taking place. It could be a matter of lack of expression at the genomic level; perhaps ER-negative tumors have a deletion in the moesin gene or a rearrangement that destroys it or prevents its expression. That could be sorted out with a restriction length fragment polymorphism study. In addition, moesin may be deleted at the transcriptional level. A formal promoter analysis of the moesin gene has not been done. I think it would be interesting to know whether ER turned off the moesin gene. Recently in our laboratory, or about 3 years ago, an ER factor was cloned that has been shown to turn on the ER gene. The ER gene, which is a transcription factor, will then turn on several other genes: progesterone receptor and PS-2. When originally designing this experiment, I thought it would be interesting to know whether there was a gene that was turned off by the ER. A logical way of looking at that would be to look at tumors that fail to express the ER gene. I think a formal promoter analysis might shed light on whether moesin is down-regulated at the transcriptional level. Also I think immunohistochemistry might help address the last question. Dr William G. Cance (Chapel Hill, NC). I think it is a fascinating gene linking the actin cytoskeleton, and the link to invasion is critical. Do you have antibodies that work in paraffin and have you done any studies on ductal carcinoma in situ (DCIS), for example, comparing DCIS with minimally invasive tumors to the more invasive tumors and correlated them with ER positivity? Dr Carmeci. We have not done those studies. We do have antibodies to moesin that would likely work. To look at in situ hybridization in DCIS lesions and then invasive carcinoma in the same lesion might provide some insight. I think it would also be interesting to look at tumors that had metastases in the same patient that went on to convert to hormone unresponsiveness and see whether they then also went on to fail to express the moesin gene.
Background. Estrogen receptor (ER)–positive breast carcinomas possess a less aggressive phenotype than ER-negative breast carcinomas. We hypothesize that a set of genes exists that is expressed only in ER-negative breast carcinomas, which account for the more malignant phenotypic characteristics of these tumors. Methods. We have used a new technique of polymerase chain reaction select suppression subtractive hybridization to identify genes that are expressed only in ER-negative carcinomas. Results. Seventy-one cDNA clones generated by suppression subtractive hybridization were screened by Northern blot analysis with RNA from ER-positive MCF7 and ER-negative MDA-MB-231 breast carcinoma cell lines. Fifteen clones were differentially expressed in MDA-MB-231 cells. Five of these 15 clones were consistently found to be associated with the ER-negative phenotype in a panel of eight breast carcinoma cell lines. Sequence analysis demonstrated that three of these clones were derived from vimentin and two clones from moesin. Western blot analysis with antihuman moesin antibody confirmed that moesin protein was overexpressed in ER-negative breast carcinoma cell lines but absent from ER-positive breast carcinomas. Moesin mRNA was examined in a panel of 29 primary breast carcinomas with semiquantitative reverse transcriptase–polymerase chain reaction. Moesin expression was found to be decreased significantly in ER-positive compared with ER-negative tumors (P < .01). Conclusions. Vimentin and moesin are differentially expressed in association with the ER-negative breast cancer phenotype. Moesin is a membrane/actin filament protein involved in dynamic restructuring of the cell surface and filopodia, a cell structure needed for cell adhesion and motility. Moesin may play a role in the invasiveness and pattern of metastasis characteristic of ER-negative breast cancers. (Surgery 1998;124:211-17.) From the Departments of Surgery and Pathology, Stanford University, Stanford, Calif.
ESTROGEN RECEPTOR (ER) EXPRESSION IS AN important prognostic marker in breast cancer. Patients with tumors that express ER (ER positive) have longer disease-free intervals and improved survival compared with patients with tumors that demonstrate minimal ER expression (ER negative).1,2 In addition, ER-positive tumors are more likely to respond to hormonal therapy.3 However, the association between ER expression and hormone Supported in part by National Institutes of Health grant R29 CA63251 and Department of the Army grant DAMD 17-94-J4353. Drs Carmeci and Thompson were funded by National Research Service Award grant F32 CA69715 and F32 CA69751, respectively. Dr Weigel is supported in part by a Clowes Career Development Award from the American College of Surgeons. Presented at the Fifty-ninth Annual Meeting of the Society of University Surgeons, Milwaukee, Wis, Feb 12-14, 1998. Reprint requests: Ronald J. Weigel, MD, PhD, Department of Surgery, Stanford University, MSOB, Room X300, Stanford, CA 94305-5414. Copyright © 1998 by Mosby, Inc. 0039-6060/98/$5.00 + 0 11/6/90368
response is not absolute because 30% to 45% of ER-positive tumors do not respond to hormone therapy and 9% to 15% of ER-negative tumors demonstrate hormone responsiveness.3 Although the relationship between ER expression, hormone response, and prognosis has been studied extensively, the molecular basis for this association remains elusive. We hypothesize that the clinical differences between ER-positive and ER-negative breast carcinomas are the result of variations in the pattern of gene expression characteristic of these two tumor phenotypes. Several lines of evidence support the hypothesis that hormone-responsive and -unresponsive cancers demonstrate differences in transcriptional regulation. The abundance of ER mRNA varies by several orders of magnitude when ER-positive and ER-negative tumors are compared.4,5 Tumors with abundant ER mRNA tend to be ER positive; progesterone receptor–positive tumors have a lower nuclear grade and are more often lobular or well-differentiated ductal carcinomas than are SURGERY 211
212 Carmeci et al.
tumors with low ER mRNA levels.5 Studies of breast carcinoma cell lines have also demonstrated that ER expression is regulated by transcription of the ER gene.6,7 Analyses of the ER gene promoter have demonstrated differences in transcriptional activity comparing ER-positve and ER-negative cell lines, and a number of transacting factors have been implicated in the regulation of the ER gene.8-10 It is certainly plausible that transcriptional mechanisms regulating ER gene expression may also control expression of a set of genes characteristic of the ER phenotype. Further evidence supporting this hypothesis is the identification of several genes, including GPCR-Br, ICERE-1, DEME-6, vimentin, and GATA-3, which are not regulated by ER but demonstrate a pattern of expression that correlates directly or inversely with the ER phenotype.11-14 The recently developed technique of suppression subtractive hybridization (SSH) has proved to be highly effective for the identification of differentially expressed genes.13 In this study SSH has been used to identify genes expressed in the ERnegative breast carcinoma cell line, MDA-MB-231, that are absent or minimally expressed in the ERpositive cell line, MCF-7. The pattern of expression of these genes was subsequently examined on a panel of breast carcinoma cell lines. These experiments demonstrated that the expression of moesin is associated with the ER-negative breast cancer phenotype. Further evidence supporting this association was developed by examining moesin expression in primary breast cancers. MATERIAL AND METHODS Cell lines. Cell lines were obtained from the American Type Culture Collection and were maintained as described previously.12 Suppression subtractive hybridization. SSH was performed with the Clontech polymerase chain reaction (PCR) select cDNA subtraction kit (Clontech Laboratories, Inc, Palo Alto, Calif), as described by the manufacturer with the following modifications and details. The starting material consisted of 2 µg MDA-MB-231 poly(A) + RNA as the tester and 2 µg MCF7 poly(A)+RNA as the driver. Primary and secondary PCR conditions were optimized to increase specificity for differentially expressed sequences. Primary PCR consisted of 25 cycles of 94° C for 10 seconds, 66° C for 15 seconds, and 72° C for 90 seconds. Primary PCR products were diluted 1:10 before secondary PCR. Secondary PCR consisted of 12 cycles of 94° C for 10 seconds, 68° C for 30 seconds, and 72° C for 90 seconds. Secondary PCR products were then subcloned into the PCR II vector with the Original
Surgery August 1998
TA Cloning Kit (Invitrogen, San Diego, Calif). Seventy-one clones were selected and screened by Northern blot analysis for differential expression between MDA-MB-231 and MCF7 cell lines. Northern blot analysis. Two micrograms of poly (A)+RNA isolated from the cell lines (MCF7, T-47D, MDA-MB-361, ZR-75-1, BT-20, MDA-MB-231, HBL100, and HeLa) were electrophoresed on 1% agarose/formaldehyde gel in 1× 3-[Nmorpholino]propanesulfonic acid (MOPS) buffer and transferred to a 0.2 µm Nytran membrane (Schleicher & Schuell) with a 20× salt sodium citrate (SSC) buffer. Blots were then hybridized with cDNA probes that had been 32P labeled by random priming (Boehringer-Mannheim) in 50% formamide, 5× Denhardt’s solution, 5× (salt sodium phosphate ethylenediaminetetraacetic acid (SSPE), 1.0% sodium dodecyl sulfate (SDS), and 100 µg/mL denatured salmon sperm DNA at 42° C overnight. The membranes were washed once with 2× SSC and 0.1% SDS at 42° C for 20 minutes and subsequently twice with 0.2× SSC and 0.1% SDS at 65° C for 20 minutes. Western blot analysis. Total cellular protein was extracted from the breast cancer cell lines MCF7, T-47D, MDA-MB-231, and HBL-100. Ten micrograms of total protein was then subjected to SDSpolyacrylamide gel electrophoresis. The proteins were transferred to polyvinylidene diflouride-plus membranes (Micron Separations, Inc) with semidry electroblotting. Membranes were blocked for 1 day with a 1% (wt/vol) dried milk solution followed by incubation with the polyclonal moesin antibody15 for 1 hour at room temperature. Subsequently, the membranes were incubated with peroxidase-labeled antirabbit or antigoat immunoglobulin G and detected by enhanced chemiluminescence (Amersham Corp) per the manufacturer’s protocol. Reverse transcriptase (RT)-PCR of primary breast tumors and slot blotting. Samples of primary human breast adenocarcinomas were obtained as a gift from Dr. Feiner of the New York University Medical Center. Tumor tissue was collected fresh from mastectomy and breast biopsy specimens, snap frozen in liquid nitrogen, and stored at –80° C. Fourteen of the tumors were ER negative and 15 were ER positive as determined by standard enzyme immunoassay. Normal breast tissue RNA was obtained from mastectomy specimens in an area free of tumor. Total RNA was isolated from each tumor sample as described previously.12 The total RNA from each tumor sample was analyzed by RT-PCR for ER and β-actin messages as described previously.12 Each sample was analyzed for moesin expression
Surgery Volume 124, Number 2
Carmeci et al. 213
Fig 1. Northern blots of SSH cDNA clones. Fifteen cDNA clones generated by SSH were found to demonstrate pattern of differential expression comparing ER-negative MDA-MB-231 and ER-positive MCF7 breast carcinoma cell lines. Each panel contains RNA from two cell lines as indicated. Northern blots probed with ER, actin, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) are included as controls. Sequence analysis identified clones 29, 67, and 71 as vimentin and clones 57 and 10 as moesin. Other clones were identified as follows: clone 6, thymidylate synthase; clones 30 and 64, human prolactin receptor–associated protein; clone 47, epithelial membrane protein; and clone 50, human MAC25/prostacyclin-stimulating factor. Remaining clones did not match sequences in GenBank/EMBL database.
with PCR primers designed across the splice junction between exons 4 and 5. Primers for moesin amplification were moe-3/long 5´-CTGAGGATGTGTCCGAGGAATTGATTCAGG-3´ and moe-2/long 5´TGTTCCTCATGCCACACCTGGATCCGCTCC-3´, which generates a 280 base-pair DNA fragment. Amplification was performed as described for ER and β-actin. Initially all the PCR samples were visualized on polyacrylamide gels. Subsequently, a 5 µL aliquot from each moesin, ER, and β-actin PCR reaction was slot blotted and the blot then probed with internal oligonucleotides as described.12 The moesin transcript was detected with the internal oligonucleotide probe moesin-1,5´-GCTGTCCAGTCTAAGTATGG-3´, designed to hybridize to the 280 basepair PCR product. To quantitate the results of the slot blots, phosphorous imaging was performed and ER and moesin levels were expressed as a percentage of β-actin as described previously.12 RESULTS SSH was used to identify genes expressed in ERnegative MDA-MB-231 cells that were absent or
minimally expressed in ER-positive MCF7 breast carcinoma cells. Seventy-one cDNA clones generated by SSH were screened by Northern blot analysis. As shown in Fig 1, 15 cDNA clones were found to hybridize to mRNA, overexpressed in MDA-MB-231 compared with MCF7. Each of these cDNA inserts was sequenced. Sequence analysis demonstrated that clones 29, 67, and 71 were derived from vimentin and clones 10 and 57 were from moesin. This analysis identified a set of genes differentially expressed in an ER-negative compared with an ER-positive breast carcinoma cell line. To identify genes consistently expressed in association with the ER-negative phenotype, the expression of each of these 15 cDNA clones was examined in a panel of breast carcinoma cell lines. Moesin was expressed in the ER-negative breast carcinoma cell lines BT-20, MDA-MB-231, and HBL-100 and the additional ERnegative cell line HeLa. Moesin expression was not detected in the ER-positive cell lines MCF7, T-47D, MDA-MB-361, and ZR-75-1 (Fig 2). As reported previously, vimentin also demonstrated a pattern of
214 Carmeci et al.
Fig 2. Northern blots of moesin and vimentin. Northern blots with panel of cell lines were hybridized with probes as indicated. Panel contains four ER-positive breast cancer cell lines (MCF7, T-47D, MDA-MB-361, and ZR-75-1) and three ER-negative breast cancer cell lines (BT-20, MDAMB-231, and HBL-100) and ER-negative cell line HeLa.
Fig 3. Western blot of moesin. Western blot of protein extracts from two ER-positive cell lines (MCF7 and T-47D) and two ER-negative cell lines (MDA-MB-231 and HBL100). Moesin is identified as 78,000-dalton protein. Radixin is also identified because of cross-reactivity with talin-4.1 family member.
expression that correlated with the ER-negative phenotype.14 The remaining 10 cDNA clones failed to demonstrate a consistent pattern of expression with the ER-negative phenotype. The data in Fig 2 demonstrate the pattern of moesin mRNA expression. Moesin protein expression was examined by Western blot. As shown in Fig 3, a 78,000-dalton protein corresponding to moesin was expressed in the ER-negative cell lines
Surgery August 1998
MDA-MB-231 and HBL-100, but moesin protein was not detected in the ER-positive cell lines MCF7 and T-47D. The moesin polyclonal antibody was known to cross-react with radixin, an 80,000-dalton member of the talin-4.1 family. Radixin was also absent from MCF7 cells. These data confirm the Northern blot data and support the association between moesin expression and the ER-negative breast cancer phenotype. RT-PCR was used to examine moesin expression in a panel of primary breast cancers. Fig 4, A illustrates moesin expression in 15 ER-positive and 14 ER-negative tumors. Two normal breast tissue samples and the MDA-MB-231 and MCF7 cell lines were also included. Higher levels of moesin expression were found in the ER-negative compared with ER-positive tumors. Fig 4, B summarizes these data for all tumor samples. As reported previously, there is a statistically significantly greater abundance of ER mRNA in ER-positive tumors.5 Consistent with the cell line data, moesin expression was found to be greater in ER-negative tumors (P < .01). Taken together, these data demonstrate a correlation between moesin gene expression and the ER-negative breast cancer phenotype. DISCUSSION There is developing evidence that there are patterns of gene expression characteristic of ERpositive and ER-negative breast cancers, and many of these genes are not part of the ER-regulation pathway. The expression of vimentin and ICERE-1 (inversely correlated with estrogen receptor; Genbank/EMBL accession number AF007790) have previously been demonstrated to be associated with the ER-negative breast cancer phenotype.12,14 These results demonstrate that the expression of moesin is also associated with the ER-negative phenotype. Previous studies of moesin have demonstrated ubiquitous expression in various tissues and tumor cell lines.16 The lack of moesin expression in a panel of ER-positive breast cancer cell lines is, therefore, particularly striking and suggests a functional role for moesin in mechanisms related to hormone-responsive breast cancer. Moesin is a member of the talin-4.1 superfamily, which includes the proteins ezrin, radixin, and merlin. 15,17 Moesin contains a conserved carboxy terminal domain that binds actin and an amino terminus that is membrane associated.18 In addition, moesin is localized in filopodia and other cellular microextensions that are cell surface structures important for cell-cell and cell-substrate recognition, adhesion, and motility.15 These findings have led to the hypothesis that moesin is critical to dynamic
Surgery Volume 124, Number 2
Carmeci et al. 215
B Fig 4. A, Moesin expression in primary breast tumors. A, RT-PCR was used to obtain semiquantitative expression of moesin in primary breast carcinomas. Fifteen ER-positive and 14 ER-negative carcinomas were evaluated. Two normal breast tissue samples and MCF7 and MDA-MB-231 cell lines were also examined in parallel as indicated. Moesin and ER expression are indicated as percentage of actin. Experiments were performed in triplicate (n = 3) with SEM shown. Tumor samples are arranged in deseending order of ER mRNA abundance. B, Cumulative data for ER-positive and ER-negative tumors (phenotype designation enzyme immunoassay). Data demonstrate statistically significant increase in moesin expression in ER-negative tumors.
216 Carmeci et al.
restructuring of the cell surface by regulating binding interactions between the plasma membrane and the actin cytoskeleton. The lack of moesin expression in ER-positive breast carcinomas thus leads us to speculate that moesin may be involved in the invasiveness or metastatic potential characteristic of the ER-negative breast cancer phenotype. The association of higher levels of moesin expression with primary ER-negative breast cancers supports the findings established in our breast cancer cell line model. However, the differences in moesin expression comparing primary ER-positive and ER-negative breast tumors were not as striking as the results in cell culture. There may be several reasons for this finding. Primary tumors are composed of a heterogenous population of cells including ductal epithelium, stromal cells, endothelium, and inflammatory cells. On the other hand, breast carcinoma cell lines are composed of a homogeneous clonal population of carcinoma cells. Previous studies have demonstrated moesin expression in all of the nonductal cell types present in breast tumors.16 Therefore even if the tumors were composed mostly of tumor cells, moesin expression in other cell types present in the tumor tissue may falsely elevate the expression level of moesin. It is possible that a direct comparison of moesin expression in ER-positive and ER-negative ductal carcinoma cells may demonstrate a more significant difference than is possible by the examination of fresh tumor tissue. Alternatively, it is possible that there is a spectrum of moesin expression in cells from the same tumor, possibly reflecting differences in metastatic potential. Further studies to examine this important finding in more detail are being pursued. Only few data are available on the control of moesin expression. One recent example has been identified in the proliferative response of mesangial to glomerular injury.19 Overexpression of the homologous protein ezrin by stable transfection of LLC-PK1 cells enhances migration of the cells but also tubulogenesis induced by stimulation with hepatocyte growth factor.20 An important effect of the expression of moesin could thus be related directly to morphogenetic effects. At present one can only speculate about the functional role of moesin in hormone-unresponsive breast cancer. Moesin has been shown to be a membrane/actin filament–linking protein involved in the formation of microextensions of the cell membrane. Such structures are considered necessary for cellular adhesion and motility and depend on activation of cdc42, rac, and rho GTPases.15,17,20,21 One attractive possibility to explain expression of moesin in
Surgery August 1998
hormone-unresponsive breast cancers is that it contributes to the increased invasiveness and metastatic potential of such tumors. These properties are reflected in more extensive and dynamic cell surface activity. REFERENCES 1. Knight WA III, Livingston RB, Gregory EJ, McGuire WL. Estrogen receptor as an independent prognostic factor for early recurrence in breast cancer. Cancer Res 1977;37:466971. 2. Sigurdsson H, Baldetorp B, Borg Å, Dalberg M, Fernö M, Killander D, et al. Indicators of prognosis in node-negative breast cancer. N Engl J Med 1990;322:1045-53. 3. Jordan VC, Wolf MF, Mirecki DM, Whitford DA, Welshons WV. Hormone receptor assays: clinical usefulness in the management of carcinoma of the breast. Crit Rev Clin Lab Sci 1988;26:97-152. 4. Barrett-Lee PJ, Travers MT, McClelland RA, Luqmani Y, Coombes RC. Characterization of estrogen receptor messenger RNA in human breast cancer. Cancer Res 1987;47:6653-9. 5. Carmeci C, deConinck EC, Lawton T, Bloch DA, Weigel RJ. Analysis of estrogen receptor messenger RNA in breast carcinomas from archival specimens is predictive of tumor biology. Am J Pathol 1997;150:1563-9. 6. Weigel RJ, deConinck EC. Transcriptional control of estrogen receptor in estrogen receptor–negative breast carcinoma. Cancer Res 1993;53:3472-4. 7. Ottaviano YL, Issa J-P, Parl FF, Smith HS, Baylin SB, Davidson NE. Methylation of the estrogen receptor gene CpG island marks loss of estrogen receptor expression in human breast cancer cells. Cancer Res 1994;54:2552-5. 8. deConinck EC, McPherson LA, Weigel RJ. Transcriptional regulation of estrogen receptor in breast carcinoma. Mol Cell Biol 1995;15:2191-6. 9. McPherson LA, Baichwal VR, Weigel RJ. Identification of ERF-1 as a member of the AP2 transcription factor family. Proc Natl Acad Sci USA 1997;94:4342-7. 10. Tang Z, Treilleux I, Brown M. A transcriptional enhancer required for the differential expression of the human estrogen receptor in breast cancers. Mol Cell Biol 1997;17:1274-80. 11. Carmeci C, Thompson DA, Ring HZ, Francke U, Weigel RJ. A novel G protein–coupled receptor associated with estrogen receptor expression in breast cancer. Genomics 1997;45:607-17. 12. Thompson DA, Weigel RJ. Characterization of a gene that is inversely correlated with estrogen receptor expression (ICERE-1) in breast carcinomas. Eur J Biochem 1998; 252:169-77. 13. Kuang WW, Thompson DA, Hoch RV, Weigel RJ. Differential screening and suppression subtractive hybridization identified genes differentially expressed in an estrogen receptor–positive breast carcinoma cell line. Nucl Acids Res 1998;26:1116-23. 14. Sommers CL, Walker-Jones D, Heckford SE, Worland P, Valverius E, Clark R, et al. Vimentin rather than keratin expression in some hormone-independent breast cancer cell lines and in oncogene-transformed mammary epithelial cells. Cancer Res 1989;49:4258-63. 15. Amieva MR, Furthmayr H. Subcellular localization of moesin in dynamic filopodia, retraction fibers, and other
Surgery Volume 124, Number 2
16.
17.
18.
19.
20.
21.
structures involved in substrate exploration, and attachment, and cell-cell contacts. Exp Cell Res 1995;219:180-96. Schwartz-Albiez R, Merling A, Spring H, Möller P, Koretz K. Differential expression of the microspike-associated protein moesin in human tissues. Eur J Cell Biol 1995;67:189-98. Lankes WT, Furthmayr H. Moesin: a member of the protein 4.1-talin-ezrin family of proteins. Proc Natl Acad Sci USA 1991;88:8297-301. Pestonjamasp K, Amieva MR, Strassel CP, Nauseef WM, Furthmayr H, Luna EJ. Moesin, ezrin, and p205 are actinbinding proteins associated with neutrophil plasma membranes. Mol Biol Cell 1995;6:247-59. Hugo C, Hugo C, Pichler R, Gordon K, Schmidt R, Amieva RM, et al. The cytoskeletal linking proteins, moesin and radixin, are upregulated by platelet-derived growth factor, but not basic fibroblast growth factor in experimental mesangial prolierative glomerulonephritis. J Clin Invest 1997;97:2499-508. Crepaldi T, Gautreau A, Comoglio PM, Louvard D, Arpin M. Ezrin is an effector of hepatocyte growth factor–mediated migration and morphogenesis in epithelial cells. J Cell Biol 1997;138:423-34. Mackay DJG, Esch F, Furthmayr H, Hall A. Rho- and racdependent assembly of focal adhesion complexes and actin filaments in permeabilized fibroblasts: an essential role for ERM proteins. J Cell Biol 1977;138:927-38.
DISCUSSION Dr Paul C. Kuo (Baltimore, Md). Although you measured mRNA in your primary breast tumors, have you determined protein expression with Western blot, immunoblot, or immunohistochemistry in these primary breast tumor cell lines? Also, what happens if you reverse the tester and driver sources of mRNA in this polymerase chain reaction–select system? Last, would you speculate on the signal transduction mechanism or the pathway that prevents expression of moesisn in ER-negative cells? Dr Carmeci. First, why did we not measure protein in the primary tumors? It is a problem to look at primary tumors when specimens and the breast biopsies they come from are getting smaller. We used samples provided as a gift from the New York University Breast Tumor Bank, and they were tested for IPOX. So the ones that were labeled ER positive were positive by IPOX and the ones that were negative, as labeled in the slides, were negative by IPOX. So they were tested for tumors. They were not tested quantitatively but they were known to be ER positive/ER negative beyond just the ER message shown on the slide. We simply did not repeat that
Carmeci et al. 217
because the finite amount of tumor sample that we had was just enough to perform our RNA isolations. Second, why not reverse the tester and the driver? We did that experiment and we did find several genes that were expressed differentially in association with the ER-positive phenotype. Two of these turned out to be novel sequences that were estrogen responsive. The results of those studies are in press in Nucleic Acids Research. With regard to the last question about the mechanism preventing moesin expression in ER-negative tumors, there are a couple of things that could be taking place. It could be a matter of lack of expression at the genomic level; perhaps ER-negative tumors have a deletion in the moesin gene or a rearrangement that destroys it or prevents its expression. That could be sorted out with a restriction length fragment polymorphism study. In addition, moesin may be deleted at the transcriptional level. A formal promoter analysis of the moesin gene has not been done. I think it would be interesting to know whether ER turned off the moesin gene. Recently in our laboratory, or about 3 years ago, an ER factor was cloned that has been shown to turn on the ER gene. The ER gene, which is a transcription factor, will then turn on several other genes: progesterone receptor and PS-2. When originally designing this experiment, I thought it would be interesting to know whether there was a gene that was turned off by the ER. A logical way of looking at that would be to look at tumors that fail to express the ER gene. I think a formal promoter analysis might shed light on whether moesin is down-regulated at the transcriptional level. Also I think immunohistochemistry might help address the last question. Dr William G. Cance (Chapel Hill, NC). I think it is a fascinating gene linking the actin cytoskeleton, and the link to invasion is critical. Do you have antibodies that work in paraffin and have you done any studies on ductal carcinoma in situ (DCIS), for example, comparing DCIS with minimally invasive tumors to the more invasive tumors and correlated them with ER positivity? Dr Carmeci. We have not done those studies. We do have antibodies to moesin that would likely work. To look at in situ hybridization in DCIS lesions and then invasive carcinoma in the same lesion might provide some insight. I think it would also be interesting to look at tumors that had metastases in the same patient that went on to convert to hormone unresponsiveness and see whether they then also went on to fail to express the moesin gene.