Snail and Slug Play Distinct Roles during Breast Carcinoma Progression

Purpose: Carcinoma progression is linked to a partially dedifferentiated epithelial cell phenotype. As previously suggested, this regulation could involve transcription factors, Snail and Slug, known to promote epithelial-mesenchymal transitions during development. Here, we investigate the role of Snail and Slug in human breast cancer progression.

Experimental Design: We analyzed Snail, Slug, and E-cadherin RNA expression levels and protein localization in large numbers of transformed cell lines and breast carcinomas, examined the correlation with tumor histologic features, and described, at the cellular level, Snail and Slug localization in carcinomas using combined in situ hybridization and immunolocalization.

Results: In contrast with transformed cell lines, Slug was found to colocalize with E-cadherin at the cellular level in normal mammary epithelial cells and all tested carcinomas. Snail also colocalized at the cellular level with E-cadherin in tumors expressing high levels of Snail RNA. In addition, Snail was significantly expressed in tumor stroma, varying with tumors. Slug and Snail genes were significantly overexpressed in tumors associated with lymph node metastasis. Finally, the presence of semidifferentiated tubules within ductal carcinomas was linked to Slug expression levels similar to or above normal breast samples.

Conclusions: These results suggest that Snail or Slug expression in carcinoma cells does not generally preclude significant E-cadherin expression. They emphasize a link between Snail, Slug, and lymph node metastasis in a large sampling of mammary carcinomas, and suggest a role for Slug in the maintenance of semidifferentiated structures. Snail and Slug proteins seem to support distinct tumor invasion modes and could provide new therapeutic targets.

Breast cancer is a major cause of female mortality in the Western world. Breast cancer is characterized by the considerable histologic heterogeneity of tumors, generating a complex classification based on cell phenotype, proliferation, and invasiveness criteria. Identification of distinct classes is crucial for therapy decisions. Typically, carcinoma cells show the reorganization of cell-cell adhesion structures and cytokeratin mesh reorganization. This partial phenotype modulation evokes more complete epithelial-mesenchymal transitions found during developmental stages involving full dissociation between individualized cells (1, 2). Phenotype modulation in breast carcinomas depends on tumor type. In infiltrating lobular carcinomas (ILC), cells express an individualized phenotype associated with the loss of E-cadherin, the main epithelial cell-cell adhesion molecule. This down-regulation is linked to distinct mechanisms including mutations, deletion events (3), and transcriptional control (4). In breast invasive ductal carcinomas (IDC), the dominant type of breast cancer, most carcinoma cells still express E-cadherin, actually used as a tumor type marker (5, 6). E-cadherin is mostly membrane-linked but may also show intracytoplasmic localization. This maintenance of some level of cell-cell adhesion is not inconsistent with cell migration and invasiveness observed in IDC. It is typically described during other migratory events throughout the developmental stages and in wound healing (2).

Members of the Snail family of zinc finger proteins have been found to be directly involved in developmental epithelial-mesenchymal transition, implicating cadherin regulation (reviewed in ref. 7). Snail and Slug have also been described in several transformed cell lines to correlate negatively with E-cadherin expression levels. Both Snail and Slug genes have been found to repress E-cadherin promoter at the level of E2 boxes (8, 9). In addition, members of the Snail family have been found to be involved in apparently unrelated functions linked to cell survival (10, 11). In recent years, the roles of Snail and Slug in cancer progression have been postulated by several groups, based on distinct lines of evidence involving RNA studies, cell models, and clinical correlations (12–14). The Snail gene was found to be regulated in human breast cancer by an estrogen-dependent corepressor, MTA3 (15). Slug was also linked to the clinical progression of lung carcinoma (16) Finally, three studies in distinct mouse models have recently shown the direct involvement of Snail (17, 18) and Slug (17) in tumor progression. However, the lack of reliable antibodies has hampered the precise cellular localization of Snail or Slug in human tumors, making it difficult to assume its precise role. Therefore, no studies have yet been published comparing Snail, Slug, and E-cadherin cellular localization within human tumors. Recently, the description of a new rat monoclonal antibody targeting human Snail has finally allowed this approach (19).

Considering the converging data suggesting a role for Snail genes during breast carcinoma progression, we initiated an exhaustive screening of normal and transformed cells to detect links between Slug, Snail, E-cadherin and tumor progression status. The screening was extended to more than a hundred clinical samples of breast cancer, examined by real-time PCR, in situ hybridization, and immunolocalization using the new monoclonal antibody directed against Snail. A new and intriguing expression pattern of Snail and Slug emerged from these studies and suggests distinct functions for the two transcription factors.

Tumor samples and clinical material. The breast tissue samples included benign lesions (five fibroadenomas and two gynecomastias). Malignant tumors were composed of 56% invasive ductal and 44% invasive lobular carcinomas. Among these carcinomas, 45% had positive lymph nodes (N+) and 55% had negative lymph nodes (N−), The histologic grade was evaluated by a certified clinical pathologist (F. Bibeau) according to the Nottingham combined histologic grade: grade 1, 11%; grade 2, 53%; grade 3, 36%. Steroid receptor status was determined using a ligand-binding assay: 74% ER+ (>10 fmol/mg protein), 26% ER−, 77% PR+ (>10 fmol/mg protein), and 23% PR−. Histologic types were determined according to the last WHO classification. Normal breast tissue was obtained, upon written consent, from three donors undergoing reductive mammoplasty and from two commercial sources (Invitrogen, Carlsbad, CA and Stratagene, La Jolla, CA).

RNA extractions and reverse transcription. RNA extraction from tumor samples and cell lines was done using a Qiagen RNA extraction kit. Prior to extraction, tumor samples were pulverized with a pestle under liquid nitrogen. After extraction, RNA quality was checked by spectrometer analysis and gel migration. One microgram of total RNA was then reverse transcribed using hexanucleotides, in combination with Superscript II Reverse Transcriptase (Invitrogen).

Real-time quantitative PCR. Real-time PCR was done using an ABI SDS7000 (Applied Biosystems, Foster City, CA), in combination with SYBR Green I. Specific primers for each gene of interest were chosen using the software Primer express (Applied Biosystems). Primer sequences are reported in the supplementary data (Table 2S). All the amplifications were done using the SyberGreen master mix (Applied Biosystems) in a final volume of 25 μL using standard PCR conditions (40 cycles with an annealing/elongation step at 60°C). Results are derived from the average of at least two independent experiments or real-time quantitative PCR estimations. Gene expression was reported relative to housekeeping gene 36B4.

Statistical analysis. Statistical analysis was done using Student's t test and Pearson correlation rank. Primary data generated by quantitative PCR were expressed as the difference in the number of cycles between the studied gene and a control housekeeping gene, 36B4, during the linear amplification phase. In some cases, values were related to a specific sample [normal breast or human mammary epithelial cells (HMEC)]. These values were used for graphs, then normalized by logarithmic transformation, when necessary, in order to apply Student's t test and Pearson correlation test.

Cell culture. BRCA MZ01, BRCA MZ02, VP 267, SUM 149-1, SUM 159-1, and SUM 185-1 cell lines were cultured as described previously (20). They were cultured in DMEM supplemented with 10% FCS. HMEC were graciously provided by Dr. Stampfer (Lawrence Berkeley National Laboratory, Berkeley, CA) and cultured according to her instructions. Briefly, finite life span 184 and extended life span culture 184 A1 HMEC were cultured in serum-free MCDB 170 medium (MEGM, Clonetics Corporation, San Diego, CA) as described (21). Cells were routinely subcultured at split ratios of 1:8. All other cell lines were obtained from American Type Culture Collection (Manassas, VA) and cultured according to the manufacturer's recommendations.

Mouse. The Slug-LacZ mouse line was initially generated by in-frame insertion of the β-galactosidase gene into the zinc finger coding region of the Slug gene (22) and graciously provided by Dr. Thomas Gridley (The Jackson Laboratory, Bar Harbor, ME). Animals were treated in accordance with institutional guidelines (Institut National de la Sante et de la Recherche Medicale).

Immunolocalization. Cells were grown on glass coverslips for 48 hours before processing. Cells were fixed with methanol at −20°C for 6 minutes. For E-cadherin and plakoglobin, cells were fixed and permeabilized with formaldehyde (4%) + triton (0.05%). For immunohistochemistry, formalin-fixed material was embedded in paraffin and processed for sectioning. After rehydration (for paraffin sections) and washing in PBS, cells or tissue sections were incubated for 1 hour at room temperature with primary antibodies, as detailed in Supplementary Table S3. For Snail antibody, antigen retrieval was done using citrate buffer (90°C, 10 minutes) followed by slow cooling to room temperature. Goat secondary antibody was added to cells for a 30-minute incubation at room temperature. Alternatively, immunoreactivity was detected using the appropriate ABC kit (Vector Laboratories, Burlingame, CA). Diaminobenzidine was used as a chromagen, and slides were counterstained with hematoxylin before dehydration and mounting. Nonspecific staining was blocked by adding 10% goat serum to both primary and secondary antibodies. Some tumor sections were stained with hematoxylin-eosin-safran. Phase contrast microscopy and immunofluorescence of cell cultures was done using a Leica DM IRB inverted microscope (Leica Microsystems, Wetzlar, Germany) and images were acquired with a CoolSNAP HQ camera (Roper, Duluth, GA).

Histochemistry. To show β-galactosidase activity, Slug-lacZ heterozygous mice were sacrificed humanely by CO₂ inhalation. Tissue samples were embedded in optimum cutting temperature, frozen in liquid nitrogen, and sectioned at 10 μm. Sections were air-dried for 30 minutes, then stained for 24 hours at 37°C in a humidified chamber using the Beta-Gal staining kit (PanVera, Madison, WI). Slides were counterstained with 0.1% nuclear fast red in 5% (NH₄)2 SO₄, dehydrated, and, mounted.

In situ hybridization. Paraffin-embedded tissue was gathered and sectioned at 3 to 6 μm onto coated slides (Fisher, Hampton, NH) and kept at 4°C until used. All slides were used within 3 months of sectioning. In situ hybridization was done in accordance with a published protocol for paraffin sections (23). Probes for Slug were designed to avoid any risks of cross-detection of other members of the Snail family and have been published previously (2). The signal was detected with BM Purple AP (Roche, Basel, Switzerland).

Slug and Snail are negatively correlated with E-cadherin expression in breast cancer cell lines, but not in HMECs. We used real-time quantitative PCR to screen 30 different human breast cancer cell lines in addition to finite life span primary HMEC 184, and an immortalized variant cell line called 184 A1 (24) for Slug, Snail, and E-cadherin mRNA expression levels (Fig. 1). Globally, Slug, Snail, and E-cadherin expression levels in HMEC cells were in the same range as normal breast cells (Fig. 1, arrow). Overall, average Slug expression was found to be significantly higher than Snail expression (∼10 times), as found by a paired t test (P < 0,05). The dispersion range of expression among distinct cell lines was very wide, spanning four levels of magnitude. As expected from published work (8, 9, 25), E-cadherin displayed a statistically significant negative correlation (P < 0,05) overall with Slug and Snail when compared one to one. The Pearson correlation indices were r = 0.50 and r = 0.36, respectively.

Slug and Snail gene expression is strongly correlated with cell phenotype. In order to establish a putative link between cell phenotypes and Slug, Snail, and E-cadherin expression, we screened all cell lines for cell-cell adhesion status, a reliable marker for epithelial phenotype. Studies included immunodetection and immunolocalization for desmoplakin, plakoglobin, cytokeratin, and E-cadherin. E-cadherin expression levels were also confirmed semiquantitatively by Western blotting. We defined functional markers for cell adhesion structures. Desmosomes were identified and located by the expression of desmoplakin at the membrane level and colocalization with anchoring cytokeratin filaments. Adherens junctions were identified by plakoglobin location at the membrane level, in addition to, in most cases, E-cadherin. Based on these criteria, we classified 32 primary or transformed cell lines into four groups, categorized in Supplementary Table S1 and shown in Fig. 1B. A majority of cells (29 out of 32) expressed at least some cell-cell adhesion structures. We sorted the cell lines by increasing the expression levels of E-cadherin mRNA (Fig. 1C) to observe a putative link between Slug, Snail, E-cadherin expression levels and phenotype groups. As expected, group 1 included cells expressing the most E-cadherin. No group 1 cells expressed significantly less E-cadherin mRNA than normal breast cells. We then examined Slug expression levels in normal and transformed cell lines. When looking specifically at the transformed cell lines, we also found Slug to be a very good indicator of cell-cell adhesion status (columns without arrows in Fig. 1C). Group 1 “epithelial” transformed cells expressed very little Slug—10 to 100 times less than normal breast samples and primary HMEC levels. All group 4 “mesenchymal” cell lines expressed Slug at levels similar or above normal breast samples. Conversely, the most intriguing finding was the high level of expression of Slug among primary and immortalized HMEC. In fact, these two categories were found to express significantly more Slug than any transformed cell lines displaying desmosomes (groups 1 and 2). Cell confluency was also found to modulate E-cadherin, Slug, and Snail RNA expression levels in HMEC cells. Similar results were observed with Snail, but to a lesser extent, potentially reflecting the very low amount of Snail expression versus Slug expression in the studied cell populations.

Slug or Snail and E-cadherin expression levels are correlated in IDCs. In order to study the involvement of Slug, Snail, and E-cadherin genes during breast cancer progression, we did quantitative PCR on 128 breast tumor samples belonging to the four most representative clinical types, covering 90% of breast cancer cases: IDCs, with or without an in situ component (DCIS) and ILCs with or without an in situ component. To this group, we added seven samples of benign breast tumors (gynecomastias and fibroadenoma) and five different normal breast samples, obtained from healthy subjects. A global comparison of Slug and Snail expression patterns (Fig. 2A) and E-cadherin expression pattern (Supplementary Fig. S1) indicated that tumor expression levels were distributed evenly around normal breast levels. Distribution was relatively narrow for normal breast samples (Fig. 2A), justifying the use of the normal breast mean value as a reference level for cell lines and tumors in this study. Similar to the cell lines, the average expression level of Snail was found by a paired t test to be significantly lower than Slug (P < 0.01) or E-cadherin (P < 0.05). Normal breast samples expressed a very significant amount of Slug. As determined by paired t test, Slug expression level was not significantly down-regulated or up-regulated in any of the tumor types when compared with normal breast sample levels. In contrast, we observed that noninvasive tumors express significantly less Snail than IDC or ILC (P < 0.01). As expected from previous findings (5, 6), levels of E-cadherin were significantly decreased (P < 0.0001) in ILC when compared with IDC (P < 0.001). We found the expression levels of E-cadherin to be the highest in DCIS, significantly above normal breast sample levels (P < 0.05; Supplementary Fig. S1).

Finally, for each tumor type, mRNA levels for E-cadherin were plotted versus the mRNA levels for Slug and Snail (Fig. 2B). Analysis of 45 IDC showed a clear positive correlation between Slug and E-cadherin expression (r = 0.35; P < 0.05) and between Snail and E-cadherin mRNA expression (r = 0.57; P < 0,0001). DCIS exhibited the same positive correlation for Slug (r = 0.60; P < 0,001) and Snail (r = 0.62; P < 0.001). Conversely, no significant correlation was observed in ILC, with or without an in situ component (data not shown).

Slug and Snail are overexpressed in IDC associated with lymph node metastasis. We studied a potential link between Slug and Snail expression and lymph node invasion among breast carcinomas. When compared with IDC that were not associated with lymph node metastasis (encompassing ∼50% of our clinical samples), the average levels of Slug or Snail expression were significantly higher in IDC associated with lymph node metastasis, as determined using unpaired t test (P < 0.05). The link was also observed within the DCIS samples. The same observation was made for Snail (P < 0.05). No significant differences were found in other tumor types. We report the expression of Slug and Snail as a function of the number of invaded lymph nodes (Fig. 2C). These observations reinforce the role of Slug and Snail in overall tumor invasiveness (7, 17, 26).

Slug is mostly expressed by E-cadherin-positive carcinoma cells. In the absence of reliable antibodies for Slug, and in order to locate Slug-expressing cells in tumors, in situ hybridization on four tumor samples was done. Serial sections from two IDC samples were analyzed simultaneously by immunolocalization to compare Slug, E-cadherin, and cytokeratin expression patterns (Fig. 3A). Carcinoma cells were distinguished from stromal cells by cytokeratin labeling and were found to express E-cadherin at the cytoplasmic or membrane level as reported earlier (5, 27). We found all carcinoma cells to express significant amounts of Slug. In samples including apparently normal mammary structures or benign breast tumors (Fig. 3B, a), mammary epithelial cells were also found to express high levels of Slug. Some expression was also found in certain stromal cells in normal and tumor samples, but only in a small percentage of cells. Control sections showed no significant hybridization (Supplementary Fig. S2).

To confirm Slug expression by mammary epithelial cells at the protein level, we analyzed a Slug-LacZ mouse (22). It was shown that in this mouse model, tissues that express full-length Slug mRNA coexpress the fusion protein, as detected by β-galactosidase activity (22), suggesting that β-galactosidase activity faithfully reflects the presence of Slug protein. We analyzed mammary gland sections from heterozygous mice to find a strong Slug expression in mammary epithelial cells from tubules (Fig. 3B, b-d).

Snail is mostly expressed by carcinoma cells but also by tumor stromal cells coexpressing vimentin. We used a recently characterized Snail antibody (19) to locate Snail expression at the cellular level. Because we found an intriguing E-cadherin and Snail correlation at the RNA level in IDC, we focused on IDC. Tumor cells were identified in all cases by cytokeratin expression. We analyzed six samples and found two distinct expression patterns (Fig. 4), reflecting Snail RNA expression levels. In the first group, immunoreactivity was restricted to stromal cells (Fig. 4A-C). Tumors in this group expressed the lowest levels of Snail RNA (lower quartile of the entire IDC group). In the second group, carcinoma cells were also found in addition to stromal cells to express Snail in nuclei at various expression levels (Fig. 4D-N). Tumors in this group expressed significantly higher levels of Snail RNA than group 1 (above median value of IDC group). In these carcinoma cells, Snail was coexpressed with E-cadherin, usually located at the membrane level (Fig. 4D-I). More generally, Snail was found to be expressed in stroma in all IDC, in fibroblast-like cells expressing vimentin in most cases, but also in leukocytes expressing CD45 (Fig. 4K-L). Not all vimentin- or CD45-positive cells expressed Snail and no correlation was observed with overall Snail RNA tumor expression levels or with tumor grade. We also looked for Snail expression by endothelial cells labeled with an anti-CD31 antibody. As expected, blood and lymphatic vessels were found to express CD31, but did not show significant Snail immunoreactivity (Supplementary Fig. S3). Overall, Snail immunostaining patterns suggest that carcinoma cells are responsible for variations in Snail RNA expression levels observed in tumor samples. We did not detect any carcinoma cells, identified by cytokeratin expression, which were simultaneously negative for E-cadherin and positive for Snail.

Morphologic analysis suggests a link between Slug and the maintenance of semidifferentiated tubes in IDC. In order to investigate a putative role for Slug or Snail in tumor cells during IDC progression, we analyzed the samples for morphologic features. Using standard criteria, a certified clinical pathologist (F. Bibeau) did a blind screen of all the available IDC samples for the expression of morphologically discernable tubes within the tumor. As seen on Fig. 5, only tumors expressing levels of Slug similar or above normal breast sample levels were found to include semidifferentiated tubes. This restriction was not observed for Snail (data not shown). We also systematically checked the IDC stromal component for changes in structure, inflammatory cell infiltration, and global extent versus tumor component. We found no correlation between Slug or Snail expression and stroma variations. We conclude that Slug could be involved in the maintenance of tubule-like structures in IDC.

This report combines, for the first time, extensive in vitro and in vivo analysis of breast carcinoma expression patterns for Slug, Snail, and E-cadherin in a panel of 35 cell lines and 144 clinical breast samples, including evaluation of RNA and protein levels. In addition to in situ hybridization, we used a quantitative PCR technique to increase sensitivity and to avoid the risks of cross-detection associated with the significant similarities between members of the Snail family. Protein expression was estimated using a newly described antibody for Snail and a mouse reporter gene model for Slug. Overall, protein expression pattern was found to validate RNA expression analysis for Slug, mostly expressed by normal and transformed epithelial cells. For Snail, elevated Snail RNA expression levels were found to correlate with substantial carcinoma cell immunoreactivity. However, we also found a significant subpopulation of stromal cells coexpressing vimentin and Snail in most tumor cases, with an unsteady pattern. In addition, we found that both genes were overexpressed in IDC that metastasized to lymph nodes and that Slug seems to be required for progression or maintenance of semidifferentiated tubules within IDC.

In accordance with previous work on a smaller sampling (8, 9, 28), we found Slug and Snail to correlate negatively with E-cadherin in transformed cell lines. However, we observed the opposite in normal HMEC cells and in a large panel of breast tumor samples, with normal or transformed epithelial cells coexpressing E-cadherin and Slug or Snail. Slug and Snail factors were found in several transformed cell models to target E-boxes in E-cadherin promoter to repress E-cadherin transcription (8, 9). Similarly, we observed that induced overexpression of Slug or Snail in HMEC cells repress an E-cadherin promoter/reporter gene construct.⁴

This observation indicates that Snail or Slug, when overexpressed in HMEC, can repress discrete E-cadherin promoter elements, even though the actual gene is not significantly down-regulated. This persistence of E-cadherin underscores the role of other regulatory elements in physiologic E-cadherin transcription regulation, as determined by several authors (29, 30). Taken as a whole, our observations are not compatible with a general down-regulation of E-cadherin gene in breast carcinomas by transcription factors Snail or Slug. However, these observations do not preclude local and transient Slug or Snail overexpression at particular sites, potentially linked to E-cadherin and cytokeratin down-regulation during epithelial-mesenchymal transition phases that could occur during the initiation of invasive events (1, 31).

Slug or Snail expression levels in breast carcinomas were found to be distributed along a wide range around normal breast sample level, encompassing two orders of magnitude. We found no correlation with any carcinoma histologic subtype, suggesting that the role of Slug and Snail during tumor progression could relate to general mechanisms involved in all subtypes. Regarding Slug, our findings unveil a new morphogenetic role during tumor progression. Slug seems to be necessary for the formation or maintenance of tubular structures in ductal carcinomas. It is tantalizing to suggest Slug that could participate in the semicohesive migration mode displayed by invasive structures within IDC. Accordingly, we found a link between the low confluency state in HMEC, associated with a less differentiated and more motile phenotype, and higher expression levels of Slug and Snail, in accordance with previous reports for Slug (32). In complement, we also found that HMEC grown on Matrigel display a more differentiated phenotype, as described previously (33), and express significantly less Slug.⁴ Finally, we recently found that keratinocytes, exemplifying a distinct population of epithelial cells involved in a cohesive migration, express large amounts of Slug and E-cadherin in vivo during wound healing re-epithelialization (2). In this case, Slug seemed to be required for cell migration to occur. During breast carcinoma progression, a related migration mode could involve semidifferentiated structures such as tubule-like structures, cordons, or less organized groups of cells, and would translate into more invasive tumors, as suggested by the link between Slug expression levels and an increase in the number of invaded lymph nodes. The presence of tubules is a morphologic criteria used to estimate tumor differentiation, one of the three criteria for Scarff, Bloom, and Richardson grading in complement to nuclear pleiomorphism and mitotic count. However, only a combination of these histopathologic markers provides a reliable prognostic tool. Probably linked to tumor heterogeneity, the presence of tubules does not itself represent a prognostic factor. Consequently, we did not find any significant correlation between Slug and Snail expression levels and tumor histologic Scarff, Bloom, and Richardson grade (data not shown).

We found an intriguing down-regulation of Snail expression in our noninvasive breast tumors when compared with invasive carcinomas as well as with normal breast samples. When added to the association we found with the number of invaded lymph nodes, this observation suggests a specific role for Snail in tumor progression, potentially linked to stroma involvement. This could involve new functions of Snail as a survival and motility inducer during cancer progression, as suggested recently (34).

In conclusion, at the cellular level, we established the links between the transcription factors Snail, Slug, and E-cadherin in breast carcinomas, as compared with transformed or normal mammary epithelial cells. Modulation of Snail and Slug expression seems to reflect distinct tumor invasion modes and could provide functional classification tools to better categorize ductal carcinomas and ultimately design appropriate therapeutic approaches.

We are very grateful to M. Stampfer for providing HMEC 184 and HMEC 184 A1 cells; A. Causse, H. Fontaine, and H. Vallès for expert technical help; C. Rodriguez for help with RNA preparation; A. Rival, C. Calas, D. Chamousset, J. Jean-Marie, and B. Ligneres for contributions to the experimental work during short-term internships; and D. Kusewitt for helpful discussions.