|Home | About | Journals | Submit | Contact Us | Français|
Mena, an actin regulatory protein, functions at the convergence of motility pathways that drive breast cancer cell invasion and migration in vivo. The tumor microenvironment spontaneously induces both increased expression of the MenaINV and decreased expression of Mena11a isoforms in invasive and migratory tumor cells. Tumor cells with this Mena expression pattern participate with macrophages in migration and intravasation in mouse mammary tumors in vivo. Consistent with these findings, anatomical sites containing tumor cells with high levels of Mena expression associated with perivascular macrophages were identified in human invasive ductal breast carcinomas and called TMEM. The number of TMEM sites positively correlated with the development of distant metastasis in humans. Here we demonstrate that mouse mammary tumors generated from EGFP-MenaINV expressing tumor cells are significantly less cohesive and have discontinuous cell-cell contacts compared to Mena11a xenografts. Using the mouse PyMT model we show that metastatic mammary tumors express 8.7 fold more total Mena and 7.5 fold more MenaINV mRNA than early non-metastatic ones. Furthermore, MenaINV expression in fine needle aspiration biopsy (FNA) samples of human invasive ductal carcinomas correlate with TMEM score while Mena11a does not. These results suggest that MenaINV is the isoform associated with breast cancer cell discohesion, invasion and intravasation in mice and in humans. They also imply that MenaINV expression and TMEM score measure related aspects of a common tumor cell dissemination mechanism and provide new insight into metastatic risk.
Multiphoton-based intravital imaging has demonstrated that invasive carcinoma cells in mouse and rat mammary tumors migrate and intravasate when associated with migratory  and peri-vascular macrophages , respectively. An in vivo invasion assay has been used in mammary tumors of rats, mice, and humans to collect migratory tumor cells associated with macrophages and these cells have been expression profiled . The gene expression changes occurring in migratory and macrophage-associated tumor cells are clustered in several pathways including the motility pathways regulating EGF stimulated locomotion. Many of these genes belong to the “minimum motility machine” comprised of the cofilin/LIM kinase, N-WASP/Arp2/3 complex and Mena/capping protein pathways [4-6]. Several genes that coordinate and control the activity of these pathways are also up regulated. One of these genes is Mena. Mena protein promotes actin polymerization by interfering with the activity of inhibitory capping proteins . The anti-capping function of Mena promotes sustained actin polymerization, which is essential for directional cell movement in response to growth factors like EGF.
Mena is upregulated in rat, mouse and human mammary tumors [8-10]. Similarly, in precursor lesions to cancer of the cervix and colon, Mena expression is increased with progressive transformation [11, 12]. Mena has been used successfully as part of a marker of metastatic risk called TMEM (Tumor Micro-Environment of Metastasis) observed in formalin-fixed paraffin-embedded sections of human invasive ductal carcinomas of the breast . TMEM is an anatomical site consisting of a macrophage in direct contact with a Mena expressing tumor cell, and an endothelial cell. TMEM counts are associated with risk of metastasis in breast cancer patients independently of lymph node, ER/PR or HER2 status .
Mena, a member of the Ena/VASP family of proteins, regulates membrane protrusion and cell movement in a variety of cell types and contexts by influencing the geometry and assembly of actin filament networks [13-17]. Mena is alternatively spliced. In particular, an exon encoding a sequence of 19 amino acids inserted between the EVH1 domain and the LERER repeats generates a Mena invasion (INV) isoform (previously called Mena+++) [13, 18] and inclusion of an exon encoding a 21 amino acid insertion in the EVH2 domain produces the Mena11a isoform . Until recently, MenaINV was thought to be expressed primarily in axons of developing neurons , wheras expression of Mena11a was found in epithelial cancer cells . However, expression profiling of cells collected using the in vivo invasion assay revealed that MenaINV expression is specifically upregulated and Mena11a is downregulated in migratory/macrophage associated tumor cells compared to average primary tumor cells (APTCs) isolated from the same rat and mouse mammary tumors by FACS . Expression of MenaINV in a xenograft mouse model increases spontaneous lung metastases from mammary tumors and alters the dose-response of tumor cells to epidermal growth factor (EGF), allowing cells to respond to much lower EGF concentrations than control cells both in vivo and in vitro .
Here we demonstrate that MenaINV promotes discohesive tumor morphology in a xenograft mouse model while Mena11a expression promotes cohesive tumor morphology. We also show that increased expression of MenaINV but not Mena11a is associated with metastasis in the PyMT mouse model of breast cancer. Furthermore, we demonstrate that MenaINV expression correlates with the assembly of cancer cell intravasation sites in human breast cancers using TMEM score. It seems that both TMEM score and Mena isoform expression measure an aspect of tumor biology unrelated to standard clinical and pathological parameters because they do not correlate with tumor, size, lymph node status or ER, PR and HER2/Neu receptor expression. Thus, we suggest that MenaINV may have useful clinical applications as a prognostic marker for metastatic risk and target for therapy.
EGFP-Mena splice isoforms were subcloned into the retroviral vector packaging Murine stem cell virus-EGFP using standard techniques . MTLn3 rat adenocarcinoma cells were infected with each Mena isoform construct to create MTLn3-EGFP-Mena (referred to as Mena cells), MTLn3-EGFP-MenaINV (referred to as MenaINV cells) and MTLn3-EGFP-Mena11a cells (referred to as Mena11a cells) and used for all orthotopic injections of SCID mice. Details about the retroviral system used to generate these lines have been described previously . Monoclonal MTLn3 cells were derived from spontaneous lung metastases in a rat model 13762 , and were used in these experiments because they are known to metastasize to the lung when injected into the mammary gland of SCID mice and thus are suitable for metastatic studies . Additionally, these cells have been used to study Mena isoforms previously [17, 21]; thus we wished to be able to compare current work with work previously published. To achieve uniform forced expression, EGFP -MenaINV and -Mena11a cells were FAC sorted to a level of 4-fold overexpression of EGFP-Mena isoforms as compared to EGFP-Mena on the protein level. MTLn3-Cerulean EGFP-Mena splice isoforms were created using a lentiviral system pCCLsin.PPT.hPGK.Cerulean.pre (courtesy of Dr. Gupta, Albert Einstein College of Medicine). Cytoplasmic expression of Cerulean was used to improve visualization of tumor cells during intravital imaging. MTLn3 cells were cultured in alpha-modified minimum essential medium (MEM) supplemented with 5% fetal bovine serum (FBS) and 0.5% PenStrep (Invitrogen).
Orthotopic mammary tumors were derived by the subcutaneous injection of 1×106 EGFP-MTLn3 (referred to as GFP), -EGFP-Mena, -EGFP-MenaINV -EGFP-Mena11a, -Cerulean-EGFP-MenaINV or - Cerulean-EGFP-Mena11a cells into the mammary gland of 5–7 week-old female SCID mice (these mice are referred to as xenograft mice) . All experiments involving animals were approved by the Einstein Institute for Animal Studies.
Intravital multiphoton imaging was performed in tumors of SCID mice expressing MTLn3-cerulean-EGFP-MenaINV or cerulean-EGFP-Mena11a as described previously [23, 24] using a 20× 1.95 NA water immersion objective with correction lens. Briefly, a skin flap surgery was performed to expose the mammary tumor of an anesthetized mouse and the mouse was then placed on the microscope for imaging. Visualization of collagen within the primary tumor is made possible by the generation of a second harmonic signal which results from the reflection of light off of collagen helices .
Scoring of cohesive and discohesive tumor morphology was done using images obtained with intravital multiphoton microscopy. A minimum of 4 fields from each tumor imaged in xenograft mice were divided into quadrants. 5 mice per isoform type were evaluated (a total of 20 fields were scored per tumor type). A score of C (cohesive) indicated a quadrant containing a continuous sheet/ large cluster of tumor cells. A score of D (discohesive) indicated that > ½ of the quadrant lacked continuous sheets or clusters of cells. Data in figure 1d is represented as percentage of total fields analyzed that received a score of C versus percentage of fields that received a score of D. Images were scored double blind and the findings from each correlated with each other.
Mice were euthanized 3 weeks after carcinoma cell injection into the mammary gland. FNA was performed as previously described . In brief, cells collected with 25 gauge needles were expelled onto a glass slide and smeared. Smears were stained by standard Diff-Quick protocol . Each smear was given a score based on the ratio of the number of clusters to the approximate percentage of single cells, either <30% <60% or <90% as observed in a single low power field area (40X; 12.56 mm2). A cluster was defined as >10 cells in direct contact with each other. Each smear was also independently described as cohesive or discohesive based on the relative abundance of clusters as compared to single cells. Both methods were scored double blind and the findings from each correlated with each other.
PCR analyses were performed using SyBR Green kit and ABI 7300 sequence detector. Primers detected MenaINV and Mena11a as described previously . The data were analyzed by ABI Sequence Detection Software (Applied Biosystems Foster City, CA).
Following 4 weeks of tumor growth, five primary tumors from xenografts of each tumor type derived from injection of MTLn3-EGFP, -EGFP-Mena, -EGFP-Mena11a and –EGFP-MenaINV, cells were fixed in 10% buffered formalin and paraffin embedded. Sections 10βm thick were cut from each tumor and placed on slides for further staining. IF was done using anti-GFP (Aves Lab Cat. #1020) at 1:500 dilution with Alexa488 as a secondary, and anti-β-catenin (BD Biosciences Cat. #610153) at 1:500 dilution with Alexa594 as a secondary. Digital images were converted in ImageJ (NIH) and analyzed using macro analysis that defines fluorescence intensity starting at the cell periphery and extend into the cell interior . Primary tumor sections were imaged using the Zeiss AxioObserver.Z1 with a 10× 1.4 NA Plan Apo objective with the apotome and processed using ImageJ (NIH).
IHC for E-cadherin was done using commercially available monoclonal anti-E cadherin antibody (Dako, Carpinteria, CA) at 1:25 dilution. Antigen retrieval was done in a steamer at 90°C for 30 minutes in Target Retrieval solution pH 6.0. The slides were incubated with the primary antibody for 30 minutes at room temperature and 30 minutes with a secondary antibody. E-cadherin was visualized using anti-mouse horseradish peroxidase-labeled polymer (Envision System, Dako) and DAB on an automated immunostainer (Autostainer, Dako) according to the manufacturer’s instructions. The slides were counterstained with hematoxylin using standard technique.
MTLn3 protein lysates were prepared, the samples were resolved by SDS-PAGE, transferred to nitrocellulose, blocked in odyssey blocking solution (LiCor), incubated in primary antibodies overnight at 4°C, secondary antibodies for 1 hour at room temperature, and analyzed using the Odyssey (LiCor) . Primary antibodies, anti-Mena [1:1000] , anti-Mena11a [1:5000] , anti-MenaINV [1:500] (unpublished, more information available upon request), anti-beta-actin [1:5000] (Invitrogen), were used at the indicated concentrations. Secondary antibodies, Mouse 680 (used against actin and Mena) and Rabbit 800 (used against Mena11a and MenaINV) were purchased from LiCor and imaged in separate channels. Endogenous Mena, Mena11a and MenaINV were detected at lower molecular weights as compared to EGFP-Mena11a and EGFP-MenaINV.
Transgenic animals with mammary gland-specific expression of Polyoma middle T (PyMT) antigen in the FVB-C3H/B6 background were obtained from the Albert Einstein College of Medicine mouse repository. FNA was performed with a 25 gauge needle on animals anesthetized with 5% isoflurane and sacrificed by cervical dislocation. The initial FNA biopsy was used for quality assessment using standard Diff-Quick stain. Subsequent biopsies were used for qRT-PCR analysis. Staging of H&E stained tumor sections from formalin fixed, paraffin embedded tissue was done as previously described .
Lumpectomy and mastectomy specimens received at the Albert Einstein College of Medicine/ Montefiore Medical Center, Moses and Weiler Divisions for pathological examination were used for FNA-based tissue collection under institutional IRB approval. Four to five FNA aspiration biopsies were performed on grossly visible lesions using 25 gauge needles. The adequacy of the sample was assessed by standard Diff-Quick protocol . Only samples composed of 90% of either benign or malignant epithelial cells, as determined by standard pathologic characteristics , were used in the study.
At the time of routine microscopic examination of the lesions on which FNA biopsies had been performed, an appropriate area of the tumor suitable for TMEM analysis was identified by low power scanning. The following criteria were used: high density of tumor, adequacy of tumor, lack of necrosis or inflammation, and lack of artifacts such as retraction or folds. TMEM staining and assessment were done as described previously . Briefly, ten digital images were acquired at 400x total magnification. Using the “circle” tool available in Photoshop, all TMEM are “marked” using circles, and the “marked” images are saved as separate files. The total TMEM count for each image was tabulated, and the counts from all ten images are then summed to give a final TMEM density for each patient sample, expressed as the number of TMEM per 10 high power fields (400x total magnification). Two pathologists independently scored all fields. The scores for each case were averaged for the two pathologists to yield a final score used in the data analysis.
Similarly to TMEM scoring, ten digital images were acquired at 400x total magnification for each xenograft tumor and cells with complete membranous staining were marked with asterisk using Photoshop tools. The total number of cells from all ten images was summed to give a final number of cells with complete E-cadherin membranous stain for each xenograft. There were 3 xenografts per group; thus a total of 30 high power fields (400x total magnification) were analyzed per group. The results are expressed as a mean (with SEM) of number of cells with complete membranous stain per group.
FNA primarily collects loose tumor cells with very few macrophages and no endothelial cells and incurs minimal tissue damage . Thus, after the FNA procedure, the entire tumor was formalin-fixed, paraffin-embedded (FFPE) and sent for pathological examination. A representative block of FFPE tumor tissue was selected and triple immunostained for TMEMs. Therefore, the FNA sample and TMEM count were obtained from the same tumor for the entire cohort (Figure S3).
Statistical significances were determined using unpaired, two-tailed Student’s t-tests assuming equal variances and an alpha level of 0.05. Differences were considered significant if the p value was <0.05. Actual P-values are listed on graphs within each figure. For differences in Mena isoform expression between metastatic and non-metastatic PyMT tumors and the association between TMEM density or Mena isoform fold change with tumor grade, lymph node status, tumor size, ER, PR and Her2/Neu status, Wilcoxon Mann-Whitney rank-sum test was used. Given that 6 comparisons were done for human samples, the P values for determining statistical significance was set at 0.008 by applying the Bonferroni correction to the standard assumption that P < 0.05 is statistically significant. The strength of association between Mena isoform expression and TMEM density was calculated using Spearman’s correlation coefficient.
To investigate the effects of expression of Mena, MenaINV and Mena11a on primary tumor morphology and tumor cell cohesiveness, we generated orthotopic mammary tumors from MTLn3 rat adenocarcinoma cell lines forced to express Mena, MenaINV and Mena11a isoforms. Our previous studies show that invasive tumor cells spontaneously increase expression of MenaINV, and decrease expression of Mena 11a in vivo . We also showed that MTLn3 mammary carcinoma cells forced to express MenaINV, are sensitized to EGF and are more metastatic . We hypothesized that the highly invasive and metastatic nature of MenaINV cells might be in part caused by decreased tumor cell cohesion [6, 33]. Likewise, the expression of Mena 11a, characteristically found in more epithelial tumor cells  and non invasive tumor cells in vivo , might promote cohesive tumor cell morphology. These considerations also suggest that FNA of tumors could be used to detect differences in Mena expression patterns. To test this, FNA samples were collected from MTLn3-derived mammary tumors. Examples of cohesive and discohesive FNA cytopathology smear patterns are shown in Figure 1a. We found that FNA smears obtained from Mena11a expressing mammary tumors were more cohesive than smears obtained from Mena and MenaINV expressing mammary tumors (Figure 1b). MTLn3 control tumors (referred to as GFP) and tumors derived from Mena and MenaINV expressing MTLn3 cells had similar discohesive smear patterns (Figure 1b). Quantification of cell clusters and discohesive areas confirmed that Mena11a tumors were significantly more cohesive than GFP, Mena or MenaINV tumors (Figure 1b).
To further characterize the effects of Mena, Mena11a and MenaINV on tumor cell cohesion within the primary tumor we used multiphoton-based intravital imaging (IVI). Within the intact primary tumor of an anesthetized mouse, Mena11a expressing tumors had well defined cohesive cell-cell contacts, whereas MenaINV expressing tumors had discontinuous cell-cell contacts and the cells were more randomly organized resulting in a discohesive appearance (Figure 1c, S1). Mena expressing tumors also demonstrated more discohesive appearance as compared to GFP tumors but did not show as drastic a trend toward discohesion as that seen in MenaINV expressing tumors. All images were taken under non-bleached conditions with lower laser power of the multiphoton microscope allowing the weaker cytoplasmic fluorescence to persist. This allowed direct comparison of Mena, MenaINV, and Mena11a expressing tumor cells and demonstrates that Mena is more diffuse than the Mena11a or MenaINV isoforms (Figure 1c).
Primary tumor morphologies were scored based on cohesive and discohesive appearance in tumors expressing GFP fused to the Mena isoform indicated. Mena11a tumors had the largest percentage of cohesive fields while MenaINV tumors had the largest percentage of discohesive fields (Figure 1d).
Given the strong correlation of Mena11a expression with cohesion, and MenaINV expression with discohesion of tumor cells, we asked whether MenaINV and Mena11a isoform expression affects the integrity of cell-cell junctions using β-catenin and E-cadherin immunofluorescence and immunohistochemistry, respectively. Localization of these proteins at cell-cell contacts has been used to investigate cell junction integrity [34, 35]. Mena11a expressing mammary tumors exhibited continuous β-catenin immunofluorescence staining at cell-cell contacts while staining for β-catenin in MenaINV expressing mammary tumors was discontinuous (Figure 2a). Mena11a tumor cells also had a significantly higher intensity of β-catenin staining at cell-cell contacts (Figure 2b), and within the whole cell (Figure S2) as compared to MenaINV cells. Mena expressing tumor cells had similar localization of β-catenin at cell-cell contacts (Figure 2a) and increased expression within the whole cell (Figure S2) as compared to GFP control tumors. In addition to a reduction in β-catenin staining, tumor cells from MenaINV expressing tumors also showed a dramatic reduction in continuous E-cadherin staining at cell-cell junctions as compared to that seen in GFP and Mena11a expressing tumors (Figure 2c). E-cadherin staining pattern was similar to β–catenin; predominantly discontinuous in MenaINV tumors when compared to GFP, Mena or Mena11a expressing tumors (Figure 2c and d). We found significantly lower numbers of cells with complete membranous staining in MenaINV tumors than in GFP control, Mena or Mena11a expressing tumors (Figure 2d), consistent with our finding that Mena11a promotes a cohesive epithelial morphology while MenaINV supports a discohesive morphology in vivo.
To rule out if the observed effects of MenaINV and Mena11a cells on invasion are due to compensatory expression of the different Mena isoforms we performed western blots on each cell line to check for expression of each Mena isoform. In MenaINV cells the levels of endogenous Mena or Mena11a were unchanged as compared to GFP control cells (Figure 3). In Mena11a there were no detectable differences in Mena or MenaINV levels as compared to GFP cells (Figure 3).
To assess how cohesion and Mena isoform expression correlate with metastatic status we studied PyMT mouse mammary carcinoma. These tumors histologically recapitulate progression of human ductal carcinoma of the breast from ductal hyperplasia and ductal carcinoma in situ to invasive ductal carcinoma . Invasive ductal carcinomas in PyMT tumors can be additionally stratified into histologically distinct subcategories of early and late carcinomas. Unlike in humans where cancer morphology does not reflect tumor metastatic potential, late PyMT cancers are associated with tumor cell dissemination and lung metastases [31, 36, 37]. Thus we assessed the FNA smear pattern and Mena isoform expression in early and late PyMT tumors to determine if Mena isoform mRNA expression correlates with the smear pattern and metastatic outcome. Smear patterns obtained from early PyMT mouse mammary tumors were predominantly cohesive, while those obtained from late, metastatic tumors were predominantly discohesive (Figure 4a-i and 4a-ii). Quantitative real time PCR (qRT-PCR) of FNA samples shows a spontaneous 8.7-fold increase in pan-Mena expression (determined with primers that identify all Mena isoforms = pan-Mena) and 7.5-fold increase in MenaINV expression in late carcinomas when compared with early carcinoma, while Mena11a amplicons were decreased by 70% in late carcinomas (Figure 4b). Thus, the Mena isoform expression pattern in cells collected by FNA from late, metastatic tumors matches that of the invasive cells collected by the in vivo invasion assay while the Mena expression pattern from early carcinomas matches that of average primary tumor cells (APTCs) isolated from the same rat and mouse mammary tumors by FACS (2). This finding suggests that the pattern of Mena isoform expression in FNA samples may reflect invasive potential of FNA collected cancer cells.
The correlation of spontaneously increased MenaINV and decreased Mena11a expression with tumor progression and metastatic outcome in transgenic PyMT mouse mammary tumors, and our recent findings showing that MenaINV, but not Mena11a, expression increases tumor cell motility, intravasation and dissemination in xenograft mouse mammary tumors [17, 21], led us to ask if the altered expression of these isoforms correlates with an increased risk of metastasis in humans.
Since our study is prospective we used TMEM score as a measure of increased metastatic risk. As previously shown TMEMs are visualized by a triple immunostain to detect the direct contact of macrophages, Mena expressing carcinoma cells and endothelial cells in formalin fixed paraffin embedded tissues . The number of TMEM sites has been shown previously to be significantly higher in tumors that produce distant metastases . TMEM was developed based on studies that used multiphoton intravital imaging of mammary tumors in PyMT mice showing breast carcinoma cells intravasating at sites along blood vessels where peri-vascular macrophages are in direct contact with tumor cells and endothelial cells [38, 39]. Examples of TMEMs (marked by circles) are shown in two human invasive ductal carcinomas as identified in histological sections (Figure 5a). TMEM was also identified in xenograft mammary tumors derived from injection of MTLn3-MenaINV cells (Figure 5b).
We hypothesize that MenaINV expressing tumor cells are involved in the assembly of TMEM and therefore human tumors with numerous TMEM intravasation sites will have a higher proportion of discohesive, invasive and migratory tumor cells expressing MenaINV compared to tumors with few TMEM sites. To investigate this we collected cells from 40 patients’ invasive ductal carcinomas (IDC) and 5 fibroadenomas by FNA.
The expression of Mena, Mena11a and MenaINV at the protein level was assessed by western blotting. As shown in 4 randomly chosen FNA samples (Figure S4), Mena isoform expression is detectable at the protein level, but only with long exposure. While these blots establish that the Mena isoforms are expressed at the protein level in FNA samples, the number of cells collected in FNA samples was variable and not sufficient for western blot analysis of correlations with TMEM count. Thus, the quantification of Mena isoform expression was determined by PCR for all FNA samples to be correlated with TMEM count.
Therefore, PCR primers specific to the Mena isoforms of interest were used with all 40 FNA samples and the results were expressed as a fold change of Mena isoform mRNA expression in IDC, compared to the mean level of Mena isoforms mRNA obtained from 5 fibroadenomas. TMEM counts were scored in tumor samples obtained from the same cohort of 40 patients, and were correlated with the fold change in Mena isoform mRNA expression (Figure 5c). The tissue collection methods are outlined in Figure S3. The Spearman’s correlation coefficient for the association between MenaINV fold change and TMEM count was 0.78 (p= 10−6). An inverse association between Mena11a and TMEM was also expected but was not found (r = −0.23, p= 0.14).
We also evaluated the collective fold change of all Mena isoforms using primers that amplify all Mena isoforms (pan-Mena). The expression of pan-Mena did not correlate with TMEM score (r= −0.15, p= 0.37) (Figure 6). These results indicate that MenaINV is the isoform that is correlated with TMEM count and therefore with the metastatic risk in breast carcinoma.
These same tumor samples were also classified according to the modified Bloom Richardson scale. Patients’ age, tumor size, tumor grade, lymph node status, estrogen, progesterone and HER2/Neu receptor status were documented along with MenaINV, Mena11a expression level and TMEM score (Table S1). Median values as well as 5th and 95th percentiles of MenaINV, Mena11a expression levels and TMEM scores for all recorded tumor characteristics stated above was calculated (Table 1). We found statistically significant difference in TMEM score between the tumors of low and high grade (p = 0.004). However, no statistically significant difference in Mena isoform mRNA fold change among the tumors of different grades was found. Likewise, there was no statistically significant difference in TMEM score and Mena isoform mRNA fold change among tumors of different size, lymph node status, ER, PR and HER2/Neu expression, suggesting that both TMEM score and Mena isoform mRNA expression reflect mechanisms of tumor invasion/ progression independent of most currently used clinical and pathological parameters.
These data indicate that the expression of MenaINV mRNA correlates with TMEM score, and are consistent with the PyMT mouse tumor results showing that expression of MenaINV, but not Mena11a, correlates with metastasis (Figure 4b). They are also consistent with the results of our recent xenograft experiments documenting that MenaINV is the isoform associated with increased intravasation, dissemination and higher rates of lung metastasis .
Animal models indicate that carcinoma cells located in invasion-inducing microenvironments undergo transient and sometimes stable epigenetic changes similar to those that drive morphogenetic cell movements in the developing embryonic organ [2, 40-42]. One of the changes is the increased expression of the actin regulatory protein MenaINV and decreased expression of Mena11a . Here we have found that changes in isoform expression from Mena11a to MenaINV correlate with the loss of epithelial cell-cell contacts and the emergence of a discohesive cell population in the primary tumor. Additionally, previous studies have reported an increase in cell motility in vivo and lung metastases in orthotopic mammary tumors derived from MenaINV expressing tumor cells . Thus, the discohesive primary tumor morphology and discontinuous cell-cell adhesion contacts observed in MenaINV tumors supports the enhanced migratory phenotype previously reported .
We demonstrate that expression of Mena11a promotes the maintenance of epithelial cell-cell contacts and a cohesive cell population within a mammary tumor. This is consistent with recent findings showing that expression of Mena11a delays but does not prevent metastatic progression in MTLn3 xenograft mammary tumors . These results suggest that the expression of Mena11a in tumor cells will support their retention in cohesive regions of tissue with relatively stable cell-cell junctions (Figure 7). Thus, the FNA smears from the tumors with the high proportion of Mena11a expressing cells show mostly cohesive clusters likely reflecting a less malignant clinical outcome.
Previous studies have shown that MenaINV expressing carcinoma cells have increased movement in vivo  (Figure 7-Box i). Sensitivity of tumor cells to a particular chemoattractant will determine the direction and speed of their movement within the tissue . Previous studies also show that MenaINV expressing cells are hypersensitive to EGF during in vivo invasion . This sensitivity could therefore promote migration toward, and association with EGF producing perivascular macrophages, resulting in intravasation (Figure 7-Box iii). This hypothesis is supported by previous observations of carcinoma cell migration toward blood vessels and intravasation in mouse mammary tumors . We propose that a high proportion of single cells in FNA smears, and their expression of MenaINV, is likely to reflect the presence of a migratory, and intravasation competent cell population, and therefore an adverse clinical outcome.
Our results show that the presence of MenaINV expressing carcinoma cells in FNA samples is correlated with the presence of the anatomical structure called TMEM [9, 17]. As previously mentioned, TMEM is defined as the direct contact of a perivascular macrophage, an endothelial cell and Mena expressing tumor cell (Figure 7-Box ii). In recent findings we show that MenaINV expressing carcinoma cells have dramatically increased intravasation . Thus, it is possible that the expression of MenaINV can provide an advantage to tumor cells in their ability to find macrophages when located in microenvironments with low concentrations of EGF due to the increased sensitivity of MenaINV tumor cells to EGF . These findings indicate that increased MenaINV expression may affect metastatic potential of a particular breast tumor by contributing to macrophage-dependent cell motility throughout the primary tumor (Figure 7 Box i), and intravasation at TMEM sites (Figure 7-Box ii and iii). Indeed, our findings in PyMT mice indicate that MenaINV promotes metastatic progression.
We show here that human invasive ductal carcinomas with high MenaINV expression in FNA samples correlate with a high TMEM score. However, expression of Mena11a does not correlate with TMEM score and this is consistent with the recent findings that Mena11a expressing tumor cells have no effect on intravasation or metastasis relative to GFP control cells . The observations in human tumors also suggest that expression of the Mena11a isoform does not affect the late stages of metastasis. These results support the involvement of the MenaINV isoform in assembly of TMEM, and intravasation in human breast carcinomas.
We recently reported a case-control study demonstrating that TMEM density is associated with increased risk of metastasis of invasive ductal carcinomas of the breast . That study demonstrated that TMEM count correlates with tumor grade supporting the fact that low grade tumors rarely metastasize. We also found a significant difference in TMEM score between low and high tumor grades in our cohort (p= 0.004). In both the previous and current studies, we did not observe any association between TMEM count and ER, PR, HER2/Neu expression or lymph node metastasis. Since about 10-15% of patients develop metastatic disease within 3 years of diagnosis [4, 5] we would expect about 5 cases from our cohort of 40 to be metastatic within the same time frame. Indeed, only 5 cases from our cohort (15%) have TMEM values around or above the 90th percentile (107.6) (Table S1 and Figure 5c). Three of those cases also have MenaINV around or above the 90th percentile (5.64).
In summary, our results indicate that carcinoma cells with an elevated MenaINV expression are discohesive (Figure 5 Box i), while those with elevated Mena11a expression are more cohesive. Our results also suggest that MenaINV expressing cells assemble TMEM, and intravasate at TMEM sites (Figure 5 Box ii and iii). It will be important to determine how epigenetic changes may affect isoform expression and if these changes act through control of alternative splicing. Studies are underway to investigate the role of alternative splicing mechanisms in control of Mena isoform expression (FG, unpublished). Future studies will investigate molecular and biochemical mechanisms of action of the MenaINV and Mena11a isoforms, as well as their utility as predictors of breast cancer outcome and as targets for therapy.
We would like to thank Drs. Jeffery Segall, Diane Cox, Antonia Patsialou, and Daqian Sun for stimulating discussion and helpful suggestions. We also thank Drs. Sasis Sirikanjanapong, Jason Moss, and Zhong, as well as research associate Mrs. Felicia Juliano for assistance with FNA biopsy procedures, Dr. Jaya Sunkara for assistance with E-cadherin staining, and Dr. Olena Dorokhova for help with RNA extraction and cDNA synthesis. Many thanks to David Entenberg, and Jenny Tadros for their technical support, Einstein histopathology, Analytical Imaging Facilities, and Koch Institute for Microscopy, Histology and Flow Cytometry sorting core facilities for their services. Grant support provided by CA100324 (SG, YW), CA113395 (MHO, JSC), CA126511 (MHO, JSC), CA150344 (ETR), AECC9526-5267 (MHO, SG), Ludwig Fund postdoctoral fellowship (MB), GM58801 and funds from the Ludwig center at MIT (FBG), ICBP grant U54 CA112967 (FBG).