Caveolin-1 expression is elevated in claudin-low mammary tumor cells

Caveolin-1 (Cav-1) is a member of the caveolin family which consists of Cav-1, Cav-2 and Cav-3. Caveolins function as scaffolding proteins and are found in plasma membrane invaginations known as caveolae. Cav-1 and Cav-2 are co-expressed and found in a number of cell types including endothelial cells, adipocytes, and type I alveolar cells while Cav-3 is expressed in skeletal, cardiac and smooth muscle cells [1‐4]. In normal cells caveolins regulate signal transduction, cholesterol metabolism, and vesicular transport.

Most of the work performed on Cav-1 in breast cancer has been directed towards the stromal cells. Loss of Cav-1 in stromal cells surrounding breast cancers has been associated with disease recurrence, metastasis, tamoxifen resistance and poor clinical outcome [5, 6]. It has been suggested that loss of Cav-1 in tumor-associated fibroblasts induces aerobic glycolysis in these cells resulting in increased production of pyruvate and lactate. The adjacent tumor cells would then utilize these substrates for ATP production. This phenomenon has been termed "Reverse Warburg Effect" [7‐10]. A modification of this hypothesis suggests that loss of stroma Cav-1 increases autophagy in the tumor microenvironment leading to recycled nutrients that can be utilized by the tumor cells and protection of the tumor cells from apoptosis and autophagy [11‐14].

There have been some studies investigating Cav-1 levels in breast tumor cells. In two independent studies Cav-1 expression was found to be lower in human breast cancer cell lines than normal mammary tissue [15, 16]. In addition, overexpression of Cav-1 in human breast cancer cell lines inhibited proliferation and soft agar colony formation [16]. However, studies by Heiser et al. [17] and Elsheikh et al. [18] showed that Cav-1 was frequently observed in basal but not luminal, breast cancer cells. Therefore, Cav-1 may be expressed only in certain breast cancer subtypes.

Cav-1 null mice have been created and display mammary epithelial hyperplasia [19‐21]. Since Cav-1 expression is lost in all cells in these mice it is difficult to determine whether the hyperplasia results from Cav-1 loss in the epithelial cells, stromal cells or both. To examine the effects of Cav-1 loss directly on mammary epithelial cells, mammary epithelial cells from Cav-1 null mice were transplanted into the mammary fat pad of wild type mice. Cav-1 deficient mammary epithelial cells displayed increases in proliferation, terminal end bud area, and mammary ductal thickness [21].

Our group became interested in Cav-1 following the analysis of DNA microarray results that compared mammary tissue from wild type mice to tumor tissue from transgenic mice that express elevated levels of the type I insulin-like growth factor receptor (IGF-IR) in mammary epithelial cells. Cav-1 was significantly downregulated (~37-fold) in the mammary tumor tissue compared to normal mammary tissue [22]. Interestingly, tumors that recurred following IGF-IR downregulation in the mammary tumors acquired a claudin-low phenotype and genotype and expressed Cav-1 at significantly higher levels (~14-fold) than the initial mammary tumor [22]. To study these differences further, Cav-1 expression was evaluated in mammary tumor tissue and mammary tumor cell lines. We found that the differences in Cav-1 expression were a result of tumor cell Cav-1 rather than stromal Cav-1. Epithelial mammary tumors expressed low levels of Cav-1 while claudin-low mammary tumors expressed much higher levels of Cav-1. These findings were confirmed in two murine mammary epithelial tumor cell lines and two murine mammary claudin-low cell lines in that the claudin-low cell lines expressed elevated levels of Cav-1 compared to the epithelial ones. Therefore, our findings suggest that Cav-1 expression can be used to differentiate mammary tumors with epithelial characteristic from those with mesenchymal or more specifically, claudin-low characteristics.

DNA microarray analysis was used to compare three sets of murine tissues; (i) adult, wild type mammary tissue (WT), (ii) primary mammary tumors induced by transgenic IGF-IR overexpression which present with an epithelial morphology (PMT; primary mammary tumor), and (iii) mammary tumors that resume growth following IGF-IR downregulation. These tumors that arise following IGF-IR downregulation express low levels of the IGF-IR and present with a spindle shaped morphology (RST; recurrent spindle tumor). The partial characterization of these tumors has been presented in [24] while characterization of the IGF-IR transgenic mice (MTB-IGFIR mice) has been presented in [23].

The DNA microarray analysis indicated that Cav-1 was reduced approximately 37-fold in the PMT tissue compared to WT mammary tissue [22]. As Cav1 is expressed in adipocytes, endothelial cells, myoepithelial cells and fibroblasts but not epithelial cells [1‐4] this finding was not completely surprising as the PMTs are composed primarily of epithelial cells [23]. Interestingly, Cav-1 expression was somewhat restored in the RST tissue. Cav-1 expression as determined by the DNA microarray was approximately 14-fold higher in the RST samples compared to the PMT samples [22]. To confirm the DNA microarray findings, western blotting was performed and these blots showed that the levels of Cav-1 protein were highest in normal mammary tissue from WT mice, reduced in the PMTs and somewhat restored in the RSTs (Figure 1).

To determine whether Cav-1 was being expressed in the tumor cells and/or the stromal cells, immunohistochemistry for Cav-1 was performed. In the PMT tissue sections adipocytes and cells lining blood vessels stained strongly for Cav-1 protein as did basal cells in normal mammary ducts (Figure 2A; arrow). The tumor cells themselves primarily expressed only low levels of Cav-1 protein (Figure 2A; T) however some regions of the PMT had clusters of cells that stained positive for Cav-1 (Figure 2B, arrowhead). Therefore, most of the tumor cells in PMTs express low levels of Cav-1 however, pockets of tumor cells expressing high levels of Cav-1 can be found within some of these tumors. PMTs in the MTB-IGFIR transgenic mice are not typically surrounded by fibroblasts. In a few tumors where fibroblasts could be found adjacent to the mammary tumor, these fibroblasts displayed little or no staining for Cav-1 (Figure 2A, B). In the RST tissue sections, Cav-1 protein could be detected at moderate (Figure 2C) or high (Figure 2D) levels in the tumor cells themselves. The level of Cav-1 staining was variable between different RSTs and within individual RSTs. Fibroblasts associated with RSTs could not be readily detected and thus the level of Cav-1 in RST associated fibroblasts could not be determined. The levels of Cav-1 were also examined in lung metastases that developed in the MTB-IGFIR transgenic mice. Approximately 30% of MTB-IGFIR transgenic mice develop lung metastases. In general, the lung metastases expressed very low levels of Cav-1 compared to the surround lung tissue (Figure 2E). One mouse had a number of large lung metastases and these metastases had pockets of Cav-1 positive cells (Figure 2F, arrowhead).

The pockets of Cav-1 positive cells observed in the PMT samples appeared to be associated with regions that had undergone some squamous differentiation. These regions were reported previously [23] and were shown to contain cells expressing cytokeratins 5 and 14. To determine whether Cav-1 positive cells also expressed cytokeratin 5 and 14, immunohistochemistry for Cav-1, cytokeratin 5 and cytokeratin 14 was performed on serial sections of two distinct mammary tumors. Although Cav-1 failed to completely co-localize with either cytokeratin 5 or cytokeratin 14, most of the Cav-1 positive cells were found in regions of the tumor were there was an abundance of cytokeratin 5 or cytokeratin 14 positive cells (Figure 3). Little or no Cav-1 was observed in regions of the tumor where the cells were negative for either cytokeratin 5 or 14. This suggests that Cav-1 is expressed at low levels in tumor epithelial cells but at higher levels in regions associated with basal/mesenchymal tumor cells.

Several cell lines have been established from the PMT and RST tumors of the MTB-IGFIR transgenic mice. These lines have recently been characterized [25]. Two of the cell lines (RM11A and RJ348) have characteristics of claudin-low mammary tumors while a cell line, RJ345 has characteristics similar to luminal tumors [25]. 4T1 cells are a commercially available murine mammary epithelial cell line [27] that has previously been shown to have characteristics similar to the RJ345 cells [25]. Western blotting and immunofluorescence for Cav-1 protein and qRT-PCR for Cav-1 mRNA revealed that tumor cells themselves can express Cav-1 and the claudin-low mammary tumor cell lines (RJ348 and RM11A) expressed higher levels of Cav-1 protein and mRNA than the epithelial tumor cell lines (Figure 4 and Table 1).

Most of the current research on Cav-1 in breast cancer has demonstrated that loss of Cav-1 expression in the stromal cells surrounding the tumor is associated with early tumor recurrence and poor clinical outcome [5, 6]. It has been proposed that loss of stromal Cav-1 induces oxidative stress within the fibroblasts which in turn results in an increased secretion of lactate and pyruvate by the stromal fibroblast and/or autophagy in the stromal compartment [13, 14, 28]. In both instances, the stromal cells provide essential fuel for tumor growth. Loss of stromal Cav-1 has also been associated with epithelial to mesenchymal transition (EMT) of mammary epithelial cells. It was shown that conditioned media from Cav-1 null mammary stromal fibroblasts could induce EMT of normal mammary epithelial cells [29]. EMT is a program where epithelial tumor cells acquire genetic and phenotypic characteristics of mesenchymal cells and these cells show enhanced invasive capacity and potentially stem cell characteristics [30, 31].

However, Cav-1 may also directly affect tumor cells as transformation of NIH-3T3 cells with oncogenes such as H-Ras is associated with decreased Cav-1 expression and re-expression of Cav-1 can inhibit anchorage-independent growth of transformed NIH-3T3 cells. This finding, coupled with our observation that Cav-1 was significantly elevated in our DNA microarray comparing claudin-low mammary tumors to our primary, IGF-IR induced epithelial mammary tumors led us to examine the expression of Cav-1 in the tumor cells themselves. Claudin-low breast cancers are a subtype of human breast cancer characterized by low levels of epithelial genes and high levels of mesenchymal genes as well as genes involved in the immune response and those associated with stem cells [32, 33].

In our MTB-IGFIR transgenic model [23, 24], Cav-1 was found to be expressed in adipocytes and smooth muscle cells in the mammary gland but the cells of mammary tumors induced by IGF-IR overexpression expressed considerably less Cav-1. Most of the epithelial cells of the mammary tumors expressed little or no Cav-1 protein. However, there were small pockets of cells within some of the IGF-IR induced mammary tumors. Interestingly, these pockets of Cav-1 cells were in close proximity to cells within the tumor expressing high levels of cytokeratin 5 or cytokeratin 14 which were frequently were associated with ghost cells. In contrast, mammary tumors that develop following IGF-IR downregulation and acquire a claudin-low genotype express moderate to high levels of Cav-1 in the tumor cells themselves. This pattern of Cav-1 expression was maintained in mammary tumor cell lines derived from MTB-IGFIR transgenic tumors with epithelial or claudin-low characteristics. Therefore, Cav-1 levels were higher in claudin-low tumor cells than epithelial tumor cells.

This association of Cav-1 with claudin-low cells supports the findings of Bailey and Liu [34] who observed that induction of EMT in embryonic carcinoma cells or murine mammary epithelial cells was associated with elevated Cav-1 expression. In addition, Orlichenko et al. observed that decreased Cav-1 expression or function resulted in elevated E-cadherin levels in MDCK epithelial cells [35]. Moreover, Savage et al. [36] found that Cav-1 was expressed at high levels in basal-like and metaplastic breast tumors while Prat et al. [33] and Haakensen et al. [37] have observed higher levels of Cav-1 in human claudin-low breast tumors or cell lines.

These findings are all consistent with our findings since we found that IGF-IR-induced mammary tumors and murine mammary tumor cell lines 4T1 and RJ345 that expressed high levels of E-cadherin and low levels of mesenchymal genes also expressed low levels of Cav-1. In contrast, IGF-IR independent mammary tumors and the murine mammary tumor cell lines RJ348 and RM11A that expressed very low levels of E-cadherin and had gene expression patterns similar to claudin-low tumors also expressed considerable amounts of Cav-1. Therefore, our study supports the findings that claudin-low mammary tumors express more Cav-1 than epithelial mammary tumors. Cav-1 may be useful with other markers in identifying/distinguishing claudin-low human breast tumors.

In general, lung metastases expressed low levels of Cav-1 protein while normal lung alveolar cells expressed much higher levels of Cav-1. The alveolar expression of Cav-1 was not unexpected as Cav-1 has been reported to be expressed in type I pneumocytes [38]. This finding suggests that Cav-1 is not important in growth of tumors at secondary sites. Although we have not extensively characterized the lung metastases these lesions do appear to have an epithelial morphology and thus if the tumor cells in the metastases are primarily epithelial we would not expect them to express high levels of Cav-1. A search of the literature did not reveal any papers describing the expression of Cav-1 in breast metastatic lesions.

Our findings suggest that Cav-1 expression is dependent of the subtype of mammary tumor being investigated. It appears that epithelial mammary tumors express low levels of Cav-1 but mammary tumors that acquire claudin-low characteristics express Cav-1. Further investigations are required to determine whether Cav-1 can be used as a marker and/or a therapeutic target for claudin-low breast cancers.