Morphologic characterization of the DCIS cell lines ETCC-006 and ETCC-010, compared to MCF10A, MCF7, MDA-MB-231 and MCF10DCIS.com
The primary goal of our work was to further characterize the recently developed ETCC-006 and ETCC-010 DCIS cell lines [14]. Before comparing the molecular details of those cell lines, we first examined overall cell morphology by comparing ETCC-006 and ETCC-010 to another DCIS cell line, MCF10DCIS.com, a normal-like breast epithelial cell line, MCF10A, and two breast cancer cell lines, MCF7 and MDA-MB-231.
Interestingly the two ETCC DCIS cell lines showed evident morphologic differences with each other, despite being isogenic and derived from the same tumor; yet, they displayed some striking similarities with the other cell lines examined. We studied their morphologic differences while growing the cells in a culture dish (Fig. 1A) and with cells grown on coverslips, followed by fixation and staining with hematoxylin and eosin (H & E staining) (Fig. 1B). Cells were also grown onto coverslips and fixed, prior to β-actin indirect immunofluorescence and DAPI staining of cell nuclei (Fig. 2). β-actin is a highly conserved and expressed protein, implicated in the control of cell growth and migration [30]. It is important for the maintenance of the cellular architecture; therefore, we chose it as an additional marker to visualize differences in cell morphology. In Figs. 1 and 2, differences in size, shape and organization can be noticed in the panel of cells examined.
Similar to MCF-7, MCF10DCIS.com cells displayed high cellular contact among cells, with cells forming clusters with each other. This suggests that those cells are highly dependent on cellular contacts for growth. Although cell size was not quantified, observation under the microscope suggests that MCF10DCIS.com cells are of similar size as MCF-7 cells.
Considering the ETCC lines, ETCC-006 cells seem to show the largest cell size of all cells analyzed in this study, with many cytoplasmic projections observed. Moreover, ETCC-006 cells were phenotypically striking because of their apparent flatness when visualized under the microscope; thus, observation under bright field microscopy proved to be more difficult than for other cell lines (Fig. 1A).
ETCC-010 cells were derived from the same DCIS tumor as ETCC-006, yet they show little morphologic similarities to each other. Instead, ETCC-010 cells displayed a high degree of resemblance to MDA-MB-231 cells (Figs. 1B and 2). ETCC-010 cells were elongated, with the nucleus occupying most of the available space within the cell. The difference in morphology between ETCC-006 and ETCC-010 is interesting since they both have the same genetic background, highlighting tumor heterogeneity within DCIS lesions [31].
Molecular subtype of DCIS cell lines
In breast cancer care, testing of the ER, PR and HER2 receptor status is routinely done to assess the prognosis and the treatment that will be given to the patient [1]. In breast cancer research, the ER, PR and HER2 receptor status of cell lines is verified to determine which breast cancer subtype those cell lines represent. Before analyzing the ER, PR and HER2 receptor status of the DCIS cell lines, we confirmed the published molecular subtype of the control cell lines by immunoblot analysis. MCF-10A and MDA-MB-231 are known to have no expression of ER, PR and HER2. MCF-7 is a hormone receptor-positive breast cancer cell line, so it expresses ER and PR and modest levels HER2 [4] Those results were confirmed at both protein and RNA levels (Fig. 3).
MCF10DCIS.com has the same genetic background as MCF-10AT, which was derived from MCF-10A cells through oncogenic transformation with an activated HRAS gene [8, 32]; therefore, it is not surprising that similar to MCF-10A, this cell line shows little expression of ER at the protein and RNA level. PR was detected at the protein level (Fig. 3A and B) but not at the RNA level (Fig. 3C). Since levels of ER and PR in MCF10DCIS.com are low, those receptors are likely not functionally relevant. High levels of HER2 were detected in MCF10DCIS.com, explained by the presence of two gene copies of ERBB2 [33, 34].
While ETCC-006 showed no expression of ER, PR and HER2, we detected very low levels of ER and no expression of PR and HER2 at protein and RNA level in ETCC-010 cells. These results contradict the previous work by Yong et al. (2014) [14]; however, our work shows directly that HER2 is not expressed in these two DCIS cell lines and agrees with previous work showing that treatment with Herceptin (monoclonal antibody directed against HER2) was not cytotoxic to ETCC cell lines [14]. Taken together, our work indicates that ETCC-006 and ETCC-010 are functionally similar, based on receptor status, to triple-negative breast cancer cell lines.
Analysis of the gene expression of Oncotype DX DCIS markers in DCIS cell lines
Since we found that ETCC-006 and ETCC-010 display different characteristics to what was published by Yong et al. (2014) [14], we decided to characterize the cell lines using previously suggested DCIS markers. However, defining DCIS using a small set of known markers has been difficult [35], mainly because medical research on DCIS is more limited compared to that of IDC.
The Oncotype DX DCIS score is a 12-gene panel that generates predicted 10-year risk of local DCIS and invasive recurrence following treatment by breast conserving surgery [36]. It includes five gene markers of proliferation – MKI67 (Ki67), STK15 (AURKA), BIRC5 (survivin), CCNB1 (cyclin B1), MYBL2 (MYB proto-oncogene like 2); PGR (progesterone receptor), GSTM1 (glutathione S-transferase M1) and five reference genes. While it is not relevant to calculate a DCIS score for cell lines, we decided to assess the expression levels of those markers in the DCIS cell lines, ETCC-006 and ETCC-010, compared to the other cell lines in our panel.
Supplemental Figure 1 shows the results of qRT-PCR to analyse the expression of MKI67, STK15, BIRC5, CCNB1, MYBL2, GSTM1 and PGR. MCF-7 cells show high expression of the proliferation markers (Panel A) along with high expression of GSTM1 and PGR (Panel B). MDA-MB-231 cells show high expression of four proliferation markers, very low expression of PGR and reduced GSTM1 expression compared to MCF-7 cells. Regarding MCF10DCIS.com, we observed lower expression of the proliferation markers, when compared to the breast cancer cell lines, and low GSTM1 and PGR expression. It is important to note that while PGR RNA levels appeared low as determined by qRT-PCR; PR protein was detectable by immunoblotting (Fig. 3). Meanwhile, ETCC-006 showed high expression of four proliferative markers and low expression of PGR and GSTM1. ETCC-010 displayed a similar trend to MCF10DCIS.com cells, with comparable levels of MKI67, STK15 and BIRC5 expression. However, ETCC-010 showed slightly higher expression of GSTM1 and lower expression of PGR, compared to ETCC-006. Analysis of this small panel of genes suggests that differences in gene expression observed between the ETCC DCIS lines likely reflects the cellular heterogeneity of the original tumor.
Genome-wide comparison of the ETCC-006 and ETCC-010 DCIS model to MCF-10A
In parallel to analyzing the molecular subtype of our cells, we performed RNA sequencing (RNAseq) of ETCC-006 and ETCC-010 cell lines. RNAseq allows the identification of transcripts present in the ETCC-006 and ETCC-010 cells at a genome-wide level. By comparing the level of expression of transcripts in ETCC-006 and ETCC-010 to those in MCF-10A cells, we aimed to identify if genes in some cancer-related pathways showed altered expression. Before performing RNAseq, we analyzed the quality of the total RNA extracted from our cell lines using bioanalyzer traces (Agilent) to calculate an RNA integrity (RIN) value that is indicative of the level of degradation of the RNA. RIN values range from 0 (degraded) to 10 (intact). We obtained RIN values of 10 for RNA extracted from ETCC-006 and ETCC-010, each with two replicates. This indicated that the RNA was of good quality and could be used for library preparation.
After RNA sequencing, reads were trimmed and filtered for quality with Trimmomatic [21] and aligned to the human reference genome (hg19) with RNA STAR [22]. Gene expression was quantified with FeatureCounts [23] and differential expression was performed using DESeq2 [24]. We compared the transcriptome of the ETCC lines to that of MCF-10A, using an available dataset, as detailed in the Methods. After the differential expression, we excluded genes with a p-value over 0.05. Figure 4A is a Venn diagram of the number of genes with higher expression in ETCC-006 and ETCC-010 in orange and lower expression in ETCC-006 and ETCC-010 in blue. In total, 235 and 246 genes were over-expressed in ETCC-006 and ETCC-010, with 99 genes in common; while 194 and 240 genes were under-expressed in ETCC-006 and ETCC-010, with 91 genes in common (Fig. 4A). All genes identified from our RNAseq analysis are listed in Additional file 1.
Next, reactome pathway-based analyses were performed using an R/Bioconductor package, ReactomePA [25], using the lists of over-expressed genes in the ETCC DCIS cell lines to identify enriched pathways (Fig. 4B and D). The most enriched pathways common to ETCC-006 and ETCC-010 related to cell proliferation: mitotic G1-G1/S phases, G1/S transition, unwinding of DNA (S phase), etc. Interestingly, integrin-cell surface interaction gene products were enriched in both cell lines compared to MCF-10A. Since those regulate cell motility, it possibly indicates that ETCC-006 and ETCC-010 have migratory abilities. ReactomePA also allowed us to generate String analyses of the genes implicated in the pathways mentioned (Fig. 4C and E). Proliferation in both cell lines seems to rely mostly on commonly upregulated genes (CCND1, MCM family, MYBL2, etc.). CCND1 is an important regulator of the G1/S transition [37], the MCM proteins constitute a family of protein involved in the regulation of DNA replication [38], and MYBL2 is a transcription factor involved in the regulation of the G1/S transition [39]. Interestingly, an enrichment of “signalling by VEGF” was observed in ETCC-006 (Fig. 5B), and VEGF (vascular endothelial growth factor) signalling has been shown to be upregulated in DCIS [40]. Overall, most enriched pathways in ETCC-006 and ETCC-010 are related to DNA replication and proliferation.
Gene ontology analyses of ETCC-006 and ETCC-010 compared to MCF10A, MCF10DCIS.com, MCF7 and MDA-MB-231
To expand our comparison of ETCC DCIS RNAseq data, we decided to analyze the ETCC-006 and ETCC-010 datasets to RNAseq data from MCF10A, MCF10DCIS.com, MCF7 and MDA-MB-231 cell lines from Klijn et al. (2015) [18]. This dataset is particularly useful due to its inclusion of the MCF10DCIS.com cell line, as this cell line is not present in other breast cancer cell line RNAseq datasets [41]. For this analysis, RNAseq data was aligned to the latest human genome reference sequence, GRCh38, and gene ontology analysis was performed using GOseq [28]. GOseq is an application specifically designed for performing gene ontology (GO) analysis on RNAseq data and has the advantage of eliminating biased results due to overrepresentation of long and highly expressed transcripts [28]. In Supplemental Figure 2, panels for ETCC-006 and ETCC-010 are shown at the top; while panels for MCF10DCIS.com, MCF7 and MDA-MB-231 cells are below. All cell lines had distinct GO profiles, with ETCC-006 showing enrichment of gene sets involved in extracellular organization and ETCC-010 with ion transport. For ETCC-010, we found the link to ion transport intriguing, as DCIS in patients is often diagnosed by the presence of microcalcifications in radiographs [42]. Similar to our previous observations, these two isogenic DCIS cell lines are distinct from each other and different to the previously characterised MCF10DCIS.com line [9]. Meanwhile from our analysis of RNAseq data from MCF7 cells, gene sets involved in transcription and RNA regulation were enriched; whereas, MDA-MB-231 cells displayed differential expression of genes linked to metabolism (Supplemental Figure 2).
Although our pathway analysis and GO enrichment of gene sets are distinct, we did find overlap in ETCC-006 cells between the COL4A1/A2 (collagen type IV) genes within the receptor tyrosine kinase STRING analysis (Fig. 4C) and the GO analysis (Supplementary Figure 2), in which extracellular organization genes were enriched. Since type IV collagens are major components of the basement membrane [43] and their overexpression is linked to cancers of the brain, stomach and liver [44,45,46], expression and/or modulation of COL4A1/A2 genes might be part of the molecular changes associated with DCIS.
Monitoring of cell proliferation and survival in DCIS cell lines, ETCC-006 and ETCC-010
To confirm the observation from the pathway enrichment analysis that proliferation is promoted in the ETCC lines, cell growth was monitored to determine population doubling time (PDT) using trypan blue staining. Figure 5A shows proliferation curves for each of the cell lines examined. MCF-10A, MCF-7 and MDA-MB-231 showed different rates of proliferation with PDT of 20 h, 24.4 h and 36.2 h respectively. The rate of proliferation of MCF10DCIS.com was comparable to that of MCF-10A, with 20.5 h PDT. ETCC-006 and ETCC-010 showed intermediate proliferative rates to that of MCF-7 and MDA-MB-231, with 30.7 h and 29.7 h respectively. Interestingly, while most cell lines seemed to undergo a lag phase after plating, the ETCC-010 growth profile displayed high proliferative abilities even at low cell numbers (Fig. 5A).
To extend our characterization of the ETCC cell lines, we used the clonogenic assay to monitor cell survival by testing the cells’ ability to go through cell division with no limit in time [47, 48]. The clonogenic assay reflects cell survival in vitro by measuring the staining intensity of colonies formed with crystal violet. Following growth and replenishing cultures with fresh media, cells were fixed after 10 days and colonies were stained (Fig. 5B). Compared to MCF7 cells, MDA-MB-231 and ETCC-006 cells showed reduced colony formation, which was quantitated by crystal violet staining intensity (Fig. 5C). For ETCC-006, the reduction in signal intensity might be explained by the characteristic flatness of the cells that makes them unsuitable for crystal violet staining. Interestingly, while MCF-7 and MDA-MB-231 showed distinct colonies, all DCIS cell lines appeared more like a monolayer, with an observable reduction in colony formation particularly for the ETCC DCIS lines (Fig. 5C).
Assessment of migration and anchorage-independent growth potential of the DCIS cell lines
Since the integrin cell surface interactions pathway was enriched in ETCC-006 and ETCC-010 (Fig. 4B and D) and the GO analysis indicated gene set enrichment in extracellular structure organization (Supplemental Figure 2), we monitored the migratory abilities of the DCIS cell lines using a wound healing assay. Wound width and wound area was examined after removal of the culture insert at time zero hours (T0), 5 h after insert removal (T5), and then at nine and 12 h post-insert removal (T9 and T12). Using bright-field microscopy, we observed wound closure within 12 h in MCF-10A, MCF-7 and ETCC-006 cells (Fig. 6A - C). In ETCC-010, after 12 h the wound was close to closure. MCF10DCIS.com and MDA-MB-231 showed slower wound closure, a wound measurement over still over 100 μm after 12 h.
Anchorage-independent growth, a property of most malignant cells to grow independently of a solid surface, was evaluated by a soft agar colony formation assay [49]. As expected, breast cancer cell lines, MCF-7 and MDA-MB-231, showed the highest ability to form colonies in soft agar. The three DCIS cell lines MCF10DCIS.com, ETCC-006 and ETCC-010 showed similar abilities to form colonies in agar, but with numbers intermediate to MCF-10A (lowest) and MCF-7 or MDA-MB231 (highest) (Fig. 6D and E).
ETCC-006 and ETCC-010 DCIS cell lines express markers of epithelial-mesenchymal transition
Taken together, our data has shown that the DCIS cell lines ETCC-006 and ETCC-010 show enrichment in migration pathways (Fig. 4), extracellular organization genes (Supplemental Figure 2) and have migration abilities (Fig. 6). Next we sought to consider whether those cell lines have undergone epithelial to mesenchymal transition (EMT). EMT is a common event in cancer progression where epithelial cells lose their epithelial phenotype, characterized by tight and adherent cells, to gain a mesenchymal phenotype (Fig. 7A). Moreover, during EMT, cells lose expression of epithelial markers, such as E-cadherin, to gain expression of mesenchymal markers (vimentin, N-cadherin, etc.). Along with cell motility [50], activation of EMT programs contribute to the events leading to metastasis [50,51,52].
To examine the level of expression of key EMT markers, whole cell lysates from each cell line were prepared. Vimentin, E-cadherin, N-cadherin, β-catenin and β1-integrin expression was analyzed by immunoblotting. High expression of vimentin was observed in ETCC-006 and ETCC-010 at similar levels to those observed in MDA-MB-231; while low expression of vimentin was detected in MCF10DCIS.com at level comparable to MCF-10A (Fig. 7B and C).
Most strikingly, while MCF10DCIS.com cells expressed E-cadherin, we could not detect expression of E-cadherin in ETCC-006 and ETCC-010. Instead, expression of N-cadherin was observed in ETCC-006 and ETCC-010 lysates. β-catenin is a transcription factor induced by the WNT pathway that stimulates transcription of genes involved in the EMT [53], and its expression is associated with poor prognosis in breast cancer [54]. In our analysis, β-catenin protein levels were elevated in MCF-10A, MCF10DCIS.com and ETCC-006, while levels were more balanced in MCF-7, MDA-MB-231 and ETCC-010 cells. β1-integrin is a transmembrane protein, whose function is to bind extracellular matrix (ECM) [55] and regulate a cell’s ability to invade [56]. β1-integrin was detected at high levels in MCF-10A, MDA-MB-231 and ETCC-006, with ETCC-010, MCF10DCIS.com and MCF-7 showing lower protein levels.
Overall, the co-expression of β-catenin, vimentin, β1 integrin and N-cadherin suggests that ETCC-006 and ETCC-010 have undergone an EMT transition. Although MCF10DCIS.com cells express some markers of EMT, overall it is more representative of cells with an epithelial phenotype.
Signalling profile and cell cycle proteins in DCIS cell lines
Since signalling by receptor tyrosine kinases was most enhanced in ETCC-006 cells according to our RNAseq data and ReactomePA analysis (Fig. 4), and ETCC-006 and ETCC-010 have undergone EMT (Fig. 7), we next decided to investigate levels of some of proteins in major cell signalling pathways. We measured protein levels of two receptor tyrosine kinases, Epidermal Growth Factor Receptor (EGFR) and Insulin-like Growth Factor 1 Receptor (IGF-1R), both critical in the processes leading to metastasis in breast cancer [57, 58]. To do this, cell lysates were prepared using cells grown in normal conditions. Total levels of IGF-1R, EGFR and the active phosphorylated-EGFR were examined by immunoblotting (Fig. 8A and B). Low protein levels of EGFR were detected in MCF-7, with the other cell lines showing similar levels of EGFR. In MCF-10A cells, phosphorylation of EGFR was detected; whereas, p-EGFR was not detected in other cell lines under normal conditions. IGF-1R showed high levels in MCF-7 and low levels in MDA-MB-231 cells. The DCIS cell lines showed similar levels of IGF-1R, with intermediate levels to those observed in MDA-MB-231 and MCF-7 cells.
Common downstream targets of EGFR and IGF-1R include the Ras/Raf/MEK/ERK and PI3K/ AKT pathways, known to stimulate proliferation, migration and EMT when activated after being phosphorylated [53]. While total levels of ERK were similar across DCIS cell lines (Fig. 8C and D, normalized to GAPDH protein), phosphorylated ERK (p-ERK) levels seem slightly higher in MDA-MB-231, MCF10DCIS.com and ETCC-006. Similarly, total AKT levels were similar across cell lines; however, phosphorylated AKT (p-AKT) showed high disparity (normalized to β-actin). Low p-AKT was observed in MCF-7 and MDA-MB-231 cells. Higher levels of p-AKT were observed in MCF-10A, ETCC-006 and ETCC-010, while MCF10DCIS.com showed consistently high levels of p-AKT.
Next, we aimed to investigate whether AKT is constitutively activated in MCF10DCIS.com cells. To do this, we compared the phosphorylation levels of AKT under normal and serum starvation conditions (Fig. 8E, F and G). In MCF7, MDA-MB-231, ETCC-006 and ETCC-010 cells, we found that p-AKT is lost after serum starvation conditions. We confirmed the constitutive activation of AKT in MCF10DCIS.com cells, as p-AKT levels remain unaffected under serum starvation conditions. This was also observed in MCF10A cells (Fig. 8E) and could reflect variability in PI3K levels [59].
Since high levels of cyclin D1 are associated with DCIS [60], we examined the levels of cyclin B1 and cyclin D1, two regulatory proteins in the cell cycle (Fig. 8H and I). While the cyclin B1 level seemed slightly lower in the DCIS cell lines compared to the control cell lines, cyclin D1 showed high expression in MCF-7 and ETCC-006, in agreement with our RNAseq results (Fig. 4).