Aberrant promoter methylation may be responsible for the control of CD146 (MCAM) gene expression during breast cancer progression*

The CD146 (also known as MCAM, MUC-18, Mel-CAM) was initially reported on in 1987, as a protein crucial for melanoma invasion. Recently, it has been confirmed that CD146 is involved in progression and poor overall sur- vival of many other cancers, including breast cancer. Im-portantly, in independent studies, CD146 was reported to be a trigger of epithelial to mesenchymal transition in breast cancer cells. The goal of our current study was to verify possible involvement of an epigenetic mechanism behind regulation of the CD146 expression in breast cancer cells, as it has been previously reported for prostate cancer. First, we analysed the response of breast cancer cells, varying in the initial CD146 mRNA and protein content, to an epigenetic modifier, 5-aza-2-deoxy- cytidine, and subsequently the methylation status of CD146 gene promoter was investigated, using direct bisulfite sequencing. We observed that treatment with a demethylating agent led to induction of CD146 expres- sion in all analysed breast cancer cell lines, both at the mRNA and protein levels, which was accompanied by an elevated expression of selected mesenchymal markers. Importantly, CD146 gene promoter analysis showed aberrant CpG island methylation in 2 out of 3 studied breast cancer cells lines, indicating epigenetic regulation of the CD146 gene expression. In conclusion, our study revealed for the first time that aberrant methylation may be involved in expression control of CD146, a very po-tent EMT inducer in breast cancer cells. Altogether, the data obtained may provide basis for novel therapies, as well as diagnostic approaches enabling sensitive and very accurate detection of breast cancer cells.


INTRODUCTION
Death of breast cancer patients is mainly caused by metastasis, transforming the locally confined disease into a disseminated and usually incurable one (Felipe Lima et al., 2016). Metastasis can be defined as spreading of cancerous cells into distinct organs, followed by formation of a secondary tumor site. At the molecular level, this process is accompanied by actin cytoskeleton rearrangement and attenuation of cell-cell and cell-extracellular matrix adhesive interactions, which altogether resembles the Epithelial to Mesenchymal Transition (EMT), a morphogenetic process observed during development (Brabletz et al., 2018;Nieto et al., 2016). EMT is driven by mesenchymal transcription factors, including Snail, Slug, Zeb, and Twist, which are responsible for altering the epithelial transcriptional profile into mesenchymal one (Nieto et al., 2016;Felipe Lima et al., 2016). Breast cancer cells undergoing EMT become highly motile and invasive, which is especially apparent in the most aggressive estrogen-negative, progesterone-negative, human epidermal growth factor receptor-negative subtype (ER−/PR−/HER2−), defined as a triple-negative breast cancer (TNBC) (Khaled & Bidet, 2019;Felipe Lima et al., 2016). TNBC diagnosed patients have a relatively poor prognosis and cannot be subjected to endocrine therapy or therapies directed against human epidermal growth factor receptor type 2 (HER2) (Schneider et al., 2008). Interestingly, recent publication and meta-analysis, covering a high number of solid tumors, revealed a significant association of CD146 protein expression, EMT and poor survival of cancer patients (Zeng et al., 2017). Importantly, independent studies reported that CD146 is highly expressed in TNBC and in metastatic breast cancer, which is contrary to normal tissue and benign tumors (de Kruijff et al., 2018;Garcia et al., 2007;Jang et al., 2015;Zabouo et al., 2009). Moreover, preclinical in vitro and in vivo studies revealed that aberrantly overexpressed CD146 is sufficient to induce acquisition of mesenchymal phenotype in breast cancer cells (Zeng et al., 2012;Imbert et al., 2012). Nevertheless, although overexpressed CD146 is considered an important oncogene in breast carcinogenesis, there is still lack of information about potential regulation of the CD146 gene expression in breast cancer cells. Despite the fact that CD146 is overexpressed in cancer cells, its amplification or mutation has been excluded so far (Wang & Yan, 2013). In the study presented here, by using breast cancer cell lines as a model, we revealed that CD146 gene promoter is aberrantly methylated in breast cancer cells and the Vol. 66, No 4/2019 619-625 demethylating agent, 5-aza-2-deoxycytidine, can trigger its expression. According to our knowledge, this is the first study suggesting a role of epigenetics behind the CD146 expression in breast cancer cells. Noteworthy, this finding not only sheds new light on regulation of CD146 expression during breast carcinogenesis, but also provides an important rationale for novel therapeutic strategies in future. If confirmed with primary samples, it may contribute to a significant improvement of the diagnostic process, allowing early and very sensitive detection of breast cancer cells.

R2
Database and statistical analysis of publically available data. The R2 database (http://r2.amc.nl) is a simple to use website-related tool for analysis and data visualization created at the Department of Oncogenomics in the Academic Medical Center (AMC) in Amsterdam, Netherlands. It gives an opportunity to perform different analyses based on well annotated datasets. In case of the study presented here, we selected tool "Correlate Genes" in the panel of 10 breast cancer data sets in order to check correlation between CD146 and the panel of mesenchymal markers. Bonferroni corrected p-value of the Spearman correlation coefficient was used to show that the panel of mesenchymal marker encoding genes significantly correlated with CD146 in breast cancer patients. Since the 9 mesenchymal markers were compared with CD146, p value below 0.0055 was considered as significant (0.05/9). The ten breast cancer data sets chosen for analysis included two sets (GSE7396 and GSE46563) composed of only lymph node negative patients and eight data sets (GSE1456, GSE12276, GSE2109, GSE3494, GSE102484, GSE29271, GSE69031, GSE36771) in which lymph node negativity was not the criterion for patient selection. Since the number of genes correlated with CD146 was lower in data sets with confirmed lymph node negativity in comparison to other data sets, we applied cluster analysis (nearest neighbor, Euclidean distance) algorithm to verify if these two data set are indeed different from all the others (Statistica, TIBCO Software Inc.). For the purpose of analysis, the non-significant Spearman correlation coefficient was set as zero.
RNA isolation and cDNA synthesis. Total RNA was isolated from harvested cells using Gene Matrix Universal RNA kit (EURx, Gdansk, Poland). Na-noDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) was used to determine concentration and quality of the isolated RNA. Total RNA (1 µg) was used for cDNA synthesis according to the manufacturer's protocol (EURx, Gdansk, Poland).
Reverse transcription polymerase chain reaction (RT-PCR). For PCR reaction, the Color OptiTaq PCR Master Mix (2x) (EURx, Gdansk, Poland) was used according to the manufacturer's protocol. Sequences of reverse and forward primers used in this study are presented in Table 1. For CD146, HPRT1, SNAI1, TWIST, ZEB1 and SNAI2, the PCR reactions were carried out as previously described (Dudzik et al., 2019). For CAD-HERIN1 and CADHERIN2, the conditions were as following: initial denaturation at 95°C for 5 min; followed by 30 cycles: 30 s at 95°C, 30 s at 58°C and 30 s at 72°C; and final extension at 72°C for 10 min. For MMP2 and MMP9 the following conditions were applied: initial denaturation at 94°C for 3 min; followed by 30 cycles: 45 s at 94°C, 60 s at 59°C and 60 s at 72°C; and final extension at 72°C for 5 min. PCR product was visualized on 1.5% agarose gel stained with ethidium bromide and photographed with Bio-Rad ChemiDoc™ XRS+ System (Bio-Rad, Hercules, CA, USA). HPRT1 was used as internal control to ensure equal sample loading. In case of MDA-MB-231 cells, the RT-PCR analysis of CD146 was performed according to a modified, adjusted PCR protocol in which 25 cycles were used instead of 30 cycles, whereas all other conditions remained unchanged.
The CD146 and HPRT1 primers were manufactured by IBB PAN (Warsaw, Poland). All other primers were purchased from Sigma-Aldrich (St. Louis, MO, USA). The images of the gels were captured using Bio-Rad ChemiDoc™ XRS+ System (Bio-Rad, Hercules, CA, USA) and subsequently analyzed by means of publicly available ImageJ software. All values were normalized to the HPRT1 signal.

The relation between CD146 and mesenchymal profile in breast cancer cell lines
In spite of a number of studies confirming the relationship between CD146 and tumor invasiveness, poor prognosis or mesenchymal features of breast cancer cells, a recent work by de Kruijff et al. questioned the existence of EMT-CD146 association in a group of analysed breast cancer patients (de Kruijff et al., 2018). Although these authors indeed reported CD146 as a prognostic factor for metastasis free and overall survival in a univariable analysis, they did not find a correlation between CD146 and mesenchymal markers at the level of mRNA expression (de Kruijff et al., 2018). In order to perform an independent verification, we analysed the relation between CD146 expression and the panel of mesenchymal markers in 10 independent transcriptomic data sets of breast cancer patients. The patients' data sets selected for this analysis included 2 data sets containing breast cancer patients with confirmed lymph node negativity (lymph node negative data sets) and 8 data sets in which patients with lymph node negative and positive status were combined (non-lymph node negative data sets). As shown in Table 2, at least 6 out of 9 mesenchymal markers were significantly correlated with CD146 in non-lymph node negative data sets, whereas in lymph node negative datasets we have found only 2 mesenchymal markers correlated with CD146 in the GSE7396 data set, and only one (vimentin) correlated with CD146 in the GSE46563 data set. In fact, vimentin was the only gene correlated Table 2. Correlation analysis between CD146 and nine selected mesenchymal markers in 10 independent transcriptomic data sets of breast cancer patients. White rectangle indicates lack of significance, grey rectangle indicates a significant Spearman correlation coefficient below 0.5, black rectangle indicates a significant Spearman correlation coefficient above 0.5. Bonferroni corrected p-value below 0.0055 was considered as statistically significant.
with CD146 in all analysed data sets, regardless of the lymph node status. In order to verify if two lymph node negative data sets are different from the others, cluster analysis was performed. This approach further confirmed that two lymph node negative data sets clustered together as being different from all the others. Thus, the results of our analysis clearly demonstrated, based on transcriptomic data, the existence of association between CD146 expression, mesenchymal profile and the progression of breast cancer.

The effect of 5-aza-2-deoxycytidine on expression of CD146 in breast cancer cell lines
In order to check whether CD146 expression is regulated at the transcriptional level by an epigenetic modifier, 5-aza-2-deoxycytidine, as it has been shown previously in prostate cancer cell lines (Dudzik et al., 2019), two epithelial breast cancer cell lines (MCF7 and T47D) (Dai et al., 2017) and a mesenchymal one (MDA-MB-231) (Dai et al., 2017) were cultured with and without 5-aza-2-deoxycytidine (10 μM) for 6 days, as previously described, followed by CD146 expression analysis at the mRNA and protein level. As expected, basal expression of CD146 was low in the MCF7 and T47D cancer cell lines with epithelial characteristics, and high in the MDA-MB-231 cell line with mesenchymal expression profile. As shown in Fig. 1, CD146 expression was apparently induced in the MCF7 and T47D cells, whereas in MDA-MB-231 its high basal expression disabled correct assessment of expression difference between the control and 5-aza-2-deoxycytidine treated cells. Thus, in order to verify if high CD146 expression in MDA-MB-231 is still further inducible, we performed an expression analysis according to an adjusted, modified PCR protocol, as described in the Materials and Methods section. Interestingly, in MD-MB-231, the CD146 expression was also significantly induced, regardless of the high basal level and the mesenchymal characteristics of these cells (Fig. 1B). As for the protein analysis (Fig. 1C), the demethylating compound had significantly induced CD146 expression in all analysed cell lines. This effect was clearly visible in the epithelial lines with very low basal expression of CD146, as well as in the MDA-MB-231 cells, where the relatively high basal expression was still apparently inducible, in accordance with the data obtained at the mRNA level. Altogether, this data suggest that epigenetic silencing plays an important role in the control of CD146 expression in breast cancer cells with epithelial characteristics (MCF7 and T47D), whereas in mesenchymal breast cancer cell line (MDA-MB-231) expression of CD146 is apparently not fully unleashed and still partially trapped by the epigenetic mechanism.

The effect of 5-aza-2-deoxycytidine on induction of the mesenchymal profile
To determine if CD146 gene expression increase is associated with a mesenchymal profile induction, we analysed the influence of an epigenetic modifier, 5 aza-2-deoxycytidine (10 µM), on expression panel of mesenchymal markers: Slug (SNAI2), Twist1 (TWIST1), Zeb1 (ZEB1), Snail (SNAI1), N-cadherin (CADHERIN 2), Vimentin (VIM), matrix metalloproteinase-2 (MMP2) and matrix metalloproteinase-9 (MMP9), and an epithelial marker E-cadherin (CADHERIN 1). The mesenchymal markers subjected to this analysis were selected based on the expression correlation studies performed in breast cancer patients ( Table 2). As shown in Fig. 2A, treatment of breast cancer cell lines with the demethylating agent resulted in gene expression changes at the mRNA level for several mesenchymal markers, such as SANI1,  . As for the epithelial marker, CADHERIN 1, the basal mRNA expression was high in cell lines with epithelial characteristics and low in MDA-MB-231. Of note, in all three independent experiments, we observed an increase of CADHERIN 1 expression at the mRNA level in the MDA-MB-231 cells, regardless of the induction of mesenchymal marker expression (SNAI1 and TWIST1) in these cells, which are the well known transcription factors inhibiting the CADHERIN 1 gene promoter. Concerning protein analysis, E-cadherin was not visibly induced in MDA-MB-231, opposite to N-cadherin, whose expression was reproducibly enhanced after treatment with 5-aza-2-deoxycytidine, in two independent experiments. As for vimentin, it was apparently expressed only in the MDA-MB-231 cell line and not further induced by treatment with 5-aza-2-deoxycytidine. Undoubtedly, however, the fact that an epigenetic mechanism appears to control expression of such an important oncogene in breast cancer seems to be of high importance. To further determine if 5-aza-2-deoxycytidine, resulting in CD146 induction, may as well trigger the changes in morphology of epithelial breast cancer cells and/or induce the protein expression changes at the level of a single cell, we performed an immunofluorescence analysis of E-cadherin and Vimentin in the MCF7 and T47D cells cultured for 6 days with and without 5-aza-2-deoxycytidine. As determined in Fig. 3, we neither observed an apparent changes in cell morphology, nor, in accordance with the Western blot results, an alteration in the E-cadherin and Vimentin cell content.

The analysis of CD146 gene promoter methylation in breast cancer cell lines
Since presence of the CpG island in CD146 gene promoter has been already described and its proper location reported by our group (Kocemba et al., 2016), we subjected the CD146 promoter region, including the CpG island encompassing the transcriptional start site in exon 1, to methylation analysis using direct bisulfitesequencing (BS) of PCR products. DNA isolated from human fibroblasts and in vitro  methylated DNA was modified and subsequently used in BS sequencing as unmethylated and methylated control, respectively, to validate our experimental set-up. BS analysis revealed that two (MCF7 and MDA-MB-231) out of the three breast (Fig. 4) cancer cell lines tested (MCF7, T47D and MDA-MB-231) displayed hypermethylation of CpG island in the CD146 promoter area, whereas T47D was methylation-free. Notably, BS analysis revealed heterogeneous methylation in MCF7 and MDA-MB-231, suggesting clonal variation in the methylation pattern of the CpG island area in these cell lines. Importantly, aberrant methylation detected in the CpG island of the CD146 gene (Fig. 4) was present in the promoter area encompassing exon one, the region reported previously as crucial for epigenetic silencing of gene expression (Brenet et al., 2011).

DISCUSSION
To our knowledge, this is the first study showing methylation of the CD146 promoter region in breast cancer cells, and concomitantly suggesting that an epigenetic mechanism may be important in expression control of this relevant metastasis related-oncogene. In our previous paper, using prostate cancer cell lines, we have shown that the CD146 gene is induced by a demethylating compound, 5 aza-2-deoxycytidine, however, analysis of the CpG island region in those cells did not confirm presence of promoter methylation (Dudzik et al., 2019). Of note, the link between epigenetics and expression of the CD146 gene was reported previously in prostate cancer patients but the authors of that study incorrectly localized the CpG island in the CD146 gene promoter (Liu et al., 2008), and in consequence misinterpreted the results suggesting, contrary to the dogma, that presence of the CpG island methylation correlates with increase in the CD146 gene expression in prostate cancer cells (Liu et al., 2008). This mistake was pointed out in a Letter to the Editor published in The Prostate in 2016 (Kocemba et al., 2016), also determining proper localization of the CpG island in the CD146 gene promoter (Kocemba et al., 2016). So far, in literature, the relation between DNA methylation in the promoter area (regardless of presence of a CpG island) and gene expression inactivation has been well confirmed (Han et al., 2011;Smith & Meissner, 2013). It has been verified that canonical promoter methylation is associated with expression of imprinted genes, the process of X chromosome inactivation and tissue specific gene regulation (Smith & Meissner, 2013;Urbano et al., 2019). On the other hand, CpG islands observed around transcription start sites in 50% of gene promoters are methylation free in non-cancerous cells, regardless of the expression state of the gene of interest (Jones, 2012). However, de novo hypermethylation of canonical CpG islands has been observed in cancer cells in association with inactivation of gene expression (Jones & Baylin, 2007). Although we observed the CpG island methylation of the CD146 promoter in 2 out of 3 analysed breast cancer cell lines, it does not have to be the only and unique methylated region which controls CD146 expression, since in the T47D breast cancer cell line and three prostate cancer cell lines published previously (Dudzik et al., 2019), the CD146 expression was induced by demethylating treatment, while the analysed CpG island was methylation free. Importantly, recent studies revealed that methylation in the upstream and downstream CpG island shores may inhibit gene transcription in cancer cells, whereas the CpG island itself can remain methylation free (Rao et al., 2013;Irizarry et al., 2009). Even more, in a recent publication, Skvortsova et al. revealed that methylation present in the CpG island shores, particularly in the breast cancer cells, is involved in transcriptional silencing of gene expression. These data clearly confirm that promoter methylation in silenced genes in cancer does not have to be confined only to the CpG island (Skvortsova et al., 2019). To verify exactly which promoter part must be demethylated to unleash the CD146 expression would be definitely of high importance. Whereas overexpressed CD146 is considered to be oncogenic in breast cancer cells, as reported by independent studies (Zeng et al., 2012;de Kruijff et al., 2018;Garcia et al., 2007;Zabouo et al., 2009;Imbert et al., 2012), the methylation-based silencing of its expression in these cells seems to be quite surprising. According to our hypothesis, at the beginning of the disease CD146 may be targeted for aberrant promoter methylation/or it has been already methylated in consequence of tissue specific epigenetic silencing, whereas the loss of methylation in advanced tumour unleashes CD146 expression, leading to metastasis. The fact that CD146-dependent induction of a mesenchymal profile has been already described in the literature (Imbert et al., 2012;Zeng et al., 2012;Zabouo et al., 2009) allows us to speculate that increase in expression of mesenchymal related genes may result from CD146 induction, triggered by an epigenetic modifier. On the other hand, we cannot exclude that the epigenetic modifier, in parallel to the increase in CD146 expression, directly induces mesenchymal profile in the breast cancer cell lines, as the EMT induction in consequence of 5-aza-2-deoxycytidine application has been already reported for breast cancer cells (Su et al., 2018). Definitely, more research is needed to elucidate the independent role of CD146, expression triggered by an epigenetic modifier, in EMT induction in breast cancer cells. Interestingly, in a mesenchymal cell line (MDA-MB-231), we observed an increase in the epithelial marker, E-cadherin, at the mRNA level, which can be the direct consequence of demethylating agents, since epigenetic silencing of E-cadherin expression by aberrant promoter methylation has been reported previously in cancer cells of mesenchymal characteristics and what is more, exactly in MDA-MB-231. Thus, it can be concluded that our results for CADHERIN 1 in MDA-MB-231 simply indicate a proper experimental setup for the demethylating treatment. It is also important to mention that Imbert et al., after overexpression of CD146 in the MCF7 cell line, reported lack of alteration in TWIST1, SNAI1, MMP2 and MMP9 expression and an increase in CADHERIN 2 and VIM (Imbert et al., 2012), whereas changes in the mesenchymal markers' profile were completely different in our current study, which is most probably the consequence of combined action of CD146 and the demethylating agent. Undoubtedly, however, showing that expression of CD146 is controlled by an epigenetic mechanism seems to be the most important message from our study. Moreover, our analysis also suggests coexistence of cells with methylated and unmethylated CD146 gene promoter, which indicates that alterations in the CD146 gene promoter methylation may reflect the process of clonal selection during breast cancer progression. In this context, the aspect of epigenetic control of CD146 expression takes on a particular significance in research centred on application of 5-aza-2-deoxycytidine in treatment of breast cancer patients, and should be considered in planning therapy combined with epigenetic modifiers. It is especially important, since this kind of therapy for breast cancer patients has been currently investigated (Connolly et al., 2017). Our data undoubtedly requires further verification to potentially reveal the scenario of methylation changes in the CD146 gene promoter during development of breast cancer progression, as we cannot exclude that CD146 acts as a tumour suppressor at the initial stage of carcinogenesis, as suggested by the study of Shih and others (Shih et al., 1997), and turn into an oncogene in the advance stage, when the transcriptional profile of cancer cells is significantly altered. Overall, our study provides strong basis for further research on epigenetic regulation of CD146, which can significantly contribute to novel therapies and/or development of DNA methylation-based assay for sensitive detection of breast cancer cells.