|Home | About | Journals | Submit | Contact Us | Français|
Breast cancers expressing estrogen receptor α (ERα) are often more differentiated histologically than ERα-negative tumors, but the reasons for this difference are poorly understood. One possible explanation is that transcriptional co-factors associated with ERα determine the expression of genes which promote a more differentiated phenotype. In this study, we identify one such cofactor as coactivator associated arginine methyltransferase 1 (CARM1), a unique co-activator of ERα that can simultaneously block cell proliferation and induce differentiation through global regulation of ERα-regulated genes. CARM1 was evidenced as an ERα co-activator in cell-based assays, gene expression microarrays, and mouse xenograft models. In human breast tumors, CARM1 expression positively correlated with ERα levels in ER+ tumors but was inversely correlated with tumor grade. Our findings suggest that co-expression of CARM1 and ERα may provide a better biomarker of well-differentiated breast cancer. Further, our findings define an important functional role of this histone arginine methyltransferase in re-programming ERα-regulated cellular processes, implicating CARM1 as a putative epigenetic target in ER-positive breast cancers.
In normal breast tissue, estrogen receptor α (ERα) regulates growth and development of the mammary gland by regulating the balance between cell proliferation and differentiation. This balance is deregulated in cancer. Enhanced ERα proliferative action contributes to the initiation and progression of breast cancer (1) by promoting cell cycle progression, in particular S phase entry (2, 3). Microarray analyses using breast cancer cell lines have revealed that a majority of ERα target genes are involved in metabolism and cell cycle regulation (4-6). ERα is expressed in nearly 70% of breast cancers. Interestingly, ERα-positive tumors are more histologically well-differentiated (7). ERα decreases in high-grade tumors (8) and the presence of ERα serves as a hallmark of differentiation and predictor of low aggressiveness and favorable disease-free survival (9, 10). The protective effect of ERα raises the possibility that ERα functions to regulate both proliferation and differentiation in breast cancer cells, albeit with the balance tilted towards proliferation. Cell proliferation and differentiation are two mutually exclusive processes. Forced differentiation of primary tumors with therapeutic compounds can inhibit proliferation (11). Differentiation therapy such as all-trans retinoic acid was successfully used in treating acute promyelocytic leukemia. However, this strategy is not widely applied to breast carcinomas because breast tumors are more heterogeneous. Moreover, how ERα regulates the balance of proliferation and differentiation is not well-understood.
ERα regulates transcription through recruitment of multiple co-factors (12). Although ERα coactivators share the common feature of activating ERα in transcriptional assays, to date, no ERα coactivator has been reported to promote differentiation in breast cancer cells. CARM1 was originally identified as a steroid receptor coactivator which activates transcription of ERα target genes (13, 14). Furthermore, loss of CARM1 in the mouse embryo leads to abrogation of the estrogen response and reduced expression of some ER-target genes (15), highlighting the significance of CARM1 in ER regulated processes. CARM1 is a multifunctional protein engaged in a variety of cellular processes including gene expression (16), coupling of transcription and mRNA processing (17), regulating protein stability (18), tissue development (15). However, the function of CARM1 in regulating cell differentiation or proliferation is contradictory and appears to be context dependent. CARM1 is required for differentiation of adipocytes (19), myocytes (20), pulmonary epithelial cells (21) and early thymocyte progenitor cells (22). In contrast, CARM1 was implicated in cancer cell proliferation and was shown to regulate the expression of E2F1 and cyclin E1, factors promoting cell cycle progression (16, 23). Thus, functions of CARM1 in ERα-dependent breast cancer require further elucidation.
Here we report an extensive study of the biological function of CARM1 in ERα-regulated processes in breast cancer cells using both gain-of-function and loss-of-function approaches.
All cell lines were purchased from ATCC and used within 6 months. MCF7-tet-on-shCARM1 was generated by two steps: First we synthesized shRNA-encoding oligo DNA “CGGCGAGATCCAGCGGCAC” targeting human CARM1 and cloned the sequence into pLVTHM plasmid (24). MCF7 cells (ATCC HTB-22) were sequentially infected with the lenti-KRAB and pLVTHM-CARM1shRNA vectors followed by selection of single clones by western blots. MCF7-tet-on-CARM1 was generated by co-transfecting pTRE-tight CARM1 plasmid with pBabe-puromycin vectors at a ratio of 10:1 into MCF7-tet-on cells (Clontech cat#630918, generous gift of Elaine Alarid) followed by selection with puromycin. MCF7-CARM1 has been described previously (25). MDA-MB-231 (ATCC HTB-26) or ZR-75 (ATCC CRL-1500) cells were infected with retrovirus derived from pLNCX (Clontech), pLNCX-CARM1 and pSIREN-Q-shCARM1 vectors to obtain pooled clones. The shRNA-encoding oligo DNA “CAGCGTCCTCATCCAGTTC” targeting human CARM1 was cloned into pSIREN-Q (Clontech) vector. All cells were maintained in DMEM containing 10% FBS and seeded in phenol red free media with 5% of stripped FBS for experiments.
All primary invasive breast tumors used in the present study were obtained from the Manitoba Breast Tumor Bank (MBTB, Cancer Care Manitoba and University of Manitoba [http://www.umanitoba.ca/institutes/manitoba_institute_cell_biology/MBTB/Index4.htm]. The MBTB operates with approval from the Research Ethics Board of the Faculty of Medicine, University of Manitoba. Tissue collection and selection of samples for constructing TMAs has been reported before (26, 27). Although 450 cases were represented on the original ER+ TMAs due to exhaustion of tumor cores from previous use of the TMAs, or incomplete data for some cases, the number (n) of tumors analyzed for some of the markers was less than 450. Immunohistochemistry scores (IHC-scores) derive from assessment of both staining intensity (scale 0, 1, 2, 3) and percentage of positive cancer cells (0-100%). These two scores were multiplied to generate an IHC or H-score with a range of 0-300 as previously described (27). Only nuclear staining was scored. Briefly, serial sections (5 μm) of the TMAs were stained with anti-CARM1 antibodies using an automated tissue immunostainer (Discovery Staining Module, Ventana Medical Systems, AZ, USA). The dilution of the primary CARM1 Ab applied initially to the slides was 1:150, incubated for 1 hr at 42 °C, on the Ventana Discovery Staining module, using CC1 buffer and the Mild and Standard protocols for Antigen Retrieval (AR). The VIEW DAB kit and reagents from Ventana Medical Systems were used. Slides were viewed and scored using standard light microscopy.
Cell growth assays, real-time quantitative PCR, cell cycle profiling, CARM1 antibody characterization, mouse xenograft experiments and microarray gene expression analyses are described in Supplemental Methods and Materials.
One-way ANOVA and ANOVA-single factor analysis was applied and P value <0.05 is regarded as statistically significant.
MCF7 breast cancer cells stably over-expressing CARM1, MCF7-CARM1, were generated (25). Compared to parental cells carrying empty vector (MCF7-vector), MCF7-CARM1 grows at a slower rate as measured by MTT assays (Figure 1A). The altered growth rate was not due to the super-physiological amounts of CARM1 in MCF7-CARM1, because compared to MCF7-vector controls, MCF7-CARM1 cells have only a two-fold increase of CARM1 expression (Figure 1B). Consistent with the growth phenotype, the expression of the CDK inhibitors p21cip1 and p27kip1 was elevated in MCF7-CARM1 treated with E2 (Figure 1B). Also the expression of p21cip1 and p27kip1 was stimulated by E2 in a time-dependent manner in MCF7-CARM1 cells (Supplemental Figure 1) but not in parental MCF7 cells, suggesting that ERα and CARM1 are involved in regulating their expression. The effect of CARM1 on anchorage independent cell growth was determined using soft-agar assays. E2 stimulates colony formation of MCF7-vector cells; in contrast, no colonies were formed in soft agar with MCF7-CARM1 (Figure 1C). This result suggests that overexpressing CARM1 in MCF7 may inhibit anchorage-independent growth. In contrast to MCF7, no growth effects were detected by over-expressing or knocking down CARM1 in MDA-MB-231, an ERα negative breast cancer cell line (Figures 1D and 1E). Consistent with it being ERα-negative, the growth rate of MDA-MB-231 was E2-independent (Figure 1D). Similarly, overexpressing CARM1 exhibits no growth effect on MDA-MB-468, another ERα-negative breast cancer cell line (Supplemental Figure 2), supporting the notion that the growth inhibitory effect of CARM1 in MCF7 is ERα-dependent. The growth inhibitory effect of CARM1 was further validated in another ERα-positive breast cancer cell line ZR-75 (Supplemental Figure 3). p21cip1 has been reported to induce cell cycle arrest as well as to induce cell differentiation in various carcinomas (28, 29). The findings that p21cip1 expression is increased by E2 in the presence of exogenous CARM1 (supplemental Figure 1) raises the possibility that CARM1 may inhibit breast cancer growth by modulating key ERα-target genes involved in cell cycle control and differentiation.
To eliminate the possibility that the growth effects of CARM1 in MCF7-CARM1 cells could be attributed to additional changes during retroviral integration events, we generated two inducible MCF7 stable cell lines: one over-expresses CARM1 (MCF7-tet-on-CARM1) and the other expresses CARM1 shRNA (MCF7-tet-on-shCARM1) under the control of a tetracycline-inducible promoter. These stable cell lines serve as gain-of-function and loss-of-function cell culture models for studying the effects of CARM1 in estrogen-dependent breast cancer growth. Cells were pre-incubated with Dox for four days to induce or knockdown CARM1 expression, followed by E2 treatment for 24 hrs. With either cell line, E2 alone has no significant effect on CARM1 expression at both mRNA and protein level (Figures 2A and 2B). Dox was able to increase CARM1 expression in MCF7-tet-on-CARM1 cells by two-fold (Figure 2A) and reduce CARM1 to >90% in MCF7-tet-on-shCARM1 cells (Figure 2B). E2 has no additional effect on CARM1 expression compared to Dox alone when both are present. The two cell lines were employed to measure cell growth using MTT assays under four treatment conditions: vehicle, E2, Dox, or combination of Dox and E2 for four time points (24, 48, 72, and 96hrs). As expected, E2 treatment significantly increases MCF7 cell growth starting from day 2 (p value <0.001) (Figure 2C). Over-expression of CARM1 by Dox treatment alone decreased MCF7 cell growth (Figure 2C). Statistical analysis of three independent experiments suggested that overexpression of CARM1 by Dox treatment significantly repressed E2-induced cell growth in two individual clones, clone 7 (Figure 2C) and clone 13 (Supplemental Figure 4). This is in contrast to the non-statistically significant effect of Dox upon E2-induced cell growth in MCF7-tet-on-shCARM1 cells (p>0.05) (Figure 2D) and a CARM1 stable knockdown MCF7 (MCF7-shCARM1) cell line expressing shRNA targeting a different sequence of human CARM1 (Supplemental Figure 5, p=0.04).
The main proliferative action of E2 in breast cancer is to promote cell cycle progression during G1/S transition (3). Since CARM1 can induce expression of p21cip1 and p27Kip1, which are negative regulators of the cell cycle, and inhibit E2-dependent growth, we determined whether CARM1 would interfere with E2-induced cell cycle progression. MCF7-tet-on-CARM1 cells were pre-incubated with Dox for 4 days, followed by E2 treatment for 24 hrs. FACS analysis of MCF7-tet-on-CARM1 using propidium iodide labeling showed that E2-induced S phase entry was inhibited by overexpressing CARM1 (Figure 2E). This result was validated by BrdU labeling (Figure 2F). Although E2 and E2+Dox both increased S phase entry as compared to that of the vehicle (p-value <0.001 and 0.0015, respectively), results from three independent experiments showed that the percentage of S phase entry induced by Dox+E2 was significantly decreased compared to E2 treatment alone (p= 0.0013), indicating that overexpression of CARM1 decreased E2 induction of S phase entry. In contrast, in MCF7-tet-on-shCARM1, Dox+E2 treatment displayed no difference in S phase entry compared to E2 alone and both treatment groups induced S phase entry compared to the vehicle treatment (p=0.0014). In either MCF7-tet-on-CARM1 or MCF7-tet-on-shCARM1 cells, Dox alone had no significant effect on S phase entry (Figure 2F). These data suggest that overexpression of CARM1 can inhibit E2-stimulated cell growth through modulating cell cycle, whilst loss of CARM1 could not further accelerate E2-stimulated growth within 4 days of treatment.
In addition to the growth inhibitory effects of CARM1, we noticed that MCF7 cells stably over-expressing CARM1 displayed a distinct cell morphology from that of MCF7-vector cells (Figure 3A) and exhibited increased cell adhesion (requires longer trypsin treatment time). Next we investigated desmoplakin 1 (DSP1) expression, a known differentiation marker of epithelial cells that plays an essential role in maintaining cell adhesion and differentiation (30, 31) and a CARM1 target gene identified in this study. Three independent experiments showed that E2 significantly decreased DSP1 mRNA, which was reversed by overexpressing CARM1 in MCF7-tet-on-CARM1 cells (Figure 3B). In addition, induction of two additional differentiation markers, GATA3 and E-cadherin, by overexpressing CARM1 was observed in MCF7-tet-on-CARM1 (Figure 3C) by western blots. These data suggested that CARM1’s growth inhibitory function may be accompanied by the induction of cell differentiation.
Since CARM1 inhibits E2-dependent growth of MCF7 cells and induces a morphology change, we determined the global effect of CARM1 on E2-dependent ERα gene signature using microarray analyses of CARM1 gain-of-function and loss-of-function cell lines treated with vehicle or E2. MCF7 inducible cells were treated under four conditions: DMSO, Dox, E2, and E2 plus Dox. The gene signature as calculated by fold change was normalized to vehicle control (Figure 4A). Microarray analysis of MCF7-tet-on-CARM1 reveals that E2 up-regulated expression of 313 genes and down-regulated 157 genes (p<0.05, fold change≥1.6). Overexpression of CARM1 drastically altered E2-regulated gene signatures. ~16% of E2-induced genes including cell cycle regulators (e.g. c-myc) (Figure 4B) was inhibited. The most profound effect of CARM1 overexpression on E2 dependent signature was to relieve the repression of ~56% of E2-repressed genes (Figure 4C) (p<0.05, fold change≤0.6 compared to vehicle). To our knowledge, CARM1 is the only coactivator which affects global expression of E2-repressed genes. Interestingly, gene ontology of the affected genes suggested that the majority of E2-repressed, CARM1-activated genes are involved in cell differentiation and development (Figure 4C). The ability of CARM1 to inhibit E2-dependent growth and S phase entry as well as to modulate E2-dependent genes involved in cell cycle progression, cell differentiation and development supports a role of CARM1 in modulating the programming of E2-dependent cellular processes (i.e., regulating the balance between cell differentiation and proliferation).
Since CARM1 has putative effects on ERα-dependent proliferation and differentiation, we applied qRT-PCR to validate the effect of CARM1 overexpression on six differentially expressed genes identified by microarray. p21cip1 and p27kip1 are known to inhibit breast cancer growth (32). Cyclin G2 is an ERα target gene and a negative regulator of cell cycle (33). Among genes involved in cell differentiation, GATA-3 is an ERα target gene and pro-differentiation marker of breast cancer (8, 34). MAZ is a transcriptional factor (35) and KRTAP10.12 is a potential pro-differentiation marker. As shown in Figure 4D, E2 alone significantly decreased cyclin G2 and KRTAP10.12 mRNA but not p21cip, p27kip1, MAZ and GATA-3 mRNA after 4 hour treatment. However, overexpression of CARM1 relieved E2 repression of cyclin G2 and KRTAP10.12, and significantly induced p21cip1, p27kip1, MAZ and GATA-3 regardless of E2 (Figure 4D) at mRNA level. Consistently, the protein levels of GATA-3, E-cadherin (Figure 3C) and p21cip1 and p27kip1 (Fig. 1B) were also increased by CARM1 overexpression and E2 treatment. These results validate our microarray data and reinforce the hypothesis that CARM1 may antagonize the proliferative action of estrogen in breast cancer cells by activating multiple cell cycle negative regulators and pro-differentiation genes. It is worth noting that p21cip1 induction requires both CARM1 overexpression and E2 treatment. In contrast, overexpressing CARM1 alone is sufficient to induce genes such as p27kip1, suggesting that CARM1 may regulate some genes in hormone deprived conditions.
The global effects of CARM1 on ER-target genes were next examined in the loss-of-function model, MCF7-tet-on-shCARM1 under aforementioned conditions. The heat map of the fold-change gene signature relative to vehicle control is shown in Figure 5A and additional description can be found in supplemental methods. Using Agilent array platform herein, CARM1 shRNA expression up-regulated 62 genes and down-regulated 2122 genes (p<0.05, fold change≥1.6 and ≤0.6 compared to vehicle, respectively). E2 treatment up-regulated 780 genes and down-regulated 5099 genes (p<0.05, fold change≥1.6). Interestingly, the genes affected by loss-of-CARM1 largely overlapped with those affected by E2 in wild-type cells. Further microarray analysis showed that 65% of genes activated by knocking-down CARM1 are also activated by E2 (Figure 5B), and 75% of genes repressed by CARM1 knockdown are also repressed by E2 (Figure 5C). Gene ontology of genes affected by Dox and E2 treatment also overlap (Figure 5B, C). Among those genes, a majority are involved in metabolism, development, protein binding and gene expression (Figures 5B, C). These data further support the notions that CARM1 is a global regulator of E2-responsive genes in breast cancer cells and profoundly affects estrogen-mediated processes. We also validated the effect of loss of CARM1 on p21cip1, p27kip1, cyclin G2, MAZ, GATA-3, and KRTAP10.12 mRNA expression. Loss of CARM1 significantly repressed p21cip1, p27kip1, cyclin G2, MAZ, GATA-3, KRTAP10.12 and DSP1 at mRNA levels, similar to E2’s effect in MCF7-tet-on-shCARM1 (Figure 5D and supplemental Figure 6A). In agreement with the mRNA results, cyclin G2, GATA-3, and E-cadherin (Supplemental Figure 6B) were decreased at protein levels with the loss of CARM1. Since both cyclin G2 (33) and GATA-3 (36) are ERα target genes, CARM1 may antagonize E2 action via ERα during re-programming of ERα- dependent differentiation and proliferation processes. Fold changes of key cell cycle regulators and genes involved in cell differentiation in MCF7-tet-on-shCARM1 are listed in Table S1. Overall, our data suggest that loss of CARM1 induces gene signatures resembling those affected by E2, and CARM1 is a regulator of E2-dependent, key cell cycle progression and differentiation genes. Collectively, the microarray analyses using CARM1 gain- and loss-of-function cell models reveal that CARM1 is a unique ER coactivator that profoundly affects the balance of genes involved in cellular differentiation and proliferation (i.e. inhibit cell growth and promote cell differentiation).
To examine the effects of CARM1 in vivo, we transplanted MCF7-tet-on-CARM1shRNA cells in nude mice. The design of the xenograft experiment is shown in Figure 6A, representing one of triplicate experiments. We first validated that the growth of xenografted tumors was E2-dependent because no growth or only tiny tumors developed in the negative control group not receiving estrogen. Tumors collected from mice engrafted with MCF7-tet-on-shCARM1 cells and receiving Dox showed a reduction of CARM1 expression at mRNA and protein levels (Figure 6B). Knocking-down CARM1 increased the size of E2-induced tumors (Figure 6C) and was associated with a modest increase in BrdU labeling. The differential rate of BrdU labeling for xenografted tumors was further increased in mice receiving a higher dose of E2 and that was associated with higher mitotic index (Figures 6D and 6E). All the data suggest that knocking-down CARM1 enhances E2-dependent proliferation of breast cancer cells in vivo. Since CARM1 inhibits E2 dependent growth by modulating negative cell cycle regulators p21cip1, p27kip1, and cyclin G2 and pro-differentiation genes, we examined the relationship between p21cip1 and E-cadherin, a differentiation marker, in E2-induced xenografted tumors. A direct correlation was observed between p21cip1 and E-cadherin expression in tumors derived from xenografts (Figure 6F), suggesting inhibition of cell growth and induction of differentiation are coherent processes in ERα-positive tumors.
Our rabbit polyclonal CARM1 antibody was determined to be specific because it detects both nuclear and cytoplasmic CARM1 in normal breast tissues and breast tumors while exhibiting no activity towards mouse embryonic fibroblasts derived from CARM1 knock-out mice (MEF-/-) (Supplemental Figure 7). CARM1 expression was determined by IHC in ER+ breast tumor tissue microarrays (TMAs) available in the Manitoba Breast Tumor Bank (MBTB) (26, 37). Statistically significant correlations between ERα expression as determined by IHC (n = 310, spearman r = 0.324, P < 0.0001) and tumor grade (n = 328, spearman r = -0.159, P = 0.004) were found. Significantly higher CARM1 expression as determined by IHC score was found in tumors with higher ERα expression compared to those with lower ERα expression (Figure 7A). Significantly higher CARM1 expression was found in lower grade (3, 4) tumors as well (Figures 7B and 7C). In addition, CARM1 expression was positively correlated with ERα levels in ER+, node negative human breast tumors, p <0.0001 (Supplemental Figure 8A). We also found an inverse correlation between CARM1 expression and tumor grade in ER+, node negative human breast tumors, p < 0.0398 (Supplemental Figure 8B). Collectively, the findings from clinical samples support a role of CARM1 in regulating ERα-dependent differentiation in ERα-positive tumors.
In most cases, proliferation and differentiation are inversely coupled: repression of proliferation is a prerequisite for initiation of differentiation (11). In many cell types, however, cell cycle arrest is necessary but not sufficient for differentiation. CARM1 appears to be a unique ER-coactivator regulating both processes. Over-expression of CARM1 in MCF7 cells results in inhibition of E2-dependent growth through inhibition of the G0/G1 transition to S-phase. This is in part due to up-regulation of key negative cell cycle regulators such as p21cip1, p27kip1, and cyclin G2. Inhibition of E2 dependent cell growth by CARM1 is accompanied by morphological changes characteristic of a more differentiated phenotype and induction of multiple differentiation markers such as GATA-3 and MAZ. This finding is supported by previous reports that CARM1 can promote cell differentiation in other systems (19-21). Nonetheless, regulation of cell differentiation by CARM1 appears to be cell-type and context dependent. In mouse embryo and embryonic stem cells, CARM1 was shown to elevate expression of key pluripotency genes and delay their response to differentiation signals (38).
In contrast to growth inhibition by CARM1 overexpression, knocking down CARM1 in MCF7 did not alter E2-dependent cell growth in cell culture nor did it affect E2-induced S phase entry. This observation contradicts the conclusion by Frietze et al. that CARM1 increases growth of MCF7 cells. The discrepancies may be due to the transient transfection of CARM1 siRNA throughout the cell cycle study by Frietze et al. (16). Moreover, the authors measured the percentage of cells in S+G2+M phase without distinguishing the percentage of cells in S phase. Also, in consistent with the observation of O’Brien et al (21), we did not observe change of E2F1 with CARM1 knock down, in contrast to Frietze et al. (16). Interestingly, and in contrast to cells grown in culture, knocking down CARM1 enhanced E2-induced xenograft tumors. This may be due to increased breast cancer cell interaction with the microenvironment which plays essential roles in promoting tumor growth in animals.
The growth inhibitory effect of CARM1 is unique from that of SRCs. Knocking down SRC2 and SRC3 but not SRC1 inhibits growth of MCF7 cells and decreases cyclin D1 expression (39). Overexpression of SRC3 also increases breast cancer cell proliferation and invasiveness. Likewise, SRC-1 promotes breast tumor metastasis and inhibits tumor cell differentiation (40). Thus, the ERα-dependent, growth inhibitory effect of CARM1 is unlikely to be mediated through SRC-1, 2 and -3.
Cell cycle genes that are regulated by E2 or loss of CARM1 include cyclin D1, c-myc, cyclin G2, cyclin L1, cyclin T2, p21cip1, p27kip1, p130 and Rb (Table S1). E2 treatment alone significantly represses cyclin G2 (33), which is reversed by overexpressing CARM1. Cyclin D1 is a well-known E2-induced ERα target gene, however, its expression is not affected by overexpression of CARM1 in the presence of E2, yet knocking down CARM1 upregulates cyclin D1 in MCF7 cells (Table S1). C-myc is upregulated by E2 alone or loss of CARM1 (Table S1) but is not affected by depletion of any of the p160 coactivators in MCF7 cells (39). Thus, the mechanism of CARM1 regulation of cell cycle regulators is complex and only partially depends on the p160 coactivators.
Microarray gene expression analyses reveal that approximately 16% of E2-activated genes were repressed by CARM1, consistent with the repressive effects of CARM1 on some ER-target genes (41). The mechanism of CARM1-mediated repression is unclear. The major effect of CARM1 overexpression was to relieve E2-repressed genes. CARM1 methyltransferase activity may be responsible for the activation since we observed increased H3R17Me2 mark on p21cip1 promoter upon CARM1 induction in MCF7-tet-on-CARM1 (data not shown), consistent with a recent publication that CARM1 is recruited to p21cip1 promoter (42). Whether CARM1 regulates ERα-target genes via an epigenetic mechanism remains to be determined. Nonetheless, global ERα transcriptional regulation by CARM1 leads to induction of many E2 repressed genes associated with differentiation.
Consistent with this finding, knocking-down CARM1 shares over 65% of the E2 gene signature. The majority of CARM1, E2-regulated genes are involved in gene expression, metabolism, cell cycle and differentiation. Knocking down CARM1 leads to up-regulation of positive cell cycle regulators (e.g. c-myc) and down-regulation of negative cell cycle regulators (e.g. cyclin G2). This result suggests that loss of CARM1 function may lead to the acquisition of a proliferative phenotype resembling estrogen stimulation of breast cancer. Further, knocking-down CARM1 also modulates genes involved in cell differentiation. For example, combination of CARM1 shRNA and E2 treatment significantly reduced the level of PPARγ, which induces terminal differentiation of breast cancer (43). Loss of CARM1 also significantly decreases KRTAP10.12, an E2-repressed gene involved in keratin filament formation and potentially in cell differentiation processes (44, 45). Collectively, either loss of CARM1 or E2 treatment significantly inhibits expression of various differentiation markers (Table S1). Overall, the gene expression data from CARM1 gain- and loss-of-function models suggest that CARM1 plays an important role in regulating ERα target genes in differentiation and proliferation.
Evidence for a functional interplay of ERα and CARM1 was explored in human breast cancer specimens. A direct correlation was observed between CARM1 and ERα in ER+ tumors. Higher ERα expression is associated with less aggressive and more differentiated tumors, and ER status is known to inversely correlate with histological grade (46). Our observation contradicts an earlier report that CARM1 is overexpressed in grade III breast tumors (23). The difference could result from analysis of RNA vs. protein and the sample size. In the study by El Messaoudi et al., CARM1 was only analyzed at the RNA level in 81 human breast tumors, while we analyzed CARM1 protein level in over 300 human breast tumors.
Histological grade using the Nottingham method, integrates scores from glandular differentiation, nuclear morphology and mitotic counts (47, 48) and higher grade is significantly associated with poor outcome and survival. The inverse correlation of CARM1 expression and tumor grade found in ER+ breast cancer cases together with enhanced tumor volume in CARM1 knock-down breast cancer xenografts in animal models support an association of low levels of CARM1 with less well differentiated, high grade breast cancers, and is consistent with the hypothesis that CARM1 inhibits breast cancer progression in ERα positive tumors. Our results suggest that co-expression of ERα and CARM1 together may serve as a better biomarker of well differentiated breast cancers.
ERα is believed to regulate growth and differentiation through balanced interaction with cofactors. This study reports an unexpected biological function of the ER-coactivator CARM1 in breast cancer. The hallmark of CARM1 action might be due to global modulation of E2-regulated genes, leading to re-programming of cell proliferation and differentiation. To our knowledge, CARM1 is the only ER coactivator that is able to simultaneously block cell proliferation and induce differentiation. Since CARM1 has histone modification activity, inducing differentiation of breast cancer cells by up-regulating CARM1 activity may be therapeutically effective in breast cancer.
We thank Elaine Alarid for providing MCF7-tet-on cells. We thank Sanghyuk Chung, Amy Cole, Ruth Sullivan, and Yunhong Zan for the technical help, and Sujun Hua (Univ. of Chicago), Flow cytometry and Histology lab facility for assistance. We also thank Lin-feng Chen for comments and Erin Shanle for editing. This work is supported by NCI grant CA125387 and Shaw Scientist Award from Greater Milwaukee Foundation to W.X. and in part by grants to LCM from the Canadian Institutes of Health Research (CIHR) and the Canadian Breast Cancer Research Alliance (CBCRA). KH is supported by Uehara Memorial Foundation. We acknowledge the strong support of the Cancer Care Manitoba Foundation (CCMF) for facilities at MICB.