Epigenetic Upregulation of MAGE-A Isoforms Promotes Breast Cancer Cell Aggressiveness

Simple Summary Breast cancer is a heterogeneous disease that has complex causes and mechanisms of development. Currently, patient treatment options depend on the breast cancer molecular subtype, which is classified based on the presence or absence of hormone receptors and HER2. However, this classification system has limitations in terms of predicting responsiveness to anticancer drugs and patient outcomes. In this study, we present a new approach to classifying molecular breast cancer subtypes: it is based on changes in histone modifications in the promoter region of the MAGEA12 locus, which we found related closely to MAGEA12 expression and MAGEA12-associated malignancy of breast cancer cells. Abstract After decades-long efforts to diagnose and treat breast cancer, the management strategy that has proved most successful to date is molecular-subtype-specific inhibition of the hormone receptors and HER2 that are expressed by individual cancers. Melanoma-associated antigen (MAGE) proteins comprise >40 highly conserved members that contain the MAGE homology domain. They are often overexpressed in multiple cancers and contribute to cancer progression and metastasis. However, it remains unclear whether the biological activity arising from MAGE gene expression is associated with breast cancer subtypes. In this study, we analyzed the RNA-sequencing (RNA-seq) data of 70 breast cancer cell lines and found that MAGEA12 and MAGEA3 were highly expressed in a subset of these lines. Significantly, MAGEA12 and MAGEA3 expression levels were independent of hormone receptor expression levels but were closely associated with markers of active histone modifications. This indicates that overexpression of these genes is attributable to epigenetic deregulation. RNA-seq of MAGEA12-depleted cells was then used to identify 382 candidate targets of MAGEA12 that were downregulated by MAGEA12 depletion. Furthermore, our gain-of-function experiments showed that MAGEA12 overexpression promoted aggressive behaviors of malignant breast cancer cells, including enhancing their cell migration and invasion. These changes were associated with increased epigenetic deregulation of the MAGEA12 signature genes. Thus, MAGEA12 may play an important role in breast cancer malignancy. Taken together, our findings suggest that MAGEA12 could be a promising therapeutic target in breast cancer, and its overexpression and epigenetic changes could serve as subtype classification biomarkers.

MAGEA3 represses apoptosis by regulating BAX, BIM, and p21, and induces cell proliferation [40,41]. MAGEA12 also induces the ubiquitination/degradation of p21, thereby promoting cell-cycle progression and apoptosis [42]. However, the roles that MAGE-A gene products play in breast cancer have not yet been established.
To date, the heterogeneity of breast cancer cells has significantly hampered the development of an optimal treatment for all types of breast cancer. New biomarkers are needed to help predict responsiveness to treatment regardless of which hormone receptors are present. In this study, we assessed whether differential MAGE gene expression could be used to generate a new classification of breast cancer subtypes and whether MAGE family members could serve as new therapeutic targets for breast cancer.

Cell Lines and Cell Culture
The human breast cancer cell lines SKBR3, MCF7, and MDAMB231 were purchased from Korean Cell Line Bank (Seoul, Korea). MDAMB468 was obtained from Sapporo Medical University and tested DNA fingerprinting analysis using STR (short-tandem repeat) markers (Korean Cell Line Bank). MDAMB468, SKBR3, and MCF7 cells were maintained as monolayer cultures in Dulbecco's modified eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. MDAMB231 cells were maintained as monolayer cultures in Roswell Park Memorial Institute (RPMI)-1640 medium supplemented with 10% FBS and 1% penicillin-streptomycin. All cells were grown at 37 • C in a humidified atmosphere containing 5% CO 2 .

Virus Infection for the Stable Cell Line
Cells were seeded in six-well plates and incubated until they reached 50-70% confluence. For infections, the virus (empty vector and MAGEA12-overexpressed) was 1:1 diluted in culture media and added 1 ug/mL polybrene (Sigma-Aldrich, Burlington, VT, USA, TR-1003-G). Next, 24 h later, cells were detached and reseeded in new plates with a final concentration of 2 ug/mL puromycin (Sigma-Aldrich, P9620). Puromycin selection was continued to enrich the population of 80-90%. At this point, cells were diluted and maintained with puromycin.

Western Blot Analysis
Protein was extracted using an RNA/protein extraction kit (MACHEREY-NAGEL) according to the manufacturer's instructions. The proteins in lysates were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on 8-10% gels; after which, proteins were transferred to polyvinylidene fluoride (PVDF) membranes. After blocking in phosphate-buffered saline (PBS) containing 1% skim milk and 1.0% Tween-20 (PBST), the membranes were incubated overnight at 4 • C with primary antibodies against MAGE-A (Santa Cruz, sc-20034) and β-actin (Bethyl Laboratories, Montgomery, TX, USA, A300-491A) that were diluted in PBST. The membranes were then washed with PBST and incubated with horseradish peroxidase-conjugated secondary antibodies (diluted in PBST) at RT for 1 h. Immunoreactive proteins were visualized using chemiluminescent reagents (ATTO, Taito, Tokyo, Japan) and detected using an Amersham Imager 600 (GE Healthcare, Chicago, IL, USA).

Proliferation Assay
For proliferation assays of MDAMB231 and MCF7 cells, 5 × 10 4 cells were resuspended in 2 mL of growth medium and seeded in six-well plates. After 72 h of incubation, the number of cells in each well was counted every 2 days using the trypan blue exclusion method. The viability of MDAMB468 and SKBR3 cells was determined by first, respectively, resuspending 4 × 10 4 and 5 × 10 4 cells in 100 µL of growth medium and then seeding the cells in 96-well plates. The cells were grown for the specified period. The luminescence signals were measured using the CellTiter-Glo reagent (Promega, G9241) as per the manufacturer's protocol.

Migration and Invasion Assay
Transwell migration and invasion assays were performed on 24-well plates containing 8 µm pore polycarbonate membrane inserts (Falcon, Corning, NY, USA, 353097). For invasion assays, the inserts were coated with diluted Matrigel (Corning, Corning, NY, USA, 354230). For all transwell assays, cell suspensions were seeded in the upper chamber and incubated for 1-2 days; after which, the inserts were fixed and stained with crystal violet.

ChIP-qPCR
Chromatin immunoprecipitation (ChIP) assays were performed as described previously [44,45]. In brief, 5 × 10 6 cells were treated with 1% formaldehyde (Sigma, F8775) at RT for 10 min to crosslink DNA, and the reaction was stopped by adding 0.125 M glycine and incubating the mixture at RT for 10 min. The cells were then pelleted by centrifugation, lysed, and subjected to 10 cycles of sonication; after which, sonicated chromatin was immunoprecipitated with 1 µg of antibody and 10 µL of Protein A beads. Crosslinks in the eluate from washed immunocomplexes were reversed by incubating with proteinase K; after which, the immunoprecipitated DNA was purified with a QIAquick PCR purification kit (QIAGEN, Hilden, Germany, 28106) using 50 µL of elution buffer. qRT-PCR was performed on the diluted ChIP DNA using SYBR Green mix and the following primer pairs (5 -3 ): human negative control region, TCC TAT TCA AGT CCT TCC TCC A (forward) and TGC AAA ACA TAT GAA ACA CAA GC (reverse); human EFNA1 promoter region, GGG ACA GGA AGC CAT GAG TA (forward), and GGA GGT GGG TAA GGA AGA GG (reverse).

RNA-Seq Library
Total RNA was extracted using a Direct-zol RNA prep kit (Zymo, Irvine, CA, USA, R2071), and RNA-seq libraries were constructed using the NEXTflex rapid directional mRNA-seq kit (PerkinElmer, Waltham, MA, USA, NOVA-5138-11). In brief, 5-10 µg of purified RNA was poly-A-selected and fragmented with fragmentation enzyme. Following first-and second-strand synthesis from a template of poly-A-selected fragmented RNA, other procedures, from adenylation to PCR amplification, were performed according to the RNA-seq library construction steps provided by the manufacturer.

ChIP-Seq Library Construction
ChIP-seq libraries were constructed using the NEXTflex ChIP-seq kit (PerkinElmer, NOVA-5143-02) according to the manufacturer's instructions. In brief, 40 µL of purified ChIP DNA was end-repaired and size-selected (250-300 bp) using AMPure XP beads. Other procedures, from adenylation to PCR amplification, were performed according to the ChIP-seq library construction steps provided by the manufacturer. The quality of the ChIP-seq library was assessed with a bioanalyzer using a high-sensitivity chip (Agilent), which confirmed that the average size of ChIP-seq libraries ranged from 250 to 350 bp. For multiplexing, equal molar quantities of libraries were combined by considering sequencing depth per sample (20-40 million reads per library). ChIP-seq libraries were sequenced using an Illumina NextSeq platform with single-end reads of 76 bases.

Data Analysis
All raw and analyzed data were accessed in gene expression omnibus (GEO). The accession numbers are series GSE85158 (samples GSM2258722, GSM2258731, GSM2258732, GSM2258794, GSM2258802, GSM2258805, GSM2258848, GSM2258856, GSM2258858, GSM2258884, GSM2258892, and GSM2258894); and series GSE96860 (samples GSM2545229, GSM2545230, GSM2545231, GSM2545232, GSM2545245, GSM2545246, GSM2545247, GSM2545248, GSM2545257, GSM2545258, GSM2545259, GSM2545260, GSM2545265, GSM2545266, GSM2545267, and GSM2545268). The RNA-seq source data of 70 breast cancer cell lines were obtained from specific reference [46]. Data quality was checked using FastQC, and adapter sequences were removed using Cutadapt. Up to this point, ChIP-seq and RNA-seq followed the same process. Thereafter, the ChIP-seq data were mapped to the reference genome using Bowtie2 and normalized tag count and visualized using HOMER. By contrast, the RNA-seq data were mapped to the reference using STAR, and bam files were converted to tdf files using igv tools. The data were loaded and viewed using integrative genomics viewer (IGV) (version 2.4.13).

A Subset of Breast Cancer Cells Expresses Both MAGEA12 and MAGEA3 in a Subtype-Independent Manner
To investigate the expression patterns of the MAGE-A isoforms in breast cancer, we analyzed the publicly available RNA-seq data of 70 breast cancer cell lines ( Figure 1a and Table S1) [46]. This analysis confirmed that in general, each cell line expressed a single MAGE-A isoform, with the most common isoform being MAGEA12. However, a subset of breast cancer cell lines, including MDAMB468 and SKBR3, expressed both MAGEA12 and MAGEA3. Overall, 19 breast cancer cell lines expressed MAGEA12, and 13 of these also expressed MAGEA3. The RNA-seq data also indicated that the most highly expressed MAGE-A isoform in breast cancer cell lines was MAGEA12, followed by MAGEA3 (Figure 1b). We also showed that the expression levels of these two genes in breast cancer cells correlated strongly (Figure 1c). This suggests that the expression of MAGEA12 and MAGEA3 is co-regulated. Significantly, the four breast cancer subtypes did not differ in terms of MAGEA12 and MAGEA3 expression patterns ( Figure 1d). Next, we examined the expression of MAGEA12 and MAGEA3 in normal tissues and various cancer cell lines by using genotype-tissue expression (GTEx) and Cancer Cell Line Encyclopedia (CCLE), respectively. While both MAGEA12 and MAGEA3 were weakly expressed in normal tissues, including the breast, they were generally upregulated to varying degrees in multiple cancers (Figure 1e,f). We also conducted an overall survival analysis of 986 breast cancer patients in the TCGA database whose MAGEA12 and MAGEA3 expression in their tumors had been measured. When the patients were stratified according to whether MAGEA12 or MAGEA3 expression was high or low, the patients with high expression of these genes had a significantly worse prognosis (Figure 1g,h). Thus, our data suggest that the aberrant expression of MAGEA12 and MAGEA3 genes may be useful for classifying and predicting malignant breast cancer phenotypes.  (Figure 1e,f). We also conducted an overall survival analysis of 986 breast cancer patients in the TCGA database whose MAGEA12 and MAGEA3 expression in their tumors had been measured. When the patients were stratified according to whether MAGEA12 or MAGEA3 expression was high or low, the patients with high expression of these genes had a significantly worse prognosis (Figure 1g,h). Thus, our data suggest that the aberrant expression of MAGEA12 and MAGEA3 genes may be useful for classifying and predicting malignant breast cancer phenotypes.  Kaplan-Meier analysis of the overall survival of 986 breast cancer patients who were stratified into two groups depending on whether MAGEA12 (g) and MAGEA3 (h) expression was 'high' (red) and 'low' (blue). The stratification was conducted with an autoselected best cutoff.

Chromatin Modifications at MAGE-A Gene Loci in Breast Cancer Cell Lines
To determine whether the altered expression of the MAGE-A isoforms in breast cancer cells is due to epigenetic changes, we used chromatin immunoprecipitation sequencing (ChIP-seq) to assess the changes in the histone modifications H3K4me3, H3K27ac, and H3K79me2 in different breast cancer cell lines. These histone modifications correspond to actively transcribed genomic regions. We focused on four cell lines that expressed high (MDAMB468; TNBC and SKBR3; HER2+) or low (MDAMB231; TNBC and MCF7; Luminal) levels of both MAGEA12 and MAGEA3. We found massive enrichment of H3K4me3, H3K27ac, and H3K79me2 at the MAGEA12 locus in the MDAMB468 and SKBR3 cell lines compared to the MDAMB231 and MCF7 cell lines (Figure 2a). By contrast, the CETN2 locus close to the MAGEA loci did not accumulate these histone modifications in any of the four cell lines. To determine whether these differences in chromatin modification at the MAGEA12 locus were associated with different levels of MAGEA12 gene expression, we assessed the RNA-seq data relating to the MAGEA12 locus by using the IGV genome browser (Figure 2b). Tag enrichment obtained through RNA-seq was clearly capable of distinguishing between the cell lines that expressed high or low MAGEA12, which is consistent with the accumulation patterns of the histone markers. Quantitative RT-PCR (qRT-PCR) analysis then confirmed that the mRNA expression of MAGEA12 and MAGEA3 was higher in the MDAMB468 cells and, to a lesser extent, in SKBR3 cells, than in the MDAMB231 and MCF7 cells (Figure 2c). Western blotting analysis using an antibody that recognizes both MAGEA12 and MAGEA3 further supported the qRT-PCR results ( Figure  2d). These results indicate that MAGEA12 and potentially MAGEA3 are upregulated at the transcriptional level via epigenetic changes in a subset of breast cancer cell lines.
actively transcribed genomic regions. We focused on four cell lines that expressed high (MDAMB468; TNBC and SKBR3; HER2+) or low (MDAMB231; TNBC and MCF7; Luminal) levels of both MAGEA12 and MAGEA3. We found massive enrichment of H3K4me3, H3K27ac, and H3K79me2 at the MAGEA12 locus in the MDAMB468 and SKBR3 cell lines compared to the MDAMB231 and MCF7 cell lines (Figure 2a). By contrast, the CETN2 locus close to the MAGEA loci did not accumulate these histone modifications in any of the four cell lines. To determine whether these differences in chromatin modification at the MAGEA12 locus were associated with different levels of MAGEA12 gene expression, we assessed the RNA-seq data relating to the MAGEA12 locus by using the IGV genome browser (Figure 2b). Tag enrichment obtained through RNA-seq was clearly capable of distinguishing between the cell lines that expressed high or low MAGEA12, which is consistent with the accumulation patterns of the histone markers. Quantitative RT-PCR (qRT-PCR) analysis then confirmed that the mRNA expression of MAGEA12 and MAGEA3 was higher in the MDAMB468 cells and, to a lesser extent, in SKBR3 cells, than in the MDAMB231 and MCF7 cells (Figure 2c). Western blotting analysis using an antibody that recognizes both MAGEA12 and MAGEA3 further supported the qRT-PCR results ( Figure 2d). These results indicate that MAGEA12 and potentially MAGEA3 are upregulated at the transcriptional level via epigenetic changes in a subset of breast cancer cell lines.

Identification of MAGEA12 Signature Genes in Breast Cancer
To explore the function of MAGEA12 in breast cancer, we used small interfering RNA (siRNA) specific for this gene (siMAGEA12) to suppress the endogenous MAGEA12 expression in the cell lines that expressed MAGEA12 at high levels (i.e., MDAMB468 and SKBR3) and then subjected these cells to RNA-seq ( Figure 3 and Table S2). First, we assessed the effect of the siRNA treatment on the expression of MAGEA12 at 24, 48, and 72 h (Figure 3a). In MDAMB468 cells, MAGEA12 expression decreased by more than 60% at each time point: at 48 and 72 h, expression was reduced by~80-90%. In SKBR3 cells, MAGEA12 expression was reduced by~90% at all time points. The siMAGEA12 treatment had similar effects on MAGEA12 protein levels at 72 h time point. RNA-seq then revealed which genes were downregulated in MDAMB468 and SKBR3 cells 72 h after siMAGEA12 treatment (Figure 3b). In total, 1257 and 870 genes were downregulated in Cancers 2021, 13, 3176 9 of 18 MDAMB468 and SKBR3, respectively; of these, 382 were downregulated in both cell lines and thus constituted the MAGEA12 signature genes (Figure 3c). The expression of all 382 MAGEA12 signature genes gradually decreased in a time-dependent manner after siMAGEA12 treatment (Figure 3d). This indicates that these genes are primary downstream targets of MAGEA12.
Next, to determine the functional relevance of these signature genes, we conducted a gene set enrichment analysis (GSEA) on five MAGEA12-high and five MAGEA12-low cell lines. This showed significant enrichment of the 382 signature genes in the cell lines with high MAGEA12 expression and low expression of these genes in the MAGEA12-low cell lines (Figure 3e). GSEA further indicated that 34 leading-edge genes were among the genes with high enrichment scores. In addition, this analysis confirmed that the expression of these 34 leading-edge genes was increased in most of the MAGEA12-high cell lines, unlike the MAGEA12-low cell lines (Figure 3f). However, in one MAGEA12-high (ZR75-1), the expression of the 34 leading-edge genes was similar to that in the MAGEA12-low cell lines. Thus, in a subset of breast cancer cell lines, the 34 leading-edge genes may be regulated in a MAGEA12-independent manner. The qRT-PCR analysis of MAGEA12depleted MDAMB468 and SKBR3 cells validated the MAGEA12-dependent expression of the aforementioned leading-edge genes, including FA2H, ALPP, C2orf48, DSC2, FBLN1, KCNC4, and EFNA1 (Figure 3g). We also found that the expression of MAGEA3 and MAGEA6 was reduced by MAGEA12 knockdown, suggesting the master regulatory role of MAGEA12. These genes, and other leading-edge genes, were also generally expressed at much lower levels in the MAGEA12-low cell lines than in the MAGEA12-high cell lines (Figure 3h). To confirm that the above results are an on-target effect of MAGEA12 siRNA, we synthesized custom siRNA and successfully reproduced the results in Figure 3g,h ( Figure S1). Taken together, these results suggest that 382 MAGEA12-regulated signature genes participate in breast cancer.

Effect of MAGEA12 Silencing and Overexpression on Breast Cancer Cell Aggressiveness
We then asked whether the putative MAGEA12-regulated genes contributed to the characteristics of breast cancer cells by analyzing the proliferation, invasion, and migration of the MAGEA12-knockdown breast cancer cells (Figure 4). Downregulation of MAGEA12 did not affect the viability of MDAMB468 or SKBR3 cells (Figure 4a). However, suppressing MAGEA12 expression decreased the migratory and invasive capacities of both cell lines, notwithstanding the differences between the lines in terms of migration and invasion (Figure 4b,c). In addition, we showed that MAGEA12 silencing reduced breast cancer aggressiveness by using a 3D culture system that can identify aggressive characteristics based on a cell shape classification [32][33][34]. Thus, siRNA-mediated MAGEA12 knockdown not only dramatically reduced the expression of MAGEA12, but it also changed the morphology of the cell clusters from their usual aggressive grape-like phenotype to the less aggressive rounded/mass cluster phenotype (Figure 4d). Cancers 2021, 13, x FOR PEER REVIEW 10 of 19  This suggests that MAGEA12 affects cell-cell adhesion and thus may be involved in the aggressiveness of breast cancer cells. These phenotypic results are consistent with a gene ontology (GO) analysis of the MAGEA12 signature genes, which showed that the most prominent terms that were associated with these genes were substrate adhesion-dependent cell spreading, regulation of cytoskeleton organization, and cell migration (Figure 4e). Notably, the expression of the cell-migration-associated genes of MDAMB468 and SKBR3 decreased in a time-dependent manner after siMAGEA12 treatment (Figure 4f). These results suggest that MAGEA12 may help regulate the aggressiveness of breast cancer cells.
To confirm that MAGEA12 functions relate to breast cancer aggressiveness, we induced MDAMB231 and MCF7 cells, which do not express MAGEA12, to stably overexpress MAGEA12 ( Figure 5). qRT-PCR and Western blot analyses indicated that these cells showed a marked increase in MAGEA12 expression (Figure 5a). Interestingly, GSEA of the 382 MAGEA12 signature genes in the MAGEA12-overexpressing cells showed that the relative expression of MAGEA12 increased significantly (Figure 5b), which further verifies that the expression of the 382 signature genes is regulated by MAGEA12 expression. As expected, the overexpression of MAGEA12 did not increase the proliferation of MDAMB231 or MCF7 cells (Figure 5c) but did enhance their migratory and invasive abilities ( Figure  5d). We also showed that the MAGEA12-overexpressing MCF7 cell line expressed high levels of MAGEA12 in even 3D cultures, and that this was associated with a change in their morphology. These cells formed grape-like clusters, which is consistent with their potential origin from metastatic tumor cells (Figure 5e). The phenotypic consequences of MAGEA12 overexpression were supported by the upregulation of the leading-edge genes and genes related to cell migration as well as MAGEA3 and MAGEA6 (Figure 5f). Taken together, these results support the notion that increased expression of MAGEA12 may contribute to the aggressiveness of breast cancer cells. This suggests that MAGEA12 affects cell-cell adhesion and thus may be involved in the aggressiveness of breast cancer cells. These phenotypic results are consistent with a gene ontology (GO) analysis of the MAGEA12 signature genes, which showed that the most prominent terms that were associated with these genes were substrate adhesiondependent cell spreading, regulation of cytoskeleton organization, and cell migration (Figure 4e). Notably, the expression of the cell-migration-associated genes of MDAMB468 and SKBR3 decreased in a time-dependent manner after siMAGEA12 treatment ( Figure  4f). These results suggest that MAGEA12 may help regulate the aggressiveness of breast cancer cells.
To confirm that MAGEA12 functions relate to breast cancer aggressiveness, we induced MDAMB231 and MCF7 cells, which do not express MAGEA12, to stably overexpress MAGEA12 ( Figure 5). qRT-PCR and Western blot analyses indicated that these cells showed a marked increase in MAGEA12 expression (Figure 5a). Interestingly, GSEA of the 382 MAGEA12 signature genes in the MAGEA12-overexpressing cells showed that the relative expression of MAGEA12 increased significantly (Figure 5b), which further verifies that the expression of the 382 signature genes is regulated by MAGEA12 expression. As expected, the overexpression of MAGEA12 did not increase the proliferation of MDAMB231 or MCF7 cells (Figure 5c) but did enhance their migratory and invasive abilities ( Figure 5d). We also showed that the MAGEA12-overexpressing MCF7 cell line expressed high levels of MAGEA12 in even 3D cultures, and that this was associated with a change in their morphology. These cells formed grape-like clusters, which is consistent with their potential origin from metastatic tumor cells (Figure 5e). The phenotypic consequences of MAGEA12 overexpression were supported by the upregulation of the leadingedge genes and genes related to cell migration as well as MAGEA3 and MAGEA6 ( Figure  5f). Taken together, these results support the notion that increased expression of MAGEA12 may contribute to the aggressiveness of breast cancer cells.

FOXA1 Is a Candidate Transcription Factor That Regulates MAGEA12 Signature Genes
To determine which transcription factors regulate the expression of the MAGEA12 signature genes, we subjected the promoter regions of the MAGEA12 signature gene loci to motif analysis (Figure 6a). OCT2, FOXL2, OCT11, FOXA3, OCT4, FOXA1, and NF-κB binding motifs were found in the promoters with high significance. To determine whether these transcription factors are expressed in breast cancer cells, we analyzed the RNA-seq data of 70 breast cancer cell lines. This showed that forkhead box A1 (FOXA1) was expressed at much higher levels in these lines than the other candidate transcription factors (Figure 6b). Notably, FOXA1 is expressed specifically in breast cancer cells and is associated with open chromatin and ERα expression [47,48]. We then asked whether FOXA1 regulates the 382 MAGEA12 signature genes by using ChIP-seq for FOXA1. This showed that 255 of the 382 MAGEA12 signature genes contained FOXA1-binding motifs in their promoter region (Figure 6c and Table S3). One of these was the gene for Ephrin A1 (EFNA1), which is involved in adhesion and migration [49] and was one of the 34 leading-edge genes as well as in association with cell migration (Figures 3g, 4f and S1). Notably, our analysis showed that its promoter region contained FOXA1-binding motifs that bore the H3K4me3 and H3K27ac modifications (Figure 6d). In addition, when we divided the 70 breast cancer cell lines into those that had high or low FOXA1 expression, we found that EFNA1 expression

FOXA1 Is a Candidate Transcription Factor That Regulates MAGEA12 Signature Genes
To determine which transcription factors regulate the expression of the MAGEA12 signature genes, we subjected the promoter regions of the MAGEA12 signature gene loci to motif analysis (Figure 6a). OCT2, FOXL2, OCT11, FOXA3, OCT4, FOXA1, and NF-κB binding motifs were found in the promoters with high significance. To determine whether these transcription factors are expressed in breast cancer cells, we analyzed the RNAseq data of 70 breast cancer cell lines. This showed that forkhead box A1 (FOXA1) was expressed at much higher levels in these lines than the other candidate transcription factors (Figure 6b). Notably, FOXA1 is expressed specifically in breast cancer cells and is associated with open chromatin and ERα expression [47,48]. We then asked whether FOXA1 regulates the 382 MAGEA12 signature genes by using ChIP-seq for FOXA1. This showed that 255 of the 382 MAGEA12 signature genes contained FOXA1-binding motifs in their promoter region (Figure 6c and Table S3). One of these was the gene for Ephrin A1 (EFNA1), which is involved in adhesion and migration [49] and was one of the 34 leading-edge genes as well as in association with cell migration (Figure 3g, Figure 4f and Figure S1). Notably, our analysis showed that its promoter region contained FOXA1-binding motifs that bore the H3K4me3 and H3K27ac modifications (Figure 6d). In addition, when we divided the 70 breast cancer cell lines into those that had high or low FOXA1 expression, we found that EFNA1 expression was significantly higher in the lines with high FOXA1 expression ( Figure 6e). These results suggest that the MAGEA12 signature genes may be regulated by FOXA1.
Cancers 2021, 13, x FOR PEER REVIEW 13 of 19 was significantly higher in the lines with high FOXA1 expression (Figure 6e). These results suggest that the MAGEA12 signature genes may be regulated by FOXA1.

Chromatin Modifications in MAGEA12 Signature Genes Parallel MAGEA12 Expression Levels
We next examined whether the level of MAGEA12 expression affected the chromatin modifications in, and the promoter activity of, the 382 MAGEA12 signature genes by using the H3K4me3 ChIP-seq data of the four cell lines that expressed MAGEA12 at high (MDAMB468 and SKBR3) or low (MDAMB231 and MCF7) levels. This analysis showed that 352 (92%) of the MAGEA12 signature genes were enriched for H3K4me3 in the promoter region (Figure 7a and Table S4). Moreover, 58 (16%) of these showed positive changes in the H3K4me3 pattern that corresponded with the MAGEA12 expression levels. Specifically, the promoter regions of these 58 genes had higher levels of the H3K4me3, H3K27ac, and H3K79me2 markers in the cell lines with high MAGEA12 expression compare to the cell lines with low MAGEA12 expression (Figure 7b), exemplified by EFNA1 (Figure 7c). It had higher levels of the active markers H3K4me3, H3K27ac, and H3K79me2 in the MAGEA12-high breast cancer cell lines (Figure 7c). Notably, a visual representation of RNA-seq data showed a higher density of mapped reads at the EFNA1 gene structure in the high MAGEA12-expressing cell lines compared in the MAGEA12-low cell lines (Figure 7c). Next, to confirm that enrichment of H3K4me3 was affected by MAGEA12 expression levels, we treated the high MAGEA12-expressing MDAMB468 cell line with siMAGEA12 and performed H3K4me3 ChIP-qPCR. This showed that the siMAGEA12treated cells had lower levels of H3K4me3 in the EFNA1 promoter region relative to the levels in IPT, a negative control region (Figure 7d). To confirm that these chromatin changes affected EFNA1 gene expression, we performed a qRT-PCR analysis (Figure 7e). This showed that EFNA1 expression was reduced by up to 70% in the MAGEA12-knockdown cells compared with the control cells. Furthermore, transcription of some of the

Chromatin Modifications in MAGEA12 Signature Genes Parallel MAGEA12 Expression Levels
We next examined whether the level of MAGEA12 expression affected the chromatin modifications in, and the promoter activity of, the 382 MAGEA12 signature genes by using the H3K4me3 ChIP-seq data of the four cell lines that expressed MAGEA12 at high (MDAMB468 and SKBR3) or low (MDAMB231 and MCF7) levels. This analysis showed that 352 (92%) of the MAGEA12 signature genes were enriched for H3K4me3 in the promoter region ( Figure 7a and Table S4). Moreover, 58 (16%) of these showed positive changes in the H3K4me3 pattern that corresponded with the MAGEA12 expression levels. Specifically, the promoter regions of these 58 genes had higher levels of the H3K4me3, H3K27ac, and H3K79me2 markers in the cell lines with high MAGEA12 expression compare to the cell lines with low MAGEA12 expression (Figure 7b), exemplified by EFNA1 ( Figure  7c). It had higher levels of the active markers H3K4me3, H3K27ac, and H3K79me2 in the MAGEA12-high breast cancer cell lines (Figure 7c). Notably, a visual representation of RNA-seq data showed a higher density of mapped reads at the EFNA1 gene structure in the high MAGEA12-expressing cell lines compared in the MAGEA12-low cell lines ( Figure  7c). Next, to confirm that enrichment of H3K4me3 was affected by MAGEA12 expression levels, we treated the high MAGEA12-expressing MDAMB468 cell line with siMAGEA12 and performed H3K4me3 ChIP-qPCR. This showed that the siMAGEA12-treated cells had lower levels of H3K4me3 in the EFNA1 promoter region relative to the levels in IPT, a negative control region (Figure 7d). To confirm that these chromatin changes affected EFNA1 gene expression, we performed a qRT-PCR analysis (Figure 7e). This showed that EFNA1 expression was reduced by up to 70% in the MAGEA12-knockdown cells compared with the control cells. Furthermore, transcription of some of the other 34 leading-edge genes was also decreased when MAGEA12 expression was downregulated ( Figure S2).
Taken together, these results suggest that MAGEA12 regulates the expression of its target genes in breast cancer cells by inducing histone modifications. other 34 leading-edge genes was also decreased when MAGEA12 expression was downregulated ( Figure S2). Taken together, these results suggest that MAGEA12 regulates the expression of its target genes in breast cancer cells by inducing histone modifications.

Discussion
Hormone receptor and HER2-based breast cancer therapy can be successful, but its efficacy is limited by the heterogeneity and complexity of breast cancer cells. This has led to the idea that an approach that transcends the hormone receptor concept is needed, particularly one that is based on a compatible biomarker of breast cancer cell aggression and resistance. In this study, we provided evidence that suggests MAGEA12 expression and histone alteration of its locus in the genome are aggressiveness-related markers in breast cancer.

Discussion
Hormone receptor and HER2-based breast cancer therapy can be successful, but its efficacy is limited by the heterogeneity and complexity of breast cancer cells. This has led to the idea that an approach that transcends the hormone receptor concept is needed, particularly one that is based on a compatible biomarker of breast cancer cell aggression and resistance. In this study, we provided evidence that suggests MAGEA12 expression and histone alteration of its locus in the genome are aggressiveness-related markers in breast cancer.
Although various cancers demonstrate MAGE-A family gene overexpression and the accumulation of their encoded proteins [19][20][21][22][23][24][25][26][27][28], the specific MAGE-A gene products that are functionally relevant in breast cancer remain largely undetermined. To address this, we first analyzed the RNA-seq data of 70 breast cancer cell lines to determine the breast-cancer-related expression patterns of the MAGE-A gene family members. While some of the cell lines showed little or no expression of any MAGE-A family genes, most expressed one or two of the MAGE-A family genes. MAGEA12 and MAGEA3 were the predominant isoforms that were expressed by breast cancer cells; their expression levels were also high compared with those of other genes belonging to the MAGE-A family. Notably, previous studies have reported that the expression of some MAGE-A family genes is co-regulated in various cancers [31,32]. For example, MAGEA3 and MAGEA6 are both associated with colorectal and lung cancers, and MAGEA6 and MAGEA11 are coexpressed in prostate cancer. Consistent with this, we found that MAGEA12 and MAGEA3 expression was strongly correlated in breast cancer. Significantly, we also observed that breast cancer patients with high MAGEA12 and MAGEA3 expression had a poor prognosis. Therefore, we propose that monitoring the expression of both MAGEA12 and MAGEA3 may help predict the prognosis and malignancy of breast cancer better than the expression of other members of the MAGE-A family genes.
While the expression of MAGE-A family genes is inhibited by DNA hypermethylation in somatic tissues, it has been reported that in cancer, the promoters of these genes become demethylated and their expression increases, thus promoting the growth of the cancer cells. For example, in prostate, ovarian, and colon cancer cells, the MAGEA1 and MAGEA11 promoters are hypomethylated, and their expression is increased [50,51]. However, other studies have also shown that promoter demethylation is not sufficient to increase the expression of the MAGE-A genes. Rather, it has been proposed that MAGE-A gene expression is associated with histone activation, as evidenced by the upregulation of the MAGE-A genes after treatment with a histone deacetylase inhibitor [52]. We confirmed in the present study that the MAGEA12 promoter in cell lines with high, but not low, MAGEA12 expression is enriched for H3K4me3, H3K27ac, and H3K79me2 markers. This suggests that MAGEA12 expression is regulated by chromatin changes. Moreover, the fact that histone modifications at the MAGEA12 locus correlated strongly with MAGEA12 expression levels suggests that both the expression levels of MAGEA12 and the unique characteristics of the histone markers at its locus have potential as biomarkers for classifying breast cancer cells.
It has been reported that MAGE-A gene products are involved in ubiquitination, proliferation, and apoptosis [31,41,[53][54][55]. The relationship between MAGE-A proteins and ubiquitination is well known; for example, MAGEA3 and MAGEA6 form an E3 ubiquitin ligase complex with TRIM28, which participates in the survival of cancer cells by degrading AMPKa1 [31,55]. In addition, MAGE-A proteins can bind directly to p53, thereby regulating the targets of p53 and, ultimately, the cell-cycle progression and apoptosis of cancer cells [41,54]. However, the relevant functions of MAGEA12 in breast cancer remain unclear. In the present study, we detected 382 MAGEA12 signature genes through transcriptome analyses and showed that these signature genes are associated with the malignancy and aggressiveness of breast cancer cells. We also found that MAGEA12 levels correlated with changes in breast cancer cell motility and invasion in monolayer cultures as well as the formation of cancer stem cell-like tumorspheres under 3D culture conditions. In addition, promoter analysis of the MAGEA12 signature genes showed that the transcription factor FOXA1 may regulate the expression of many of the MAGEA12 signature genes; thus, we demonstrated that FOXA1 was strongly expressed in breast cancer cell lines and an investigation of FOXA1 binding at the whole-genome level using ChIP-seq showed that FOXA1 occupied the promoters of many MAGEA12 signature genes. It has been reported that FOXA1 cooperates with estrogen receptor-α (ESRa) to regulate chromatin accessibility in breast cancer [56][57][58]. However, no link between FOXA1 and MAGEA12 has yet been reported. Our findings suggest that MAGEA12 may collaborate with FOXA1 to regulate the aggressiveness of breast cancer cells.
Finally, we propose that MAGEA12 plays a role in maintaining chromatin activation. In breast cancer cell lines that had high levels of MAGEA12 expression, the promoters of the MAGEA12 signature genes were enriched for H3K4me3. In particular, we showed that siRNA-mediated MAGEA12 knockdown reduced H3K4me3 levels at the promoter of one of the MAGEA12 signature genes, namely, EFNA1. Given that diminished H3K4me3 levels are associated with reduced expression of EFNA1, we suggest that MAGEA12 acts through an, as of yet, unknown mechanism to activate chromatin and thereby regulate the transcription of genes involved in breast cancer cell malignancy.

Conclusions
In conclusion, we found that MAGEA12 is associated significantly with aggressiveness in breast cancer regardless of the hormone receptor subtype status. Moreover, we observed that the MAGEA12-regulated signature genes are involved in breast cancer cell migration and invasion and that the regulation of MAGEA12 expression could play an important role in determining the shape of aggressive breast cancer cells. In addition, we showed that MAGEA12 could regulate the expression of signature genes via chromatin modifications. These results suggest that the overexpression of MAGEA12 may contribute to the metastasis of breast cancer cells and that histone modifications that are regulated by MAGEA12 could be potential markers of breast cancer aggressiveness.