ELOVL2: a novel tumor suppressor attenuating tamoxifen resistance in breast cancer

To comprehensively understand the molecular mechanism of tamoxifen resistance (TamR) acquisition by epigenetically regulated genes, it is essential to identify pivotal genes by genome-wide methylation analysis and verify their function in xenograft animal model and cancer patients. The MCF-7/TamR breast cancer cell line was developed and a genome-wide methylation array was performed. The methylation and expression of ELOVL2 was validated in cultured cells, xenografted tumor tissue, and breast cancer patients by methylation-specic PCR, qRT-PCR, Western blot analysis, and immunohistochemistry. Deregulation of ELOVL2 and THEM4 was achieved using siRNA or generating stable transfectants. Tam sensitivity, cell growth, and apoptosis were monitored by colorimetric and colony formation assay and ow cytometric analysis. Pathway analysis was performed to generate networks for the differentially methylated genes in the MCF-7/TamR cells and for the differentially expressed genes in the ELOVL2-overexpressing cells. The human epithelial breast cancer cell line MCF-7 was obtained from the American Type Culture Collection (ATCC, VA, USA) and was cultured under a humidied condition at 37 °C, 5% CO 2 in RPMI 1640 medium (Gibco BRL, Carlsbad, CA, USA) containing 10% fetal bovine serum (Capricorn, Germany) and 2% penicillin/streptomycin (Capricorn). All cells were used within 12 passages after resuscitation of stocks. The MCF-7/TamR cells were generated by culturing MCF-7 cells in the presence of 4-hydroxytamoxifen (Tam) (Sigma-Aldrich, St. Louis, MO, USA) in complete RPMI 1640 medium. The cells were continuously exposed to increasing concentrations of Tam up to 160 nM over a period of 22 weeks, during which the medium was changed twice a week. 50 Effect of THEM4 on recovery of Tam sensitivity. THEM4 is downregulated via a siRNA in MCF-7/TamR (f) and ELOVL2-overexpressing MCF-7/TamR cells (g). Sensitivity to Tam is examined by colony formation assay. Representative images from three independent assays are shown. h Effect of THEM4 on cell proliferation is examined by a dye-based CCK assay. i Effect of THEM4 on TamR is examined by exposing the cells to Tam after downregulating the gene with siRNA. All the assays are performed in triplicates, and the result is depicted as mean ± SE.


Abstract Background
To comprehensively understand the molecular mechanism of tamoxifen resistance (TamR) acquisition by epigenetically regulated genes, it is essential to identify pivotal genes by genome-wide methylation analysis and verify their function in xenograft animal model and cancer patients.

Methods
The MCF-7/TamR breast cancer cell line was developed and a genome-wide methylation array was performed. The methylation and expression of ELOVL2 was validated in cultured cells, xenografted tumor tissue, and breast cancer patients by methylation-speci c PCR, qRT-PCR, Western blot analysis, and immunohistochemistry. Deregulation of ELOVL2 and THEM4 was achieved using siRNA or generating stable transfectants. Tam sensitivity, cell growth, and apoptosis were monitored by colorimetric and colony formation assay and ow cytometric analysis. Pathway analysis was performed to generate networks for the differentially methylated genes in the MCF-7/TamR cells and for the differentially expressed genes in the ELOVL2-overexpressing cells.

Results
Genome-wide methylation analysis in the MCF-7/TamR cells identi ed elongation of very-long chain fatty acid protein 2 (ELOVL2) to be signi cantly hypermethylated and downregulated, which was further veri ed in the tumor tissues from TamR breast cancer patients (n = 28) compared with those from Tamsensitive (TamS) patients (n = 33) (P < 0.001). Immunohistochemical analysis of tissues from cancer patients showed lower expression of ELOVL2 in the TamR than TamS tissues. Growth of the MCF-7/TamR cells overexpressing ELOVL2 was retarded in cell culture and also in xenograft tumor tissue.
Strikingly, ELOVL2 attenuated resistance to Tam up to 70% judged by the colorimetric and colony formation assay and xenograft mouse model. ELOVL2 contributed to the recovery of Tam sensitivity by regulating a group of genes in the AKT and ERα signaling pathways, e.g., THEM4, which plays crucial roles in drug resistance.

Conclusions
ELOVL2 was hypermethylated and downregulated in TamR breast cancer patients compared with TamS patients. ELOVL2 is responsible for the recovery of Tam sensitivity. AKT-and ERα-hubbed networks are pivotal in ELOVL2 signaling, where THEM4 contributes to the relaying ELOVL2 signaling. This study implies that deregulation of a gene in fatty acid metabolism can lead to drug resistance, giving insight into the development of a new therapeutic strategy for drug-resistant breast cancer.

Background
Tamoxifen is a non-steroidal antagonist of the estrogen receptor. It has been the rst choice for adjuvant therapy in estrogen receptor-positive breast cancer as it reduces cancer recurrence and the annual mortality rate [1]. Despite the obvious bene ts, 40% of breast cancer patient show cancer recurrence 5-10 years after initial therapy, which is one of the major setbacks for the clinicians [2]. This is partly because of the complexity of the signaling pathways that in uence estrogen-mediated regulation in breast cancer [3,4]. Thus, identifying the key molecular markers and elucidating the molecular mechanism of drug resistance are pivotal for offering appropriate treatment options to cancer patients. During the course of tamoxifen-resistance (TamR) acquisition, cancer cells undergo cellular as well as molecular changes. A key change in these cells is increased proliferation and decreased apoptosis via BAX and BCL2 regulation [5]. In addition, the TamR cells show greater stemness phenotype by overexpressing Nanog, Oct3/4, and Sox2 [6].
Complex factors/pathways are attributed to TamR cells, including the activation of estrogen receptor (ER) signaling, up-regulation of growth factors (HER2, EGFR, FGFR, and IGF1R), alterations in RTK, a crosstalk among them, and consequently, the deregulation of the PI3K/AKT/mTOR pathway [7]. Cyclin D1/CDK4/6 complex is a target of the PI3K/AKT/mTOR pathway and has also been shown to crosstalk with the ER signaling pathway [8]. Previous studies have shown different Tam targets and their dysregulation from ER in TamR cancer, e.g., androgen receptor [9], Hedgehog signaling pathway [10], and non-coding RNAs [11], suggesting that the mechanism of TamR is far more complicated than just the modulation of ERassociated activity.
Therapeutic strategies for treating TamR cancer are under development, mainly targeting RTK pathways and PI3K/AKT/mTOR axis [12]. Further, cell cycle proteins can be targeted, for example, by using CDK4/6 inhibitors in combination with Tam [13]. Other approaches include targeting AKT pathway [14] and MYC [15] that are highly expressed in TamR cancer cells. However, clinical improvement has only been modest for these approaches till date; this may be because the affected pathways differ between patients. Thus, establishing accurate prognostic markers would hold the key to effective therapy for TamR breast cancer.
ELOVL2 is a member of the mammalian microsomal ELOVL fatty acid enzyme family, involved in the elongation of very long-chain fatty acids required for various cellular functions in mammals [16]. A study using Elovl2 −/− mice rea rmed the importance of ELOVL2 for the elongation activity in rodents [17]. A prime characteristic of the gene is that the CpG near the gene exhibits consistent age-related changes in various tissues [18]. These strong associations have led to the development of a predictor that can accurately estimate the chronological age based on the methylation levels at the speci c CpG site [19].
However, the relationship of ELOVL2 with cancer occurrence or development is unknown. Furthermore, no role of ELOVL2 in cancer drug resistance has been elucidated. In this study, ELOVL2 was identi ed to be downregulated by hypermethylation in TamR breast cancer. The involvement of ELOVL2 in the recovery of Tam sensitivity was suggested by presenting experimental evidence via in vitro as well as in vivo xenograft animal models. It was also suggested that ELOVL2 is a novel tumor suppressor by showing its lower expression in cancer and its inhibitory effect on cancer cell growth. Finally, the molecular mechanism of how ELOVL2 overcomes TamR was elucidated by revealing the signaling pathway.

Materials And Methods
Cell culture and establishment of tamoxifen-resistant MCF-7 (MCF-7/TamR) cells The human epithelial breast cancer cell line MCF-7 was obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and was cultured under a humidi ed condition at 37 °C, 5% CO 2 in RPMI 1640 medium (Gibco BRL, Carlsbad, CA, USA) containing 10% fetal bovine serum (Capricorn, Germany) and 2% penicillin/streptomycin (Capricorn). All cells were used within 12 passages after resuscitation of stocks. The MCF-7/TamR cells were generated by culturing MCF-7 cells in the presence of 4-hydroxytamoxifen (Tam) (Sigma-Aldrich, St. Louis, MO, USA) in complete RPMI 1640 medium. The cells were continuously exposed to increasing concentrations of Tam up to 160 nM over a period of 22 weeks, during which the medium was changed twice a week.

Study subjects
Solid tissues and slide-mounted formalin-xed para n-embedded (FFPE) tissue sections of tumor samples were obtained from patients who underwent surgery between 2012 and 2013 at the National Cancer Center (NCC) in Korea. TamS tissues were obtained from patients who showed a clinical response to Tam, i.e., no tumor recurrence (n = 33). TamR tissues were obtained from patients who subsequently developed TamR (de ned as disease recurrence while administering Tam; n = 28). Clinical details are presented in Table S1. All patients provided written informed consent to donate the removed tissues to NCC in Korea, and samples were obtained according to the protocols approved by the Research Ethics Board of NCC.

Generation of stable cell lines
Lentiviral particles with control clones and human ELOVL2 ORF clones containing C-terminal mGFP tag were purchased from OriGene (Rockville, MD, USA). MCF-7 and MCF-7/TamR cells were seeded at a density of 5 × 10 3 cells/well in a 96-well plate 1 day before transduction. The next day, the cells were infected with lentivirus for 4 h in the presence of 8 µg/mL polybrene (Sigma-Aldrich), and then the medium was replaced with a fresh complete medium. After 72 h, the cells were selected using 1 µg/mL puromycin (Thermo Fisher Scienti c, Waltham, MA, USA) for 10 days.
Cell Transfection siRNAs against ELOVL2 and THEM4 were purchased from Bioneer (Daejeon, Korea), and an ELOVL2overexpressing vector was developed using the pEZ-MT02 plasmid vector (GeneCopoeia, Rockville, MD, USA) by CosmoGenetech (Seoul, Korea). All siRNAs were diluted in Opti-MEM Medium (Gibco BRL) with Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA, USA), and the mixture was incubated for 5 min. The cells were transiently transfected at nal concentrations of 20 nM or 40 nM with siRNA following the manufacturer's instructions. Overexpression vectors (2 µg) were transfected into the cells using Lipofectamine 3000 transfection reagent (Invitrogen). After 24 h of transfection, cells were harvested and used for the following experiments. All results for the optimization of transfection are demonstrated in Fig. S1.

Cell proliferation assay
The cell growth rate was monitored by colony formation assay and colorimetric assay using CCK-8 reagent (Dojindo, Kumamoto, Japan). In all, 3 × 10 3 cells/well were seeded onto a 96-well plate and cultured up to 5-7 days. Following staining with CCK-8 solution according to the provided instructions, optical densities were measured on a microplate reader (Sunrise, Tecan, Switzerland) and OD 595 was eliminated from the OD 450 . For colony formation assay, cells were seeded at a density of 3 × 10 3 cells/dish on a 60-mm culture dish. After transfection and Tam treatment, cells were maintained in a 5% CO 2 incubator (37 °C) for 14-20 days. Colonies were xed with a 7:1 mixture of methanol and acetic acid, stained using 0.2% crystal violet (Gibco BRL), and counted with ImageJ software (NIH, MD, USA).

Flow cytometric analysis
Apoptosis was analyzed using an APC Annexin V Apoptosis Detection Kit with PI (BioLegend, San Diego, CA, USA). Annexin V staining was performed for cells diluted in Annexin V binding buffer for 8 min followed by propidium iodide (PI) reagent treatment for 10 min. Samples were measured using an Accuri C6 ow cytometer (BD Biosciences, San Jose, CA, USA) with 488-nm and 640-nm lasers. To monitor Tam uptake by cells, 1 × 10 6 cells seeded in a 60-mm dish were treated with FLTX1 (Aobious, Gloucester, MA, USA, AOB4054) for 2 h at nal concentration of 10 µM. Cells were then harvested after washing with PBS, and a concentration of 1 × 10 6 cells/mL was prepared. FLTX1 uorescence was detected with Becton Dickinson FACSAria III (BD Biosciences) and analyzed with the Flowing Software 2.5 (http:// owingsoftware.btk. /).

Tam sensitivity assay
Alterations in sensitivity to Tam were measured by cytotoxicity assay. Brie y, 1 × 10 4 cells were seeded per well in a 96-well plate and transfected with recombinant cDNA-harboring plasmids and/or siRNAs. On the following day, Tam dissolved in sterile-ltered ethanol was added to cells at nal concentrations of 0, 0.05, 0.1, 0.5, and 2 µM with a nal ethanol concentration 0.1%. After 24 h, 10 µL of CCK-8 solution was added to each well and the plate was incubated for 90 min. Following this, the plate was read at 450 nm on a plate reader.

Methylation-speci c PCR (MSP) and quantitative real-time RT-PCR (qPCR)
MSP was performed to determine the methylation level of speci c CpG sites, as previously described [20]. Brie y, DNA and RNA from FFPE sections were extracted using the RecoverAll Multi-Sample RNA/DNA Work ow (Invitrogen). Total DNA and RNA were prepared using ZR-Duet DNA/RNA MiniPrep kit (Zymo research, Irvine, CA, USA) from solid tissues and cultured cells. For preparing samples for methylation analysis, the genomic DNA was treated with bisul te using a Zymo Research EZ DNA Methylation Kit (Zymo Research). Demethylation of the cytosine residues was achieved by exposing the cells to culture media containing a methyltransferase inhibitor, 5-Aza-2′-deoxycytidine (Aza) (Sigma-Aldrich), at a concentration of 5 µM for 72 h. PCR was conducted using 4-8 ng of DNA, and the yielded signals were calculated. To identify the transcript level of coding genes, cDNA was synthesized using a ReverTra Ace qPCR RT MasterMix with gDNA Remover kit (Toyobo, Osaka, Japan). qPCR analysis was conducted using KAPA SYBR FAST qPCR Kit (Kapa Biosystems, Wilmington, MA, USA) on an ABI 7300 instrument (Applied Biosystems, Foster City, CA, USA). Oligonucleotide primers were purchased from Bionics (Daejeon, Korea) (Table S2).

Tumor xenograft experiments
All mouse experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Dongguk University (No: IACUC-2017-010-1). We used 6-to 7-week-old female BALB/c nude mice (Orient Bio, Seongnam, Korea) for MCF-7 and MCF-7/TamR-derived xenograft models. The mice were anesthetized with a mixture of iso urane (Piramal Critical Care, Mettawa, IL, USA) and oxygen, and administered 17β-estradiol pellets (0.72 mg/pellet total dose; Innovative Research of America, Sonnasota, FL, USA) subcutaneously in the lateral neck area. On the next day, subcutaneous injections of 1 × 10 7 breast cancer cells resuspended with 100 µL of 1:1 mixture of PBS (Gibco BRL) and Matrigel (BD Biosciences, Bedford, MA, USA) were administered to the mice. Tumor growth was monitored weekly, and tumor volumes were calculated based on the following formula: length × width 2 × 0.5. When tumor sizes reached approximately 100 mm 3 , the mice were randomized into two groups for Tam treatment (Sigma-Aldrich). One group received intraperitoneal administration of 100 µL of 1 mg/kg Tam in corn oil (Sigma-Aldrich) and the other group was injected with a vehicle control for 5 days a week during the experiment. After 7 weeks of implantation, animals were sacri ced and tumors were harvested. The cancer tissues were xed in 4% paraformaldehyde and embedded in para n blocks for histological analysis by Logone Bio (Seoul, Korea).

Immunohistochemical staining
Immunohistochemical analysis was performed using tumor tissues of xenograft mice. To do this, para n blocks were sectioned 10 µm thick, organized into slides, and rehydrated through a graded ethanol series. Endogenous peroxidase activity in sections was ceased with 0.3% H 2 O 2 treatment for 15 min and then rabbit anti-ELOVL2 (1:400, Bioss, bs-7053R) or rabbit anti-THEM4 (1:100, Abcam, ab106435) was applied for 1 h at room temperature followed by incubation with horseradish peroxidaseconjugated anti-rabbit antibodies (Dako, Glostrup, Denmark, K4003). Liquid diaminobenzidine tetrahydrochloride (DAB) (Dako, K3468) was used as a chromogen to detect horseradish peroxidase activity. After counterstaining with Mayer's hematoxylin, immunohistochemical images were generated using panoramic MIDI scanner (3Dhistech, Budapest, Hungary). The ImageJ program (NIH) was used to pro le the DAB-positive areas of immunohistochemical images.

Statistical analysis
For microarray data, observations with adjusted P-values ≥ 0.05 were removed and were excluded from further analysis. Adjustments were made to control for false discoveries. Following adjustments, the remaining genes were de ned as differentially methylated if they displayed an increased or decreased methylation level which was equal to or higher than 0.2 compared with the control, or differentially expressed if they displayed at least a 2-fold difference compared with the control. Student's t-test was implemented to demonstrate statistical signi cance for all data from qPCR, MSP, IHC, and Western blot analysis comparing samples and control groups. Chi-squared test was used to analyze the differences in the rate of each variable for tumor tissues. Statistical analyses were conducted using SPSS for Windows, release 17.0 (SPSS Inc., Chicago, IL, USA). The results are expressed as the mean ± standard error and considered statistically signi cant at P-value < 0.05.

Results
Generation of MCF/TamR cells As a prerequisite to explore the molecular mechanism of TamR, a MCF-7/TamR cell line was established by exposing the MCF-7 cells to increasing concentrations of Tam up to 160 nM for 22 weeks (Fig. S3a). The nally developed MCF-7/TamR cells showed a higher growth rate compared with their parental cells (Fig. S3b) and also a large number of cells survived against high concentrations of Tam (0.1 and 0.5 ump) as judged by the colony formation assay (Fig. S3c). The apoptosis rate was lower in TamR cells and less affected by Tam than in the MCF-7 cells (Fig. S3d and Fig. S4), possibly explaining the higher growth and survival rate in the presence of Tam. The acquisition of TamR was further monitored by examining a few marker genes, the expressions of which were previously known to be altered in the course of TamR acquisition. Thus, EGFR, CCND1, CD146, and BCAR3 were upregulated, whereas BAG1 and IGF1 were downregulated, as previously observed [22][23][24][25], con rming appropriate generation of MCF-7/TamR cells (Fig. S3e). Notably, the level of Tam inside the TamR cells was not decreased compared with that in the parental MCF-7 cells after the two cell types were equally treated with Tam, indicating that the resistance is not caused by a net decrease of the drug transport across the plasma membrane (Fig. S3f) ELOVL2 is downregulated by DNA hypermethylation in TamR breast cancer Genome-wide methylation analysis was performed in duplicates for each sample of MCF-7 and MCF-7/TamR cells. Comparison of the two cell types showed hyper-and hypo-methylation with |Δβ| ≥ 0.2 at 331 and 94 CpG sites, respectively, corresponding to 356 unique genes (Fig. 1a). Among highly altered genes, ESR1 (Δβ = 0.35), MAGED1 (Δβ = 0.3), and RASAL1(Δβ = 0.36) were listed, alteration in methylation of which in TamR cells has been previously known [26][27][28], indicating the reliability of MCF-7/TamR cells developed in the current study. The microarray data were also veri ed by examining the expression of ve randomly selected genes from the highly altered genes via qPCR. Consequently, hypermethylated SCL19A1, SKAP1, and ELOVL2 were downregulated, whereas hypomethylated CD59 and MMP1 were upregulated (Fig. 1b), supporting the close relationship between methylation and expression. Next, the 356 genes were examined for functional inter-relatedness using the IPA software tool. The top network with the highest con dence was "Skeletal and Muscular Disorders, Cellular Assembly and Organization, Connective Tissue Development and Function" (Fig. 1c). Canonical pathway analysis identi ed "Neuroactive ligand-receptor interaction" as the predominant pathway (Fig. 1d). Disease and function annotation analysis indicated that genes associated with "Cancer" and "Cell death and survival" are frequently included (Fig. 1e).
To identify a novel and pivotal marker that contributes to the acquisition of TamR, ELOVL2 in Table S3 was selected because the gene showed signi cant hypermethylation (Δβ = 0.49), but its role in cancer or drug resistance is not yet known. Both transcripts and proteins were downregulated in the MCF-7/TamR cells compared with those in MCF-7 cells (Fig. 1b and 2a). Induction of lower methylation by treating the MCF-7/TamR cells with Aza, a methyltransferase inhibitor, upregulated ELOVL2, suggesting an epigenetic regulation of the gene (Fig. 2b). Next, the methylation and expression of the gene were examined in breast cancer tissues obtained from patients showing resistance to clinical treatment with Tam. The result indicated that ELOVL2 was hypermethylated and downregulated in the TamR cancer tissues (n = 28) compared with the Tamoxifen sensitive (TamS) tissues (n = 33) (P < 0.001) (Fig. 2c).
Immunohistochemical analysis of tissues from cancer patients showed lower expression of ELOVL2 in the TamR tissues than in the TamS tissues (Fig. 2d). The rate of distant metastasis-free survival (DMFS) for breast cancer patients, which was investigated through the GOBO database, indicates that lower rates of DMFS were observed in cancer patients with lower expression of ELOVL2, whereas higher rates of DMFS were observed in patients with a higher ELOVL2 expression (P < 0.05) (Fig. 2e).

ELOVL2 inhibits MCF-7/TamR cell proliferation and recovers Tam sensitivity
To obtain information about the role of ELOVL2 in the acquisition of TamR, its effect on cell growth and drug sensitivity recovery was examined after constructing ELOVL2-overexpressing MCF-7/TamR cells (MCF-7/TamR-ELOVL2 ORF). The growth rate of MCF-7/TamR-ELOVL2 ORF cells was retarded up to 18% compared with that of MCF-7/TamR cells (Fig. 3a). In the presence of Tam, the growth rate of the MCF-7/TamR-ELOVL2 ORF cells was further retarded, especially at Tam concentrations of < 0.5 µM (Fig. 3b). This observation was also reproduced in the cell survival experiment of colony formation assay, where a lower number of colonies were observed in the MCF-7/TamR-ELOVL2 ORF cells under Tam pressure (Fig. 3c).
Next, the effect of ELOVL2 on the recovery of Tam sensitivity was monitored in a mouse xenograft model where the cancer cells were subcutaneously injected. As a result, MCF-7/TamR cells grew faster than MCF-7 (Fig. 4a) and showed larger tumor volume in all the eight mice sacri ced 7 weeks after injection (Fig. 4b). A decrease in ELOVL2 protein expression in the tumor was observed by Western blot and immunohistochemical analysis (Fig. 4c and d). The animals implanted with MCF-7/TamR-ELOVL2 ORF showed retarded tumor growth compared with animals with MCF-7/TamR cells (Fig. 4e and f), indicating the tumor suppressive activity of ELOVL2. Furthermore, the tumors over-expressing ELOVL2 showed an increased sensitivity to Tam by representing a smaller tumor size ( Fig. 4e and f).

ELOVL2 resists Tam by suppressing the AKT pathway
To get an insight into the regulatory mechanism of ELOVL2 for drug resistance, a genome-wide expression assay was performed in duplicates for identifying target genes. A comparison of the expression pro le between MCF-7/TamR and MCF-7/TamR-ELOVL2 ORF revealed 969 genes that were signi cantly altered (expression level change > 2) (Fig. 5a). IPA analysis identi ed "Cardiovascular Disease, Cell-To-Cell Signaling and Interaction, In ammatory Response" pathway and "Behavior, Reproductive System Development and Function, Cardiac Infarction" pathway as the top rst and second networks, respectively (Fig. 5b). In accordance with this, the canonical pathway analysis (Fig. 5c) and the disease and function analysis (Fig. 5d) predicted the immune-related pathway and cancer as the top categories, respectively. Notably, AKT and ERα are placed at the center of each network, interacting directly or indirectly with many genes in the pathways, such as THEM4, BMF, and FTO in the case of AKT pathway and HYOU1, CDC42EP2, and S100A9 in the case of ERα. The expression of the genes in the two pathways were further examined by qPCR and the result con rmed the same direction of expression alteration as observed in the expression array (Fig. S5). The AKT pathway is a key pathway responsible for cell metabolism, growth and division, apoptosis suppression, and angiogenesis [29]. In particular, THEM4 is known to promote AKT phosphorylation and functions as an oncogenic molecule in breast cancer [30].

THEM4 is downregulated by ELOVL2 recovering Tam sensitivity
To elucidate the molecular mechanism of how THEM4 induced TamR in association with ELOVL2, its expression was examined in cell lines and tumor tissues from xenografted mice. THEM4 was upregulated in the MCF-7/TamR cells compared with MCF-7 cells at both RNA and protein levels (Fig. 6a).
Overexpression of ELOVL2 in the MCF-7/TamR cells set back the THEM4 expression to a lower level than that observed in the control MCF-7/TamR cells (Fig. 6b). Western blot analysis (Fig. 6c) and immunohistochemical analysis (Fig. 6d, e) also found a similar expression pro le for THEM4 in the tumor tissues of the xenografted mice, which were generated from MCF-7/TamR and MCF-7/TamR-ELOVL2 ORF. Subsequently, the association of THEM4 with ELOVL2 was tested with regard to TamR, cell growth, and apoptosis. siRNA-induced downregulation of THEM4 contributed to the recovery of Tam sensitivity as shown by the colony formation assay (Fig. 6f), and this effect was strengthened by ELOVL2 (Fig. 6g). The dye-based CCK assay using the same THEM4-siRNA and ELOVL2-overexpression strategy con rmed the results of the colony formation assay (Fig. 6h, i), suggesting an inhibitory and a stimulatory effect of ELOVL2 and THEM4, respectively, for the acquisition of TamR. However, the total apoptosis rate was not changed signi cantly, even though early and late apoptosis was decreased and increased, respectively, by the downregulation of THEM4 (Fig. S6).
Considering all the experimental ndings from this study and the previous literature, which indicate a close association of AKT and THEM4 in the signaling pathway, it is suggested that ELOVL2 contributes to the recovery of TamR by regulating pivotal genes such as THEM4 in the AKT pathway (Fig. 7).

Discussion
This study aimed at identifying epigenetically regulated marker genes responsible for TamR in breast cancer patients and then elucidating the molecular mechanism to give an insight for the prevention and treatment of TamR recurrence in cancer patients. Occasionally, drug-resistant cancer cells have shown to actively pump the drug out of the cells, reducing the net amount of drug inside cells. This fact could partly explain drug resistance. For instance, a 10-fold lower Tam concentration was found in extracts from TamR tumors than that in TamS tumors in mice [31]. However, it is controversial whether all TamR cells acquire the high e ux rate, because the P-glycoprotein, an e ux pump that is known to bind to Tam, is not expressed in these tumors. Furthermore, the TamR cells generated and used in our study showed no signi cant change in Tam transport compared with the parental MCF-7 cells. Therefore, we speculated that genetic changes may have more strongly contributed to TamR than alterations in the drug e ux rate.
ELOVL2 has consistently shown hypermethylation and downregulation in TamR cancer, indicating its potential application as an epigenetic marker for the diagnosis of TamR cancer. However, no signi cant difference of methylation or expression between normal and cancer tissues of breast was found. In normal tissue, ELOVL2 has been known to undergo hypermethylation on the promoter DNA as one ages.
The downregulation of ELOVL2 induced by the hypermethylation is not considered enough to drive cells into cancer, although suppression of ELOVL2 stimulated cell proliferation in our study. This may imply that the deregulation of ELOVL2 is crucial during the acquisition of TamR, rather than during the development of the primary cancer. Moreover, an association between ELOVL2 DNA methylation and future breast and colorectal cancer development has been observed [32]. ELOVL2 Expression has been previously known to be enhanced by ERα in breast cancer cells [33]. In this study, Tam exposure speci cally abolished ELOVL2 expression. Our microarray analysis also revealed upregulation of ERα by ELOVL2, suggesting a positive feedback mechanism for the regulation of the two genes. In addition, many ERα-regulated genes such as HYOU1 (3.8-fold decrease), S100A9 (1.2-fold decrease), and CDC42EP2 (2.4-fold decrease) were also deregulated by ELOVL2. HYOU1 is a hypoxiainduced protein and its upregulation suppresses programmed cell death, contributing to invasiveness in breast cancer [34]. S100A9 has been identi ed to be expressed by epithelial cells involved in malignancy and its expression levels are inversely correlated with ERα in breast cancer [35]. CDC42EP2 is a member of the binder of Rho GTPases (Borg) family and little is known about its role in the disease [36].
The AKT pathway is a pivotal one wherein a few TamR-related genes have been identi ed [23]. In accordance, previously identi ed AKT-regulated genes also appeared in the ELOVL2-overexpression network, such as THEM4, BMF, and FTO (Fig. 5b). Furthermore, other genes involved in the AKT pathway, including mTOR, PIK3CA, and CREB1, were shown to be downregulated by ELOVL2 (Fig. S7a). Notably, the expression of Akt as well as the phosphorylated form was downregulated in a similar ratio by ELOVL2 (Fig. S7b). Therefore, the decrease of the phosphorylated form is caused by the lowered total AKT levels, suggesting that ELOVL2 regulates AKT and not p-AKT. The two networks closely communicating each other by sharing a few common genes. For example, NFAT5 is regulated by an estrogen-induced microRNA [37] and the regulation is mediated via PI3K/AKT-signaling pathways [38]. In addition, A100A9, a calcium-binding protein that is highly expressed in malignant breast cancer, induces a decrease of ERα in MCF-7 cell [35], and inhibits PI3K/AKT pathway in pancreatic adenocarcinoma cells [39]. Collectively, the current study proposes that ELOVL2 is an integral signaling molecule of the AKT axis in ER-positive breast cancer cells.
Further it is noteworthy that a representative aging marker gene is associated with drug resistance. ELOVL2 has shown to increase methylation at the promoter CpGs with aging, which accompanies a decreased expression with aging [18]. Considering the increased drug resistance and lower expression of ELOVL2 in cancer cells of patients, epigenetic aging could make cancer patients more vulnerable to acquisition of drug resistance. This gives us an insight on how to design the strategy for treating drugresistant cancers. Meanwhile, it should be mentioned that ELOVL2 function is not limited to epigenetic aging and drug resistance. The gene showed tumor suppressor-like activity by inhibiting cancer cell growth in cultures cancer cells as well as in the xenografted mouse model. Therefore, ELOVL2 is considered to have a wide spectrum of biological functions in addition to the fatty acid elongation activity.
Genome-wide methylation analysis has found that numerous genes were deregulated in addition to ELOVL2 in TamR cells, suggesting distortion of multiple pathways during the course of drug resistance acquisition. A genome-wide expression array also found > 1,200 genes clustered into ERα function, cell cycle regulation, transcription/translation, and mitochondrial dysfunction [40]. Therefore, to completely understand the molecular mechanisms and to conquer TamR in cancer, a further comprehensive approach is needed.

Conclusions
Altogether, a genome-wide pro le of epigenetic changes during TamR acquisition in breast cancer cells was constructed. ELOVL2 was identi ed as a marker that was hypermethylated and downregulated in TamR cancer compared with TamS cancer. ELOVL2 is responsible for the recovery of TamS, which was shown in an in vivo animal xenograft model. AKT-and ERα-hubbed networks are pivotal in ELOVL2 signaling, where THEM4 contributes to the relaying ELOVL2 signaling. This study is the rst to identify a linkage between drug resistance and a gene involved in fatty acid synthesis. Our data may give credence to elucidating the mechanism of TamR cancer and to developing its treatment strategy.

Consent for publication
Written informed consent was obtained from all patients.

Competing interest
The authors declare no competing interest.   Table S3 and their expression is examined by qPCR. c Genome-wide methylation analysis with 356 genes identi es "Skeletal and Muscular Disorders, Cellular Assembly and Organization, Connective Tissue Development and Function" pathway as the top network.
Genes hypermethylated in MCF-7/TamR are shaded in red, whereas those hypomethylated are shaded in green, with the color intensity signifying the magnitude of methylation change. Solid lines represent direct interactions, and dashed lines represent indirect interactions. Top canonical pathways (d) and disease and biofunction (e) for the genes in which methylation is signi cantly altered in MCF-7/TamR.   weeks after transplantation to obtain the tumor tissues. Expression of ELOVL2 in the xenografted tumor is examined by Western blot analysis (c) and immunohistochemical analysis (d). Three tumor sets are analyzed and the average protein expression is denoted in a bar graph. Representative images are shown.
Scale bar, 50 μm. e MCF-7/TamR cells that are stably transfected with ELOVL2-expressing cDNA or control DNA are subcutaneously injected into nude mice and Tam is administered 3 weeks after cell injection. The tumor volume is measured for 7 weeks. f At week 8, mice are sacri ced to obtain the tumor tissues (n = 6 for corn oil-treated mice; n = 4 for Tam-treated mice).

Figure 5
Highest con dence network of genes displaying altered ELOVL2 expression in MCF-7/TamR. ELOVL2 is overexpressed in MCF-7/TamR and a genome-wide expression analysis is performed. a Heatmap analysis of 969 genes that are signi cantly deregulated by ELOVL2. The data are from the microarray in duplicates. b Highest con dence network of genes displaying altered expression identi es "Cardiovascular Disease, Cell-To-Cell Signaling and Interaction, In ammatory Response" pathway and "Behavior, Reproductive System Development and Function, Cardiac Infarction" pathway as the top networks. Genes that are upregulated are shaded in red, whereas those that are downregulated are shaded in green, with the color intensity signifying the magnitude of expression change. Solid lines representing direct interactions, and dashed lines representing indirect interactions. c Top 10 canonical pathways and d disease and function annotation of the genes of which expression is signi cantly altered by ELOVL2. The most signi cant canonical pathway is "Antigen Presentation Pathway" and disease and function annotation is "Solid tumor". cells stably transfected with ELOVL2 cDNA; ORF NC; negative cDNA control. c Increased expression of THEM4 in the xenografted MCF-7/TamR but suppression by ELOVL2. Western blot analysis is performed for tumor tissues from ELOVL2 ORF and control. Immunohistochemical analysis of THEM4 in xenografted tumor tissue of MCF-7/TamR (d) and cells stably transfected with ELOVL2 cDNA (e) Scale bar, 50 μm. Effect of THEM4 on recovery of Tam sensitivity. THEM4 is downregulated via a siRNA in MCF-7/TamR (f) and ELOVL2-overexpressing MCF-7/TamR cells (g). Sensitivity to Tam is examined by colony formation assay. Representative images from three independent assays are shown. h Effect of THEM4 on cell proliferation is examined by a dye-based CCK assay. i Effect of THEM4 on TamR is examined by exposing the cells to Tam after downregulating the gene with siRNA. All the assays are performed in triplicates, and the result is depicted as mean ± SE.

Figure 7
Schematic illustration of the regulatory pathway by ELOVL2. The uptake ratio of Tam across plasma membrane in MCF-7/TamR cells is similar to that in the parental MCF-7 cells. ER and ELOVL2 crosstalk to regulate each other. ELOVL2 blocks the PI3K/AKT/mTOR pathway via inhibiting THEM4 and PI3K. In TamR cancer, ELOVL2 is downregulated by hypermethylation, resulting in loss of inactivation of AKT and