ELF5 and DOK7 regulation in anti-estrogen treated cells and tumors

During the initial expansion of the tamoxifen-selected cell lines, a few of the cell lines appeared to grow either in unusual clumps or in suspension. However, after several passages in T-75 flasks all of the tamoxifen-selected lines displayed a phenotype similar to that of MCF-7 (Fig. 1).

Blinded observers (pathologists RMJ and CNO) assessed IHC-stained slides for ER, PR and HER2. ER protein expression in the 21 tamoxifen-selected cell lines was similar to that observed in the parental MCF-7 (Figs. 2, 3a). Nineteen of the tamoxifen-selected cell lines received an Allred score of 7, and one line each received an Allred score of 5 and 6, which is only slightly higher than the score of 6 routinely assigned ER staining in MCF-7 [22]. In contrast, the protein expression for PR was substantially lower in the tamoxifen-selected cell lines than in the parental MCF-7 (Figs. 2, 3a). For progesterone receptor, six cell lines received an Allred score of 0, thirteen lines received a score of 2, and one line each received a score of 3 and 4. Allred scores of two and less are considered PR-negative. Thus, only two of the 21 cell lines were scored positive for PR and these positive scores are still considerably lower than the typical Allred score of 6 assigned for PR staining in MCF-7 [22]. HER2 protein expression was similar between the parental and tamoxifen-selected cell lines with all lines receiving a score of 0 or 1+ (Figs. 2, 3a).

Having established with IHC that the novel tamoxifen-selected cell lines were ER-positive, PR-negative (18 of 21), and HER2-negative, we wanted to examine the expression of additional genes that might play a role in endocrine resistance, and compare the expression in the novel cell lines with the expression in several previously described tamoxifen-resistant cell lines, TMX2-28, TMX2-11, and TMX2-4 [13, 23]. We began with a comparison of ESR1 mRNA levels to confirm the ER-immunostaining data. As expected, based on the protein data, the ESR1 mRNA levels of the novel tamoxifen-selected cell lines did not differ significantly from the parental MCF-7 (Fig. 3b). Interestingly, however, the ESR1 mRNA levels for the majority of the cell lines were slightly lower than that of MCF-7. In comparison, ESR1 expression is significantly higher in TMX2-11 as compared to MCF-7 (p < 0.001) and significantly lower in TMX2-28 as compared to MCF-7 (p < 0.001), confirming previous reports [13, 23].

We next compared the expression of DOK7 mRNA in MCF-7 with the expression in all of the tamoxifen-selected lines and the mass culture TMX3A. DOK7 was significantly overexpressed in eleven of the 21 novel tamoxifen-selected cell lines (p < 0.05) and was not significantly altered in any of the other cell lines or in the mass culture (Fig. 4). Interestingly, DOK7 expression was reduced, although not significantly, in TMX2-28 the only ER-negative tamoxifen-resistant cell line examined. Next, we compared ELF5 expression in MCF-7 with the expression in all of the tamoxifen-selected lines and the mass culture. ELF5 was significantly over expressed in ten of the 21 novel cell lines (p < 0.01) and not in the mass culture or the previously described cell lines (Fig. 5). Again, expression was decreased, but not significantly, in the ER-negative TMX2-28 cell line. There was limited overlap between the tamoxifen-selected cell lines with elevated DOK7 mRNA levels and those with elevated ELF5 mRNA levels; only three cell lines (clones 6, 7 and 19) had increased expression of both genes.

Given that the expression of DOK7 and ELF5 was significantly increased in roughly half of the novel tamoxifen-selected, ER–positive cell lines, and decreased (although not significantly) in the only ER-negative cell line, TMX2-28, we wanted to determine the extent to which DNA methylation in the promoter region of DOK7 and ELF5 controlled gene expression. If DNA promoter methylation was controlling expression we would expect to see a decrease in methylation associated with the up-regulation observed in the novel tamoxifen-selected cell lines, and an increase in DNA methylation associated with the down-regulation observed in the TMX2-28 cell line.

We first examined DNA methylation data available from the Illumina human methylation 450K bead chip (HM450BC) for MCF-7, TMX2-11 and TMX2-28 [12]. The HM450BC includes nine CpGs within the transcriptional start site (TSS) of DOK7, and eight CpGs within the TSS-region of ELF5. As shown in the heat maps of Fig. 6, the methylation in the TSS region of both DOK7 and ELF5 are substantially increased in TMX2-28 cells as compared to MCF-7 and TMX2-11. The increase in methylation observed in TMX2-28 cells is consistent with the lower expression of DOK7 and ELF5 in this cell line and suggests that promoter methylation in DOK7 and ELF5 may control gene expression. The results from TMX2-28 are in agreement with the inverse association between promoter methylation and gene expression previously demonstrated for ELF5 [24‐26] and DOK7 [27]. In contrast, the increased expression of both ELF5 and DOK7 observed among a subset of the ER-positive, tamoxifen-selected, cell lines cannot be attributed to a decrease in promoter methylation as compared to the parental cell line, since the parental MCF-7 methylation is already near zero at the majority of tested sites.

To confirm the methylation results observed with the HM450BC we designed pyrosequencing assays to interrogate six CpG sites within the TSS of DOK7 and three CpG sites within the TSS of ELF5. Figure 7a shows the pyrosequencing results for MCF-7, TMX2-4, TMX2-11, TMX2-28, the mass culture TMX3A and several TMX3A clonal cell lines for DOK7. As expected, only the ER-negative TMX2-28 cell line has increased DNA methylation. Average DOK7 methylation in TMX2-28 cells was 35.7 %, whereas average methylation of MCF-7 was 3.3 % and TMX3A was 4.2 %. Pyrosequencing of all 21 novel tamoxifen-selected cell lines showed no differences in DNA methylation in DOK7 as compared to the parent MCF-7 (6 representative tamoxifen-selected cell lines are shown).

Likewise, increased DNA methylation of ELF5 was observed only for TMX2-28 (Fig. 7b). Average DNA methylation of ELF5 in TMX2-28 was 74.9 % compared to MCF-7 at 5.5 %, and TMX2-4, 2-11 and TMX3A (mass culture) at 4.5, 5.75, and 6 %, respectively. Additionally, all 21 tamoxifen-selected clonal cell lines, six of which are shown in Fig. 7b, had low methylation of ELF5.

We next examined DNA methylation in primary and second/recurrent breast tumors from three groups of women (Table 1) to determine the extent to which the tamoxifen-related changes observed in cell culture also occurred in the tumors of women treated with anti-hormonal therapy. Primary and recurrent tumors were categorized by PR status because PR expression is an indication of ER activity [28]. Furthermore, in the majority of novel tamoxifen-selected clones described in this report, decreases in PR expression were accompanied by increases in ELF5 and DOK7 expression. As shown in Table 1, all 16 women who had a PR-positive primary tumor received hormonal therapy, and 10 of the 16 are known to have received tamoxifen (primary tumors from groups 1 and 2: tamoxifen 10; aromatase inhibitors 2; type of hormonal therapy unknown 4). Additionally, 3 of the 8 women with PR-negative primary tumors (group 3) received hormonal therapy because their primary tumor was ER-positive.

Comparison of DOK7 promoter methylation across the primary and second tumors of women in the three groups showed no differences (Fig. 8a); median mean methylation for the 9 CpGs sites within the TSS was below 10 % in all groups. In contrast, analysis of the 8 CpGs in the TSS of ELF5 showed wide differences among groups (Fig. 8b). First, ELF5 was significantly more methylated in PR-positive tumors than in PR-negative tumors (β = .58 and .39, respectively; t = 4.97; df = 45; p < 0.001). The lower methylation of ELF5 in PR-negative tumors is consistent with the increased expression of ELF5 observed in ten of our tamoxifen-selected cell lines, all of which had decreased levels of PR. Second, there was a trend for primary tumors from women treated with anti-estrogen therapy to have higher mean ELF5 methylation as compared to the woman’s second/recurrent tumor (β = .55 and .49, respectively; paired t = 1.53; df = 19; p = 0.144). This is the first demonstration of a decrease in ELF5 methylation in human breast tumors occurring after anti-estrogen treatment, and is consistent with a report by Kalyuga et al. showing an increase in ELF5 expression in luminal cell lines that acquired resistance to tamoxifen [8]. Of particular interest is the difference between PR-positive primary tumors in women of groups 1 and 2 (Table 1). PR-positive primary tumors from women who had second PR-positive tumors had significantly higher mean methylation than did PR-positive primary tumors from women whose second tumor was PR-negative (β = .64 and .51, respectively; t = 3.54; df = 14; p = 0.003). To further examine the relationship between ELF5 methylation, PR status, and tumor occurrence, cluster analysis was performed with GenomeStudio using a correlation matrix (Fig. 9). Groups from women whose first and recurrent tumors were both PR-positive clustered together (far right). Groups from first and recurrent tumors of women where the recurrent tumor was PR-negative clustered separately. The ELF5 methylation of PR-positive primary tumors in women who went on to have a PR-negative recurrence was clearly distinct from the ELF5 methylation of other PR-positive tumors. This indicates the possibility of stratification of women for tailored treatment in the early stages of disease.

Widespread resistance to endocrine therapies limits their usefulness and reduces patient remission. Previous studies have identified tamoxifen-induced cellular changes. Here we characterize novel clonal cell lines whose diversity could represent drug resistant cancers in women. All of the tamoxifen-selected lines remained ER positive, which is consistent with results from past experiments [7, 13], yet most of the lines lost expression of PR. We show that many of these cell lines had significantly elevated DOK7 and ELF5 gene expression compared to MCF-7 cells not treated with tamoxifen. Promoter methylation of DOK7 and ELF5 in the MCF-7 control cell culture was extremely low, and therefore, a further decrease in promoter methylation of DOK7 and ELF5 was not observed to associate with the increased expression seen in the tamoxifen-selected cell lines.

However, we found the promoter region of ELF5 to be highly methylated in primary PR-positive tumors and that methylation was vastly reduced in second tumors occurring after anti-estrogen treatment. This finding is consistent with Kalyuga et al. [8] who found higher expression of ELF5 to correlate with resistance to anti-estrogens. Importantly, we also observed significantly higher levels of ELF5 methylation in the subset of primary, PR-positive tumors that would go on to recur as PR-negative, indicating an opportunity for the development of more personalized treatment in the early stages of disease. To our knowledge this is the first report of differential ELF5 methylation in paired primary and recurrent breast tumors and these results deserve to be replicated in a larger study.

MCF-7 cells were purchased from the American Type Culture Collection (ATCC) and grown as previously described [23], with tamoxifen added at 1 μM. Briefly, cells were maintained at 37 °C with 5 % CO₂. Cells were grown in Falcon T-75 flasks, given fresh media twice per week, and split once or twice per week when 70–90 % confluency was obtained. The tamoxifen-selection was conducted in triplicate for six months resulting in three populations (referred to as mass cultures), one of which, TMX3A, was used for generating clonal lines. Cloning by limiting dilution was performed to assess the diversity of cells within the tamoxifen-selected mass culture. Tamoxifen-selected TMX3A cells were seeded at one cell/well and allowed to grow in 96-well plates for 6–7 weeks. Twenty-one cell lines were selected (referred to as clonal cell lines), expanded and cryopreserved.

RNA isolation was performed using Tri-reagent (Molecular Research Center, Inc. Cat. No. E8875). DNA was isolated from frozen cells using the QIAmp DNA Mini Kit (Qiagen Cat. No. 51304). RNA and DNA were quantified using a NanoDrop 8000 (Thermo Scientific). cDNA was created from isolated RNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) with added RNasin^® Ribonuclease Inhibitor (Promega, Ref. N2511). Quantitative RT-PCR (qRT-PCR) was performed on this cDNA using FastStart Universal SYBR Green Master mix (Roche, Cat. No. 04 913 850 001). Primers used for qRT-PCR are listed in Table 2. Primers are designed to span an exon–exon junction, therefore DNaseI treatment of the RNA was not necessary.

500 ng of DNA was bisulfite treated using the EZ DNA methylation-Lightning Kit (Zymo Research Cat # D5030). Bisulfite-treated DNA was pyrosequenced using the Pyromark Q24 system (Qiagen) and manufacturers protocol (Qiagen), and gene methylation levels were detected as cysteine to thymidine ratios at specific loci. Primers used for pyrosequencing are listed in Table 3. For DOK7, the region interrogated lies within the transcriptional start site region, from chromosome 4: 3434598 to 3434634 (NCBI build36 coordinates). For ELF5, the region interrogated lies within the transcriptional start site region from chromosome 11 (−1): 34491972 to 37791933 (NCBI build 36 coordinates).

Cells were seeded onto poly-l-lysine coated-slides (Polysciences #22247) at a density of 200,000 cells/mL/slide and fixed the following day using 100 % ice-cold methanol. Slides were stained on a BenchMark Ultra platform using the UltraView Universal DAB Detection Kit with previously optimized antibodies for ER (Ventana CONFIRM anti estrogen receptor SP1 Rabbit monoclonal primary antibody), PR (Ventana CONFIRM anti progesterone receptor 1E2 Rabbit monoclonal primary antibody) and HER2 (Ventana PATHWAY anti HER-2/neu antibody 4B5 Rabbit monoclonal antibody). Stained slides were dehydrated in ethanol and xylene and coverslips were added manually. Slides were examined by a pathologist and ER and PR were recorded as Allred scores [29], HER2 was scored positive if greater than 30 % of the cells showed 3+ membrane staining; scores of 0–2 were considered negative [22].

Collection and processing of human breast tumors was conducted with IRB approval from Baystate Medical Center and the University of Massachusetts as previously described [30]. Breast tumor samples purified using the BiOstic FFPE tissue DNA isolation kit (Mo Bio, Carlsbad, CA) were sent to the core facility at the University of Southern California for HM450 BeadChip (Illumina) analysis and detailed results are in preparation (Williams in prep). Twenty-four matched pairs of primary and second/recurrent breast tumors for which methylation data and hormone receptor status were known and for which patient treatment data, and age were available were selected. We use the term second and recurrent interchangeably throughout the manuscript. It is unknown whether the second tumor is a true new second tumor or a recurrence of the first tumor. Nineteen of the 24 women received anti-estrogen therapy after resection of the first tumor, i.e. prior to resection of the second tumor (Table 1).

Q RT-PCR data represent biological duplicates for each cell line run in technical duplicate. The data were normalized to hypoxanthine-guanine phosphoribosyltransferase (HPRT) as a reference gene. One-way ANOVAs with Bonferroni post hoc tests were run to determine significance at p < 0.05. For presentation purposes, each value was normalized to the average MCF-7 value for the gene of interest. Data were then log transformed to provide the expression levels relative to MCF-7, with values greater than 0 indicating higher expression, and values below 0 indicating reduced expression. The average of 4 biological replicates was used to determine the standard deviation. Error bars on Figs. 3, 4, 5 and 7 represent the standard deviation. Data were analyzed using GraphPad Prism (version 3.02, La Jolla, CA). GenomeStudio Methylation Module (v.1.9, Illumina, San Diego, CA) was used to analyze the β values of the methylation data obtained from the HM450 BeadChip. Statistical comparisons between groups were conducted with paired and unpaired t-tests provided in the Data Analysis ToolPak of Excel (Microsoft Office Professional plus 2013).