Epigenetic Dysregulation of KCNK9 Imprinting and Triple-Negative Breast Cancer

Simple Summary Genomic imprinting is an inherited form of parent-of-origin specific epigenetic gene regulation that is dysregulated by poor prenatal nutrition and environmental toxins. Here, we showed that KCNK9 is imprinted in breast tissue and identified the differentially methylated region (DMR) controlling its imprint status. Hypomethylation at the DMR, coupled with biallelic expression of KCNK9, occurred in 63% of triple-negative breast cancers (TNBC). The association between hypomethylation and TNBC status was highly significant in African-Americans (p = 0.006), but not in Caucasians (p = 0.70). The high frequency of KCNK9 DMR hypomethylation in TNBC and non-cancerous breast tissue from high-risk women provides evidence that hypomethylation of the KNCK9 DMR/TASK3 overexpression may provide a new target for prevention of TNBC. Abstract Genomic imprinting is an inherited form of parent-of-origin specific epigenetic gene regulation that is dysregulated by poor prenatal nutrition and environmental toxins. KCNK9 encodes for TASK3, a pH-regulated potassium channel membrane protein that is overexpressed in 40% of breast cancer. However, KCNK9 gene amplification accounts for increased expression in <10% of these breast cancers. Here, we showed that KCNK9 is imprinted in breast tissue and identified a differentially methylated region (DMR) controlling its imprint status. Hypomethylation at the DMR, coupled with biallelic expression of KCNK9, occurred in 63% of triple-negative breast cancers (TNBC). The association between hypomethylation and TNBC status was highly significant in African-Americans (p = 0.006), but not in Caucasians (p = 0.70). KCNK9 hypomethylation was also found in non-cancerous tissue from 77% of women at high-risk of developing breast cancer. Functional studies demonstrated that the KCNK9 gene product, TASK3, regulates mitochondrial membrane potential and apoptosis-sensitivity. In TNBC cells and non-cancerous mammary epithelial cells from high-risk women, hypomethylation of the KCNK9 DMR predicts for increased TASK3 expression and mitochondrial membrane potential (p < 0.001). This is the first identification of the KCNK9 DMR in mammary epithelial cells and demonstration that its hypomethylation in breast cancer is associated with increases in both mitochondrial membrane potential and apoptosis resistance. The high frequency of hypomethylation of the KCNK9 DMR in TNBC and non-cancerous breast tissue from high-risk women provides evidence that hypomethylation of the KNCK9 DMR/TASK3 overexpression may serve as a marker of risk and a target for prevention of TNBC, particularly in African American women.


Introduction
Epigenetic adaptations in response to in utero nutritional and environmental factors are hypothesized to play an important role in developmental plasticity and human disease susceptibility [1][2][3]. Diet-derived methyl donors and co-factors are necessary for the synthesis of S-adenosylmethionine (SAM), the methyl group donor for DNA methylation. Thus, environmental factors that alter early nutrition and/or SAM production can potentially influence adult disease risk by altering CpG methylation at critically important, epigenetically labile regulatory regions [4,5].
Genomic imprinting is an inherited form of parent-of-origin dependent epigenetic gene regulation that renders autosomal genes functionally haploid in a species, developmental stage, and tissue dependent manner [6,7]. There is evidence that epigenetic modifications in the genome link environmental exposures to adult disease susceptibility [5,[8][9][10][11], including cancer [12][13][14][15]. Moreover, imprinting can be dysregulated not only in somatic cells, but also in germ cells, potentially affecting offspring never subject to the parental exposure [2,8]. Since imprinted genes are frequently clustered and coordinately regulated by differentially methylated regions (DMRs), changes in a single DMR can disrupt the expression of more than one imprinted gene [16,17]. Disease susceptibility due to epigenetic deregulation also has specific windows of vulnerability, including embryogenesis, puberty, pregnancy, and old age [5,[18][19][20].
In a computational model, KCNK9 was predicted to be regulated by imprinting [21]. KCNK9 and its gene product, TASK3, is of interest for human health studies, as overexpression is strongly tied to cancer. KCNK9 is maternally expressed in the human brain, as well as in the mouse brain [21,22]. The KCNK9 gene encodes for the pH sensitive potassium channel protein, TASK3. TASK3 is present at the plasma membrane and regulates membrane depolarization in response to acidosis via inhibition of the background potassium-current [23,24]. Inactivation of the expressed maternal copy of KCNK9 results in Birk-Barel syndrome [25]. Overexpression of TASK3 in cell lines promotes tumor formation and hypoxia-resistance [23,24,26]. Blocking the TASK3 channel protein, either chemically or by mutation, reduces cell proliferation and increases apoptosis, by unknown mechanisms [26,27]. However, TASK3 has been observed to also be localized in the mitochondria, in addition to the plasma membrane [28][29][30][31], and inhibition of TASK3 function has been shown to lead to mitochondrial dysfunction [29][30][31].
TASK3 is overexpressed in >40% of breast cancers, but genomic amplification of KCNK9 only accounts for TASK overexpression in <10% of breast cancers [24]. KCNK9 is known to be epigenetically regulated. Consequently, we hypothesized that overexpression of TASK3 protein (in the absence of KCNK9 duplication) could be due to epigenetic dysregulation, specifically the loss of parental silencing methylation or imprinting.
In this study, we investigated the imprinting status of KCNK9 in normal and malignant mammary epithelial cells. We demonstrated that it is monoallelically expressed and identified an associated DMR and regulatory DMR chromatin structure. In both cancer cells and mammary epithelial cells from high-risk women, we observed hypomethylation of the KCNK9 DMR. Hypomethylation of the KCNK9 DMR increased TASK3 protein expression and resulted in increased mitochondrial membrane potential and apoptosisresistance. DMR hypomethylation, and an increase in mitochondrial membrane potential, was observed most frequently in triple negative breast cancer (TNBC). antibodies were sc-2313 (anti-rabbit) and sc-2020 (anti-goat) (Santa Cruz Biotechnology), used at a dilution of 1/20,000. 5-Aza-2 -deoxycytidine treatment: MCF10A (data shown) and HMEC15 (data not shown) cells were treated with 1.0 or 2.5 µM of 5-aza-2 -deoxycytidine (Sigma Aldrich, St. Louis, MO, USA) or vehicle control for up to 72 h, with fresh medium added after 24 h. Cells were harvested every 24 h for determining DNA methylation and ∆Ψ M . Expression of TASK3 was detected via western blot. Experiments were performed in triplicate. Total RNA was also extracted for KCNK9 mRNA quantitation and cell growth was measured using an MTT assay. See Appendix A Materials and Methods for details.
Apoptosis and measurement of ∆Ψ M : Apoptosis was assessed by measuring intracellular caspase-3 activity. ∆Ψ M in cell lines was detected using JC-1 (Life Technologies). ∆Ψm was also determined in aspirated RPFNA cells from patients. The cells were collected as described below. An aspirate was removed from the subject and was analyzed immediately after collection. See Appendix A Materials and Methods for details.
Mitochondrial localization of TASK3: Transient expression of FLAG-tagged TASK3 in MCF10A cells was detected by immunofluorescence using anti-FLAG antibody (Sigma Aldrich) according to the manufacturer's instructions. The expression of FLAG-tagged TASK3 in the mitochondria was assessed by SDS-PAGE and western analysis of purified mitochondria (see below).
Isolation of mitochondria: High purity mitochondria were isolated using a Qproteome mitochondrial isolation kit, according to the manufacturer's instructions (Qiagen).
Human subjects and tissue collection: Tissue collection protocols were approved by the Human Subjects Committee and the Institutional Review Board at Duke University. Highrisk women, or women with cancer, were sequentially recruited by a research coordinator in order of presentation. All women provided informed consent to participate in this study. Demographic data were collected by patient interview. Definition of high-risk is described in detail in Appendix A Materials and Methods. All tissues underwent pathology review by a pathologist who was blinded to the results of this study.
RPFNA of breast epithelial cells and Masood Cytology Score: RPFNA was performed by our published methods [32,36]. Masood Cytology Score and epithelial cell count for duplicate RPFNA samples were assigned by a sample blinded, single dedicated cytopathologist as previously reported [36].
Statistical analysis: Assessment of the association of breast cancer subtype with methylation status was performed using Fisher's exact test. Stratified contingency tables and Fisher's exact tests were used to assess the association of breast cancer subtype with methylation status by race. Association was estimated for total hypomethylation, partial hypomethylation, normal, and hypermethylated status versus ER+, HER2+, and TNBC (and also categorized as hypomethylated total versus not, and TNBC yes versus not). The association of Masood Cytology Index Score and R/G ratio versus methylation status in high-risk women was completed using linear models with robust standard errors to control for the correlation of multiple measures per patient. Least squares means are reported with paired comparisons across methylation status levels as well as 95% confidence intervals of mean differences. Adjustment of confidence intervals and p-values for the two comparisons to normal methylation status was based on a Bonferroni adjustment [37]. Statistical analysis was completed in SAS v9.4 (SAS Institute, Cary, NC, USA) and in Stata v12 (StataCorp LP, College Station, TX, USA).

KCNK9 Exhibits Monoallelic Expression in Breast Tissue
Monoallelic expression of KCNK9 in normal mammary epithelial cells was tested using the single nucleotide polymorphism, rs2615374; this was the same polymorphism used to identify monoallelic expression in the brain [21]. Using matched DNA/RNA Cancers 2021, 13, 6031 5 of 27 samples extracted from adult breast tissue, monoallelic expression was observed in two samples heterozygous for rs2615374 ( Figure 1A). used to identify monoallelic expression in the brain [21]. Using matched DNA/RNA samples extracted from adult breast tissue, monoallelic expression was observed in two samples heterozygous for rs2615374 ( Figure 1A).   Figure 1B).

Identification of the KCNK9 Imprint DMR
Analysis of the canonical CpG island covering the KCNK9 promoter region, transcription factor binding sites, transcription start site, and first exon identified no allelespecific methylation (Table A1). Expanding the search by a bioinformatic approach, using ENCODE data and cross-species sequence conservation, we identified a KCNK9 upstream region (KCNK9-US1) that had characteristics of an imprint regulatory sequence ( Figure  1B). Approximately 20 kb upstream of the KCNK9 start site is a 500 bp region Boxed region-area of transcription factor binding from ENCODE data ( Figure 1B).

Identification of the KCNK9 Imprint DMR
Analysis of the canonical CpG island covering the KCNK9 promoter region, transcription factor binding sites, transcription start site, and first exon identified no allelespecific methylation (Table A1). Expanding the search by a bioinformatic approach, using ENCODE data and cross-species sequence conservation, we identified a KCNK9 upstream region (KCNK9-US1) that had characteristics of an imprint regulatory sequence ( Figure 1B). Approximately 20 kb upstream of the KCNK9 start site is a 500 bp region (chr8:140,732,400-140,732,870 GRCh37/hg19) containing areas of strong cross-species conservation, DNase hypersensitivity, and binding of multiple transcription factors [38][39][40]. Immediately adjacent to the conserved sequence is a cluster of CpG sites, but the sequence containing these dinucleotides has no cross-species conservation [39]. Quantitative methy-lation analysis by Sequenom MassArray identified intermediate methylation (20-70%) for five consecutive CpG sites within this adjacent region in brain, breast, liver, and testis from eight different adults (Table 1). These differentially methylated CpG sites are not within the ENCODE defined DNase hypersensitive region, nor the majority of the transcription factor binding regions [38]. Sequencing of PCR clones shows continuity of methylated or unmethylated cytosines for this region in DNA from the adult brain, with the adult breast showing a higher average methylation level ( Figure 1C). While the putative KCNK9-US1 DMR has no sequence similarity to mouse, human and mouse have comparable numbers of CpG sites and spacing of these sites [40]. Humans have 10 CpG sites in 385 bp with 53% GC content, for an observed:expected CpG ratio of 0.37. Mice have 12 CpG sites in 291 bp with a 51% GC content, for an observed:expected CpG ratio of 0.64. Clone sequencing of this region in mouse liver identified a DMR comparable to that observed in humans, with five CpG sites showing strong contiguous methylation (sites a-d, and f) ( Figure 1C).

Methylation of the KCNK9 DMR Regulates Chromatin Structure
The ENCODE data suggested open regulatory chromatin near the KCNK9-US1 DMR, based on observed DNase hypersensitivity and transcription factor binding ( Figure 1B). NOMe-seq was used to determine the relationship between CpG methylation and chromatin structure at the DNase hypersensitive and transcription factor binding sites [33,41]. Chromatin structure dependent in vitro methylation by bacterial GpC methyltransferase identified 10 consecutive GpC dinucleotides over~130 bp as consistently methylated or demethylated in the conceptus brain ( Figure 1D). Clone sequencing showed an inverse relationship between this region of differential open/closed chromatin and methylation at DMR CpG sites 3 and 4 ( Figure 1D). This strong correlation was observed in conceptus brain tissue, where KCNK9 mRNA expression is highest, with a lesser amount of open chromatin seen in conceptus kidney and liver (data not shown).

Methylation Status of the KCNK9 DMR Does Not Correlate with Age
We used bisulfite sequencing to determine (1) the consistency of KCNK9-US1 DMR methylation in individuals without breast cancer and (2) whether KCNK9-US1 DMR hypomethylation increased with increasing age. Genomic DNA from six women with no breast cancer history, ages 21 to 83 years, showed consistent methylation for each of the five CpG sites for five individuals, while one 81-year-old individual had marginal hypomethylation at sites 3 to 5 ( Figure 2).   Analysis of KCNK9 DMR methylation in invasive cancer biopsy specimens indicated that the most significant differences in methylation between brain and invasive cancer biopsies were for sites 3 to 5. Sites 3 to 5 were the most highly correlated to chromatin structure by NOMe-seq. NOMe-seq of three invasive cancers with DMR hypomethylation indicated open chromatin was linked to DMR hypomethylation ( Figure 3A). The open chromatin position in tumor cores (GRCh37/hg19 Chr8:~140732560-140732690) is shifted from that observed in the conceptus brain (Chr8:~140732450-140732630), but is within the ENCODE defined region containing transcription factor binding sites ( Figure 1B). Expression analysis by cDNA sequencing of rs2615374 in heterozygous core and epithelial samples showed expression of both alleles, i.e., LOI, in 36% (4/11) of the breast tumors informative for rs2615374 ( Figure 3B). As discussed below, methylation was variable in noncancerous breast tissue from high-risk women, with some individuals showing hypomethylation at sites 3-5, and others showing hypomethylation at sites 1-5 (for a representative analysis of two women, see Figure 3C).   Analysis of KCNK9 DMR methylation in invasive cancer biopsy specimens indicated that the most significant differences in methylation between brain and invasive cancer biopsies were for sites 3 to 5. Sites 3 to 5 were the most highly correlated to chromatin structure by NOMe-seq. NOMe-seq of three invasive cancers with DMR hypomethylation indicated open chromatin was linked to DMR hypomethylation ( Figure 3A). The open chromatin position in tumor cores (GRCh37/hg19 Chr8:~140732560-140732690) is shifted from that observed in the conceptus brain (Chr8:~140732450-140732630), but is within the ENCODE defined region containing transcription factor binding sites ( Figure 1B). Expression analysis by cDNA sequencing of rs2615374 in heterozygous core and epithelial samples showed expression of both alleles, i.e., LOI, in 36% (4/11) of the breast tumors informative for rs2615374 ( Figure 3B). As discussed below, methylation was variable in non-cancerous breast tissue from high-risk women, with some individuals showing hypomethylation at sites 3-5, and others showing hypomethylation at sites 1-5 (for a representative analysis of two women, see Figure 3C). with 0, 1.0, or 2.5 µM 5AzC there was, respectively, a 3.2 and 3.4-fold increase in TASK3 protein relative to T0 control ( Figure 4C,D; Figure 4D is a quantitation of results in Figure  4C, including triplicate blots not shown in this figure). There was a statistically significant increase in mitochondrial membrane potential as measured by JC1 R/G ratio after 24 and 48 h treatment of MCF10A cells with 1.0 and 2.5 µM 5AzC; similar results were observed for HMEC-15 strain treated with 1.0 and 2.5 µM 5AzC for 48 h ( Figure 4E).   with 2.5 µ M 5AzC for 48 h, there was a 3.2-fold increase in TASK3 protein; after treatment of HMEC-15 with 0, 1.0, or 2.5 µ M 5AzC there was, respectively, a 3.2 and 3.4-fold increase in TASK3 protein relative to T0 control ( Figure 4C,D; 4D is a quantitation of results in 4C, including triplicate blots not shown in this figure). There was a statistically significant increase in mitochondrial membrane potential as measured by JC1 R/G ratio after 24 and 48 h treatment of MCF10A cells with 1.0 and 2.5 µ M 5AzC; similar results were observed for HMEC-15 strain treated with 1.0 and 2.5 µ M 5AzC for 48 h ( Figure 4E).  DMR in MCF10A. (B) Cell proliferation (viability) was tested using the MTT assay at 0-72 h; MCF10A cells and HMEC-15 normal cell strain were treated with 0 (blue line), 1.0 (red line), and 2.5 µ M (green line) 5-AzC. Experiments were performed in triplicate. (C) Expression of TASK3 protein analyzed by western blotting at 0 and 48 h in MCF10A after treatment with 2.5 µ M 5AzC and HMEC-15 normal cell strain after treatment with 1.0 and 2.5 µ M 5AzC. GAPDH is used as the loading control. (D) Relative TASK3 expression in MCF10A after treatment with 2.5 µ M 5AzC for 0 and 48 h and HMEC-15 normal breast cell strain after treatment with 1.0 and 2.5 µ M 5AzC for 0 and 48 h. TASK3 expression was normalized to 0 h treatment (Control). Expression is for the western blots performed in triplicate for panel Figure 4C. Significant levels relative to 0 h control: *** p < 0.001. (E) JC-1 R/G ratio in MCF10A cells treated with 0, 1.0, and 2.5 µ M 5AzC for 0, 24, and 48 h normalized to 0 h treatment (Control). JC-1 R/G ratio in HMEC-15 normal cell strain was treated with 0, 1.0, and 2.5 µ M 5AzC for 0 and 48 h normalized to 0 h treatment (Control). Significant levels relative to 0 h control: * p < 0.05, ** p < 0.01, *** p < 0.001. Experiments were performed in triplicate. The uncropped blots and molecular weight markers are shown in Supplementary Materials.   Significant levels relative to control: * p < 0.05, ** p < 0.01. (E) Caspase-3 activity, a measure of apoptosis initiation, normalized to DNA content in untreated MCF10A cells (Control); cells treated with transfection reagent alone (Txf); cells transfected with empty FLAG tag plasmid (Vec); and cells transfected with TASK3-FLAG tag plasmid (TK3). Cells were exposed for 4 h to either 0.0 µg/mL (grey bar) or 0.50 µg/mL (black bar) of staurosporine. Significant levels relative to cells not exposed to staurosporine: * p < 0.05, ** p < 0.01. The uncropped blots and molecular weight markers are shown in Supplementary Materials. Transient transfection using a FLAG-tagged construct was used to overexpress TASK3 ( Figure 5B) in the TASK3-low expressing MCF10A cells ( Figure 5A). The overexpressed FLAG-tagged TASK3 localized to the mitochondria as demonstrated by western analysis ( Figure 5B). Immunohistochemistry demonstration of the FLAG-tagged construct is shown in Figure 5C.

Exogenous Expression of TASK3 Increases
We tested whether TASK3 expression had an impact on mitochondrial membrane potential (∆Ψ M ) and apoptosis-sensitivity using a knock-in/knock-out approach ( Figure 5D,E and Figure 6). Overexpression of wild-type TASK3 significantly increased ∆Ψ M relative to controls in the TASK3-low expressing cells, MCF10A (p = 0.0003), and SUM225 (p = 0.0037) ( Figure 6C). These observations are also consistent with ∆Ψ M after treatment of the TASK3low expressing MCF10A cells with 5-aza-2 -deoxycytidine ( Figure 4C).
The LXSN retroviral vector was used to (1) stably express the dominant-negative TASK3-G95E construct in the TASK3-expressing cells, HEK293, MDA231, and DKAT ( Figure 6A) and to (2) overexpress wild-type TASK3 in TASK3-low expressing cell line SUM225 ( Figure 6B). To ensure that our results were neither cell line-nor constructdependent, we also tested the FLAG-tagged TASK3 construct expressed in MCF10A cells above in Figure 5. The voltage sensitive dye JC-1 was used to test ∆Ψ M relative to TASK3 expression and function. In TASK3-G95E-expressing cells, ∆Ψ M was significantly decreased relative to controls in HEK293 (p = 0.0076) and MDA231 (p < 0.0001) ( Figure 6C). These results show that TASK3 expression regulates ∆Ψ M .
Apoptosis sensitivity was tested by treatment with staurosporine. Staurosporine is known to promote apoptosis and is used as an apoptosis-control. Expression of TASK3-G95E significantly increased apoptosis-sensitivity of HEK293 (p = 0.006), MDA231 (p = 0.0091), and DKAT (p = 0.0003) cells ( Figure 6D), while overexpression of TASK3 in MCF10A (p = 0.0016) and SUM225 (p = 0.0067) cells resulted in apoptosis-resistance relative to controls, as measured by caspase-3 activity, an indicator of the early stages of apoptosis ( Figures 5E and 6D). These results are consistent with a study by Nagy et al. who showed that, in melanoma, TASK3 regulated apoptosis and mitochondrial function [30].

High-Risk Sample Set
In Tables 2 and 3, we performed detailed methylation sequencing in biopsies obtained in a highly annotated group of high-risk women. Table 3 includes subjects from Table 2; subjects were only counted once. All women were followed by Dr. Victoria Seewaldt while at Duke University and were part of her high-risk cohort. High-risk was defined as (1) greater than 20% lifetime risk for breast cancer or (2) the presence of a deleterious germline mutation (e.g., mutation in BRCA1, BRCA2, etc.). is shown in Figure 5C.

Hypomethylation of the KCNK9 DMR Was Most Frequently Observed TNBC
We tested for hypomethylation at the KCNK9-US1 DMR in 53 primary human breast cancers ( Table 2). TNBCs had the highest frequency of DMR hypomethylation. There was not a significant association between breast cancer subtype and DMR methylation status, with hypomethylation present in 27% of HER2+ cancer, 39% of ER+ cancer, and 63% of TNBC (p = 0.11), with borderline significant association in TNBC, p = 0.06.

KCNK9 DMR Hypomethylation Is Observed in Our Dataset More Frequently in High-Risk African-American Women with TNBC
Hypomethylation at the KCNK9 DMR in TNBC tumors was (1) much more frequently observed in African-Americans with TNBC than in European-American with TNBC and 2) was highly significant in African-Americans, but not in European-Americans (Table 2).
We also attempted to test for hypomethylation of KCNK9 DMR imprinting in a methylation dataset from cBioportal.org, generated from Illumina 450 chip analysis of CpG sites (1500 bp upstream of the transition start site of KCNK9). Unfortunately, the KCNK9 imprint DMR lies significantly upstream of this region, and we were not able to test whether hypomethylation of KCNK9 DMR methylation occurred at a higher frequency in African-American women in this expanded dataset.

Hypomethylation of the KCNK9 DMR Is Observed in Both TNBC and Non-Cancerous Breast Tissue, but Not in WBCs Hypomethylation of the KCNK9 DMR Methylation Is Observed in Non-Cancerous Tissue
To determine the utility of KCNK9-US1 DMR methylation as an early breast cancer diagnostic, we tested whether hypomethylation was detectable prior to cancer diagnosis in high-risk women in morphologically normal breast tissue, or WBCs in women with invasive breast cancer. Hypomethylation of the KCNK9-US1 DMR methylation was assessed in matched WBCs and non-cancerous breast tissue from 15 high-risk women and 18 women with invasive cancer (Table 3). Of the 14 women for whom RPFNA DNA was available for both breasts, 14% (2/14) of the women had KCNK9-US1 DMR partial or full hypomethylation in both samples; 57% (8/14) showed partial or full DMR hypomethylation in one or both breast samples (Table 3). Methylation was variable, with some individuals showing hypomethylation at sites 3-5, and others showing hypomethylation at sites 1-5 ( Figure 3C). The observed hypomethylation of the KCNK9-US1 DMR in breast tissue, but not in WBC, is consistent with epigenetic alterations at the KCNK9 locus occurring after embryonic implantation and the establishment of three fetal germ layers (i.e., mesoderm, ectoderm, and endoderm).

KCNK9 DMR Hypomethylation Is Not Associated with the Degree of Cytologic Abnormality, but Is Associated with Increased Mitochondrial Membrane Potential
The relationships between KCNK9-US1 DMR methylation and ∆Ψ M (measured by JC-1 R/G ratio) or Masood Score (a measure of cytologic abnormality) [36] were investigated in aspirated human mammary epithelial cells. Quantitative DNA methylation analysis was performed and KCNK9-US1 DMR methylation was classified as methylation (75-125% methylation), partial hypomethylation (20-74%), or hypomethylation (0-20% methylation). Mammary epithelial cell cytology was classified by the Masood Score [36].
When samples were grouped by methylation status, KCNK9-US1 DMR methylation was not significantly associated with the Masood Score (p = 0.3) ( Figure 7A). In contrast, ∆Ψ M was significantly higher in mammary epithelial cells with a hypomethylated KCNK9-US1 DMR than in those where it was methylated (p < 0.001) ( Figure 7B). Thus, hypomethylation of the KCNK9-US1 DMR predicts ∆Ψ M independent of the degree of cytologic abnormality.

KCNK9 DMR Hypomethylation Is Not Associated with the Degree of Cytologic Abnormality, but Is Associated with Increased Mitochondrial Membrane Potential
The relationships between KCNK9-US1 DMR methylation and ΔΨM (measured by JC-1 R/G ratio) or Masood Score (a measure of cytologic abnormality) [36] were investigated in aspirated human mammary epithelial cells. Quantitative DNA methylation analysis was performed and KCNK9-US1 DMR methylation was classified as methylation (75-125% methylation), partial hypomethylation (20-74%), or hypomethylation (0-20% methylation). Mammary epithelial cell cytology was classified by the Masood Score [36].
When samples were grouped by methylation status, KCNK9-US1 DMR methylation was not significantly associated with the Masood Score (p = 0.3) ( Figure 7A). In contrast, ΔΨM was significantly higher in mammary epithelial cells with a hypomethylated KCNK9-US1 DMR than in those where it was methylated (p < 0.001) ( Figure 7B). Thus, hypomethylation of the KCNK9-US1 DMR predicts ΔΨM independent of the degree of cytologic abnormality.

Discussion
Loss of normal imprinting (1) occurs due to poor prenatal nutrition or exposure to heavy metals (e.g., cadmium, lead, arsenic) and (2) is linked to obesity, autism, and cancer [1,2,5,[8][9][10][11][12][13][14][15]. The KCNK9 gene-product TASK3 is a pH-regulated, potassium channel membrane protein that we, and others [29][30][31], show regulates mitochondrial membrane potential and apoptosis. This study provides the first demonstration that KCNK9 is imprinted and monoallelically expressed in mammary epithelial cells. It also identifies a DMR that likely regulates imprinting at this locus in human breast tissue. In addition to the DMR identification, we demonstrated, by NOMe-Seq, a region of differential chromatin structure related to the methylation status of the DMR. The relationship between DNA methylation and condensed chromatin structure is consistent with a model in which methylation at the DMR silences gene expression by impacting chromatin accessibility and preventing transcription factor binding. This model also supports an epigenetic mechanism for KCNK9/TASK3 overexpression in breast cancer.
Hypomethylation of the KCNK9 DMR was observed more frequently in African-American women with TNBC (p = 0.006) and less frequently in European-American women with TNBC (p = 0.70). Hypomethylation of the KCNK9 DMR was observed concurrently in TNBC and normal-appearing adjacent breast tissue. Abnormal KCNK9 imprinting was associated with increased mitochondrial membrane potential in live TNBC cells and non-cancerous mammary epithelial cells from high-risk women (p < 0.001). These results are consistent with a study by Nagy et al. who showed that, in melanoma, TASK3 regulates apoptosis and mitochondrial function [30]. The finding that KCNK9-US1 DMR hypomethylation occurs preferentially in African-Americans suggests that KCNK9/TASK3 may provide a new target for prevention of TNBC.
While 50% of Ashkenazi European women with TNBC have a germline mutation of BRCA1, only 20% of African-American women with TNBC have a BRCA1 mutation [42]. This indicates that other mechanism(s) beyond germline mutation of BRCA1 are responsible for the etiology of TNBC in African-American women. African-American women experience disparities in income, access to care, and an unequal burden of environmental exposures [43]. Given that imprinting is dysregulated by poor nutrition and environmental toxicants, our findings provide a potential mechanistic link between disparities and TNBC in African-American women who do not have germline BRCA1 mutations. A limitation of this study is that it was conducted in a single institution in a restricted number of women; multi-institutional testing with an expanded test set and validation set is required to validate KCNK9/TASK3 as a potential risk biomarker.
The hypomethylation at the KCNK9 DMR was observed in both non-cancerous and cancerous breast tissue, but it is rare in the WBC of at-risk individuals. These findings indicate that epigenetic alterations occurring at the KCNK9 locus do not typically form at the time of fertilization and implantation. In contrast, they are consistent with alterations occurring in later epigenetically vulnerable developmental windows. These windows include during tissue differentiation, early childhood, or puberty. Identification of hypomethylation of the KCNK9 DMR in both breasts of at-risk individuals is indicative of a relatively early developmental change with large spatial distribution. Hypomethylation seen in only one breast would indicate epigenetic alteration occurring later in development, in a more specific location or cell type. The epialleles identified here are quantifiable markers for association studies between the environmental factors and the critical exposure timing that contribute to breast cancer risk.
The DMR methylation and NOMe-Seq chromatin data provide intriguing targets for future studies to better understand the origins and progression of TNBC. This will include further investigation of DMR methylation, chromatin structure, transcription factor binding, and gene/protein expression in this aggressive form of breast cancer. As TNBC has rapid-onset, aggressive growth, and resistance to treatment, patient survival could be improved through better classification of risk status, early detection, and better treatment targets. Furthermore, since genomic imprinting is dysregulated by poor nutrition and exposure to environmental toxicants, the results here provide support for the importance of good nutrition and a healthy environment in the prevention of TNBC.
Other recent work into the epigenetic regulation of KCNK9 has identified long distance cis-interactions between the promoter CpG island of KCNK9, which was found to be hypomethylated, and the PEG13 DMR [44]. As mentioned previously, parent-of-origin specific methylation in the KCNK9 CpG island was not detected (Table A1), so any parentalspecific regulation by this interaction would be due solely to the PEG13 DMR. Thus, it will be important to determine how CpG methylation and chromatin structure in the KCNK9 promoter influence long-range interactions with PEG13. Furthermore, it will be of interest to determine whether interactions between the KCNK9-US1 DMR and the KCNK9 promoter are involved in gene regulation. Such increased understanding of the epigenetic regulation of KCNK9 will be of great value in determining the role that expression of this gene has in the development and progression of cancers and in developing new treatment methods.
Our studies highlight the importance of therapeutically targeting TASK3. While there has been a recent emphasis on precision medicine, it is unclear whether the benefits of precision medicine will impact women of color. Here we identify a target that has specific promise for African-American women. TASK3 can be targeted by natural products such as hydroxy-α-sanshool (active agent in Szechuan peppers) [45]. Most promising, a recent study showed that antibodies targeting the KCNK9 protein inhibited 410.4 cell breast cancer cell metastasis in the mouse [46]. The high frequency of abnormal KCNK9 imprinting in both TNBC and adjacent non-cancerous breast tissue provides evidence that KNCK9 may serve both as a target for precision therapy and chemoprevention of TNBC in African-American women.

Conclusions
TASK3 is overexpressed in >40% of breast cancers, but genomic amplification of KCNK9 is reported to occur in <10% of breast cancers [24], indicating that epigenetic mechanisms may play a key role in TASK3 overexpression. Herein, we show that KCNK9 is imprinted and monoallelically expressed in mammary epithelial cells. We identify (1) a DMR that likely regulates imprinting at this locus and (2) by NOMe-Seq a region of differential chromatin structure that is determined by methylation status of the DMR. The relationship between DNA methylation and condensed chromatin structure is consistent with a model in which methylation at the DMR silences gene expression by impacting chromatin accessibility and preventing transcription factor binding.
Less than 20% of African-American women with TNBC have a BRCA1 mutation [42]. This indicates that other mechanism(s) beyond germline mutation of BRCA1 may be responsible for the etiology of TNBC in African-American women. Given that imprinting is dysregulated by poor nutrition and environmental toxins, our findings provide a potential mechanistic link between disparities and TNBC in African-American women who do not have germline BRCA1 mutations. Hypomethylation at the DMR, coupled with biallelic expression of KCNK9, occurred in 75% of TNBC. The association between hypomethylation and TNBC status was highly significant in African-Americans (p = 0.006), but not in Caucasians (p = 0.70). KCNK9 hypomethylation was also found in non-cancerous tissue from women at high risk of developing breast cancer. The high frequency of abnormal KCNK9 imprinting in both TNBC and adjacent non-cancerous breast tissue provides evidence that KNCK9 has the potential to serve both as a target for precision therapy and chemoprevention of TNBC in African-American women.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Materials and Methods
Methylation Primers: Primers used for analysis of the KCNK9-US1 upstream methylation site were KCNK9-US1f, KCNK9-US1r, KCNK9-US2f, and KCNK9-US2r, which were used for DMR screening. Clone amplification and sequencing were performed using KCNK9-US0 and KCNK9-US4. Methylation analysis: CpG methylation was measured by analysis of bisulfite converted DNA. Sodium bisulfite modification of DNA was performed using EpiTect Bisulfite Kits (Qiagen) according to the manufacturer's instructions, converting unmethylated cytosines to uracil. Quantitative methylation for DMR identification was measured by Sequenom MassARRAY with EpiTYPER analysis software (Sequenom, San Diego, CA, USA). Sequenom EpiDesigner software was used for primer design, incorporating a T7 promoter site to forward primers, and tags to the reverse primers, to balance the annealing temperatures. PCR was performed using HotStarTaq (Qiagen). PCR products were processed using Sequenom MassCLEAVE reagents to deactivate unincorporated dNTPS, transcribe the amplicons from the T7 promoters, and cleave the transcripts with RNAse A. Incorporation of dTTP in the transcription reactions restricts RNase A cleavage to cytosines. The fragmented transcripts were spotted onto SpectroCHIPS for mass spectrometry analysis on the MassARRAY instrument. EpiTYPER software identified fragments based on predicted mass, calculating ratios for fragments of different mass resulting from conversion/non-conversion of cytosines at CpG sites, reporting this ratio as the methylation fraction.
Semiquantitative methylation analysis in screening patient samples for methylation changes was completed using dye terminator sequencing (BigDye, Applied Biosystems, ThermoFisher Scientific, Waltham, MA, United States) on amplicons of the DMR region. For CpG sites 3 and 4, signal intensities of the overlapping C and T peaks were compared to classify the methylation level: C intensity at 75-125% of the T signal was designated "full methylation" (+), C intensity ranging from undetectable to 20% of the T signal was designated "full hypomethylated" (−), peak intensity from 20-74% was designated "partial hypomethylation" (+/−), and peak intensity greater than 125% was designated "hypermethylated" (++).
Contiguous and allele specific methylation were determined by sequencing the clone's bisulfite DNA PCR products. Amplicons were cloned using a TOPO-TA cloning kit, with the pCR4 sequencing vector and Top 10 chemically competent cells (Life Technologies). Single transformed colonies were chosen and the inserts amplified and sequenced by BigDye 3.1 sequencing reagents (Applied Biosystems).
DNA extraction: Total DNA was isolated from RPFNA samples and enzymatically dispersed normal mammary epithelial cell samples by washing and resuspending the cells in 500 µL 10 mM Tris (pH 7.4), 10 mM NaCl, 3 mM MgCl 2 , and 0.1 mM EDTA [47,48]. The cells were then incubated overnight with 50 µL proteinase K (Qiagen) at 55 • C, followed by phenol-chloroform extraction and ammonium acetate/isopropanol precipitation of the DNA with glycoblue coprecipitant (Life Technologies).
Total DNA was isolated from conceptus tissues using buffer ATL, proteinase K, and RNase A (Qiagen), followed by phenol-chloroform extraction and ethanol precipitation.
NOMe-Seq: For in vitro methylation, tissues were homogenized in a nuclei homogenization buffer (approximately 50 mg tissue in 500 µL 10 mM Tris (pH 7.4), 10 mM NaCl, 3 mM MgCl 2 , and 0.1 mM EDTA) plus protease inhibitors (Protease Inhibitor Cocktail and PMSF, Sigma-Aldrich). Samples were kept on ice for 10 min, Igepal (Sigma-Aldrich) was added to 1%, samples were vortexed, and nuclei collected by centrifugation. Nuclei were then resuspended in 500 µL M.cviPI reaction buffer and incubated with 100 U M.cviPI at 37 • C for 15 min (New England Biolabs, Ipswich, MA, USA). Then, 50 µL proteinase K (Qiagen) was added and samples incubated overnight at 55 • C. Samples were extracted with phenol:chloroform:isoamyl alcohol and DNA precipitated with ammonium acetate and isopropanol.
Monoallelic expression analysis: Determination of mono-and biallelic gene expression was performed by cDNA sequencing. Commercially obtained breast genomic DNA (Biochain, Newark, CA, USA) was obtained for analysis. Tissues were genotyped at rs2615374 using Applied Biosystems components and equipment, according to the manufacturer's standard protocols; TaqMan Made to Order Assay C_2776566_20, TaqMan Genotyping Master Mix, and amplification and analysis were performed on an Applied Biosystems 7900HT Fast Real-Time PCR System (Applied Biosystems). Informative samples heterozygous at rs2615374 were selected for expression analysis. Commercially purified, matched DNAs and RNAs from normal breast tissue were purchased from Biochain (Hayward, CA, USA). RNAs from tissues were purified using RNA-Stat 60 (Tel-Test, Inc., Friendswood, TX, USA), according to the manufacturer's instructions. Reverse transcription was performed using Quantitect-RT kits (Qiagen), according to the manufacturer's modified protocol for gene specific primers, using primer TM1R at a final concentration of 0.7 µM. TM1R is based on M1R from Luedi et al. [21], using the M1R transcript specific sequence with the addition of a non-genomic sequence as an amplification tag. Using primers specific to this non-genomic tag controls against amplification of contaminating genomic DNA or aberrantly primed anti-sense RNA.
KCNK9 cDNA was amplified from brain tissue with a single round of amplification using TM1R and M1F primers. Amplification from breast tissue required two rounds of amplification with nested primers, using the TwistDX recombinase amplification kit (TwistDX, Ltd., Cambridge, UK). The first round used primers TM1R and KCDX-1F primers, and the second, primers KCDX-3F and KCDX-3R. All primer pairs used span the~85 kb single intron of KCNK9 ( Figure 1B).
Sequencing of the amplicons was performed using ABI BigDye 3.1 (Applied Biosystems), primers M1R and M1F for brain cDNA, and primers KCDX-3R and KC-S2 for breast cDNA.
Cell Lines: Cell lines were obtained from either the American Type Culture Collection (Manassas, VA, USA) or the Duke Cell Culture Facility (Durham, NC, USA). Cells were validated by short tandem repeat DNA profile analysis and used within 15 passages. MDA231 and HEK293 cells were grown in Dulbecco's modified eagle medium (DMEM) and the media was supplemented with 10% fetal bovine serum. SUM149 and SUM225 cells were grown in F12 media supplemented with 5% fetal bovine serum, 10 mM HEPES, 1 µg/mL of hydrocortisone (Sigma-Aldrich), and 5 µg/mL of human insulin. SUM190 cells were grown in the same media without fetal bovine serum. HMEC15 cells (Lonza, Walkersville, MD, USA) were immortalized with human telomerase. The hTERT immortalized HMEC15 and DKAT cells were grown in HuMEC media supplemented with bovine pituitary extract and HuMEC supplement. MCF10A cells were grown in F12/DMEM media supplemented with 5% horse serum, 20 ng/mL EGF, 10 µg/mL insulin, 0.5 µg/mL hydrocortisone, and 100 ng/mL cholera toxin (Sigma-Aldrich). All media and supplements were from GIBCO (Life Technologies) unless otherwise noted. All cell lines were tested for mycoplasma prior to the initiation of these studies.
SDS-PAGE and western analysis: SDS-PAGE was performed using Novex precast 4-12% Bis-TRIS gels and the MES running buffer system (Life Technologies). Separated proteins were transferred to a PVDF membrane (BioRad, Hercules, CA, USA), the membrane blocked with 5% BSA in 0.1% TTBS and then incubated with the primary antibody indicated in blocking solution overnight at 4 • C. The membrane was then washed with 0.1% TTBS, incubated for 1 h with secondary antibody, rewashed with 0.1% TTBS, and imaged with a Kodak IS 2000MM (Carestream, Norwalk, CT, USA) using Pierce SuperSignal Dura reagent (Thermo Fisher).
KCNK9 mRNA quantitation: The HMEC15 cell line was seeded at a density of 4.2 × 10 4 cells/well in a six-well plate with complete media. After 24 h, cells were treated with 1.0 or 2.5 µM of 5-aza-2 -deoxycytidine or vehicle for up to 72 h. After treatment, cells were harvested for real-time qPCR analysis. Total RNA was isolated using the RNeasy Mini Kit (Qiagen) and included an on column DNase digestion treatment according to the manufacturer's protocol. Total RNA was converted to cDNA using the qPCRBIO cDNA synthesis kit (PCR Biosystems, Wayne, PA, USA) according to the manufacturer's protocol. QPCR was performed using the Apex qPCR Green Master Mix (Apex Bioresearch) on the QuantStudio 5 Real-Time PCR system (Applied Biosystems). Experiments were performed three times. MTT assay: Cells were seeded at a density of 2.5 × 10 3 cells/well (five wells/treatment) in a 96-well plate with complete media. After 24 h, cells were treated with 1.0 or 2.5 µM of 5-aza-2 -deoxycytidine (Sigma-Aldrich) or vehicle control for up to 72 h with fresh medium added after 24 h. After treatment, cell supernatant was replaced with MTT (.5 mg/mL, Invitrogen) and incubated for an additional 4 h at 37 • C. MTT was replaced with DMSO and incubated for 20 min at room temperature on a shaker. Absorbance was measured at 560 nm and 750 nm using a microplate reader. Experiments were performed three times.
Apoptosis Assays: Cells were plated on day one and allowed to attach and resume growth. On day zero, cells were treated with staurosporine (0.25 µg/mL or 0.50 µg/mL) (Sigma-Aldrich), or an equal volume of dimethylsulfoxide (DMSO), and incubated for 4 h at 37 • C. Subsequently, the media was removed, and the cells were incubated for 30 min with 300 µL lysis buffer (AnaSpec, Fremont, CA, USA). Cells were then scraped and the supernatant collected and centrifuged at 2500× g for 10 min at 4 • C. The supernatant was assayed for caspase-3 activity by mixing 170 µL of supernatant with 57 µL of caspase-3 substrate solution (AnaSpec), and then incubating the mixture for 1 h at room temperature with shaking. Following incubation, fluorescence was measured with a Shimadzu 1501 fluorimeter (Shimadzu USA, Columbia, MD, USA) at λ ex = 490 nm and λ em = 520 nm with a 20 nm bandwidth. Experiments were performed in triplicate.
∆Ψ M analysis of cell lines and live RPFNA cytology: The relationships between KCNK9-US1 DMR methylation and ∆Ψ M (measured by JC-1 R/G ratio) or Masood Score (a measure of cytologic abnormality) [36] were investigated in aspirated human TNBC cells from seven women and non-cancerous mammary epithelial cells from ten high-risk women. Seven women, including TNBC and high-risk women, were BRCA1 mutation carriers; bilateral breast tissue was available from 59% (10/17) of these women (Table A2). Quantitative DNA methylation analysis was performed and KCNK9-US1 DMR methylation was classified as methylation (75-125% methylation), partial hypomethylation (20-74%), or hypomethylation (0-20% methylation). Mammary epithelial cell cytology was classified by the Masood Score [36]. One-way ANOVA was used to determine the association between KCNK9-US1 DMR methylation status, and the continuous variables of ∆Ψ M and Masood Score.
On day -1, cells were seeded in a black 96 well microtiter plate at 20,000 cells per well in triplicate. On day 0, the media was removed and replaced with media containing 0.5 ug/mL JC-1 (ThermoFisher), and then the cells were incubated in the dark for 30 min at 37 • C. Next, the JC-1 containing media was removed, the cells were washed three times with warm PBS, and 200 mL of warm PBS was added to each well. Epithelial cells were identified by morphology. Red (λ ex = 525 nm, λ em = 580-640 nm) and green (λ ex = 490 nm, λ em = 510-570 nm) fluorescence were measured, and the ratio of red to green fluorescence was determined as a measure of ∆Ψ M .
Human subjects and tissue collection: Informed consent was obtained from women at high risk for breast cancer who were scheduled for a breast surgical procedure in the operating room. RPFNA samples were obtained from the patient and transferred from the operating room to the laboratory within 5 min and immediately processed for analysis. RPFNA cytology samples were washed with phosphate buffered saline (PBS) and resuspended in 1 mL DMEM. Then, they were incubated with either JC-1 (final concentration: 1 µg/mL) or vehicle control at 37 • C for 30 min. After incubation, the cells were centrifuged, and the staining solutions were decanted. Each sample was washed twice with warm PBS and resuspended in 200 µL of warm PBS. The cell suspension was transferred to a black flat-bottom 96-well plate and measured as above. Epithelial cells were identified by morphology. DAPI (4 ,6-diamidino-2-phenylindole) was used as a control and its fluorescence was measured (λ ex = 365 nm, λ em = 410-460 nm).
High risk was defined by (1) known BRCA mutation carrier, (2) a Gail Model Risk of >1.66% over five years, or (3) a strong family history of breast and/or ovarian cancer. A strong family history of breast and/or ovarian cancer risk was defined as having at least one of the following: (a) two or more first-degree relatives with breast cancer before the age of 50 years, (b) one first degree relative with bilateral breast cancer or ovarian cancer, (c) two or more first-degree relatives with breast cancer, (d) one first degree relative and two or more second or third degree relatives with breast cancer, (e) one first-degree relative with breast cancer and one or more relatives with ovarian cancer, (f) two second or third degree relatives with either breast cancer and one or more with ovarian cancer, (g) one second or third degree relative with breast cancer and two or more with ovarian cancer, or (h) three or more second or third degree relatives with breast cancer.
To control for hormonal effects on mammary cell proliferation, menstruating women were aspirated between days 1 and 12 of their cycle. The breast was anesthetized with 5 mL of 1% lidocaine, immediately adjacent to the areola, at~3 and 9 o'clock positions. Eight to ten aspirations were performed per breast for random sampling of epithelial cells. After the aspiration, cold packs were applied to the breasts for 10 min, and both breasts were bound in kerlex gauze for 12 to 24 h. Epithelial cells were pooled and placed in modified CytoLyt (Cytyc Co., Boxborough, MA, USA) with 1% formalin for 24 h. Cells from the right and left breast were processed independently, so as to obtain one specimen per aspirated breast. Epithelial cells were split into two samples, with half designated for cytology and half designated for DNA extraction.
Cytology preparations were also given a semiquantitative index score through evaluation by the Masood cytology index. As previously described, cells were given a score of one to four points for each of six morphologic characteristics that include cell arrangement, pleomorphism, number of myoepithelial cells, anisonucleosis, nucleoli, and chromatin clumping.  Methylation levels were measured by Sequenom MassArray with a possible value range from 0.0 to 1.0 (i.e., hypomethylated to hypermethylated); primers and amplicon positions are listed (GRCh37/hg19). Measurements were made in conceptus kidney (C Kd), liver (C Lv), brain (C Br) and adult breast (Ad Brst), liver (Ad Lv), and brain (Ad Br) tissues. The number of CpG sites listed per amplicon were those readable by MassArray analysis, with mean, standard deviation, median, and range of methylation given for each tissue. (ND) Not Determined.